Evaluating AD/DA loops by means of Audio Diffmaker

[guze]

I did the test with a audiofire8 (non adat CS4272) and these were the results:

Corr Depth: 27,7 dB (L), 29,5 dB (R)

[didier.brest]

Edirol FA-66 added.

Mytek 8 x 192
-1,002sec, 0,450dB (L), 0,553dB (R)..Corr Depth: 35,3 dB (L), 36,9 dB (R)

Prism Orpheus
-1sec, 0,075dB (L), 0,064dB (R)..Corr Depth: 30,8 dB (L), 32,4 dB (R)

RME Fireface 400
-36,94usec, 0,298dB (L), 0,311dB (R)..Corr Depth: 30,2 dB (L), 31,8 dB (R)

Echo Audiofire 8 (non adat CS4272)
Corr Depth: 27,7 dB (L), 29,5 dB (R)

Yamaha Steinberg MR816X
-2,101msec, 1,104dB (L), 0,605dB (R)..Corr Depth: 19,1 dB (L), 21,0 dB (R)

Edirol FA-66
-8,95msec, 1,015dB (L), 0,605dB (R)..Corr Depth: 15,6 dB (L), 17,1 dB (R)

RME Babyface
1,555usec, 0,352dB (L), 0,414dB (R)..Corr Depth: 9,2 dB (L), 10,8 dB (R)

Wanted: Apogee, Crane Song, Focusrite, Forsell, Lake People, Lavry, Metric Halo, Universal audio, Weiss etc. And other models from the brands already here.

[lamixer]

Hello Didier:

Here is the single loop through on the 192 for your results set.

Thanks.

[didier.brest]

Thanks. But it is not the same original file. I'm using the one from MainTime.

[lamixer]

Ah. I see. Well, someone else would have to do it I guess as I've sold my 192 recently!

[johnnyc]

How was this done exactly as there are 2 original files? I don't have pro tools so sorry if that's what the problem is.

[didier.brest]

I used only the symphonic clip.

[AndG]

Hi Everyone
Tc Electronic Impact Twin
-2.286msec, 0.524dB (L), 0.817dB (R)..Corr Depth: 32.0 dB (L), 33.6 dB (R)


Link to file:
http://dl.dropbox.com/u/15405661/impcttwin.wav

[didier.brest]

Thanks.
I added it to my list and for each file I added the left and right RMS levels of the time-aligned gain-balanced difference file provided by Audio DiffMaker, measured in Wavelab.

Mytek 8 x 192 (MainTime)
-1,002sec, 0,450dB (L), 0,553dB (R)..Corr Depth: 35,3 dB (L), 36,9 dB (R) Difference: -55.2 dBFS (L) -56.0 dBFS (R)

TC Electronics Impact Twin (AndG)
-2,286msec, 0,524dB (L), 0,817dB (R)..Corr Depth: 32,0 dB (L), 33,6 dB (R) Difference: -51.7 dBFS (L) -52.7 dBFS (R)

Prism Orpheus (MainTime)
-1sec, 0,075dB (L), 0,064dB (R)..Corr Depth: 30,8 dB (L), 32,4 dB (R) Difference: -51.1 dBFS (L) -52.4 dBFS (R)

RME Fireface 400 (didier.brest)
-36,94usec, 0,298dB (L), 0,311dB (R)..Corr Depth: 30,2 dB (L), 31,8 dB (R) Difference: -49.8 dBFS (L) -50.8 dBFS (R)

Echo Audiofire 8, non adat CS4272 (guze)
Corr Depth: 27,7 dB (L), 29,5 dB (R)

Yamaha Steinberg MR816X (didier.brest)
-2,101msec, 1,104dB (L), 0,605dB (R)..Corr Depth: 19,1 dB (L), 21,0 dB (R) Difference: -38.6 dBFS (L) -40.0 dBFS (R)

Edirol FA-66 (didier.brest)
-8,95msec, 1,015dB (L), 0,605dB (R)..Corr Depth: 15,6 dB (L), 17,1 dB (R) Difference: -34.4 dBFS (L) -35.4 dBFS (R)

RME Babyface (didier.brest)
1,555usec, 0,352dB (L), 0,414dB (R)..Corr Depth: 9,2 dB (L), 10,8 dB (R) Difference: -28.8 dBFS (L) -29.8 dBFS (R)


Wanted: Apogee, Crane Song, Focusrite, Forsell, Lake People, Lavry, Metric Halo, Universal audio, Weiss etc. And other models from the brands already here.

