Details

Parent Category: Products

Category: Digital-to-Analog Converter Measurements

Created: 01 December 2023

Link: reviewed by Matt Bonaccio on SoundStage! Hi-Fi on February 15, 2024

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The Ferrum Audio Wandla was conditioned for 30 minutes at 0dBFS/4Vrms out (2Vrms unbalanced output) into 200k ohms before any measurements were taken.

The Wandla offers one analog line-level input (RCA) and several digital inputs (S/PDIF over coaxial RCA and optical, AES-EBU, I2S over HDMI, and asynchronous over USB). There are both unbalanced (RCA) and balanced (XLR) outputs. For the purposes of these measurements, unless otherwise stated, the following inputs were evaluated using the balanced XLR outputs: digital coaxial S/PDIF (RCA), analog unbalanced (RCA). Comparisons were made between unbalanced (RCA) and balanced (XLR) line outputs, and lower THD and an extra 6dB of gain were seen using the XLR outputs (FFTs for different configurations can be seen in this report).

Unless noted, the measurements were performed with the stock power supply. Where noted, additional measurements were performed with the optional Ferrum Audio Hypsos power supply.

The Wandla offers two different types of volume control: analog and digital. Unless otherwise stated, the analog volume control was used. Using digital inputs, at the maximum volume, the Wandla can output just over 10Vrms (XLR).

There are also a few options available through the user menu. Unless otherwise stated the following was used by default:

• Filter: HQ Gauss (also available: HQ Apod, HQ Apod MP, HQ short, ESS Lin-Ph)
• Bypass: off
• Theatre bypass: off
• Output balance: C
• Digital input trim: 0dB
• Analog input gain: 0dB (-12dB to 12dB available)

Most measurements were made with a 2Vrms line-level and 0dBFS digital input with the volume set to achieve 4Vrms at the balanced output (100 for analog, 92 for digital).  

Based on the accuracy and random results of the left/right volume channel matching (see table below), the Wandla analog volume control is likely digitally controlled in the analog domain. The Wandla offers 100 volume steps from -93dB to 5.8dB for the line-level analog inputs and balanced outputs. For a 0dBFS digital input, the XLR outputs ranged from 102uVrms (volume 1) to 10.2Vrms (volume 100). All steps are in 1.0dB increments.

Volume-control accuracy (measured at preamp outputs): left-right channel tracking

Published specifications vs. our primary measurements

The table below summarizes the measurements published by Ferrum Audio for the Wandla compared directly against our own. The published specifications are sourced from Ferrum’s website, either directly or from the manual available for download, or a combination thereof. With the exception of frequency response, where the Audio Precision bandwidth was set at its maximum (DC to 1MHz), assume, unless otherwise stated, 4Vrms output (XLR) into 200k ohms and a measurement input bandwidth of 10Hz to 22.4kHz, and the worst-case measured result between the left and right channels.

Our primary measurements revealed the following using the balanced line-level analog input and digital coaxial input (unless specified, assume a 1kHz sinewave at 2Vrms or 0dBFS, 4Vrms output (XLR), 200kohm loading, 10Hz to 22.4kHz bandwidth):

Frequency response (line-level input)

In our measured frequency-response (relative to 1kHz) plot above, the Wandla is perfectly flat within the audioband (0dB at 20Hz and 20kHz). At the extremes, the Wandla is 0dB at 5Hz and 0dB at 205kHz (the maximum allowable frequency by the signal generator). Deviations between 5Hz and 200kHz are within +/-0.05dB. These data corroborate Ferrum’s claim of 10Hz to 200kHz (+/-0.1dB). The Wandla appears to be DC-coupled, as there is no attenuation at low frequencies, even at 5Hz. It is also appropriate to say that the Wandla is a very high-bandwidth audio device. In the graph above and most of the graphs below, only a single trace may be visible. This is because the left channel (blue or purple trace) is performing identically to the right channel (red or green trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Phase response (line-level input)

Above is the phase response plot from 20Hz to 20kHz for the balanced line=level input. The Wandla does not invert polarity and exhibits essentially no phase shift within the audioband.

Frequency response vs. input type

The chart above shows the Wandla’s frequency response (relative to 1kHz) as a function of input type. The dark green trace is the same (but limited to 80kHz) analog input data from the previous graph. The blue trace is for a 16-bit/44.1kHz dithered digital input signal from 5Hz to 22kHz using the coaxial input, the purple trace is for a 24/96 dithered digital input signal from 5Hz to 48kHz, and finally pink is 24/192 from 5Hz to 96kHz. The behavior at low frequencies is the same for all the digital sample rates, as well as the analog input - flat down to 5Hz. The behavior at high frequencies for all three digital sample rates is as expected, offering filtering around 22k, 48k, and 96kHz (half the respective sample rate). The 44.1kHz sampled input signal exhibits typical “brickwall”-type behavior, with a -3dB point at 20.3kHz. The -3dB point for the 96kHz sampled data is at 44kHz, and 67kHz for the 192kHz sampled data.

Frequency response vs. filter type (16/44.1; HQ Gauss, HQ Apod, HQ Apod MP)

The chart above shows the frequency response (relative to 1kHz) as a function of filter type for a 16/44.1 0dBFS digital input. The blue trace is for the HQ Gauss filter, red for HQ Apod filter, and green for HQ Apod MP filter. All three behaved the same, with brickwall-type behavior and a -3dB point at 20.3kHz.

Frequency response vs. filter type (16/44.1; HQ short, ESS Lin-Ph)

The chart above shows the frequency response (relative to 1kHz) as a function of filter type for a 16/44.1 0dBFS digital input. The blue trace is for the HQ Short filter, red for ESS Lin-Ph filter. The HQ Short filter exhibits softer attenuation around the corner frequency, with a -3dB point just shy of 20kHz, while the ESS Lin-Ph filter exhibits more brickwall-type behavior with a -3dB point of 21kHz.

Phase response vs. filter type (16/44.1; HQ Gauss, HQ Apod, HQ Apod MP)

Above is the phase response plot from 20Hz to 20kHz as a function of filter type for a 16/44.1 0dBFS digital input. The blue trace is for the HQ Gauss filter, red for HQ Apod, and green for HQ Apod MP. Both the HQ Gauss and HQ Apod filters exhibit essentially no phase shift within the audioband. The HQ Apod MP filter is at -180 degrees at 17kHz.

Phase response vs. filter type (16/44.1; HQ short, ESS Lin-Ph)

Above is the phase-response plot from 20Hz to 20kHz as a function of filter type for a 16/44.1 0dBFS digital input. The blue trace is for the HQ Short filter, red for ESS Lin-Ph. Both the HQ Short and ESS Lin-Ph filters exhibit essentially no phase shift within the audioband.

Digital linearity (16/44.1 and 24/96 data)

The chart above shows the results of a linearity test for the coaxial digital input (the optical input performed identically) for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the balanced outputs of the Wandla. The digital input is swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output is analyzed by the APx555. The ideal response would be a straight flat line at 0dB. At -120dBFS, the 16/44.1 data overshot by only 1-2dB, while the 24/96 remained perfect. To verify how well the 24/96 data would perform down to -140dBFS, we extended the sweep in the chart below.

Here we can see that the 24/96 data only overshot the mark by +4.5/+2dB (left/right) at -140dBFS. This is an exceptional digital linearity test result.

Impulse response vs. filter type (24/44.1; HQ Gauss, HQ Apod, HQ Apod MP) (24/48 data)

The graph above shows the impulse response for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period then back to digital silence, measured at the balanced outputs of the Wandla. The blue trace is for the HQ Gauss filter, red for HQ Apod and green for HQ Apod MP. The HQ Gauss and HQ Apod filters behave like typical symmetrical sinc functions. The HQ Apod MP filter shows no pre-ringing but long post-ringing.

Impulse response vs. filter type (16/44.1; HQ short, ESS Lin-Ph) (24/48 data)

The graph above shows the impulse response for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period then back to digital silence, measured at the balanced outputs of the Wandla. The blue trace is for the HQ Short filter, red for ESS Lin-Ph. Both filters behave like typical symmetrical sinc functions, but with less pre-/post-ringing compared to the HQ Gauss and HQ Apod filters above.

J-Test (coaxial input)

The chart above shows the results of the “-Test test for the coaxial digital input measured at the line-level balanced output of the Wandla. J-Test was developed by Julian Dunn the 1990s. It is a test signal—specifically, a -3dBFS undithered 12kHz squarewave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz squarewave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sinewave (the main peak). In addition, an undithered 250Hz squarewave at -144dBFS is mixed with the signal. This test file causes the 22 least significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc). Note that the alternating peaks are in the test file itself, but at levels of -144dBrA and below.  The test file can also be used in conjunction with artificially injected sinewave jitter by the Audio Precision, to show how well the DAC rejects jitter.

The coaxial S/PDIF input of the Wandla shows essentially a perfect J-Test result, with no peaks (other than signal peaks: 12kHz and 250/750Hz) visible above the -160dBrA noise floor.

J-Test (optical input)

The chart above shows the results of the J-Test test for the optical digital input measured at the balanced outputs of the Wandla. The results are the same as the coaxial input.

J-Test (AES-EBU input)

The chart above shows the results of the J-Test test for the AES-EBU digital input measured at the balanced outputs of the Wandla. The results are the same as the coaxial and optical inputs.

J-Test (coaxial, 2kHz sinewave jitter at 100ns)

The chart above shows the results of the J-Test test for the coaxial digital input measured at the line-level output of the Wandla, with an additional 100ns of 2kHz sinewave jitter injected by the APx555. The results are as the pristine J-Test would predict—visible but highly suppressed (-140dBrA) peaks at the 10kHz and 14kHz positions.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (HQ Gauss filter, coaxial input)

The chart above shows a fast Fourier transform (FFT) of the Wandla’s balanced outputs with white noise at -4dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1, using the HQ Gauss filter. The steep roll-off around 20kHz in the white noise spectrum shows the use of a brick-wall type reconstruction filter. There are no obvious aliased images within the audio band. The primary aliasing signal at 25kHz is also completely suppressed, while the second and third distortion harmonics (38.2, 57.3kHz) of the 19.1kHz are below -110dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (HQ Apod MP filter, coaxial input)

The chart above shows a fast Fourier transform (FFT) of the Wandla’s balanced outputs with white noise at -4dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1, using the HQ Apod filter. We see essentially the same FFT as with the HQ Gauss filter above.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (HQ Apod MP filter, coaxial input)

The chart above shows a fast Fourier transform (FFT) of the Wandla’s balanced outputs with white noise at -4 dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1, using the HQ Apod MP filter. We see essentially the same FFT as with the HQ Gauss and HQ Apod MP filters above.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (HQ Short filter, coaxial input)

The chart above shows a fast Fourier transform (FFT) of the Wandla’s balanced outputs with white noise at -4dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1, using the HQ Short filter. Here the roll-off around 20kHz in the white-noise spectrum is shallower compared to the first three filters.  There is also a little bit of attenuation in the main peak at 19.1kHz, not quite reaching 0dBrA. There are no obvious aliased images within the audioband. The primary aliasing signal at 25kHz is also completely suppressed, while the second and third distortion harmonics (38.2, 57.3kHz) of the 19.1kHz are below -110dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (ESS Lin-Ph filter, coaxial input)

The chart above shows a fast Fourier transform (FFT) of the Wandla’s balanced outputs with white noise at -4 dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1, using the ESS Lin-Ph filter. We see essentially the same FFT as with the HQ Short filter above.

THD ratio (unweighted) vs. frequency vs. load (analog)

The chart above shows THD ratios at the balanced line-level output into 200k ohms (blue/red) and 600 ohms (purple/green) for an analog 2Vrms input as a function of frequency. The 200k ohms and 600 ohms data are close but not identical. THD ratios are very low, from 0.0001% from 20Hz to 1kHz for the 200k ohm load, then rising up to 0.0004% at 20kHz. Into 600 ohms, between 20Hz and 1kHz, THD ratios were roughly 5dB higher.

THD ratio (unweighted) vs. frequency vs. load (digital 24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms (blue/red) and 600 ohms (purple/green) for a 24/96 0dBFS input as a function of frequency. The 200k ohms and 600 ohms data are close but not identical. THD ratios are extraordinarily low, from 0.00006% to 0.0001% from 20Hz to 1kHz for the 200k ohm load, then rising up to 0.0002% at 20kHz. Into 600 ohms, between 20Hz and 2kHz, THD ratios were roughly 5dB higher.

THD ratio (unweighted) vs. frequency vs. load (digital 24/96, Hypsos power supply)

The chart above shows THD ratios at the balanced line-level output into 200k ohms (blue/red) and 600 ohms (purple/green) for a 24/96 0dBFS input as a function of frequency. This time the Hypsos power supply was used. The pink/orange traces are with the Hypsos using the maximum 30V output, while purple/green is the standard 24V, both into 600 ohms. We find that using the upgraded power supply had no effect on THD ratios into 600 ohms.

THD ratio (unweighted) vs. frequency vs. sample rate (16/44.1 and 24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms for a 16/44.1 (blue/red) dithered 1kHz signal at the coaxial input and a 24/96 (purple/green) signal, as a function of frequency. The 16/44.1 THD ratios were higher, although this is due to the higher 16-bit noise floor and limited averaging times for each measurement. The long FFTs (below) show that 16/44.1 and 24/96 data yield the same THD results. THD ratios are extraordinarily low for the 24/96 data, from 0.00006% to 0.0001% from 20Hz to 1kHz for the 200k ohm load, then rising up to 0.0002% at 20kHz.

THD ratio (unweighted) vs. output (analog)

The chart above shows THD ratios measured at the balanced outputs of the Wandla as a function of output voltage for the unbalanced line-level input. THD values start at 0.1% at 1mVrms, down to a low of just below 0.0001% at 3Vrms, then a steep rise past 5Vrms to the 1% THD mark at 20Vrms.

THD+N ratio (unweighted) vs. output (analog)

The chart above shows THD+N ratios measured at the balanced outputs of the Wandla as a function of output voltage for the unbalanced line-level input. THD+N values start at 1% at 1mVrms, down to a low of just below 0.0003% at 5-7Vrms, then a steep rise past 5Vrms to the 1% THD mark at 20Vrms.

THD ratio (unweighted) vs. output (16/44.1 and 24/96)

The chart above shows THD ratios measured at the balanced outputs of the Wandla as a function of output voltage for the digital coaxial S/PDIF input, swept from -90dBFS to 0dBFS, with the volume control at maximum. Blue/red traces are for 16/44.1 data, and purple/green for 24/96. For the 16/44.1 data, THD values start at 3%, and predictably, reach their low at the maximum output voltage of about 10Vrms, at 0.0003%. For the 24/96 data, THD ratios ranged from 0.2% down to 0.0001% at the maximum output voltage. The 16/44.1 THD ratios were higher, although this is due to the higher 16-bit noise floor and limited averaging times for each measurement. The long FFTs (below) show that 16/44.1 and 24/96 data yield the same THD results.

THD+N ratio (unweighted) vs. output (16/44.1 and 24/96)

The chart above shows THD+N ratios measured at the balanced outputs of the Wandla as a function of output voltage for the digital coaxial S/PDIF input, swept from -90dBFS to 0dBFS, with the volume control at maximum. Blue/red traces are for 16/44.1 data, and purple/green for 24/96. For the 16/44.1 data, THD+N values start at 20%, and predictably, reach their low at the maximum output voltage of about 10Vrms, at 0.002%. For the 24/96 data, THD ratios ranged from 1.5% down to 0.00015% at the maximum output voltage.

