Listening Comparison Test: DSP Phono vs SUT

[watchnerd]

Below are two vinyl rips for listening comparison / digital file analysis. I'd love to get the feedback and insights of the collected community.

Source: Michell Gyro SE turntable / Jelco SA-750D tonearm / AT33EV cartridge
Medium: Side A of the Analogue Productions 45 RPM remastering of Getz / Gilberto
ADC Process: 24 bit / 96khz Devialet Expert 400 built-in ADC, RIAA via built in DSP (1976 EQ curve), USB out to Macbook, captured via Audacity and exported to FLAC. No other processing, normalization, or level matching.

WARNING: The files are 24bit / 96khz FLACs, unzipped, so about 190 MB each

Scenario 1, with SUT:
K&K Audio / Lundahl LL1931 SUT, set to 1:16 ratio / 24 dB gain
Loading: 183 ohms reflected (calculated, not measured), no additional resistors added
Input: Devialet set to MM mode, 4.7mv max input

File: https://drive.google.com/file/d/15Kgr17kfVf4Kg0ht0zWkig_XCL38VLAC/view?usp=sharing



Scenario 2, no SUT:
Straight into Devialet
Loading: 200 ohms (set per Devialet configuration UI, default for AT33EV preconfig)
Input: Devialet set to MC mode, 0.3mv max input

https://drive.google.com/file/d/1e0g-W4E1wlLz0xhbCZ-AK4-cYfC1OiKR/view?usp=sharing


The original rationale for getting a SUT was to reduce hiss / improve SNR of the phono stage. But I'm curious if others subjectively perceive or analytically deduce anything else going on.

Any comments or insights welcome!

 

[watchnerd]

Below are the spectrum plots of the first 109 seconds, FWIW:

Scenario 1:

Scenario 2:

 

[Soniclife]

Have you put the files through delta wave?

 

[watchnerd]

Have you put the files through delta wave?

I don't know what you're referring to.

 

[Soniclife]

https://deltaw.org/

There is a thread here on it as well.

It's simple to use, and great. Should be very good for this.

 

[watchnerd]

https://deltaw.org/

There is a thread here on it as well.

It's simple to use, and great. Should be very good for this.

Doesn't look like there is a Mac version...

 

[LTig]

Below are two vinyl rips for listening comparison / digital file analysis. I'd love to get the feedback and insights of the collected community. [..]
The original rationale for getting a SUT was to reduce hiss / improve SNR of the phono stage. But I'm curious if others subjectively perceive or analytically deduce anything else going on.

Any comments or insights welcome!

I had a quick listen, using a Sennheiser HD800 fed by RME ADI-2 PRO fs (with 6kHz EQ), played bit correct by VLC player on Linux.

Well, the first rational has been reached. Although the preamp hiss in scenario S2 is lower than the vinyl noise I seem to hear the vinyl noise at 5:30 with more definition than in S1, so it's an improvement.

Subjectively S1 has more charm and is smoother in the mids and highs (or whatever my ears are able to detect in this range), S2 on the other hand may have a bit more resolution. Can't say much about soundstage using headphones.

I prefer S1 for both hiss and subjective presentation. It would be great to have a digital source of the same piece for comparison though.

 

[watchnerd]

I had a quick listen, using a Sennheiser HD800 fed by RME ADI-2 PRO fs (with 6kHz EQ), played bit correct by VLC player on Linux.

Well, the first rational has been reached. Although the preamp hiss in scenario S2 is lower than the vinyl noise I seem to hear the vinyl noise at 5:30 with more definition than in S1, so it's an improvement.

Subjectively S1 has more charm and is smoother in the mids and highs (or whatever my ears are able to detect in this range), S2 on the other hand may have a bit more resolution. Can't say much about soundstage using headphones.

I prefer S1 for both hiss and subjective presentation. It would be great to have a digital source of the same piece for comparison though.

I more or less have similar subjective impressions: SUT seems to have less vinyl noise (why, I don't know), a bit less resolution, and something different going on in the highs. But it's pretty subtle.

Noise floor aside, S2 may be more transparent to the source -- but, hey, it's vinyl. If full transparency to the master is the goal, listen to the high rez digital version.