[schtim]

Hi Didier,

Here is the RME HDSP 9632

parameters: -1.564usec, 0.463dB (L), 0.461dB (R)..Corr Depth: 31.4 dB (L), 33.0 dB (R)

Thanks for this test!

https://www.yousendit.com/download/d...VnNuSlJFQlE9PQ

[didier.brest]

Thanks.
It has been added.

Mytek 8 x 192 (MainTime)
-1,002sec, 0,450dB (L), 0,553dB (R)..Corr Depth: 35,3 dB (L), 36,9 dB (R) Difference: -55.2 dBFS (L) -56.0 dBFS (R)

TC Electronics Impact Twin (AndG)
-2,286msec, 0,524dB (L), 0,817dB (R)..Corr Depth: 32,0 dB (L), 33,6 dB (R) Difference: -51.7 dBFS (L) -52.7 dBFS (R)

RME HDSP 9632 (schtim)
-1.564usec, 0.463dB (L), 0.461dB (R)..Corr Depth: 31.4 dB (L), 33.0 dB (R) Difference: -51.2 dBFS (L) -52.1 dBFS (R)

Prism Orpheus (MainTime)
-1sec, 0,075dB (L), 0,064dB (R)..Corr Depth: 30,8 dB (L), 32,4 dB (R) Difference: -51.1 dBFS (L) -52.4 dBFS (R)

RME Fireface 400 (didier.brest)
-36,94usec, 0,298dB (L), 0,311dB (R)..Corr Depth: 30,2 dB (L), 31,8 dB (R) Difference: -49.8 dBFS (L) -50.8 dBFS (R)

Echo Audiofire 8, non adat CS4272 (guze)
Corr Depth: 27,7 dB (L), 29,5 dB (R)

Yamaha Steinberg MR816X (didier.brest)
-2,101msec, 1,104dB (L), 0,605dB (R)..Corr Depth: 19,1 dB (L), 21,0 dB (R) Difference: -38.6 dBFS (L) -40.0 dBFS (R)

Edirol FA-66 (didier.brest)
-8,95msec, 1,015dB (L), 0,605dB (R)..Corr Depth: 15,6 dB (L), 17,1 dB (R) Difference: -34.4 dBFS (L) -35.4 dBFS (R)

RME Babyface (didier.brest)
1,555usec, 0,352dB (L), 0,414dB (R)..Corr Depth: 9,2 dB (L), 10,8 dB (R) Difference: -28.8 dBFS (L) -29.8 dBFS (R)

[Kaigen]

Originally Posted by didier.brest

Thanks.
It has been added.

Mytek 8 x 192 (MainTime)
-1,002sec, 0,450dB (L), 0,553dB (R)..Corr Depth: 35,3 dB (L), 36,9 dB (R) Difference: -55.2 dBFS (L) -56.0 dBFS (R)

TC Electronics Impact Twin (AndG)
-2,286msec, 0,524dB (L), 0,817dB (R)..Corr Depth: 32,0 dB (L), 33,6 dB (R) Difference: -51.7 dBFS (L) -52.7 dBFS (R)

RME HDSP 9632 (schtim)
-1.564usec, 0.463dB (L), 0.461dB (R)..Corr Depth: 31.4 dB (L), 33.0 dB (R) Difference: -51.2 dBFS (L) -52.1 dBFS (R)

Prism Orpheus (MainTime)
-1sec, 0,075dB (L), 0,064dB (R)..Corr Depth: 30,8 dB (L), 32,4 dB (R) Difference: -51.1 dBFS (L) -52.4 dBFS (R)