Intermodulation distortion vs. generator level (SMPTE, 60Hz:4kHz, 4:1, 16/44.1, 24/96)

The chart above shows intermodulation distortion (IMD) ratios measured at balanced output for 16/44.1 (blue/red) input data and 24/96 input data (purple/green), from -60dBFS to 0dBFS. Here, the SMPTE IMD method was used, where the primary frequency (F1 = 60Hz) and the secondary frequency (F2 = 7kHz) are mixed at a ratio of 4:1. The SMPTE IMD analysis results consider the second (F2 ± F1) through the fifth (F2 ± 4xF1) modulation products. The 16/44.1 data yields IMD ratios from 2% down to 0.003% at 0dBFS. The 24/96 data yields IMD ratios from 0.2% down to 0.0004% near 0dBFS.

FFT spectrum – 1kHz (XLR output,  line-level input)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced outputs for the unbalanced line-level input. We see that the signal’s second harmonic, at 2kHz, is just below -120dBrA or 0.0001%, and -130dBrA, or 0.00003%, at the third (3kHz) harmonic. The subsequent signal harmonics are around the very low -140dBrA, or 0.00001%, level. Below 1kHz, we see a hint of a peak at 60Hz, but at -150dBrA, or 0.000003%.

FFT spectrum – 1kHz (RCA output, line-level input)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the unbalanced outputs for the unbalanced line-level input. We see that the main difference is that the RCA outputs yield higher THD compared to the XLR outputs, with the second (2kHz) harmonic at the -110dBrA, or 0.0003%, level.

FFT spectrum – 1kHz (digital input, 16/44.1 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced outputs for the coaxial digital input, sampled at 16/44.1. We see that both the second (2kHz) and third (3kHz) signal harmonics are just above the -130dBrA, or 0.00003%, level. The noise floor is much higher due to the 16-bit depth limitation.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced outputs for the coaxial digital input, sampled at 24/96. We see that both the second (2kHz) and third (3kHz) signal harmonics are just above the -130dBrA, or 0.00003%, level. With the lower noise floor due to the 24-bit depth, we can see higher-order signal harmonics just above and below the -140dBrA, or 0.00001%, level. There as zero noise-related peaks to be seen above the -155dBrA noise floor.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS, digital volume control)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced outputs for the coaxial digital input, sampled at 24/96, but using the digital volume control (set to the same value of 92 to achieve 4Vrms at the output). Although this FFT is still relatively clean and similar to the FFT above where the analog volume control was used, it’s not as clean. Signal harmonics here reach almost -120dBrA, or 0.0001%, at 3/5/7kHz. There is no difference in terms of power-supply-related noise.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS, RCA output)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the unbalanced outputs for the coaxial digital input, sampled at 24/96. We a similar FFT compared to the 24/96 balanced output FFT, but for a higher second (2kHz) signal harmonic peak here at almost -110dBrA, or 0.0003%.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS, Hypsos power supply)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced outputs for the coaxial digital input, sampled at 24/96, but with the external Hypsos power-supply. We can see that this FFT is essentially identical to the 24/96 FFT above using the stock power-supply and balanced output.

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is the FFT for a 1kHz -90dBFS dithered 16/44.1 input sinewave stimulus at the coaxial digital input, measured at the balanced outputs. We see the 1kHz primary signal peak, at the correct amplitude, and no signal or noise related harmonic peaks above the -140dBrA noise floor.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is the FFT for a 1kHz -90dBFS dithered 24/96 input sinewave stimulus at the coaxial digital input, measured at the balanced outputs. We see the 1kHz primary signal peak, at the correct amplitude, and no signal or noise related peaks above the -155dBrA noise floor.

FFT spectrum – 50Hz (line-level input)

Shown above is the FFT for a 50Hz input sinewave stimulus measured at the balanced outputs for the unbalanced line-level input. The X axis is zoomed in from 40 Hz to 1kHz, so that peaks from noise artifacts can be directly compared against peaks from the harmonics of the signal. The most predominant (non-signal) peaks are that of the signal’s second (100Hz) harmonic at -120dBrA or 0.0001%, and the third signal harmonic (150Hz) at -130dBrA or 0.00003%. A small power-supply-related peak can be seen at 60Hz at -140dBrA, or 0.00001%, but this is inherent to the Audio Precision sinewave generator.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, line-level input)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced outputs for the unbalanced line-level input. The input RMS values are set at -6.02dBrA so that, if summed for a mean frequency of 18.5kHz, would yield 4Vrms at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is at -120dBRA, or 0.0001%, while the third-order modulation products, at 17kHz and 20kHz, are at around -130dBrA, or 0.00003%. This is a very clean IMD result.

Intermodulation distortion FFT (line-level input, APx 32 tone)

Shown above is the FFT of the balanced outputs of the Wandla with the APx 32-tone signal applied to the analog unbalanced input. The combined amplitude of the 32 tones is the 0dBrA reference, and corresponds to 4Vrms. The intermodulation products—i.e., the “grass” between the test tones—are distortion products from the amplifier and are around the extremely low -150dBrA, or 0.000003%, level.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, coaxial digital input, 16/44.1)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced outputs for the digital coaxial input at 16/44.1. We find that the second-order modulation product (i.e., the difference signal of 1kHz) cannot be seen above the -140dBrA, or 0.00001%, noise floor, while the third-order modulation products, at 17kHz and 20kHz, are at -130dBrA, or 0.00003%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, coaxial digital input, 24/96)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced outputs for the digital coaxial input at 24/96. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is at -140dBrA, or 0.00001%, while the third-order modulation products, at 17kHz and 20kHz, are at -130dBrA, or 0.00003%. This is an exquisitely clean IMD result.

Square-wave response (10kHz)

Above is the 10kHz squarewave response using the analog line-level input, at roughly 4Vrms at the balanced output. Due to limitations inherent to the Audio Precision APx555 B Series analyzer, this graph should not be used to infer or extrapolate the Wandla’s slew-rate performance. Rather, it should be seen as a qualitative representation of its extremely extended bandwidth. An ideal squarewave can be represented as the sum of a sinewave and an infinite series of its odd-order harmonics (e.g., 10kHz + 30kHz + 50kHz + 70kHz . . .). A limited bandwidth will show only the sum of the lower-order harmonics, which may result in noticeable undershoot and/or overshoot, and softening of the edges. The Wandla’s reproduction of the 10kHz square wave is essentially perfect, with sharp corners and no overshoot/undershoot.

Diego Estan
Electronics Measurement Specialist

Details

Parent Category: Products

Category: Digital-to-Analog Converter Measurements

Created: 15 October 2023

Link: reviewed by Roger Kanno on SoundStage! Simplifi on October 15, 2023

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The Bluesound Node X was evaluated as a digital-to-analog converter and conditioned for 30 minutes at 0dBFS (2.1Vrms out) into 100k ohms before any measurements were taken.

The Node X offers one combination digital-optical (S/PDIF) and analog 1/8″ TRS input. There is a 1/4″ TRS headphone output on the front of the unit. There is a digital volume control for the headphone and line-level outputs. There are also tone controls and a subwoofer output that can be turned on using the accompanying BluOS app. The app also offers full bass control with adjustable low/high pass filters. For the analog input, our standard 2Vrms level was replaced with 1Vrms, because at 2Vrms, the Node X’s ADC was nearing overload and random excessive noise was observed at the output. This is consistent with the behavior we have noted with other Bluesound products.

The analyzer’s input bandwidth filter was set to 10Hz–22.4kHz for all measurements, except for frequency response (DC to 1 MHz), FFTs (10Hz to 90kHz), and THD vs. frequency (10Hz to 90kHz). The latter to capture the second and third harmonic of the 20kHz output signal.

The Node X digital volume control ranges from -80 to 0dB, in steps ranging from 1 to 4dB. Channel-to-channel deviation proved excellent, at 0.001dB throughout the range.

Volume-control accuracy (measured at line-level outputs): left-right channel tracking

Primary measurements

Our primary measurements revealed the following using the digitall input and the line-level outputs (unless specified, assume a 1kHz sinewave at 0dBFS, 200k ohms loading, 10Hz to 22.4kHz bandwidth):

Our primary measurements revealed the following using the digital input and the headphone output (unless specified, assume a 1kHz sinewave at 0dBFS, 300 ohms loading, 10Hz to 22.4kHz bandwidth):

Frequency response vs. sample rate (16/44.1, 24/96, 24/192, analog)

The plot above shows the Node X’s frequency response (relative to 1kHz) as a function of sample rate. The blue/red traces are for a 16-bit/44.1kHz dithered digital input signal from 5Hz to 22kHz, the purple/green traces are for a 24/96 dithered digital input signal from 5Hz to 48kHz, and finally orange/pink represents 24/192 from 5Hz to 96kHz. The cyan plot is for the analog input. It’s obvious from the response that incoming analog signals are sampled at 44.1kHz. There is also a slight roll-off (-0.3dB) from 5–10Hz that is not present for the digital input. The behavior at low frequencies is the same for all digital sample rates—perfectly flat down to 5Hz. The behavior at high frequencies for all three digital sample rates is as expected, offering sharp filtering around 22, 48, and 96kHz (half the respective sample rate), although the 24/192 data shows softer attenuation around the corner frequency. The -3dB point for each sample rate is roughly 21, 46, and 91.5kHz, respectively. In the graph above and most of the graphs below, only a single trace may be visible. This is because the left channel (blue, purple, or orange trace) is performing identically to the right channel (red, green, or pink trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Frequency response (bass and treble, 24/96)

Above are two frequency-response plots (relative to 1kHz) for the digital input (24/96), measured at the analog outputs, with the treble/balance controls set at both minimum and maximum. They show that the Node X will provide a maximum gain/cut of approximately 6dB at 20Hz and 20kHz.

Frequency response (bass management, 24/96)

Above are two frequency-response plots for the digital input (24/96), measured at the subwoofer output and left/right analog outputs, with the crossover set at 120Hz. The Node X crossover uses a slope of 18dB/octave, and the subwoofer output is flat down to 5Hz.

Digital linearity (16/44.1 and 24/96 data)

The graph above shows the results of a linearity test for the digital input for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the line-level outputs of the Node X. The digital input was swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output was analyzed by the APx555. The ideal response is a straight flat line at 0dB (i.e., the amplitude of the digital input perfectly matches the amplitude of the measured analog output). The 24/96 input data is essentially perfect down to -120dBFS, while the 16/44.1 input data performed well, only over-responding by 3/2dB (left/right) at -120dBFS. The 24/96 data yielded such superb results that we extended the sweep down to . . .

. . . -140dBFS. Above we see that even at -140dBFS, the Node X is only overshooting by 1 to 3dB with 24/96 data. This is an exemplary linearity test result.

Impulse response

The graph above shows the impulse responses for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s), for one sample period then back to digital silence, fed to the digital input, measured at the analog outputs, for the left channel only. We can see that Node X DAC reconstruction filter exhibits symmetrical pre- and post-ringing as seen in a typical sinc function.

J-Test (optical input)

The plot above shows the results of the J-Test test for the coaxial digital input measured at the analog outputs of the Node X. J-Test was developed by Julian Dunn the 1990s. It is a test signal—specifically,  a -3dBFS undithered 12kHz squarewave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz squarewave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sinewave (the main peak). In addition, an undithered 250Hz squarewave at -144dBFS is mixed with the signal. This test file causes the 22 least significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc). Note that the alternating peaks are in the test file itself, but at levels of -144dBrA (fundamental at 250Hz) down to -170dBrA for the odd harmonics. The test file can also be used in conjunction with artificially injected sinewave jitter by the Audio Precision, to also show how well the DAC rejects jitter.

The optical digital input shows an average-to-mediocre J-Test result, with several peaks at the -130dBrA level and below clearly visible throughout the audioband. This is an indication that the Node X may be sensitive to jitter.

J-Test (optical input, 2kHz sinewave jitter at 100ns)

The plot above shows the results of the J-Test test for the optical digital input measured at the line-level output, with an additional 100ns of 2kHz sinewave jitter injected by the APx555, which would manifest as sideband peaks at 10kHz and 14kHz, without perfect jitter immunity. The results show no visible sidebands at 10kHz and 14kHz, and essentially the same J-Test result as seen above without the injection of jitter. The Node X DAC lost sync with the signal when roughly 600ns of jitter was added to the test signal.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone

The plot above shows a fast Fourier transform (FFT) of the Node X’s line-level output with white noise at -4dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the optical digital input, sampled at 16/44.1 (purple/green). There is a steep rolloff above 20kHz in the white-noise spectrum, characteristic of a brick-wall-type filter. There are no imaged aliasing artifacts in the audioband above the -135dBrA noise floor, except for a very small peak at roughly 11kHz at -130dBrA from the left channel. The primary aliasing signal at 25kHz is at -80dBrA.

THD ratio (unweighted) vs. frequency vs. load (24/96)

The chart above shows THD ratios at the line-level output into 100k ohms (blue/red) and 600 ohms (purple/green) as a function of frequency for a 24/96 dithered 1kHz signal at the optical input. The 100k and 600 ohms data are extremely close throughout the audioband (3-5dB higher for the 600-ohm load at the frequency extremes), which is an in indication that Node X’s outputs are robust and can handle loads below 1k ohms with no difficultly. THD ratios into 100k ohms ranged from 0.0004% from 20Hz to 300Hz, then up to 0.01% at 20kHz.

THD ratio (unweighted) vs. frequency vs. sample rate (16/44.1, 24/96)

The chart above shows THD ratios at the line-level output into 100k ohms as a function of frequency for a 16/44.1 (blue/red) and a 24/96 (purple/green) dithered 1kHz signal at the optical input. THD ratios were identical and ranged from 0.0004% from 20Hz to 300Hz, then up to 0.01% at 20kHz.

THD ratio (unweighted) vs. output (16/44.1, 24/96)

The chart above shows THD ratios measured at the line-level output as a function of output voltage for the optical input into 100k ohms from -90dBFS to 0dBFS at 16/44.1 (blue/red) and 24/96 (purple/green). The 24/96 data outperformed the 16/44.1 data at low output voltages by roughly 10dB, with a THD range from 0.5% at 200uVrms to 0.0006-0.001% at 0.5 to 2.1Vrms, while the 16/44.1 ranged from 3% down to the same 0.0006-0.001% at 0.5 to 2.1Vrms. The difference in THD ratios is owed to the lower noise floor with 24/96 data—the analyzer cannot measure/assign a THD ratio below the noise floor.

THD+N ratio (unweighted) vs. output (16/44.1, 24/96)

The chart above shows THD+N ratios measured at the line-level output as a function of output voltage for the optical input into 100k ohms from -90dBFS to 0dBFS at 16/44.1 (blue/red) and 24/96 (purple/green). The 24/96 outperformed the 16/44.1 data throughout by roughly 10dB, with a THD+N range from 6% down to 0.001% at 1–2Vrms, while the 16/44.1 ranged from 20% down to 0.002% at the maximum output voltage of 2.1Vrms.

Intermodulation distortion vs. generator level (SMPTE, 60Hz:4kHz, 4:1; 16/44.1, 24/96)

The chart above shows intermodulation distortion (IMD) ratios measured at the line-level output for 16/44.1 (blue/red) input data and 24/96 input data (purple/green), from -60dBFS to 0dBFS. Here, the SMPTE IMD method was used, where the primary frequency (F1 = 60Hz) and the secondary frequency (F2 = 7kHz) are mixed at a ratio of 4:1. The SMPTE IMD analysis results consider the second (F2 ± F1) through the fifth (F2 ± 4xF1) modulation products. The 16/44.1 data yields IMD ratios from 2% down to 0.003% at 0dBFS. The 24/96 data yields IMD ratios from 0.2% down to 0.001% at 0dBFS. The difference here again is likely due to the lower noise floor with 24/96 data.