 

[RayDunzl]

@pkane

Tried to load the files above into DW

It crashes - just, poof, gone


DeltaWave v1.0.37, 2019-07-28T18:31:53.6417024-04:00
Reference: ipanema1.flac[L] 47722304 samples 96000Hz 24bits, stereo, MD5=00
Comparison: ipanema2.flac[L] 47165248 samples 96000Hz 24bits, stereo, MD5=00
Settings:
Gain:True, Remove DC:True
Non-linear Gain EQ:False Non-linear Phase EQ: False
EQ FFT Size:65536, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
Correct Drift:True, Precision:30
Non-Linear drift Correction:False
Upsample:True, Window:Hann
Spectrum Window:Hann, Spectrum Size:524288
Spectrogram Window:Hann, Spectrogram Size:1024, Spectrogram Steps:256
Dither:False
Trim Silence:True

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -0.545dB Comparison: -0.352dB
Initial RMS values Reference: -17.75dB Comparison: -17.412dB

Null Depth=23.065dB
Cross-correlation found multiple phase inversions -- data may be too noisy or too dissimilar!
X-Correlation offset: 62011 samples
Trimming 0 samples at start and 3905 samples at the end that are below -90.31dB level

I think this is all it had to say...

---

Trimmed 90 seconds off the end of both files... It ran.

 

[pkane]

@pkane

Tried to load the files above into DW

It crashes - just, poof, gone


DeltaWave v1.0.37, 2019-07-28T18:31:53.6417024-04:00
Reference: ipanema1.flac[L] 47722304 samples 96000Hz 24bits, stereo, MD5=00
Comparison: ipanema2.flac[L] 47165248 samples 96000Hz 24bits, stereo, MD5=00
Settings:
Gain:True, Remove DC:True
Non-linear Gain EQ:False Non-linear Phase EQ: False
EQ FFT Size:65536, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
Correct Drift:True, Precision:30
Non-Linear drift Correction:False
Upsample:True, Window:Hann
Spectrum Window:Hann, Spectrum Size:524288
Spectrogram Window:Hann, Spectrogram Size:1024, Spectrogram Steps:256
Dither:False
Trim Silence:True

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -0.545dB Comparison: -0.352dB
Initial RMS values Reference: -17.75dB Comparison: -17.412dB

Null Depth=23.065dB
Cross-correlation found multiple phase inversions -- data may be too noisy or too dissimilar!
X-Correlation offset: 62011 samples
Trimming 0 samples at start and 3905 samples at the end that are below -90.31dB level

I think this is all it had to say...

---

Trimmed 90 seconds off the end of both files... It ran.

Sounds like an out of memory error because of the file size. DW tries to catch and report all errors but an out of memory condition can be nasty as even error reporting may fail.

 

[LTig]

I more or less have similar subjective impressions: SUT seems to have less vinyl noise (why, I don't know)

It's not the vinyl noise which is better. There is just less preamp noise riding on it. You will notice this with better or virgin vinyl (less vinyl noise) although this recording is already very good in this regard.

 

[watchnerd]

It's not the vinyl noise which is better. There is just less preamp noise riding on it. You will notice this with better or virgin vinyl (less vinyl noise) although this recording is already very good in this regard.

Thanks for the explanation, that seems to make sense. And, as you said, at least the goal of lowering the noise floor seems to have been achieved by the SUT, even if it is introducing some other elements to the sound (at least with this loading).

 

[Blumlein 88]

I more or less have similar subjective impressions: SUT seems to have less vinyl noise (why, I don't know), a bit less resolution, and something different going on in the highs. But it's pretty subtle.

Noise floor aside, S2 may be more transparent to the source -- but, hey, it's vinyl. If full transparency to the master is the goal, listen to the high rez digital version.

I hear more noise in version 2 prior to the music, and none prior to the music in 1.

1 sounds almost echoing or slightly ringing, and sounds too soft in the treble. 2 sounds much more like it would be transparent to the recording. I've not measured them yet, but just listened.

 

[watchnerd]

I hear more noise in version 2 prior to the music, and none prior to the music in 1.

1 sounds almost echoing or slightly ringing, and sounds too soft in the treble. 2 sounds much more like it would be transparent to the recording. I've not measured them yet, but just listened.

Thanks for taking the time to listen.

Thankfully, the initial goal of reducing noise by adding 24 dB of gain via the SUT seems to have worked. I'm glad I'm not the only one who hears it.