RME Fireface 400 (didier.brest)
-36,94usec, 0,298dB (L), 0,311dB (R)..Corr Depth: 30,2 dB (L), 31,8 dB (R) Difference: -49.8 dBFS (L) -50.8 dBFS (R)

Echo Audiofire 8, non adat CS4272 (guze)
Corr Depth: 27,7 dB (L), 29,5 dB (R)

Yamaha Steinberg MR816X (didier.brest)
-2,101msec, 1,104dB (L), 0,605dB (R)..Corr Depth: 19,1 dB (L), 21,0 dB (R) Difference: -38.6 dBFS (L) -40.0 dBFS (R)

Edirol FA-66 (didier.brest)
-8,95msec, 1,015dB (L), 0,605dB (R)..Corr Depth: 15,6 dB (L), 17,1 dB (R) Difference: -34.4 dBFS (L) -35.4 dBFS (R)

RME Babyface (didier.brest)
1,555usec, 0,352dB (L), 0,414dB (R)..Corr Depth: 9,2 dB (L), 10,8 dB (R) Difference: -28.8 dBFS (L) -29.8 dBFS (R)

Amazing. The test result reflects my listening experience very well.

[jrasia]

Love to see a RME UFX thrown in the mix!

[diatonic]

Ideas on what is bottlenecking the babyface results? I figured it would be closer to the FF400

[didier.brest]

I'm also surprised. I would like that someone test another one.

[diatonic]

I am guessing it is because the BabyFace pre-amps are inline on the two analog inputs, so you are not just testing A/D but the pre-amp circuity which will be expected to have some coloration. My ears tell me the AD/DA loop is doing a good job.

[didier.brest]

I agree that listening to the baby face loop do not reveal such poor performance. Note that the MR816X (*) and the Edirol FA-66 have also a preamp ahead the converter.

(*) I di use the balanced inputs, not the unbalanced insert inputs that bypass the preamps.

[diatonic]

I heard back from RME. Their converters apply a sub-sample offset. Looking at Diffmaker's AES paper, the reference vs processed comparison mentions sample by sample comparison. Sub-sample offsets would throw this off even though the waveform is OK. Add in the pre-amp circuitry of the babyface and this may explain the numbers. Still very happy with the babyface, so only looking into this for interest's sake.

[didier.brest]

Audio Diffmaker is correcting the time offset with 1 ns accuracy, which is much less than 1/1000 of the sampling step, as shown by some time offset figures expressed in us (µs) with 3 figures after the decimal mark. A comparison sample by sample at the original sampling rate could not achieve so low difference figures. I am sure about that because before using Audi Diiffmaker I was evaluating such differences in Wavelab and I had to upsample by 32 the files (and then testing several shifts as a number of samples) to get consistent difference RMS level figures.

[Panelhead]

Originally Posted by didier.brest

Audio Diffmaker is correcting the time offset with 1 ns accuracy, which is much less than 1/1000 of the sampling step, as shown by some time offset figures expressed in us (µs) with 3 figures after the decimal mark. A comparison sample by sample at the original sampling rate could not achieve so low difference figures. I am sure about that because before using Audi Diiffmaker I was evaluating such differences in Wavelab and I had to upsample by 32 the files (and then testing several shifts as a number of samples) to get consistent difference RMS level figures.

Didier,
Can this be run D/A and then A/D to see how well the digital files match? I have a TC Impact Twin. Been looking for a deal on a Prism Orpheus. Glad I never found one. Looks like the Mytek needs to be looked for.


George

[didier.brest]

Originally Posted by Panelhead

Can this be run D/A and then A/D to see how well the digital files match?

Hi George,

it is what we are doing here: an AD-DA loop ... May be I did not catch your point ?

Didier

[Panelhead]

Originally Posted by didier.brest

Hi George,

it is what we are doing here: an AD-DA loop ... May be I did not catch your point ?

Didier

I was wondering if the test could start with a PCM file, convert to analog, and convert back to digital. Then compare the digital files instead of the two analog.
My application is ripping vinyl. The test as run is better. But I will only be running the A/D part. And using the mic pres for balanced step ups. I feel the Impact Twin will be fine in this application. I would like a Mytek anyway.