FFT spectrum – 1kHz (analog input at 1Vrms)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the line-level output into 100k ohm for the analog input, which is resampled by the Node X ADC at 16/44.1. The second (2kHz) harmonic dominates at nearly -90dBra, or 0.003%, while the third (3kHz) harmonic is at -105dBrA, or 0.0006%. There are very low-level power-supply-related noise peaks to the left of the main signal peak around the -130dBrA, or 0.00003%, level. Also visible are the 43.1kHz and 45.1kHz IMD peaks associated with the 44.1kHz sample rate.

FFT spectrum – 1kHz (digital input, 16/44.1 at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the line-level output into 100k ohm for the optical digital input, sampled at 16/44.1. The signal harmonic profile is similar but lower in amplitude to the FFT above, which would include artifacts of the Node X’s ADC. The second (2kHz), third (3kHz), and fifth (5kHz) signal harmonics dominate at the -100 to -110dBrA level, or 0.001 to 0.0003%. The noise floor is also lower from 10Hz to 50Hz.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 100k ohm for the optical digital input, sampled at 24/96. Due to the increased bit-depth, the noise floor is lower compared to the 16/44.1 FFT above, at a very low -150dBrA. We see signal harmonics are essentially the same as the 16/44.1 FFT above. With the lower noise floor, noise-related harmonics are easier to see, and are actually a bit higher than the 16/44.1 FFT above, reaching nearly -120dBrA, or 0.0001%.

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the line-level output into 100k ohm for the optical digital input, sampled at 16/44.1 at -90dBFS. We see the signal peak at the correct amplitude, and no signal harmonics above the noise floor within the audioband.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the line-level output into 100k ohm for the optical digital input, sampled at 24/961 at -90dBFS. We see the signal peak at the correct amplitude, and extremely low-level signal harmonics (2/3/4kHz) at and below -150dBrA, or 0.000003%. The 60Hz power-supply fundamental peak can be seen at -135dBrA, or 0.00002%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 16/44.1)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the line-level output into 100k ohms for the optical input at 16/44.1. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 2.1Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz), is at -120dBrA, or 0.0001%, and the third-order modulation products, at 17kHz and 20kHz, are at -95dBrA, or 0.002%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 24/96)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the line-level output into 100k ohms for the optical input at 24/96. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 2.1Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz), is at -120dBrA, or 0.0001%, and the third-order modulation products, at 17kHz and 20kHz, are at -95dBrA, or 0.002%.

Diego Estan
Electronics Measurement Specialist

Details

Parent Category: Products

Category: Digital-to-Analog Converter Measurements

Created: 15 September 2023

Link: reviewed by Roger Kanno on SoundStage! Hi-Fi on September 15, 2023

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The Technics SL-G700M2 was evaluated as a digital-to-analog converter via the digital inputs and conditioned for 30 minutes at 0dBFS (2.2Vrms out) into 200k ohms before any measurements were taken.

The SL-G700M2 offers three digital inputs: one coaxial S/PDIF (RCA), one optical S/PDIF, and one USB. There are two line-level outputs (balanced XLR and unbalanced RCA) and one headphone output (1/4″ TRS labelled “PHONES”. There is a digital volume control for the headphone and line-level outputs. Comparisons were made between unbalanced and balanced line level outputs, no appreciable differences were seen in terms of THD and noise, but 1kHz FFTs are provided for both balanced and unbalanced outputs.

The SL-G700M2 offers a few features and settings. The following are the default settings used for the coaxial input, balanced line-level outputs, using a 0dBFS input, unless otherwise specified:

• Analog output level: fixed
• MQA processing: off
• Coherent Processing: on (forces a dedicated reconstruction filter)
• Filter: Mode 1, Mode 2, and Mode 3 are available (when Coherent Processing is off). These were evaluated for different parameters, such as phase, frequency, and impulse response (as indicated in the graphs below).

The analyzer’s input bandwidth filter was set to 10Hz to 22.4kHz for all measurements, except for frequency response (DC to 1MHz), FFTs (10Hz to 90kHz), and THD vs frequency (10Hz to 90kHz). The latter to capture the second and third harmonics of the 20kHz output signal.

The SL-G700M2’s digital volume control ranges from -99 to 0dB, in steps of 0.5dB. Channel-to-channel deviation proved average, at around 0.19dB throughout the range.

Volume-control accuracy (measured at line-level outputs): left-right channel tracking

Primary measurements

Our primary measurements revealed the following using the coaxial input and the balanced line-level outputs (unless specified, assume a 1kHz sinewave at 0dBFS, 200k ohms loading, 10Hz to 22.4kHz bandwidth):

*due to very low noise of DUT, analyzer self-noise has been removed from measurement to more accurately report value

Our primary measurements revealed the following using the coaxial input and the headphone output (unless specified, assume a 1kHz sinewave at 0dBFS, 300 ohms loading, 10Hz to 22.4kHz bandwidth):

Frequency response vs. sample rate (16/44.1, 24/96, 24/192; Coherent Processing)

The plot above shows the SL-G700M2’s frequency response as a function of sample rate. The blue/red traces are for a 16bit/44.1kHz dithered digital input signal from 5Hz to 22kHz, the purple/green traces are for a 24/96 dithered digital input signal from 5Hz to 48kHz, and finally orange/pink represents 24/192 from 5Hz to 96kHz. The behavior at low frequencies is the same for the digital input—perfectly flat down to 5Hz. The behavior at high frequencies for all three digital sample rates is as expected, offering sharp filtering around 20k, 42k, and 82kHz (less than half the respective sample rate), although the 24/192 data shows softer attenuation around the corner frequency. The -3dB point for each sample rate is roughly 19.2, 41.7 and 81.3kHz respectively. In the graph above and most of the graphs below, only a single trace may be visible. This is because the left channel (blue, purple or orange trace) is performing identically to the right channel (red, green or pink trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Frequency response vs. filter type (16/44.1)

The plots above show frequency-response for a 0dBFS input signal sampled at 16/44.1 for Mode 1  (blue), Mode 2 (red), Mode 3 (green), and Coherent Processing (pink) filters into a 200k ohm load for the left channel only. The graph is zoomed in from 1kHz to 22kHz to highlight the various responses of the three Mode filters. We can see Mode 1 and Coherent Processing offer essentially the same frequency response, with a -3dB point at 19.2 kHz, while Mode 2 is very close, with a -3dB point at 19.7kHz. It’s worth pointing out that the “knee” for these three filters occurs just past 16kHz, a frequency many audiophiles can no longer even hear. The Mode 3 filter behaves like a typical brickwall-type filter, with a -3dB point at 21.2kHz.

Phase response vs. sample rate (16/44.1, 24/96, 24/192; Coherent Processing)

Above are the phase-response plots from 20Hz to 20kHz for the coaxial input, measured across at the balanced output, using the Coherent Processing filter setting. The blue/red traces are for a dithered 16/44.1 input at 0dBFS, the purple/green for 24/96, and the orange/pink for 24/192. We can see that the SL-G700M2 does not invert polarity, with a worst-case phase shift of -140 degrees at 20kHz for the 16/44.1 data. Phase shift at 20kHz for the 24/96 and 24/192 input data are inconsequential, at about -5 degrees.

Phase response vs. filter (16/44.1)

Above are the phase response plots from 20Hz to 20kHz for a 0dBFS input signal sampled at 44.1kHz for the Mode 1 (blue), Mode 2 (red), Mode 3 (green), and Coherent Processing (pink) filters into a 200k ohm load for the left channel only. Predictably, the brickwall filter (Mode 3) yields the highest phase shift at around -180 degrees at 20kHz. The Mode 1 and Coherent Processing filters are identical, at -140 degrees at 20kHz, while the Mode 2 filter exhibits no phase shift throughout the audioband.

Digital linearity (16/44.1 and 24/96 data)

The graph above shows the results of a linearity test for the coaxial digital input (the optical input performed identically) for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the balanced line-level output of the SL-G700M2. The digital input is swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output is analyzed by the APx555. The ideal response is a straight flat line at 0dB (i.e., the amplitude of the digital input perfectly matches the amplitude of the measured analog output). The 24/96 input data is essentially perfect down to -120dBFS, while the 16/44.1 input data performed well, only over-responding by 4/2dB (left/right channels) at -120dBFS. The 24/96 data yielded such superb results that we extended the sweep down to . . .

. . . -140dBFS. Above we see that even at -140dBFS, the SL-G700M2 is only undershooting by -1 to -3dB. This is an exemplary linearity-test result.

Impulse response vs. filter type (Mode 1, Mode 2, Mode 3, Coherent Processing)

The graph above shows the impulse responses for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s), for one sample period then back to digital silence, fed to the coaxial digital input, measured at the balanced outputs, for Mode 1 (blue), Mode 2 (red), Mode 3 (green), and Coherent Processing (pink) filters into a 200k ohm-load for the left channel only. We can see that the Mode 1 and Coherent processing filters are nearly identical, with minimal pre-ringing and some post-ringing. The Mode 3 filter has no pre-ringing, but significant post-ringing, while the Mode 2 filter exhibits symmetrical pre- and post-ringing, as seen in a typical sinc function.

J-Test (coaxial input)

The plot above shows the results of the J-Test test for the coaxial digital input measured at the balanced line level output of the SL-G700M2. J-test was developed by Julian Dunn the 1990s. It is a test signal—specifically, a -3dBFS undithered 12kHz squarewave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz squarewave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sinewave (the main peak). In addition, an undithered 250Hz squarewave at -144dBFS is mixed with the signal. This test file causes the 22 least-significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc.). Note that the alternating peaks are in the test file itself, but at levels of -144dBrA (fundamental at 250Hz) down to -170dBrA for the odd harmonics.  The test file can also be used in conjunction with artificially injected sinewave jitter by the Audio Precision, to also show how well the DAC rejects jitter.

The coaxial input shows an average to mediocre J-Test result, with two peaks at the -130dBrA level clearly visible near 11kHz and 13kHz. This is an indication that the SL-G700M2 may be sensitive to jitter.

J-Test (optical input)

The plot above shows the results of the J-Test test for the optical digital input measured at the balanced line-level output of the SL-G700M2. The optical input shows essentially the same result as the coaxial input above.

J-Test (coaxial input, 2kHz sinewave jitter at 100ns)

The plot above shows the results of the J-Test test for the coaxial digital input (the optical input behaved the same) measured at the balanced line-level output, with an additional 100ns of 2kHz sinewave jitter injected by the APx555, which would manifest as sideband peaks at 10kHz and 14kHz without perfect jitter immunity. The results show visible sidebands at 10kHz and 14kHz, but at a relatively low -125dBrA. This is further evidence of the SL-G700M2’s average jitter immunity.

J-Test (coaxial input, 2kHz sinewave jitter at 600ns)

The plot above shows the results of the J-Test test for the coaxial digital input (the optical input behaved the same) measured at the balanced line-level output, with an additional 600ns of 2kHz sinewave jitter injected by the APx555. Here sidebands are visible at 10kHz and 14kHz again, but remain relatively low at -110dBrA. With jitter above this level, the SL-G700M2 lost sync with the signal.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (Mode 1)

The plot above shows a fast Fourier transform (FFT) of the SL-G700M2’s balanced-line level output with white noise at -4 dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Mode 1 filter setting. There is a soft roll-off above 20kHz in the white-noise spectrum. There are absolutely no imaged aliasing artifacts in the audio and above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -35dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (Mode 2)

The plot above shows a fast Fourier transform (FFT) of the SL-G700M2’s balanced-line level output with white noise at -4 dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Mode 2 filter setting. There is a soft roll-off above 20kHz in the white-noise spectrum. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -25dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (Mode 3)

The plot above shows a fast Fourier transform (FFT) of the SL-G700M2’s balanced-line level output with white noise at -4 dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Mode 3 filter setting. There is a sharp roll-off above 20kHz in the white-noise spectrum showing the implementation of a brickwall-type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -100dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (Coherent Processing)

The plot above shows a fast Fourier transform (FFT) of the SL-G700M2’s balanced line-level output with white noise at -4 dBFS (blue/red) and a 19.1 kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Coherent Processing filter setting. There is a soft roll-off above 20kHz in the white-noise spectrum. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -35dBrA.

THD ratio (unweighted) vs. frequency vs. load (24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms (blue/red) and 600 ohms (purple/green) as a function of frequency for a 24/96 dithered 1kHz signal at the coaxial input. The 200k and 600 ohms data are close throughout the audioband (within 10dB from 2kHz to 20kHz), which is in indication that the SL-G700M2’s outputs are robust and can handle loads below 1k ohms with no difficultly. THD ratios into 200k ohms ranged from 0.0002% from 20Hz to 500Hz, then up to 0.002% at 20kHz.

THD ratio (unweighted) vs. frequency vs. sample rate (16/44.1, 24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms as a function of frequency for a 16/44.1 (blue/red) and a 24/96 (purple/green) dithered 1kHz signal at the coaxial input. The 24/96 data consistently outperformed the 16/44.1 data by 5-10dB from 20Hz to about 2kHz due to the inherently higher noise floor at 16 bits (i.e., the analyzer cannot assign a THD value below the noise floor). THD ratios with 16/44.1 data range from 0.0003% from 20Hz to 2kHz, then up to 0.002% at 16kHz. THD ratios with 24/96 data range from 0.0001-0.0002% from 20Hz to 2kHz, up to 0.002% at 20kHz.

THD ratio (unweighted) vs. output (16/44.1, 24/96)

The chart above shows THD ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS at 16/44.1 (blue/red) and 24/96 (purple/green). Once again, the 24/96 outperformed the 16/44.1 data, with a THD range from 0.1% at 200uVrms to 0.0001% at 0.5 to 2Vrms, while the 16/44.1 ranged from 2% down to nearly 0.0002%.

THD+N ratio (unweighted) vs. output (16/44.1, 24/96)

The chart above shows THD+N ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS at 16/44.1 (blue/red) and 24/96 (purple/green). The 24/96 outperformed the 16/44.1 data, with a THD+N range from 1% down to  0.0002% at 1.5-2Vrms, while the 16/44.1 ranged from 20% down to 0.002% at the maximum output voltage of 2.2Vrms.

Intermodulation distortion vs. generator level (SMPTE, 60Hz:4kHz, 4:1; 16/44.1, 24/96)

The chart above shows intermodulation distortion (IMD) ratios measured at balanced output for 16/44.1 (blue/red)  and 24/96 input data (purple/green), from -60dBFS to 0dBFS. Here, the SMPTE IMD method was used, where the primary frequency (F1 = 60Hz) and the secondary frequency (F2 = 7kHz) are mixed at a ratio of 4:1. The SMPTE IMD analysis results consider the second (F2 ± F1) through the fifth (F2 ± 4xF1) modulation products. The 16/44.1 data yields IMD ratios from 2% down to 0.002% at 0dBFS. The 24/96 data yields IMD ratios from 0.1% down to 0.0005% from -10 to 0dBFS.

FFT spectrum – 1kHz (digital input, 16/44.1 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced outputs into 200k ohm for the coaxial digital input, sampled at 16/44.1. We see signal harmonics at 2/3/5kHz, with the third harmonic (3kHz) dominating at -120dBra, or 0.0001%. There are also no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced outputs into 200k ohm for the coaxial digital input, sampled at 24/96. Due to the increased bit-depth, the noise floor is much lower compared to the 16/44.1 FFT at a very low -150 to -160dBrA. We see signal harmonics ranging from -120dBrA to -140dBrA, or 0.0001% to 0.00001%, all the way to 20kHz (and beyond). Here also, there are no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS, unbalanced output)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the unbalanced outputs into 200k ohm for the coaxial digital input, sampled at 24/96. We find small differences in the signal harmonic pattern here compared to the balanced inputs above. Here the second signal harmonic (2kHz) reaches -110dBrA, or 0.0003%, compared to -130dBrA, or 0.00003%, for the balanced inputs. There are also very low-level power-supply-related (or IMD) peaks on the right channel here to the left of the signal peak, from -140 to -150dBrA, that do not show up in the balanced outputs. 