As for the S1 ringing -- it might be. I didn't add any additional resistors to the SUT to dampen things and there is no Zobel network. But I agree there is something going on that makes it sound almost a bit tube-like in the treble.

What would be the best way to measure for ringing or other treble differences in the files?

 

[SIY]

What would be the best way to measure for ringing or other treble differences in the files?

Do you have a test record with a square wave?

 

[watchnerd]

I hear more noise in version 2 prior to the music, and none prior to the music in 1.

1 sounds almost echoing or slightly ringing, and sounds too soft in the treble. 2 sounds much more like it would be transparent to the recording. I've not measured them yet, but just listened.

Also, on the topic of ringing / loading:

Now I'm really curious how Devialet has implemented adjustable real-time, on the fly resistance loading via a menu....it can't be physical / electrical, I don't think. Would it be possible to emulate that in software??

 

[watchnerd]

Do you have a test record with a square wave?

I have a few test records that have sine wave sweeps and other tests, but I don't think any have a square wave. I'll check.

 

[SIY]

See how high the sweeps go if there's no square wave. If they go high enough, you'll see peaking or droop if the SUT isn't loaded optimally.

 

[Blumlein 88]

@pkane

Tried to load the files above into DW

It crashes - just, poof, gone


DeltaWave v1.0.37, 2019-07-28T18:31:53.6417024-04:00
Reference: ipanema1.flac[L] 47722304 samples 96000Hz 24bits, stereo, MD5=00
Comparison: ipanema2.flac[L] 47165248 samples 96000Hz 24bits, stereo, MD5=00
Settings:
Gain:True, Remove DC:True
Non-linear Gain EQ:False Non-linear Phase EQ: False
EQ FFT Size:65536, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
Correct Drift:True, Precision:30
Non-Linear drift Correction:False
Upsample:True, Window:Hann
Spectrum Window:Hann, Spectrum Size:524288
Spectrogram Window:Hann, Spectrogram Size:1024, Spectrogram Steps:256
Dither:False
Trim Silence:True

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -0.545dB Comparison: -0.352dB
Initial RMS values Reference: -17.75dB Comparison: -17.412dB

Null Depth=23.065dB
Cross-correlation found multiple phase inversions -- data may be too noisy or too dissimilar!
X-Correlation offset: 62011 samples
Trimming 0 samples at start and 3905 samples at the end that are below -90.31dB level

I think this is all it had to say...

---

Trimmed 90 seconds off the end of both files... It ran.

Under the process tab I Loaded the files without processing. Then ran a match and it did run. Drift error was too great and it had to use alternative drift correction. It took forever, used all 8 gig of RAM I had while running the cpu in the 90% range for several minutes. And I don't trust the results. I trimmed 10 seconds off the front and 400 off the rear and ran it. Results aren't great with RMS difference of only 15 db. There is a .5% speed difference with the 2nd file being faster.

I then took the last two minutes, corrected speed and ran that. Pretty much the same results and need alternative drift correction.

Using other software to confirm, there is this up and down frequency response difference in the two files as shown below. It would be a straight line if the FR was identical. I don't know if this is flutter causing short term speed fluctuations or if this could be some kind of ringing result of what the Devialet is doing. Probably flutter.

 

[watchnerd]

Under the process tab I Loaded the files without processing. Then ran a match and it did run. Drift error was too great and it had to use alternative drift correction. It took forever, used all 8 gig of RAM I had while running the cpu in the 90% range for several minutes. And I don't trust the results. I trimmed 10 seconds off the front and 400 off the rear and ran it. Results aren't great with RMS difference of only 15 db. There is a .5% speed difference with the 2nd file being faster.

I then took the last two minutes, corrected speed and ran that. Pretty much the same results and need alternative drift correction.

Using other software to confirm, there is this up and down frequency response difference in the two files as shown below. It would be a straight line if the FR was identical. I don't know if this is flutter causing short term speed fluctuations or if this could be some kind of ringing result of what the Devialet is doing. Probably flutter.

View attachment 30237

Thanks for doing that.

From other tests I've seen (e.g. Stereophile) of the Devialet phono stage, I doubt it's something in the ADC/DSP, as it tested quite linearly.

It could definitely be flutter. The last time I did a speed test, the speed accuracy was 0.10%..and that's just RPMs, not a flutter/wow test.

Could the SUT be ringing?