[didier.brest]

Originally Posted by Panelhead

I was wondering if the test could start with a PCM file, convert to analog, and convert back to digital. Then compare the digital files instead of the two analog.

It is exactly what we are doing here. The original PCM file is here. Anybody who wants to participate has just to digitally record the analog playback of this digital recording, then upload somewhere (30 MB is too big for uploading here) the resulting digital recording and share the download link here.

[guze]

Here is the file for the audiofire, so you can make your tests. To see how it compare with the others.
Good tests

audiofire8.wav

[didier.brest]

Thank you!

There was an accident in your loop. 8 extra samples were added, which makes impossible the time alignment both before and after this accident. The time alignment from DiffMaker is correct for the part after the accident. As a result there is a strong error before. See attached screenshots from Wavelab.
I repaired your file by removing these 8 samples and I did again the test. I got figures higher than before repair but lower than your first figures. I guess that this loop is not the one that your tested first ? My figures are included in the list here below. If you have a better loop, I am keen redoing the test.

Mytek 8 x 192 (MainTime)
-1,002sec, 0,450dB (L), 0,553dB (R)..Corr Depth: 35,3 dB (L), 36,9 dB (R) Difference: -55.2 dBFS (L) -56.0 dBFS (R)

TC Electronics Impact Twin (AndG)
-2,286msec, 0,524dB (L), 0,817dB (R)..Corr Depth: 32,0 dB (L), 33,6 dB (R) Difference: -51.7 dBFS (L) -52.7 dBFS (R)

RME HDSP 9632 (schtim)
-1.564usec, 0.463dB (L), 0.461dB (R)..Corr Depth: 31.4 dB (L), 33.0 dB (R) Difference: -51.2 dBFS (L) -52.1 dBFS (R)

Prism Orpheus (MainTime)
-1sec, 0,075dB (L), 0,064dB (R)..Corr Depth: 30,8 dB (L), 32,4 dB (R) Difference: -51.1 dBFS (L) -52.4 dBFS (R)

RME Fireface 400 (didier.brest)
-36,94usec, 0,298dB (L), 0,311dB (R)..Corr Depth: 30,2 dB (L), 31,8 dB (R) Difference: -49.8 dBFS (L) -50.8 dBFS (R)

Echo Audiofire 8, non adat CS4272 (guze)
-1,003sec, 0,840dB (L), 0,888dB (R)..Corr Depth: 26,4 dB (L), 28,2 dB (R) Difference: -39.3 dBFS (L) -39.3 dBFS (R)

Yamaha Steinberg MR816X (didier.brest)
-2,101msec, 1,104dB (L), 0,605dB (R)..Corr Depth: 19,1 dB (L), 21,0 dB (R) Difference: -38.6 dBFS (L) -40.0 dBFS (R)

Edirol FA-66 (didier.brest)
-8,95msec, 1,015dB (L), 0,605dB (R)..Corr Depth: 15,6 dB (L), 17,1 dB (R) Difference: -34.4 dBFS (L) -35.4 dBFS (R)

RME Babyface (didier.brest)
1,555usec, 0,352dB (L), 0,414dB (R)..Corr Depth: 9,2 dB (L), 10,8 dB (R) Difference: -28.8 dBFS (L) -29.8 dBFS (R)

[guze]

Hi,
Yes , i did add some time in the beggining because of the fade in/out.
I'll make a new one , I hope it looks better, or maybe i will start saving for a mytek.

http://www.mediafire.com/?abxblbggleql3ud

[schtim]

Hi,

Here are the specs for the audiofire 8 with ADAT clocked to an RME HDSP 9632.

parameters: -1.164sec, 0.128dB (L), 0.126dB (R)..Corr Depth: 25.9 dB (L), 27.5 dB (R)

https://rcpt.yousendit.com/115296821...965066308e6afb

Bought this box after seeing these test results, haha.
Only weird thing about it is that the ADAT sockets are really small and don't allow for larger ADAT plugs. The 1/4" sockets are also pretty close together. Am thinking of switching to an RME ADI-2 just to rectify this issue...