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced outputs into 200k ohm for the coaxial digital input, sampled at 16/44.1 at -90dBFS. We see the signal peak at the correct amplitude, and no signal harmonics above the noise floor within the audioband.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 24/961 at -90dBFS. We see the signal peak at the correct amplitude, and extremely low-level signal harmonics (2/3/4kHz) at and below -150dBrA, or 0.000003%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 16/44.1)

Shown above is an FFT of the intermodulation (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 16/44.1. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 2.2Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz), is perhaps barely visible above the noise floor from the right channel at -130dBrA, or 0.00003%, and the third-order modulation products, at 17kHz and 20kHz, are at -120dBrA, or 0.0001%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 24/96)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 24/96. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 2.2Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz), is at -130dBrA (right), or 0.00003%, and the third-order modulation products, at 17kHz and 20kHz, are at -115dBrA, or 0.0002%.

Intermodulation distortion FFT (coaxial input, APx 32 tone, 24/192)

Shown above is the FFT of the balanced line-level output of the SL-G700M2 with the APx 32-tone signal applied to the input. The combined amplitude of the 32 tones is the 0dBFS reference, and corresponds to 2.2Vrms into 200k ohms. The intermodulation products—i.e., the "grass" between the test tones—are distortion products from the amplifier and below the -145dBrA, or 0.000006%, level. This is a very clean IMD result.

Diego Estan
Electronics Measurement Specialist

Details

Parent Category: Products

Category: Digital-to-Analog Converter Measurements

Created: 15 July 2023

Link: reviewed by Phil Gold on SoundStage! Hi-Fi on July 15, 2023

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The Musical Fidelity M6x was conditioned for 30 minutes at 0dBFS (4Vrms out) into 200k ohms before any measurements were taken.

The M6x offers five digital inputs: one coaxial S/PDIF (RCA), two optical S/PDIF (TosLink), one AES/EBU (XLR), and one USB. There are two line-level outputs (balanced XLR and unbalanced RCA) and one headphone output (1/4″ TRS). There is a digital volume control that can be engaged for both the line-level outputs and headphone output, but was left in the Fix (fixed) default setting for all measurements, with the exception of the volume tracking table. Comparisons were made between unbalanced and balanced line-level outputs for a 24/96 0dBFS input, and aside from the 6dB extra voltage using balanced, there were no differences in THD+N. In terms of input types (USB, coaxial, optical), there were no differences in terms of THD+N at 24/96.

There is a button labeled Upsample on the M6x, which, when engaged, upsamples incoming PCM data up to 192kHz to 352.8 or 384kHz (using integer multiples of the incoming sample rate). There are eight filter settings labeled 1 through 8. All measurements below, unless otherwise stated, are for the coaxial digital input, and the balanced output, using filter 1. The eight filters are described as follows in the M6x manual:

1. Linear phase fast roll off: most common filter with clean overall suppression and excellent rejection, best for music with large transients. Provides clean, crisp highs.
2. Linear phase slow roll off: low group delay – and symmetrical input response. Less ringing than linear-phase fast roll-off (LPFR). Punchier bass than LPFR, with clean highs.
3. Minimum phase fast roll off: minimal pre-ringing preferred for imaging and soundstage. No aliasing in frequency domain. Stronger bass than linear phase, clean highs.
4. Minimum phase slow roll off: non-symmetrical filter designed to minimize pre-ringing. Strong punchy bass with good transient attacks.
5. Apodizing fast roll off: a version of linear-phase fast roll off filter optimized to improve pre-ringing.
6. Hybrid fast roll off: a combination of linear-phase and minimum phase. Fast transient attack, strong punchy bass, crisp highs.
7. Brick wall: one of the earliest designs, intended for highest suppression possible, with high delay and pre-ringing. Linear phase, crisp clean highs.
8. Oversampling bypass: the oversampling FIR filter, used for the 7 above mentioned presets, is bypassed and source data is up-sampled to 352.8/384kHz.

Note: it seemed clear comparing frequency, phase, and impulse responses for 16-bit/44.1kHz input data, as well as wideband noise FFTs, between filter 8 and any other filter with “Upsample” engaged, yielded the exact same results.

The M6x volume control has no indicator on the front panel. The volume control can be engaged by pressing the Output button on the front panel for 2 to 3 seconds to change from Fix (fixed) to Var (variable) output. When headphones are plugged in, Var is automatically selected. For a 0dBFS 1kHz input signal using the full range of the volume control will yield from a minimum of about 0.1mVrms (-90dB) to 4.1Vrms (0dB) in 1dB steps at the balanced line-level outputs, and the headphone outputs. The volume control operates in the digital domain, as every step was exactly 1dB, and the channel-to-channel deviation was exactly 0.105dB at every step, throughout the range.

Volume-control accuracy (measured at line-level outputs): left-right channel tracking

Published specifications vs. our primary measurements

The table below summarizes the measurements published by Musical Fidelity for the M6x DAC compared directly against our own. The published specifications are sourced from Musical Fidlelity’s website, either directly or from the manual available for download, or a combination thereof. With the exception of frequency response, where the Audio Precision bandwidth was set at its maximum (DC to 1MHz), assume, unless otherwise stated, the coaxial digital input (24/96 1kHz sinewave at 0dBFS), the balanced line-level or unbalanced headphone outputs into 200k ohms (line-level) and 300 ohms (headphone) using a measurement input bandwidth of 10Hz to 90kHz, and the worst-case measured result between the left and right channels.

Our primary measurements revealed the following using the coaxial input and the balanced line-level outputs (unless specified, assume a 1kHz sinewave at 0dBFS, 200k ohms loading, 10Hz to 90kHz bandwidth):

Our primary measurements revealed the following using the coaxial input and the headphone output (unless specified, assume a 1kHz sinewave at 0dBFS input, 2Vrms into 300 ohms, 10Hz to 90kHz bandwidth):

Frequency response vs. sample rate (16/44.1, 24/96, 24/192; filter 1)

The plot above shows the M6x frequency response (relative to 1kHz) as a function of sample rate. The blue/red traces are for a 16-bit/44.1kHz dithered digital input signal from 5Hz to 22kHz, the purple/green traces are for a 24/96 dithered digital input signal from 5Hz to 48kHz, and finally orange/pink represents 24/192 from 5Hz to 96kHz. The behavior at low frequencies is the same for the digital input—perfectly flat down to 5Hz. The behavior at high frequencies for all three digital sample rates is as expected, offering sharp filtering around 22, 48, and 96kHz (half the respective sample rate for each). The -3dB point for each sample rate is roughly 21, 45.7, and 70.7kHz, respectively. It is also obvious from the plots above that the 44.1kHz sampled input signal offers the most brickwall-type behavior, while the attenuation of the 96kHz and 192kHz sampled input signals approaching the corner frequencies (48kHz and 96kHz) is more gentle. In the graph above and most of the graphs below, only a single trace may be visible. This is because the left channel (blue, purple or orange trace) is performing identically to the right channel (red, green or pink trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Frequency response vs. filter type (16/44.1, filters 1 to 4)

The plots above show frequency-response for a 16/44.1 input, for filters 1 through 4, into a 200k ohm-load, for the left channel only. Filter 1 is in red, filter 2 in purple, filter 3 in green, and filter 4 in blue. The graph is zoomed in from 1kHz to 22kH, to highlight the various responses of the filters around the “knee” of the response. At 20kHz, filter 1 is at -0.35dB, filter 2 is at -3.78dB, filter 3 is at -0.38dB, and filter 4 is at -5.33dB.

Note: the filter characteristics are described under the General information section above. Our measured frequency responses match the descriptions provided by Musical Fidelity.

Phase response vs. filter type (16/44.1, filters 5 to 8)

The plots above show frequency-response for a 16/44.1 input, for filters 5 through 8, into a 200k ohm-load, for the left channel only. Filter 5 is in red, filter 6 in purple, filter 7 in green, and filter 8 in blue. The graph is zoomed in from 1kHz to 22kHz to highlight the various responses of the filters around the “knee” of the response. At 20kHz, filter 5 is at -0.15dB, filter 6 is at -12.60dB, filter 7 is at 4.31dB, and filter 8, which is 16/44.1 input data up-sampled to 352.8kHz and the ESS DAC oversampling filter disabled, is at -0.59dB. Of note, filter 5 yields up/down deviations in the frequency response by almost 0.5dB nearing 20kHz.

Note: the filter characteristics are described under General Information section above. Our measured frequency responses generally match the descriptions provided by Musical Fidelity.

Phase response vs. sample rate (16/44.1, 24/96, 24/192; filter 1)

Above are the absolute phase response plots (including group delay) from 20Hz to 20kHz for the coaxial input, measured at the balanced output for the left channel only. The blue/red traces are for a dithered 16/44.1 input at -20dBFS, the purple/green for 24/96, and the orange/pink for 24/192. We can see that the M6x does not invert polarity, with a worst-case phase shift of just under 80 degrees at 20kHz for the 16/44.1, and phase shift just above and below 20 degrees at 20kHz for the 24/96 and 24/192 input data, respectively.

Phase response vs. filter type (16/44.1, filters 1 to 4)

Above are the absolute phase response plots (including group delay) plots from 20Hz to 20kHz, for a 16/44.1 signal at the coaxial input, measured at the balanced output, for filters 1 through 4, into a 200k ohm-load, for the left channel only. Filter 1 is in blue, filter 2 in purple, filter 3 in orange, and filter 4 in green. We find the general trend of better or less severe phase shift for filters that have poorer frequency response (e.g., more attenuation at 20kHz), and vice-versa.

Phase response vs. filter type (16/44.1, filters 5 to 8)

Above are the absolute phase response plots (including group delay) from 20Hz to 20kHz for a 16/44.1 signal at the coaxial input, measured at the balanced output, for filters 5 through 8, into a 200k ohm-load, for the left channel only.  Filter 5 is in blue, filter 6 in purple, filter 7 in orange, and filter 8 in green. We find the general trend of better or less severe phase shift for filters that have poorer frequency response (e.g., more attenuation at 20kHz), and vice-versa.

Digital linearity (16/44.1 and 24/96 data)

The graph above shows the results of a linearity test for the coaxial digital input (the optical input performed identically) for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the balanced line-level output. The digital input is swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output is analyzed by the APx555. The ideal response is a straight flat line at 0dB (i.e., the amplitude of the digital input perfectly matches the amplitude of the measured analog output). The 24/96 data is perfectly linear down to -120dBFS, while the 16/44.1 data was perfect down to -100dBFS, and only +3dB (left) and +1dB (right) at -120dBFS. Because of this DAC’s exceptional linearity performance, a second measurement was performed extending down to -140dBFS . . .

. . . that shows the 24/96 data remained within 0.4dB of flat, which is a phenomenal result. The 16/44.1 data results are as expected, showing significant deviations below -120dBFS. It is worth highlighting that a linearity result that is flat down to -100dBFS for 16-bit input data is also exceptional.

Impulse response vs. filter type (24/44.1, filters 1 to 4)

The graph above shows the impulse responses for the first four filter types, for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period, then back to digital silence. Filter 1 is in blue, filter 2 in purple, filter 3 in red, and filter 4 in green.

Note: the filter characteristics are described under General information above. Our measured impulse responses generally match the descriptions provided by Musical Fidelity.

Impulse response vs. filter type (24/44.1, filters 5 to 8)

The graph above shows the impulse responses for the first 4 filter types, for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period then back to digital silence. Filter 5 is in blue, filter 6 in purple, filter 7 in red, and filter 8 in green.

Note: the filter characteristics are described under General Information above. Our measured impulse responses generally match the descriptions provided by Musical Fidelity.

J-Test (coaxial input)

The plot above shows the results of the J-Test test for the coaxial digital input measured at the balanced line-level output of the M6x. J-Test was developed by Julian Dunn the 1990s. It is a test signal—specifically, a -3dBFS undithered 12kHz squarewave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz squarewave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sinewave (the main peak). In addition, an undithered 250Hz squarewave at -144dBFS is mixed with the signal. This test file causes the 22 least significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc). Note that the alternating peaks are in the test file itself, but at levels of -144dBrA (fundamental at 250Hz) down to -170dBrA for the odd harmonics.  The test file can also be used in conjunction with artificially injected sinewave jitter by the Audio Precision, to also show how well the DAC rejects jitter.

The coaxial SPDIF input shows some of the alternating 500Hz peaks in the audioband but at very low levels, below -140dBrA, with only a few other peaks visible near -150dBrA. This is an indication that the M6x 200M should not be sensitive to jitter.

J-Test (optical input)

The optical S/PDIF input shows essentially the same result as the coax input. This is an indication that the M6x should not be sensitive to jitter. Both the coaxial and optical inputs were also tested for jitter immunity by injecting artificial sinewave jitter at 2kHz on top of the J-Test test file, which would manifest as sidebands at 10kHz and 14kHz without any jitter rejection. Jitter immunity proved exceptional, with no visible sidebands at even 1000ns of jitter level.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 1)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line-level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 1. The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of the brickwall-type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -115dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Note: the filter characteristics are described under General Information above. Our measured wideband FFT spectrum of white noise and 19.1kHz sinewave tone, generally match the descriptions provided by Musical Fidelity.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 2)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line-level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 2. We see a slower roll-off in the white-noise spectrum compared to filter 1, and slight attenuation of the 19.1kHz signal peak. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -30dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Note: the filter characteristics are described under General Information above. Our measured wideband FFT spectrum of white noise and 19.1kHz sinewave tone, generally match the descriptions provided by Musical Fidelity.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 3)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line-level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 3. The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of the brickwall-type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audio band above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -110dBrA, with subsequent harmonics of the 25kHz peak at Ir below this level.

Note: the filter characteristics are described under General information above. Our measured wideband FFT spectrum of white noise and 19.1kHz sinewave tone, generally match the descriptions provided by Musical Fidelity.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 4)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line-level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 4. We see a slower roll-off in the white noise spectrum compared to filter 1, and slight attenuation of the 19.1kHz signal peak. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -30dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Note: the filter characteristics are described under General Information above. Our measured wideband FFT spectrum of white noise and 19.1kHz sinewave tone, generally match the descriptions provided by Musical Fidelity.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 5)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line-level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 5. The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of the brickwall-type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audio band above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -100dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Note: the filter characteristics are described under General Information above. Our measured wideband FFT spectrum of white noise and 19.1kHz sinewave tone, generally match the descriptions provided by Musical Fidelity.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 6)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line-level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 6. The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of the brick-wall type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -110dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Note: the filter characteristics are described under General Information above. Our measured wideband FFT spectrum of white noise and 19.1kHz sinewave tone, generally match the descriptions provided by Musical Fidelity.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 7)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line-level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 7. The sharp roll-off above 20kHz in the white noise spectrum shows the implementation of the brickwall-type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audio band above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -105dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Note: the filter characteristics are described under General Information above. Our measured wideband FFT spectrum of white noise and 19.1kHz sinewave tone, generally match the descriptions provided by Musical Fidelity.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 8)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line-level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 8, which bypasses the ESS DAC over-sampling filter and engages up-sampling at 352.8kHz. With the up-sampling engaged, high frequency signals exhibit digital clipping at 0dBFS, which explains the all of the harmonics seen in the plot above.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (filter 8, -2dBFS)

The plot above shows a fast Fourier transform (FFT) of the M6x balanced line level output with white noise at -4dBFS (blue/red), and a 19.1kHz sinewave at -2dBFS (to avoid digital clipping) fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using filter 8, which bypasses the ESS DAC over-sampling filter and engages up-sampling at 352.8kHz. The very slow roll-off above 20kHz in the white-noise spectrum shows the lack of a reconstruction oversampling filter. There are nonetheless, absolutely no imaged aliasing artifacts in the audio band above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -20dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Note: the filter characteristics are described under General Information above. Our measured wideband FFT spectrum of white noise and 19.1kHz sinewave tone, generally match the descriptions provided by Musical Fidelity.

