• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DAC measurements using DeltaWave

manisandher

Addicted to Fun and Learning
Joined
Nov 6, 2016
Messages
800
Likes
810
Location
Royal Leamington Spa, UK
With the Christmas break, I finally managed to find some time to compare the analogue outputs of some DACs playing real music (as opposed to sine waves), using @pkane 's excellent DeltaWave nulling software.

In the next post I’ll outline the equipment and method I’ve used. But I’d like to reserve this OP for a ‘league table’ of DACs.

Here are the results to date (10 samples each):

1736084835351.png


I find the charts in DeltaWave to be more insightful.

Wadia 15 (4x PCM63-K)
1736084895144.png


SMSL SU-10 – no filter (2x ES9038Pro)
1736084934282.png


SMSL SU-10 - fast linear
1736084981626.png


Chord DAVE (custom FPGA)
1736085025273.png


The Chord DAVE actually performs the best in these tests, even though it falls far short of the SMSL SU-10 in THD+N (see https://audiosciencereview.com/forum/index.php?threads/smsl-su-10-dac-review.38415/ and https://audiosciencereview.com/forum/index.php?threads/chord-dave-review-dac-hp-amp.35974/).

Due to the requirement to synchronise clocks (to obtain accurate nulls), I’m restricted to DACs with spdif inputs. Other DACs that I have here that I’ll get around to testing eventually include:
  • Gustard X30 (4x ES9039Pro)
  • SMSL DL200 (ES90392QM)
  • Audio Synthesis DAX (2x PCM63-K)
  • Audio Synthesis DSM (UltraAnalog D20400A)
All thoughts welcome.
 
Here's the setup:

1736112536243.png


Here's the 24/44.1 'reference' file:
24_44.1 reference file.jpg

Roon
  • No DSP (bit-perfect output)
DAC
  • AES input
  • set to variable output
  • analogue output via XLR
RME
  • set to M/S-Proc
  • +13dBu (referenced to 0 dBFS)
  • clock source = spdif
  • analogue input via split XLR
Method:
  1. Play 24/44.1 -2dB 1kHz sine tone.
  2. Set DAC's digital volume control so that RME reads approx. -2dB. (DeltaWave will make required gain adjustment to 'compare' file.)
  3. Play 24/44.1 'reference' file.
  4. Capture 24/44.1 'compare' file on Tascam SD card.
  5. Repeat steps 3 and 4 ten times.
  6. Load 'reference' and 'compare' files into DeltaWave.
The DeltaWave settings are as follows:
Standard DeltaWave Setup.JPG

Where the DAC has different filter settings, I have taken measurements for each filter.

I'm sure I've forgotten to mention a bunch of details, so feel free to ask.
 

Attachments

  • 1736107297567.png
    1736107297567.png
    36 KB · Views: 77
Last edited:
Due to the requirement to synchronise clocks (to obtain accurate nulls), I’m restricted to DACs with spdif inputs.
I thought Deltawave could time align different recordings to eliminate the need for clock sync - or have I misunderstood?

Also - I can't read the scales on the charts - is it possible to include higher res versions?

Looking forward to the test method post. :D
 
Just a few very quick replies...

Could you provide numbers both with and without the Non-linear EQ feature enabled?

All measurements are currently taken with the following DeltaWave settings:
Standard DeltaWave Setup.JPG

Happy to try the non-linear EQ settings, but wouldn't want to mask anything the DAC itself is doing.

I thought Deltawave could time align different recordings to eliminate the need for clock sync - or have I misunderstood?

DeltaWave will certainly time align the two files. And yes, it has a very good "clock drift" function. However, the best nulls are still achieved by synchronising the DAC and ADC clocks.

Also - I can't read the scales on the charts - is it possible to include higher res versions?

I intend to create a separate post for each DAC, where I'll show the charts in more detail. But for now, the scales in the OP are:

- Delta Waveform: plus/minus 30dB
- Spectrum of Delta: 0dB to -160dB
- PK Metric: -60dB to <-300dB

Can you provide a magnitude and phase measurement of each in REW? That will show what is important for achieving a good null.

Unfortunately, it's been many, many years since I used REW.

If you cannot, the Delta of Spectra and Delta Phase plots can work as lower resolution substitute measurements.

Not sure how useful these are as they all seem to be very similar for all the DACs. But here are a couple for the Wadia 15:

1736091475471.png
1736091498951.png

(Note the scales.)
 
Not sure how useful these are as they all seem to be very similar for all the DACs. But here are a couple for the Wadia 15:

Because they show you where the errors are occurring. On that plot you have clear magnitude response deviation above 10 kHz. Phase is mostly flat (linear) but there is definitely some deviation. These errors can be corrected with EQ.

Put another way, what do you think affects the null more, noise or distortion that is 100 dB below the fundamental or a 4 dB magnitude response error?