Really hoping the other big names start doing this test to rival the Mytek, at least in terms of clean conversion. But I guess we can see some results in Kraku's 10x ADDA loopback test.

[didier.brest]

Updated list with the third test of the Aufiofire 8 from guze and the one from schtim

Mytek 8 x 192 (MainTime)
-1,002sec, 0,450dB (L), 0,553dB (R)..Corr Depth: 35,3 dB (L), 36,9 dB (R) Difference: -55.2 dBFS (L) -56.0 dBFS (R)

TC Electronics Impact Twin (AndG)
-2,286msec, 0,524dB (L), 0,817dB (R)..Corr Depth: 32,0 dB (L), 33,6 dB (R) Difference: -51.7 dBFS (L) -52.7 dBFS (R)

RME HDSP 9632 (schtim)
-1.564usec, 0.463dB (L), 0.461dB (R)..Corr Depth: 31.4 dB (L), 33.0 dB (R) Difference: -51.2 dBFS (L) -52.1 dBFS (R)

Prism Orpheus (MainTime)
-1sec, 0,075dB (L), 0,064dB (R)..Corr Depth: 30,8 dB (L), 32,4 dB (R) Difference: -51.1 dBFS (L) -52.4 dBFS (R)

RME Fireface 400 (didier.brest)
-36,94usec, 0,298dB (L), 0,311dB (R)..Corr Depth: 30,2 dB (L), 31,8 dB (R) Difference: -49.8 dBFS (L) -50.8 dBFS (R)

Echo Audiofire 8 with ADAT clocked to an RME HDSP 9632 (schtim)
-1.164sec, 0.128dB (L), 0.126dB (R)..Corr Depth: 25.9 dB (L), 27.5 dB (R) Difference: -45.5 dBFS (L) -46.5 dBFS (R)

Echo Audiofire 8 non adat CS4272 (guze)
-2,952msec, 0,840dB (L), 0,885dB (R)..Corr Depth: 26,4 dB (L), 28,2 dB (R) Difference: -40.6 dBFS (L) -40.6 dBFS (R)

Yamaha Steinberg MR816X (didier.brest)
-2,101msec, 1,104dB (L), 0,605dB (R)..Corr Depth: 19,1 dB (L), 21,0 dB (R) Difference: -38.6 dBFS (L) -40.0 dBFS (R)

Edirol FA-66 (didier.brest)
-8,95msec, 1,015dB (L), 0,605dB (R)..Corr Depth: 15,6 dB (L), 17,1 dB (R) Difference: -34.4 dBFS (L) -35.4 dBFS (R)

RME Babyface (didier.brest)
1,555usec, 0,352dB (L), 0,414dB (R)..Corr Depth: 9,2 dB (L), 10,8 dB (R) Difference: -28.8 dBFS (L) -29.8 dBFS (R)

[isma]

Hi didier it's Isma (from audiofanzine), i found this post, very interesting, as it could explain the babyface's results:

Originally Posted by Ulrich

Hi Robin,

did you use two seperate Myteks for the test?

In case you used only one Mytek being D/A and A/D at the same time, the test doesn't provide the same conditions for the Mytek as for the other coverters because only in this case D/A and A/D share exactly the same clock circuit, undoubtedly leading to better results.

Can you please clarify this? Thanks!

Regards,
Ulrich

As the guy from Rme said, the babyface doesn't use the same ship for DA's and AD's, (so different clock circuit,) however, the fireface does, so better results, it could be the same for the TC impact.
If Ulrich is write, it means that, to compare the fireface to the babyface, we'll need 2 different firefaces.
Or send the fireface into the babyface, then compare to your babyface's results (which could act like 2 different babyfaces), this will tell which DA's are more transparent.
Do you have both fireface 400 and babyface?
As soon as i can i'll send the loop from the duet 2 to the babyface, so we'll have:
fireface-babyface
duet 2-babyface
babyface-babyface

cheers