THD ratio (unweighted) vs. frequency vs. load (24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms (blue/red) and 600 ohms (purple/green) as a function of frequency for a 24/96 dithered 1kHz signal at the coaxial input. The 200k and 600 ohms data are nearly identical from 100Hz to 5kHz, hovering around a very low 0.0003%. At the frequency extremes, THD increased into 600 ohms vs 200k ohms, where we see 0.0005% vs. 0.0003% at 20Hz, and 0.001% vs. 0.0003% at around 20kHz.

THD ratio (unweighted) vs. frequency vs. sample rate (16/44.1, 24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms as a function of frequency for a 16/44.1 (blue/red) and a 24/96 (purple/green) dithered 1kHz signal at the coaxial input. The 24/96 data consistently outperformed the 16/44.1 data by about 3-4dB, up to 2kHz, above which, both data sets performed identically. THD values for the 24/96 data were either just above, or just below, the very low threshold of 0.0002%.

THD ratio (unweighted) vs. output (16/44.1, 24/96)

The chart above shows THD ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. Once again, the 24/96 outperformed the 16/44.1 data, with a THD range from 0.2% to nearly 0.0001% at just over 1Vrms, while the 16/44.1 ranged from 5% down to 0.0004% at 4.1Vrms. It’s important to highlight that large discrepancies in THD ratios at lower input levels are due to the inherently lower 24-bit noise floor of the 24/96 data (i.e., when no signal harmonics are measurable in an FFT, a THD value can only be assigned as low as the noise floor permits).

THD+N ratio (unweighted) vs. output (16/44.1, 24/96)

The chart above shows THD+N ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. The 24/96 outperformed the 16/44.1 data, with a THD+N range from 4% down to  0.0004%, while the 16/44.1 ranged from 50% down to 0.002% at the maximum output voltage of 4.1Vrms.

Intermodulation distortion vs. generator level (SMPTE, 60Hz:4kHz, 4:1; 16/44.1, 24/96)

The chart above shows intermodulation distortion (IMD) ratios measured at the balanced output as a function of generator input level for the coaxial input into 200k ohms from -60dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. Here, the SMPTE IMD method is used, where the primary frequency (F1 = 60Hz), and the secondary frequency (F2 = 7kHz), are mixed at a ratio of 4:1. The SMPTE IMD analysis results consider the second (F2 ± F1) through the fifth (F2 ± 4xF1) modulation products. The 24/96 outperformed the 16/44.1 data, with an IMD range from 0.05% down to  0.0005% between -10 and -5dBFS, then up to about 0.0007% at 0dBFS, while the 16/44.1 ranged from 2% down to 0.002% at the maximum output voltage of 4.1Vrms at 0dBFS.

FFT spectrum – 1kHz (digital input, 16/44.1 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 16/44.1. We see the third signal harmonic (3kHz) at -115dBrA, or 0.0002%, and subsequent odd harmonics (3, 5, 7, 9kHz) at levels below -120dBrA, or 0.0001%. No even signal harmonics are visible in the audioband above the -135dBrA noise floor. There are also no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 24/96. Due to the increased bit-depth, the noise floor is much lower compared to the 16/44.1 FFT at a very low -160dBrA. We see signal harmonics ranging from -115dBrA, or 0.0002%, at 3kHz, down to -150dBrA. With the lower noise floor, we can see even order harmonics, for example at 2kHz where the peaks (left/right) are just above and below -140dBrA, or 0.00001%. Here also, there are no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 16/44.1 at -90dBFS. We see the main peak at the correct amplitude, and no signal harmonics above the noise floor within the audioband.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 24/96 at -90dBFS. We see the main peak at the correct amplitude, and hint of signal harmonic peaks within the audioband at a vanishingly low -160dBrA, or 0.000001%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 16/44.1)

Shown above is an FFT of the intermodulation distortion (IMD) products for a 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 16/44.1. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 4.1Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is at -130dBRA, or 0.00003%, just barely peaking above the noise floor, while the third-order modulation products, at 17kHz and 20kHz, are slightly higher, reaching -125dBrA, or 0.00006%, for the left channel (the right channel peaks are barely perceptible above the -135dBrA noise floor).

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 24/96)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 24/96. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 4.1Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is just below -130dBRA, or 0.00003%, while the third-order modulation products, at 17kHz and 20kHz are slightly higher, right around -130dBrA.

Diego Estan
Electronics Measurement Specialist

Details

Parent Category: Products

Category: Digital-to-Analog Converter Measurements

Created: 01 April 2023

Link: reviewed by Roger Kanno on SoundStage! Hi-Fi on April 1, 2023

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The Rotel Diamond Series DT-6000 was conditioned for 30 min at 0dBFS (4.3Vrms out) into 200k ohms before any measurements were taken.

The DT-6000’s primary function is that of a CD player; however, because it offers digital inputs, the DT-6000 was evaluated as a standalone DAC. The DT-6000 offers three digital inputs: one coaxial S/PDIF (RCA), one optical S/PDIF (TosLink), and one USB. There are two sets of line-level outputs (balanced XLR and unbalanced RCA).

Comparisons were made between unbalanced and balanced line level outputs for a 24/96 0dBFS input, and aside from the 6dB extra voltage over balanced and a significant difference in output impedance (see primary table below), there were no appreciable differences in THD+N. In terms of digital input types (i.e., USB, coaxial, optical), THD ratios were essentially the same across all three; however, noise levels were about 10dB higher with the USB input (10Hz to 90kHz bandwidth).

Published specifications vs. our primary measurements

The table below summarizes the measurements published by Rotel for the DT-6000 compared directly against our own. The published specifications are sourced from Rotel’s website, either directly or from the manual available for download or included in the box, or a combination thereof. With the exception of frequency response, where the Audio Precision bandwidth was set at its maximum (DC to 1MHz), assume, unless otherwise stated, the coaxial digital input (24/96 1kHz sinewave at 0dBFS), and the worst- case measured result between the left and right channels.

Our primary measurements revealed the following using the coaxial input and the balanced line-level outputs (unless specified, assume a 1kHz sinewave at 0dBFS, 200k ohms loading, 10Hz to 90kHz bandwidth):

Frequency response (16/44.1, 24/96, 24/192)

The plot above shows the DT-6000’s frequency response as a function of sample rate. The blue/red traces are for a 16-bit/44.1kHz dithered digital input signal from 5Hz to 22kHz, the purple/green traces are for a 24/96 dithered digital input signal from 5Hz to 48kHz, and, finally, orange/pink represents 24/192 from 5Hz to 96kHz.

The behavior at low frequencies is the same for the different digital sample rates—perfectly flat down to 5Hz. The behavior at high frequencies for all three digital sample rates is not quite as expected and deviates from Rotel’s published specs of 20Hz-20kHz (+0dB, -0.15dB) and 10Hz-70kHz (+0dB, -3dB). Here we find slightly more high-frequency attenuation than is typical for a brickwall-type filter. At 20kHz, all three sample rates are down -0.55dB. The -3dB point for each sample rate is roughly 21.1, 45.7, and 54.2kHz, respectively. In the graph above and most of the graphs below, only a single trace may be visible. This is because the left channel (blue, purple or orange trace) is performing identically to the right channel (red, green or pink trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Phase response (16/44.1, 24/96, 24/192)

Above are the phase response plots from 20Hz to 20kHz for the coaxial input, measured at the balanced output. The blue/red traces are for a dithered 16/44.1 input at -20dBFS, the purple/green for 24/96, and the orange/pink for 24/192. We can see that the DT-6000 inverts polarity (i.e., -180 degrees of phase shift), with a worst-case phase shift (from the baseline -180 degrees) of about 80 degrees at 12kHz for the 16/44.1 data, 40 degrees for the 24/96 data, and less than 20 degrees for the 24/192 input data.

Digital linearity (16/44.1 and 24/96 to -120dB)

The graph above shows the results of a linearity test for the coaxial digital input (the optical input performed identically) for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the balanced line-level output. The digital input is swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output is analyzed by the APx555. The ideal response is a straight flat line at 0dB (i.e., the amplitude of the digital input perfectly matches the amplitude of the measured analog output). The 24/96 data was perfectly linear down to -120dBFS, while the 16/44.1 data was perfect down to -100dBFS, and only +2dB (left) and +2.5dB (right) at -120dBFS.

Impulse response (24/44.1)

The graph above shows the impulse responses for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period then back to digital silence into a 200k ohm-load for the left channel only. The DT-6000 does not use a typical symmetrical sinc function type filter, but rather one that exhibits no pre-ringing. Another thing to take note is how the plot first moves downward at 53.6ms, followed by an upward movement as it nears 53.7ms, which is the opposite of what is usually seen. That is another indicator that the DT-6000 inverts polarity.

J-Test (coaxial input)

The plot above shows the results of the J-Test test for the coaxial digital input measured at the balanced line-level output of the DT-6000. J-Test was developed by Julian Dunn the 1990s. It is a test signal—specifically,  a -3dBFS undithered 12kHz squarewave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz squarewave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sinewave (the main peak). In addition, an undithered 250Hz squarewave at -144dBFS is mixed with the signal. This test file causes the 22 least significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc). Note that the alternating peaks are in the test file itself, but at levels of -144dBrA (fundamental at 250Hz) down to -170dBrA for the odd harmonics. The test file can also be used in conjunction with artificially injected sinewave jitter by the Audio Precision, to also show how well the DAC rejects jitter.

The coaxial SPDIF input shows some significant peaks in the audioband, with levels reaching nearly -95dBrA. This is an indication that the DT-6000 may be sensitive to jitter.

J-Test (optical input)

The optical S/PDIF input shows a very different—and much better—response to the J-Test than the coaxial input, with the most significant peaks reaching -110dBrA.

J-Test with 10ns of injected jitter (coaxial input)

The coaxial input was also tested for jitter immunity by injecting artificial sinewave jitter at 2kHz on top of the J-Test test file, which would manifest as sidebands at 10kHz and 14kHz without any jitter rejection. Above is an FFT with jitter injected at the 10ns level, and significant peaks can be seen at -70dBrA. This demonstrates that the DT-6000 DAC is quite susceptible to jitter.

J-Test with 10ns of injected jitter (optical input)

The optical input was also tested for jitter immunity by injecting artificial sinewave jitter at 2kHz on top of the J-Test test file. As was the case for the coaxial input, the FFT above shows significant peaks at -70dBrA at the 10kHz and 12kHz positions.

J-Test with 100ns of injected jitter (coaxial input)

Above is an FFT with jitter injected at the 100ns level, and significant peaks can be seen at -50dBrA. This is further evidence that the DT-6000 DAC is susceptible to jitter.

J-Test with 100ns of injected jitter (optical input)

The optical input was also tested for jitter immunity by injecting artificial sinewave jitter at 2kHz on top of the J-Test test file. As was the case for the coaxial input, the FFT above shows significant peaks at -50dBrA at the 10kHz and 12kHz positions.

Wideband FFT spectrum of white noise and 19.1kHz sine-wave tone

The plot above shows a fast Fourier transform (FFT) of the DT-6000 balanced-line level output with white noise at -4 dBFS (blue/red) and a 19.1kHz sinewave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green). The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of a reconstruction filter with relatively steep attenuation. There are absolutely no imaged aliasing artifacts in the audioband above the -120dBrA noise floor. The primary aliasing signal at 25kHz is at -100dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

THD ratio (unweighted) vs. frequency vs. load (24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms (blue/red) and 600 ohms (purple/green) as a function of frequency for a 24/96 dithered 1kHz signal at the coaxial input. The 200k- and 600-ohms data are nearly identical up to about 2kHz, above which the 200k-ohm THD data outperformed the 600-ohm data by about 3dB from 10 to 20kHz. THD ratios are very low from 20 to 500Hz, between 0.0001 and 0.0002%. Above 500Hz, there is a steady rise in THD, up to a peak of 0.002% at 10kHz into 600 ohms.

THD ratio (unweighted) vs. frequency vs. sample rate (16/44.1 and 24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms as a function of frequency for a 16/44.1 (blue/red) and a 24/96 (purple/green) dithered 1kHz signal at the coaxial input. The 24/96 data out-performed the 16/44.1 data by about 5dB from 20Hz to 500Hz, where THD ratios were as low as 0.00015%. Above 1kHz, both THD data sets were identical, reaching a high of only 0.0015% at around 10kHz.

THD ratio (unweighted) vs. output (16/44.1 and 24/96) at 1kHz

The chart above shows THD ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. Once again, the 24/96 outperformed the 16/44.1 data, with a THD range from 1% to just above 0.0003% at 4.3Vrms, while the 16/44.1 ranged from 5% down to 0.0004% at 4.53rms. It’s important to highlight that large discrepancies in THD ratios at lower input levels are due to the inherently lower 24-bit noise floor of the 24/96 data (i.e., when no signal harmonics are measurable in an FFT, a THD value can only be assigned as low as the noise floor permits). The 24/96 data also shows a slight rise in THD between around 50mVrms and 200mVrms.

THD+N ratio (unweighted) vs. output (16/44.1 and 24/96) at 1kHz

The chart above shows THD+N ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. The 24/96 outperformed the 16/44.1 data, with a THD+N range from 10% down to 0.001%, while the 16/44.1 ranged from 50% down to 0.002% at the maximum output voltage of 4.3Vrms.

Intermodulation distortion vs generator level (SMPTE, 60Hz:4kHz, 4:1, 16/44.1, 24/96)

The chart above shows intermodulation distortion (IMD) ratios measured at the balanced output as a function of generator input level for the coaxial input into 200k ohms from -60dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. Here, the SMPTE IMD method was used, where the primary frequency (F1 = 60Hz) and the secondary frequency (F2 = 7kHz) are mixed at a ratio of 4:1. The SMPTE IMD analysis results consider the second (F2 ± F1) through the fifth (F2 ± 4xF1) modulation products. The 24/96 outperformed the 16/44.1 data, with an IMD range from 0.5% down to 0.001%, while the 16/44.1 ranged from 2% down to 0.002% at the maximum output voltage of 4.3Vrms at 0dBFS. The 24/96 data exhibited a slight rise in IMD between -20dBFS and -15dBFS.

FFT spectrum – 1kHz (digital input, 16/44.1 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 16/44.1. We see the second and third signal harmonics (2/3kHz) at -120dBrA, or 0.0001%. There are also higher-order signal harmonics at and below this level. There are no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 24/96. Due to the increased bit depth, the noise floor is much lower compared to the 16/44.1 FFT. We see signal harmonics ranging from -120dBrA (left/right), or 0.0001%, at 2/3/9kHz, and other signal harmonics down to -140dBrA, or 0.00001%. Even with the lower noise floor (-145dBrA), there are still essentially no visible low-level power-supply noise-related peaks on the left side of the main signal peak.

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 16/44.1 at -90dBFS. We see the main peak at the correct amplitude, and no signal or noise harmonics above the noise floor within the audioband.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sinewave stimulus, measured at the balanced output into 200k ohms for the coaxial digital input, sampled at 24/96 at -90dBFS. We see the main peak at the correct amplitude, and no signal or noise harmonics above the noise floor within the audioband.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 16/44.1)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 16/44.1. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 4.3Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is at -115dBRA, or 0.0002%, while the third-order modulation products, at 17kHz and 20kHz, are just above -120dBrA, or 0.0001%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 24/96)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 24/96. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 4.3Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is at -115dBRA, or 0.0002%, while the third-order modulation products, at 17kHz and 20kHz, are just above -120dBrA, or 0.0001%.