Michael
 
Nice work. Are you comparing the source file to the DAC? (what is your reference)

For DeltaWave, consider adding the log as a spoiler tag in each post and using dBr instead of dBFS for your pkmetric graphs.
 
Are you comparing the source file to the DAC? (what is your reference)

I'll go into more detail in the setup/method post, but essentially, it's a 30-second 24/44.1 (quite dynamic) segment from a jazz track.

For DeltaWave, consider adding the log as a spoiler tag in each post and using dBr instead of dBFS for your pkmetric graphs.

Thanks. Will do.
 
Put another way, what do you think affects the null more, noise or distortion that is 100 dB below the fundamental or a 4 dB magnitude response error?

Certainly the magnitude error. But that's why I like looking at the Δ spectra plots - there are clear differences well below 10kHz.
 
The Chord DAVE actually performs the best in these tests, even though it falls far short of the SMSL SU-10 in THD+N

In theory, the DAVE and SU-10 should both be transparent. It would be interesting to see how the recording from the DAVE and SU-10 compare

You can see how two different DACs here nulled out really well

while same thing here showed that the two DACs are different.
 
Happy to try the non-linear EQ settings, but wouldn't want to mask anything the DAC itself is doing.
It'd be like measuring both THD+N and THD, to give more insight into DAC behavior.

As-is null metrics would likely be significantly affected by which oversampling filter your DAC and ADC is using, and whether their in-/outputs are AC- or DC-coupled.

Therefore, for the sake of completeness, each oversampling option should be tested.

Using Phase and Level EQ would compensate for these parameters so that only noise and nonlinear distortion remain.

To be clear: both numbers are meaningful, just two sides of the same coin.
 
Last edited:
But surely, if they're being created by the DAC, then we don't want to correct them - we want to measure them, no?

At high frequency they are being created by the DAC/ADC filter behavior and at low frequency they are being created by the DAC/ADC DC high pass behavior. In general longer filters with more latency will give better nulls as they have flatter and more extended frequency response. Flatter DC response will give better low frequency nulls.

Again, the way to measure this is with a simple frequency response measurement in REW. With the frequency response measurement you can also develop a FIR filter with impulse inversion to correct the magnitude and frequency response errors. If you apply this FIR filter and re-run your Deltawave test you will see a substantial improvement in the null proving that this is the most important issue.

Another fun trick is to see how the results change by resampling to a higher sample rate and recording at a higher sample rate. If this resampling is high quality you will get better results as the DAC/ADC filter roll-off will be well above 22 kHz.

Michael
 
Last edited:
To be clear: both numbers are meaningful, just two sides of the same coin.

That's why I wanted to show some null results - lot's of THD+N measurements already out there.

As-is null metrics would likely be significantly affected by which oversampling filter your DAC and ADC is using, so for the sake of completeness, each oversampling option should be tested.

I've done this and will show the results for each DAC in a separate post. For the league table, I've taken the best result.
 
That's why I wanted to show some null results - lot's of THD+N measurements already out there.
With two sides of the same coin I meant Null tests with phase+level EQ and Null tests w/o phase+level EQ :D

Not THD+N and Null tests.

(Not that those aren't also two sides of one coin :D)
 
At high frequency they are being created by the DAC/ADC filter behavior and at low frequency they are being created by the DAC/ADC DC high pass behavior. In general longer filters with more latency will give better nulls as they have flatter and more extended frequency response. Flatter DC response will give better low frequency nulls.

Again, the way to measure this is with a simple frequency response measurement in REW. With the frequency response measurement you can also develop a FIR filter with impulse inversion to correct the magnitude and frequency response errors. If you apply this FIR filter and re-run your Deltawave test you will see a substantial improvement in the null proving that this the most important issue.

Another fun trick is to see how the results change by resampling to a higher sample rate and recording at a higher sample rate. If this resampling is high quality you will get better results as the DAC / ADC filter roll-off will be well above 22 kHz.

All good stuff Michael. Thanks.

I suppose my purpose in this thread is to show how accurately a DAC produces the music signal it is presented with. And that THD+N is perhaps not the best measure...
 
Using Phase and Level EQ would compensate for these parameters so that only noise and nonlinear distortion remain.

Sure. But people listen with the oversampling filter in place :). What we can (and will) certainly show is that some filters are way more accurate than others. I mean, look at the difference in performance between the SU-10's 'none' and 'fast linear' filters.

Also, a lot is said about 'minimum phase' filters - apparently they sound better. But null tests show that these are less accurate than linear filters.
 
I suppose my purpose in this thread is to show how accurately a DAC produces the music signal it is presented with. And that THD+N is perhaps not the best measure...

In terms of accuracy, simple frequency response will tell you the most about DAC accuracy, unless THD+N is really bad.

Have you tried using the Deltawave comparator to see if you can reliably differentiate the recordings from each other (or from the original) via listening with your "best" DAC? That would be the most interesting comparison IMO.

Michael
 
Back
Top Bottom