Diego Estan
Electronics Measurement Specialist

Details

Parent Category: Products

Category: Digital-to-Analog Converter Measurements

Created: 15 January 2023

Link: reviewed by Matt Bonaccio on SoundStage! Hi-Fi on January 15, 2023

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The RME ADI-2 DAC FS was conditioned for 30 minutes at 0dBFS (3.6Vrms out) into 200k ohms before any measurements were taken.

The ADI-2 DAC FS offers three digital inputs: one coaxial S/PDIF (RCA), one optical SPDIF (TosLink), and one USB. There are two sets of line-level outputs (balanced XLR and unbalanced RCA) and two headphone outputs (1/4″ TRS labelled “PHONES” and 1/8″ TRS labelled “IEM”). There is a digital volume control for the headphone and line-level outputs. Comparisons were made between unbalanced and balanced line-level outputs, and aside from the 6dB extra voltage over balanced, no differences were seen in terms of THD and noise.

The ADI-2 DAC FS offers a dizzying array of features and settings. The following are the default settings used for the coaxial input, balanced line-level outputs, PHONES and IEM headphone outputs, using a 0dBFS input, unless otherwise specified:

Line-level output:

• Ref level: +7dBu (3.6Vrms out over XLR)
• Auto ref level: off
• Mono: off
• Width: 1.0 (full stereo)
• M/S proc: off
• Polarity: off
• Crossfeed: off
• DA filter: SD Sharp
• De-emphasis: Auto
• Dual EQ: off
• Volume: 0dB
• Lock volume: off
• Balance: center
• Loopback to USB: off

PHONES output:

• Hi-power: on
• Auto ref level: off
• Mono: off
• Width: 1.0 (full stereo)
• M/S proc: off
• Polarity: off
• Crossfeed: off
• DA filter: SD Sharp
• De-emphasis: Auto
• Dual EQ: off
• Volume: set to output 2Vrms
• Lock volume: off
• Balance: center
• Loopback to USB: off

IEM output:

• Hi-power: off
• Auto ref level: off
• Mono: off
• Width: 1.0 (full stereo)
• M/S proc: off
• Polarity: off
• Crossfeed: off
• DA filter: SD Sharp
• De-emphasis: Auto
• Dual EQ: off
• Volume: set to output 0.5Vrms
• Lock volume: off
• Balance: center
• Loopback to USB: off

There are six digital filter settings, labelled: Short Delay (SD) Sharp, SD Slow, Sharp, Slow, Non-oversampling (NOS), and Brickwall. Here are RME’s descriptions for each:

• SD Sharp: the most linear frequency response and lowest latency
• SD Slow: causes a small drop in the higher frequency range
• Sharp and Slow: same as SD Sharp and SD Slow respectively but with higher latency but linear phase over the entire audioband
• NOS: smallest steepness and therefore affects treble more than the other filters, but offers the best impulse response
• Brickwall: sharp filtering and phase-linear

The ADI-2 DAC FS volume control ranges from -93.8dB to +6dB, in steps of 6dB to 0.5dB (most of the range is 0.5dB steps). Channel-to-channel deviation proved exceptional, at around 0.01dB throughout the range.

Volume-control accuracy (measured at line-level outputs): left-right channel tracking

Published specifications vs. our primary measurements

The table below summarizes the measurements published by RME for the ADI-2 DAC FS compared directly against our own. The published specifications are sourced from RME’s website, either directly or from the supplied manual, or a combination thereof. With the exception of frequency response, where the Audio Precision bandwidth was set at its maximum (DC to 1MHz), assume, unless otherwise stated, the coaxial digital input (24/96 1kHz sine wave at 0dBFS), the balanced line-level or unbalanced headphone outputs into 200k ohms (line-level) and 300 ohms (headphone) using a measurement input bandwidth of 10Hz to 90kHz, and the worst-case measured result between the left and right channels.

*due to very low noise of DUT, analyzer self-noise has been removed from measurement to more accurately report value

Our primary measurements revealed the following using the coaxial input and the balanced line-level outputs (unless specified, assume a 1kHz sine wave at 0dBFS, 200k ohms loading, 10Hz to 90kHz bandwidth):

Our primary measurements revealed the following using the coaxial input and the headphone output (unless specified, assume a 1kHz sine wave at 0dBFS, 300-ohm loading, 10Hz to 90kHz bandwidth, and 2Vrms output for the PHONES output, and 0.56Vrms for the IEM output): 

*due to very low noise of DUT, analyzer self-noise has been removed from measurement to more accurately report value

Frequency response (16/44.1, 24/96, 24/192 with SD Sharp filter)

The plot above shows the ADI-2 DAC FS’s frequency response as a function of sample rate. The blue/red traces are for a 16-bit/44.1kHz dithered digital input signal from 5Hz to 22kHz, the purple/green traces are for a 24/96 dithered digital input signal from 5Hz to 48kHz, and, finally, orange/pink represents 24/192 from 5Hz to 96kHz. The behavior at low frequencies is the same for the all input resolutions—perfectly flat down to 5Hz. The behavior at high frequencies for all three digital sample rates is as expected, offering sharp filtering around 22k, 48k, and 96kHz (half the respective sample rate), although the 24/192 data shows softer attenuation around the corner frequency. The -3dB point for each sample rate is roughly 21.2, 46.6, and 92.6kHz, respectively. In the graph above and most of the graphs below, only a single trace may be visible. This is because the left channel (blue, purple or orange trace) is performing identically to the right channel (red, green or pink trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Frequency response (bass and treble)

Above are frequency-response plots measured at the balanced outputs with the bass and treble controls set to maximum (blue/red plots) and minimum (purple/green plots). We see that roughly +/- 6dB of gain/cut is available for each.

Frequency response (16/44.1 with SD Sharp, SD Slow, and Sharp fiters)

The plots above show frequency responses for a 0dBFS input signal sampled at 44.1kHz for the SD Sharp (blue), SD Slow filter (red), and Sharp (green) filters into a 200k ohm load for the left channel only. The graph is zoomed in from 1kHz to 22kHz to highlight the various responses of the three filters. We can see that the SD Sharp and Sharp filters offer essentially the same frequency response, with a -3dB point at 21.2 kHz. The SD Slow filter offers a much shallower attenuation, with a -3dB point at 17.4kHz.

Frequency response (16/44.1 with Slow, NOS, and Brickwall fiters)

The plots above show frequency responses for a 0dBFS input signal sampled at 44.1kHz for the Slow (blue), NOS (red), and Brickwall (green) filters into a 200k ohm load for the left channel only. The graph is zoomed in from 1kHz to 22kHz to highlight the various responses of the three filters. We can see that the Slow filter is similar to the SD Slow filter above, but with a -3dB point at 20kHz. The NOS filter, predictably, exhibits significant high-frequency roll-off with a -0.8dB response at 10kHz, and -3.4dB at 20kHz. The Brickwall filter exhibits the most, well, brickwall-type behavior, although with a lower -3dB point (19.8kHz) than the filters shown above.

Phase response vs. sample rate (16/44.1, 24/96, 24/192 with SD Sharp filter)

Above are the phase-response plots from 20Hz to 20kHz for the coaxial input, measured at the balanced output, using the SD Sharp filter setting. The blue/red traces are for a dithered 16/44.1 input at 0dBFS, the purple/green for 24/96, and the orange/pink for 24/192. We can see that the ADI-2 DAC FS does not invert polarity, with a worst-case phase shift of 92 degrees at 13kHz for the 16/44.1, 60 degrees at 20kHz for the 24/96 input data, and just 20 degrees of phase shift at 20kHz for the 24/192 input data.

Phase response vs. filter type (16/44.1 with SD Sharp, SD Slow, and Sharp filters)

Above are the phase-response plots from 20Hz to 20kHz for a 0dBFS input signal sampled at 44.1kHz for the SD Sharp (blue), SD Slow (red), and Sharp (green) filters into a 200k ohm load for the left channel only. The SD Slow and Sharp filters yielded significantly less phase shift than the default SD Sharp filter, with -10 and +40 degrees respectively of phase shift at 20kHz.

Phase response vs. filter type (16/44.1 with Slow, NOS, and Brickwall filters)

Above are the phase-response plots from 20Hz to 20kHz for a 0dBFS input signal sampled at 44.1kHz for the Slow (blue), NOS (red), and Brickwall (green) filters into a 200k ohm load for the left channel only. Predictably, the Brickwall filter yields the highest phase shift at over +180 degrees at 20kHz, the Slow filter is at +120 degrees at 20kHz, while the NOS filter is +80 degrees at 20kHz.

Digital linearity (16/44.1 and 24/96 data)

The graph above shows the results of a linearity test for the coaxial digital input (the optical input performed identically) for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the balanced line-level output of the ADI-2 DAC FS. The digital input is swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output is analyzed by the APx555. The ideal response is a straight flat line at 0dB (i.e., the amplitude of the digital input perfectly matches the amplitude of the measured analog output). The 24/96 input data is essentially pefect down to -120dBFS, while the 16/44.1 input data performed well, only over-responding by 2.5dB at -120dBFS. The 24/96 data yielded such superb results that we extended the sweep down to . . .

Digital linearity (16/44.1 and 24/96 data)

. . . -140dBFS. Above we see that even at -140dBFS, the ADI-2 DAC FS is only undershooting by -1 to -2 dB. This is an exemplary linearity test result.

Impulse response (SD Sharp, SD Slow, Sharp filters)

The graph above shows the impulse responses for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period, then back to digital silence, fed to the coaxial digital input, measured at the balanced outputs, for the SD sharp (blue), SD Slow (red), and Sharp (green) filters into a 200k ohm load for the left channel only. We can see that the SD Sharp filter has no pre-ringing, but significant post-ringing, the SD Slow filter has only very minor post-ringing, and the Sharp filter exhibits symmetrical pre- and post-ringing, as seen in a typical sinc function.

Impulse response (Slow, NOS, Brickwall filters)

The graph above shows the impulse responses for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period, then back to digital silence, fed to the coaxial digital input, measured at the balanced outputs, for the Slow (blue), NOS (red), and Brickwall (green) filters into a 200k ohm load for the left channel only. We can see that the Slow filter has very minor pre- and post-ringing, the NOS filter is as advertised and shows a single pulse with essentially no pre- or post-ringing, and the Brickwall filter exhibits symmetrical pre/post ringing, as seen in a typical sinc function.

Impulse response (NOS filter)

We decided to investigate the impulse response of the NOS filter in more detail. The graph above shows the impulse responses for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s), for one sample period then back to digital silence, fed to the coaxial digital input, measured at the balanced outputs, for the NOS (red) filter only into a 200k ohm load for the left channel only. The graph is zoomed in to show that the NOS filter is as advertised, and yields a single pulse with essentially no pre/post ringing.

J-Test (coaxial input)

The plot above shows the results of the J-Test test for the coaxial digital input measured at the balanced line-level output of the ADI-2 DAC FS. J-Test was developed by Julian Dunn the 1990s. It is a test signal—specifically,  a -3dBFS undithered 12kHz squarewave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz squarewave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sine wave (the main peak). In addition, an undithered 250Hz squarewave at -144dBFS is mixed with the signal. This test file causes the 22 least-significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc). Note that the alternating peaks are in the test file itself, but at levels of -144dBrA (fundamental at 250Hz) down to -170dBrA for the odd harmonics.  The test file can also be used in conjunction with artificially injected sine-wave jitter by the Audio Precision, to also show how well the DAC rejects jitter.

The coaxial input shows an extremely clean J-Test result, with only minor peaks at the -150dBrA level. This is an indication that the ADI-2 DAC FS should not be sensitive to jitter through this input.

J-Test (optical input)

The plot above shows the results of the J-Test test for the coaxial digital input measured at the balanced line-level output of the ADI-2 DAC FS. The optical input shows an extremely clean J-Test result, with only minor peaks at the -150dBrA level. This is an indication that the ADI-2 DAC FS should not be sensitive to jitter through this input.

J-Test (coaxial input, 2kHz sine-wave jitter at 500ns)

The plot above shows the results of the J-Test test for the coaxial digital input (the optical input behaved similarly), measured at the balanced line-level output, with an additional 500ns of 2kHz sine-wave jitter injected by the APx555. The result remains clean with no visible sidebands at 10kHz and 14kHz (12kHz main signal +/-2kHz jitter signal). This is further evidence of the ADI-2 DAC FS’s superb jitter immunity.

J-Test (coaxial input, 2kHz sine-wave jitter at 900ns)

The plot above shows the results of the J-Test test for the coaxial digital input (the optical input behaved similarly), measured at the balanced line-level output, with an additional 900ns of 2kHz sine-wave jitter injected by the APx555. Here sidebands are visible at 10kHz and 14kHz (12kHz main signal +/- 2kHz jitter signal), but remain relatively low at -110dBrA. With jitter above this level, the ADI-2 DAC FS lost sync with the signal.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (SD Sharp filter)

The plot above shows a fast Fourier transform (FFT) of the ADI-2 DAC FS’s balanced line-level output with white noise at -4dBFS (blue/red) and a 19.1kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the SD Sharp filter setting. The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of the brick-wall-type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -105dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (SD Slow filter)

The plot above shows a fast Fourier transform (FFT) of the ADI-2 DAC FS’s balanced line-level output with white noise at -4dBFS (blue/red) and a 19.1kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the SD Slow filter setting. The slow roll-off above 20kHz in the white-noise spectrum shows the implementation of a reconstruction filter with a slow roll-off. Despite this, there are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is near -10dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (Sharp filter)

The plot above shows a fast Fourier transform (FFT) of the ADI-2 DAC FS’s balanced line-level output with white noise at -4dBFS (blue/red) and a 19.1kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Sharp filter setting. The sharp roll-off above 20kHz in the white noise spectrum shows the implementation of the brick-wall-type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -105dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (Slow filter)

The plot above shows a fast Fourier transform (FFT) of the ADI-2 DAC FS’s balanced line-level output with white noise at -4dBFS (blue/red) and a 19.1kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Slow filter setting. The slow roll-off above 20kHz in the white-noise spectrum shows the implementation of a reconstruction filter with slow roll-off. Despite this, there are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is near -10dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (NOS filter)

The plot above shows a fast Fourier transform (FFT) of the ADI-2 DAC FS’s balanced line-level output with white noise at -4dBFS (blue/red) and a 19.1kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the NOS filter setting. Due to the lack of a filter, the noise spectrum is mostly constant (i.e., un-attenuated), except at multiples of the 44.1kHz sample rate. Despite this, the imaged aliasing artifacts in the audioband at 13.2kHz is only at -120dBrA. The primary aliasing signal at 25kHz is near -5dBrA.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (Brickwall filter)

The plot above shows a fast Fourier transform (FFT) of the ADI-2 DAC FS’s balanced line-level output with white noise at -4dBFS (blue/red) and a 19.1 kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Brickwall filter setting. The very sharp roll-off above 20kHz in the white noise spectrum shows the implementation of the brick-wall-type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audioband above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at roughly -100dBrA.

THD ratio (unweighted) vs. frequency vs. load (24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms (blue/red) and 600 ohms (purple/green) as a function of frequency for a 24/96 dithered 1kHz signal at the coaxial input. The 200k- and 600-ohm data are nearly identical throughout the audioband, which is in indication that ADI-2 DAC FS’s outputs are robust and can handle loads below 1k ohms with no difficultly. The right channel does outperform the left by about 5dB from 20Hz to 1kHz; however, at these THD levels (0.00005% to 0.0001%), the differences are of absolutely no consequence. It should also be noted that these THD ratios are pushing up against the limits of the AP analyzer, which exhibits just under 0.00002% THD at 3.6Vrms in loopback mode. Above 3kHz, there is a rise in THD, up to 0.0004% at 20kHz.

THD ratio (unweighted) vs. frequency vs. sample rate (16/44.1 and 24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms as a function of frequency for a 16/44.1 (blue/red) and a 24/96 (purple/green) dithered 1kHz signal at the coaxial input. The 24/96 data consistently outperformed the 16/44.1 data by about 10dB from 20Hz to about 2kHz due to the inherently higher noise floor at 16 bits (i.e., the analyzer cannot assign a THD value below the noise floor). THD ratios with 16/44.1 data range from 0.0002% at 20Hz, down to 0.0001% at 5kHz, back up to 0.0004% at 20kHz. THD ratios with 24/96 data range from 0.00005% at 20Hz, up to 0.0004% at 20kHz.

THD ratio (unweighted) vs. output (16/44.1 and 24/96)

The chart above shows THD ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS at 16/44.1 (blue/red) and 24/96 (purple/green). Once again, the 24/96 outperformed the 16/44.1 data, with a THD range from 0.2% to 0.00005%, while the 16/44.1 ranged from 5% down to nearly 0.0002%.

THD+N ratio (unweighted) vs. output (16/44.1 and 24/96)

The chart above shows THD+N ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS at 16/44.1 (blue/red) and 24/96 (purple/green). The 24/96 outperformed the 16/44.1 data, with a THD+N range from 3% down to  0.0003%, while the 16/44.1 ranged from 40% down to 0.002% at the maximum output voltage of 3.6Vrms.

THD ratio (unweighted) vs. output (24/96) at maximum gain

The chart above shows THD ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS at 24/96 (blue/red), this time with the ADI-2 DAC FS gain set to maximum (i.e., Ref Level set +19dBu, volume set to +6dB). The THD ratios ranged from 0.1% at 0.5mVrms, down to 0.00005% at the “knee” at 7Vrms, with the 1% THD mark hit at roughly 10Vrms at the output.

THD+N ratio (unweighted) vs. output (24/96) at maximum gain

Similar to the chart above, this chart shows THD+N ratios (the addition of noise to THD) measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS at 24/96 (blue/red), with the ADI-2 DAC FS gain set to maximum (i.e., Ref Level set +19dBu, volume set to +6dB). The THD+N ratios ranged from 2% at 0.5mVrms, down to 0.0003% at the “knee” at 7Vrms, with the 1% THD+N mark hit at roughly 10Vrms at the output.

FFT spectrum – 1kHz (digital input, 16/44.1 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 16/44.1. We see no sign of signal harmonics above the -135dBrA noise floor. There are also no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 24/96. Due to the increased bit-depth, the noise floor is much lower compared to the 16/44.1 FFT, at a very low -160dBrA. We see very low signal harmonics ranging from -130dBrA, or 0.00003%, at 7kHz, down to below -150dBrA, or 0.000003%. Here also, there are no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 16/44.1 at -90dBFS. We see the signal peak at the correct amplitude, and no signal harmonics above the noise floor within the audioband.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 24/961 at -90dBFS. We see the signal peak at the correct amplitude, and no signal harmonics above the noise floor within the audioband.

Intermodulation distortion vs generator level (SMPTE, 60Hz:4kHz, 4:1, 16/44.1, 24/96)

The chart above shows intermodulation distortion (IMD) ratios measured at balanced output for 16/44.1 (blue/red) and 24/96 input data (purple/green), from -60dBFS to 0dBFS. Here, the SMPTE IMD method is used, where the primary frequency (F1 = 60Hz), and the secondary frequency (F2 = 7kHz), are mixed at a ratio of 4:1. The SMPTE IMD analysis results consider the second (F2 ± F1) through the fifth (F2 ± 4xF1) modulation products. The 16/44.1 data yields IMD ratios from 2% down to 0.002% at 0dBFS. The 24/96 data yields IMD ratios from 0.1% down to 0.0003% at 0dBFS.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 16/44.1)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sine-wave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 16/44.1. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 3.6Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz), as well as the third-order modulation products, at 17kHz and 20kHz, are not visible above the -135dBrA noise floor.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 24/96)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sine-wave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 24/96. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 3.6Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is at -135dBrA, or 0.00002%, for the left channel (right channel peak cannot be seen above -150dBrA noise floor), while the third-order modulation products, at 17kHz and 20kHz, are slightly higher, at around -120dBrA to -130dBrA, or 0.0001% to 0.00003%. This is an exceptionally clean IMD result.

Diego Estan
Electronics Measurement Specialist

Details

Parent Category: Products

Category: Digital-to-Analog Converter Measurements

Created: 01 July 2022

Link: reviewed by Gordon Brockhouse on SoundStage! Simplifi on July 1, 2022

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The Neo S was conditioned for 30 minutes at 0dBFS (4.5Vrms out) into 200k ohms before any measurements were taken.

The Neo S offers four digital inputs: one coaxial S/PDIF (RCA), one optical S/PDIF, one AES/EBU (XLR), and one USB. There are two line-level outputs (balanced XLR and unbalanced RCA) and two headphone outputs (1/8″ TRS unbalanced and 3.4mm TRRS balanced). There are also LAN (over ethernet) and Bluetooth inputs, as well as HDMI and coaxial digital outputs. There is also a digital volume control. Comparisons were made between unbalanced and balanced line-level outputs for a 24/96 0dBFS input, and aside from the 6dB extra voltage over balanced, there were no appreciable differences in THD+N. In terms of input types (USB, coaxial, optical), there were no differences in terms of THD+N at 24/96 resolution.

The Neo S offers three different digital filter settings, accessible through the touchscreen user menu. These are: Fast Rolloff, Slow Rolloff Minimum Phase, and Minimum Phase Corrected.

The Neo S volume control can provide adjustments in 0.5, 1, 2, or 3dB steps. The step value size can be selected in the user menu. The range is -60dB to 0dB. At -60dB, the output is effectively muted; at -59.5dB, the output over the balanced connectors measured 4.8mVrms; and at 0dB, the output from the balanced connectors measured 4.5Vmrs.  Using the headphone outputs does not offer any further gain/output voltage. The volume control is implemented in the digital domain, as every step was exactly 0.5dB, and the channel-to-channel deviation was exactly 0.035-0.036dB at every step, throughout the range, as seen in the table below.

Unless otherwise stated, all measurements are with the coaxial digital input, balanced outputs, the Fast Roll-Off filter, and the volume set to 0dB.

Volume-control accuracy (measured at line-level outputs): left-right channel tracking

Published specifications vs. our primary measurements

The table below summarizes the measurements published by Zidoo for the Neo S compared directly against our own. The published specifications are sourced from Zidoo’s website, either directly or from the manual available for download or included in the box, or a combination thereof. With the exception of frequency response, where the Audio Precision bandwidth is set at its maximum (DC to 1MHz), assume, unless otherwise stated, the coaxial digital input (24/96 1kHz sine wave at 0dBFS), the line-level or headphone outputs into 200k ohms (line-level) and 300/32 ohms (headphone high/low gain), using a measurement input bandwidth of 20Hz to 20kHz, and the worst-case measured result between the left and right channels.

Our primary measurements revealed the following using the coaxial input and the balanced line-level outputs (unless specified, assume a 1kHz sine wave at 0dBFS, 200k ohms loading, 10Hz to 90kHz bandwidth):

Our primary measurements revealed the following using the coaxial input and the balanced headphone output (unless specified, assume a 1kHz sine wave at 0dBFS input, 4.5Vrms into 300 ohms, 10Hz to 90kHz bandwidth):

High gain setting

Low gain setting

Frequency response (16/44.1, 24/96, 24/192)

The plot above shows the Neo S frequency response as a function of sample rate. The blue/red traces are for a 16-bit/44.1kHz dithered digital input signal from 5Hz to 22kHz, the purple/green traces are for a 24/96 dithered digital input signal from 5Hz to 48kHz, and finally orange/pink represents 24/192 from 5Hz to 96kHz. The behavior at low frequencies is the same for all sample rates—perfectly flat down to 5Hz. The behavior at high frequencies for all three digital sample rates is as expected, offering sharp filtering around 22, 48, and 96kHz (half the respective sample rate). The -3dB point for each sample rate is roughly 21, 46.2, and 92.2kHz, respectively. It is also obvious from the plots above that all three sample rates offered “brick-wall”- type behavior with the default Fast Rolloff filter. In the graph above and most of the graphs below, only a single trace may be visible. This is because the left channel (blue, purple or orange trace) is performing identically to the right channel (red, green or pink trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Frequency response (16/44.1, all three filters)

The plots above show frequency-response for a 16/44.1 input for all three filters (Fast Rolloff, Slow Rolloff Minimum Phase, and Minimum Phase Corrected) into a 200k ohm load for the left channel only. Fast Rolloff is in blue, Slow Rolloff Minimum Phase in purple, and Minimum Phase Corrected is in red. The graph is zoomed in from 1kHz to 22kHz, to highlight the various responses of the filters around the “knee” of the response. At 20kHz, the Fast Rolloff filter is at -0.03dB, the Slow Rolloff Minimum Phase filter is at -5dB, and the Minimum Phase Corrected filter is at -12.3dB.

Phase response (16/44.1, 24/96, 24/192 with Fast Rolloff filter)

Above are the phase response plots from 20Hz to 20kHz for the coaxial input, measured at the balanced output, using the Fast Rolloff filter. The blue/red traces are for a dithered 16/44.1 input at -20dBFS, the purple/green for 24/96, and the orange/pink for 24/192. We can see that the NEO S does not invert polarity, with a worst-case phase shift of just under 160 degrees at 20kHz for the 16/44.1 data, and within +/-20 degrees or so for the 24/96 and 24/192 input data.

Phase response (16/44.1, all three filters)

Above are the phase response plots from 20Hz to 20kHz for a 16/44.1 signal at the coaxial input, measured at the balanced output, for all three filters (Fast Rolloff, Slow Rolloff Minimum Phase, and Minimum Phase Corrected) into a 200k ohm load for the left channel only. Fast Rolloff is in blue, Slow Rolloff Minimum Phase in purple and Minimum Phase Corrected in red. We see that both the Slow Rolloff Minimum Phase and Minimum Phase Corrected filters exhibit far less phase shift between 5kHz and 20kHz than the Fast Rolloff filter.

Digital linearity (16/44.1 and 24/96 to -120dB)

The graph above shows the results of a linearity test for the coaxial digital input (the optical input performed identically) for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the balanced line-level output. The digital input is swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output is analyzed by the APx555. The ideal response is a straight flat line at 0dB (i.e., the amplitude of the digital input perfectly matches the amplitude of the measured analog output). The 24/96 data is perfectly linear down to -120dBFS, while the 16/44.1 data was perfect down to -100dBFS, and only +3dB (left) and +1dB (right) at -120dBFS. Because of this DAC’s exceptional linearity performance, a second measurement was performed extending down to -140dBFS, plotted in the chart below.

Digital linearity (16/44.1 and 24/96 to -140dB)

This shows the 24/96 data remained within 0.5dB or so of flat to -140dB, which is a phenomenal result. The 16/44.1 data results are as expected, showing significant deviations below -120dBFS. But it is worth highlighting that a linearity result that is flat down to -100dBFS for 16-bit input data is also exceptional.

Impulse response (24/44.1, all three filters)

The graph above shows the impulse responses for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period then back to digital silence for all three filters (Fast Rolloff, Slow Rolloff Minimum Phase, and Minimum Phase Corrected) into a 200k ohm load for the left channel only. Fast Rolloff is in blue, Slow Rolloff Minimum Phase in purple, and Minimum Phase Corrected in red. The default Fast Rolloff filter exhibits a typical sinc function, with symmetrical pre- and post-ringing behavior. The Slow Rolloff Minimum Phase filter exhibits no pre-ringing and very little post-ringing, where the Minimum Phase Corrected filter is somewhere in between the other two.

J-Test (coaxial input)

The plot above shows the results of the J-Test test for the coaxial digital input measured at the balanced line-level output of the Neo S. J-Test was developed by Julian Dunn the 1990s. It is a test signal—specifically, a -3dBFS undithered 12kHz square wave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz square wave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sine wave (the main peak). In addition, an undithered 250Hz square wave at -144dBFS is mixed with the signal. This test file causes the 22 least-significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc). Note that the alternating peaks are in the test file itself, but at levels of -144dBrA (fundamental at 250Hz) down to -170dBrA for the odd harmonics.  The test file can also be used in conjunction with artificially injected sinewave jitter by the Audio Precision, to also show how well the DAC rejects jitter.

The coaxial S/PDIF input shows some of the alternating 500Hz peaks in the audio band but at low levels; below -130dBrA. This is an indication that the Neo S should not be sensitive to jitter.

J-Test (optical input)

The optical S/PDIF input shows essentially the same result as with the coaxial input above.

J-Test with 10ns of injected jitter (coaxial input)

Both the coaxial and optical inputs were also tested for jitter immunity by injecting artificial sine wave jitter at 2kHz on top of the J-Test test file, which would manifest as sidebands at 10kHz and 14kHz without any jitter rejection. Above is an FFT with jitter injected at the 10ns level, and peaks can be seen at -125dBrA. This demonstrates that the Neo S DAC’s jitter rejection is not as robust as the J-test result alone would have indicated. Only the coaxial input is shown because the optical input showed basically the same result.

J-Test with 100ns of injected jitter (coaxial input)

Both the coaxial and optical inputs were also tested for jitter immunity by injecting artificial sine wave jitter at 2kHz on top of the J-Test test file, which would manifest as sidebands at 10kHz and 14kHz without any jitter rejection. Above is an FFT with jitter injected at the 100ns level, and peaks can be seen at -105dBrA. This demonstrates that the Neo S DAC’s jitter rejection is not as robust as the J-Test result alone would have indicated. Again, the optical input showed pretty much the same result.

Wideband FFT spectrum of white noise and 19.1kHz sine-wave tone (Fast Rolloff filter)

The plot above shows a fast Fourier transform (FFT) of the Neo S balanced-line level output with white noise at -4dBFS (blue/red) and a 19.1kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Fast Rolloff filter. The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of the brick-wall type reconstruction filter. There are absolutely no imaged aliasing artifacts in the audio band above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -85dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Wideband FFT spectrum of white noise and 19.1kHz sine-wave tone (Slow Rolloff Minimum Phase filter)

The plot above shows a fast Fourier transform (FFT) of the Neo S balanced-line level output with white noise at -4dBFS (blue/red) and a 19.1kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Slow Rolloff Minimum Phase filter. The roll-off above 20kHz in the white-noise spectrum is shallower than what is seen with the filters above and below. There are absolutely no imaged aliasing artifacts in the audio band above the -135dBrA noise floor. The primary aliasing signal at 25kHz is at -35dBrA, with subsequent harmonics of the 25kHz peak at or below this level.

Wideband FFT spectrum of white noise and 19.1kHz sine-wave tone (Minimum Phase Corrected filter)

The plot above shows a fast Fourier transform (FFT) of the Neo S balanced-line level output with white noise at -4dBFS (blue/red) and a 19.1 kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green), using the Minimum Phase Corrected filter. The FFT is very similar to the one for the Fast Rolloff filter.

THD ratio (unweighted) vs. frequency vs. load (24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms (blue/red) and 600 ohms (purple/green) as a function of frequency for a 24/96 dithered 1kHz signal at the coaxial input. The 200k and 600 ohms data are nearly identical; however, the right channel, at 0.0002% and below, did outperform the left channel, which was at 0.0003%. In either case, these are extremely low levels of THD. These data also demonstrate that the Neo S’s line-level outputs are robust and can handle lower impedance loads.

THD ratio (unweighted) vs. frequency vs. sample rate (16/44.1 and 24/96)

The chart above shows THD ratios at the balanced line-level output into 200k ohms as a function of frequency for a 16/44.1 (blue/red) and a 24/96 (purple/green) dithered 1kHz signal at the coaxial input. The 24/96 data (right channel) consistently outperformed the 16/44.1 data by about 5dB. Still, all THD values are very low, between 0.0005% and 0.00015%.

THD ratio (unweighted) vs. output (16/44.1 and 24/96) at 1kHz

The chart above shows THD ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. Once again, the 24/96 outperformed the 16/44.1 data, with a THD range from 0.2% to nearly 0.0001% (right channel) at 4.5Vrms, while the 16/44.1 ranged from 5% down to 0.0004% at 4.5Vrms. It’s important to highlight that large discrepancies in THD ratios at lower input levels are due to the inherently lower 24-bit noise floor of the 24/96 data (i.e., when no signal harmonics are measurable in an FFT, a THD value can only be assigned as low as the noise floor permits). The 24/96 data also shows a slight rise in THD between around 100mVrms and 1Vrms.

THD+N ratio (unweighted) vs. output (16/44.1 and 24/96) at 1kHz

The chart above shows THD+N ratios measured at the balanced output as a function of output voltage for the coaxial input into 200k ohms from -90dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. The 24/96 outperformed the 16/44.1 data, with a THD+N range from 3% down to  0.0005%, while the 16/44.1 ranged from 50% down to 0.002% at the maximum output voltage of 4.5Vrms.

Intermodulation distortion vs generator level (SMPTE, 60Hz:4kHz, 4:1, 16/44.1, 24/96)

The chart above shows intermodulation distortion (IMD) ratios measured at the balanced output as a function of generator input level for the coaxial input into 200k ohms from -60dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. Here, the SMPTE IMD method was used, where the primary frequency (F1 = 60Hz) and the secondary frequency (F2 = 7kHz) are mixed at a ratio of 4:1. The SMPTE IMD analysis results consider the second (F2 ± F1) through the fifth (F2 ± 4xF1) modulation products. The 24/96 outperformed the 16/44.1 data, with an IMD range from 0.06% down to  0.0005% between -10 and -5dBFS, then up to about 0.001% at 0dBFS, while the 16/44.1 ranged from 2% down to 0.002% at the maximum output voltage of 4.5Vrms at 0dBFS. The 24/96 data exhibited a slight rise in IMD between -30dBrA and -15dBrA.

FFT spectrum – 1kHz (digital input, 16/44.1 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the balanced output into 200k ohm for the coaxial digital input, sampled at 16/44.1. We see the third signal harmonic (3kHz) at -110/-120dBrA (left/right), or 0.0003/0.0001%. The second and fifth signal harmonics for the left channel are visible at -125dBrA, or 0.00006%, just above the noise floor. There are also no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the balanced output into 200k ohms for the coaxial digital input, sampled at 24/96. Due to the increased bit-depth, the noise floor is much lower compared to the 16/44.1 FFT. We see signal harmonics ranging from -110/-120dBrA (left/right), or 0.0003/0.0001% at 3kHz, down to -140dBrA, or 0.00001%. With the lower noise floor, we can see higher even order harmonics, for example at 4 and 6kHz where the peaks are just below -140dBrA, or 0.00001%. Here we see low level peaks on the left side of the main signal peak, at -130dBrA, or 0.00003%, and below.

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the balanced output into 200k ohms for the coaxial digital input, sampled at 16/44.1 at -90dBFS. We see the main peak at the correct amplitude, and no signal harmonics above the noise floor within the audio band.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the balanced output into 200k ohms for the coaxial digital input, sampled at 24/96 at -90dBFS. We see the main peak at the correct amplitude, and power-supply related harmonics at 60Hz, 180Hz, 300Hz, etc., at -130dBrA, or 0.00003%, and below.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 16/44.1)

Shown above is an FFT of the intermodulation (IMD) products for an 18kHz + 19kHz summed sine-wave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 16/44.1. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 4.5Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) cannot be seen above the -130dBRA, or 0.00003%, noise floor, while the third-order modulation products, at 17kHz and 20kHz, are just above (left) and below (right) -120dBrA, or 0.0001%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 24/96)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sine-wave stimulus tone measured at the balanced output into 200k ohms for the coaxial input at 24/96. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 4.5Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is at a vanishingly low -140dBRA, or 0.00001%, while the third-order modulation products, at 17kHz and 20kHz, are just above (left) and below (right) -120dBrA, or 0.0001%.

Diego Estan
Electronics Measurement Specialist

Details

Parent Category: Products

Category: Digital-to-Analog Converter Measurements

Created: 01 June 2022

Link: reviewed by Dennis Burger on SoundStage! Access on June 1, 2022

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The iFi Audio Zen One Signature was conditioned for 30 min at 0dBFS (2.1Vrms out) into 100k ohms before any measurements were taken.

The Zen One Signature offers four digital inputs: one coaxial S/PDIF (RCA), one optical S/PDIF, one USB, and Bluetooth. There are two line-level outputs (balanced 3.4mm TRRS and unbalanced RCA) and one digital output (coaxial, over the same RCA connector used for the coaxial digital input). Comparisons were made between unbalanced and balanced line-level outputs for a 24/96 0dBFS input, and aside from the 6dB extra voltage over balanced, there were virtually no differences in THD+N and dynamic range. In terms of input types (USB, coaxial, optical), there were no differences in terms of THD+N.

Unless otherwise stated, all measurements are with the coaxial digital input and unbalanced outputs.

Published specifications vs. our primary measurements

The table below summarizes the measurements published by iFi Audio for the Zen One Signature compared directly against our own. The published specifications are sourced from iFi’s website, either directly or from the manual available for download or included in the box, or a combination thereof. With the exception of frequency response, where the Audio Precision bandwidth is set at its maximum (DC to 1MHz), assume, unless otherwise stated, the coaxial digital input (24/96 1kHz sine wave at 0dBFS), the unbalanced line-level output into 100k ohms (line-level), using a measurement input bandwidth of 10Hz to 90kHz, and the worst-case measured result between the left and right channels.

Our primary measurements revealed the following using the coaxial input and the unbalanced line-level outputs (unless specified, assume a 1kHz sine wave at 0dBFS, 100k ohms loading, 10Hz to 90kHz bandwidth):

Frequency response (16/44.1, 24/96, 24/192)

The plot above shows the Zen One Signature frequency response as a function of sample rate. The blue/red traces are for a 16-bit/44.1kHz dithered digital input signal from 5Hz to 22kHz, the purple/green traces are for a 24/96 dithered digital input signal from 5Hz to 48kHz, and, finally, orange/pink represents 24/192 from 5Hz to 96kHz. The behavior at low frequencies is the same for the three sample rates—essentially perfectly flat down to 5Hz. The behavior at high frequencies for all three digital sample rates is as expected, offering filtering around 22k, 48k, and 96kHz (half the respective sample rate). The -3dB point for each sample rate is roughly 21.1, 46.2, and 83kHz, respectively. It is also obvious from the plots above that the 44.1kHz sampled input signal offers the most “brick-wall”-type behavior, while the attenuation of the 96kHz and 192kHz sampled input signals approaching the corner frequencies (48kHz and 96kHz) is more gentle. In the graph above and most of the graphs below, only a single trace may be visible. This is because the left channel (blue, purple or orange trace) is performing identically to the right channel (red, green or pink trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Digital linearity (16/44.1 and 24/96 data)

The graph above shows the results of a linearity test for the coaxial digital input (the optical input performed identically) for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the unbalanced line-level output. The digital input is swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output is analyzed by the APx555. The ideal response is a straight flat line at 0dB (i.e., the amplitude of the digital input perfectly matches the amplitude of the measured analog output). The 24/96 data is only -1dB at -120dBFS, while the 16/44.1 data was perfect down to -100dBFS, and only +2.5dB (left) at -120dBFS. This is an excellent result.

Impulse response

The graph above shows the impulse response for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period then back to digital silence. We see a typical sinc function response.

J-Test (coaxial input)

The plot above shows the results of the J-Test test for the coaxial digital input measured at the unbalanced line-level output of the Zen One Signature. The J-Test was developed by Julian Dunn the 1990s. It is a test signal, specifically a -3dBFS undithered 12kHz square wave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz square wave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sinewave (the main peak). In addition, an undithered 250Hz square wave at -144dBFS is mixed with the signal. This test file causes the 22 least significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc). Note that the alternating peaks are in the test file itself, but at levels of -144dBrA (fundamental at 250Hz) down to -170dBrA for the odd harmonics.  The test file can also be used in conjunction with artificially injected sine-wave jitter by the Audio Precision, to also show how well the DAC rejects jitter.

The coaxial S/PDIF input shows worst case peaks at -130dBrA. This is an indication that the Zen One Signature should not be sensitive to jitter.

J-Test (optical input)

The optical S/PDIF input shows essentially the same result as with the coaxial input. Both the coaxial and optical inputs were also tested for jitter immunity by injecting artificial sine-wave jitter at 2kHz on top of the J-Test file, which would manifest as sidebands at 10kHz and 14kHz without any jitter rejection. Jitter immunity proved exceptional, with no visible sidebands at even 1000ns of jitter level.

Wideband FFT spectrum of white noise and 19.1kHz sine-wave tone

The plot above shows a fast Fourier transform (FFT) of the Zen One Signature unbalanced line-level output with white noise at -4 dBFS (blue/red), and a 19.1 kHz sine wave at 0dBFS fed to the coaxial digital input, sampled at 16/44.1 (purple/green). The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of the brick-wall-type reconstruction filter. There are a few imaged aliasing artifacts in the audio band, the most predominant at 17kHz at -110dBrA. The primary aliasing signal at 25kHz is at -80dBrA, with subsequent harmonics of the 25kHz peak slightly above this level.

THD ratio (unweighted) vs. frequency vs. load (24/96)

The chart above shows THD ratios at the unbalanced line-level output into 100k ohms (blue/red) and 600 ohms (purple/green) as a function of frequency for a 24/96 dithered 1kHz signal at the coaxial input. There’s a left/right THD imbalance into 100k ohms from 30Hz to 5kHz or so, with the left channel (blue) outperforming the right channel (red) by as much as almost 10dB. Into 600 ohms, the difference in THD between left and right was much smaller, about 3dB. In general, higher THD ratios were observed into 600 ohms, ranging from roughly 0.005% at low frequencies to 0.04% at 20kHz. This compared to the 100k ohm data, which ranged from as low as 0.001% (left channel from 100-300 Hz), to 0.04% at 20kHz.

THD ratio (unweighted) vs. frequency vs. sample rate (16/44.1 and 24/96)

The chart above shows THD ratios at the unbalanced line-level output into 100k ohms as a function of frequency for a 16/44.1 (blue/red) and a 24/96 (purple/green) dithered 1kHz signal at the coaxial input. All data tracked closely, with THD ratios ranging from 0.002% at 20Hz, down to 0.001% from 100-300Hz, then up to 0.04% at 20kHz. The exception is the right channel at 24/96, which yielded THD values almost 10dB higher from 100Hz to 500Hz or so.

THD ratio (unweighted) vs. output (16/44.1 and 24/96) at 1kHz

The chart above shows THD ratios measured at the unbalanced output as a function of output voltage for the coaxial input into 100k ohms from -90dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. The 24/96 outperformed the 16/44.1 data at lower output levels, with a THD range from 1.5% to nearly 0.0005% at 0.2-0.3Vrms, while the 16/44.1 data ranged from 4% down to 0.001% at 1Vrms. It’s important to highlight that large discrepancies in THD ratios at lower input levels are due to the inherently lower 24-bit noise floor of the 24/96 data (i.e., when no signal harmonics are measurable in an FFT, a THD value can only be assigned as low as the noise floor permits). At output voltages above 0.4Vrms or so, again we see the right channel with higher THD values than the left.

THD+N ratio (unweighted) vs. output (16/44.1 and 24/96) at 1kHz

The chart above shows THD+N ratios measured at the unbalanced output as a function of output voltage for the coaxial input into 100k ohms from -90dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. The 24/96 outperformed the 16/44.1 data, with a THD+N range from 30% down to  0.003%, while the 16/44.1 ranged from 40% down to 0.004% at the maximum output voltage of 2.1Vrms.

Intermodulation distortion vs generator level (SMPTE, 60Hz:4kHz, 4:1, 16/44.1, 24/96)

The chart above shows intermodulation distortion (IMD) ratios measured at the unbalanced output as a function of generator input level for the coaxial input into 100k ohms from -60dBFS to 0dBFS with 16/44.1 (blue/red) and 24/96 (purple/green) dithered input data. Here, the SMPTE IMD method is used, where the primary frequency (F1 = 60Hz), and the secondary frequency (F2 = 7kHz), are mixed at a ratio of 4:1. The SMPTE IMD analysis results consider the second (F2 ± F1) through the fifth (F2 ± 4xF1) modulation products. The 24/96 outperformed the 16/44.1 data, with an IMD range from 0.5% down to  0.003% between -15 and 0dBFS, although, again, the right channel performed worse (almost 10dB) at these higher generator levels. The 16/44.1 data ranged from 2% down to roughly 0.005% at the maximum output voltage of 2.1Vrms at 0dBFS, with the left channel slightly outperforming the right above -10dBFS.

FFT spectrum – 1kHz (digital input, 16/44.1 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the unbalanced output into 100k ohms for the coaxial digital input, sampled at 16/44.1. We see the second signal harmonic (2kHz) at -95dBrA, or 0.002%, and subsequent harmonics (3, 4, 5, 6, 7kHz, etc.) at descending lower levels from -100dBrA, or 0.001%, down to below -120dBrA, or 0.0001%. There are no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the unbalanced output into 100k ohms for the coaxial digital input, sampled at 24/96. We consistently see the right channel signal harmonic peaks, 5-10dB higher than the left peaks. The left channel signal harmonic peaks essentially match what was measured at 16/44.1 (shown above). There are no power-supply noise peaks to speak of to the left of the main signal peak.

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the unbalanced output into 100k ohm for the coaxial digital input, sampled at 16/44.1 at -90dBFS. We see the main peak at the correct amplitude, and no signal harmonics above the noise floor within the audio band.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is the fast Fourier transform (FFT) for a 1kHz input sine-wave stimulus, measured at the unbalanced output into 100k ohm for the coaxial digital input, sampled at 24/96 at -90dBFS. We see the main peak at the correct amplitude, and a hint of signal harmonic peak within the audio band at 3kHz at a very low -140dBrA, or 0.00001%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 16/44.1)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sine-wave stimulus tone measured at the unbalanced output into 100k ohms for the coaxial input at 16/44.1. The input dBFS values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 2.1Vrms (0dBrA) at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is just below -90dBRA, or 0.0003%, while the third-order modulation products, at 17kHz and 20kHz, are slightly higher, reaching -85dBrA, or 0.0006%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, 24/96)

Shown above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sine-wave stimulus tone measured at the unbalanced output into 100k ohms for the coaxial input at 24/96. The input dBFS values are set at -6.02dBFS, so that, if summed for a mean frequency of 18.5kHz, would yield 2.1Vrms (0dBrA) at the output. Here again, we find higher distortion peaks for the right channel compared to the left, by about 5dB. Otherwise, the FFT looks essentially the same at the 16/44.1 IMD FFT above.

Diego Estan
Electronics Measurement Specialist