Play & Record 10 times - AD/DA loop back test

[kraku]

I've been wondering:
When people compare audio interfaces, they just record one signal into their DAW with different audio interfaces and then try to hear those incredibly small detail differences.

Wouldn't it be much better idea to take one stereo signal and then play & record it 10 times using the same audio interface. This way you'll get 10x the conversion artifacts ---> you can hear much better which audio interface does the most damage to the signal and what this damage is.

I'd be interested in these 10x play&record loopback shootouts.

Any volunteers? I can provide RME Fireface 400 results for the interested people. First we should come up with a stereo material which should reveal the signal degradation as well as possible. Maybe a well recorded acoustic material would do?

EDIT:

All the test results people post in this thread will be mirrored in this URL, including the original reference audio file:

AD/DA 10x Loopback Test Results

Current audio interface results in the above URL:

Apogee DA16x into the AD16x (connected by X-HD)
Digidesign 192 IO
Echo Audio AudioFire 12
Echo Audio Layla 24
Focusrite Saffire LE
Metric Halo ULN 8
Motu 2408 Mk3
Mytek Digital 8x192
Prism Orpheus
RME Fireface 400
Roland MMP-2 with frequency responce matched + Drawmer M-Clock
Steinberg MR816
Universal Audio 2192

[GoldMember]

Originally Posted by kraku

I've been wondering:
When people compare audio interfaces, they just record one signal into their DAW with different audio interfaces and then try to hear those incredibly small detail differences.

Wouldn't it be much better idea to take one stereo signal and then play & record it 10 times using the same audio interface.

... using a software like RMAA, to generate the 10x AD&DA test file & to analize it, would be a nice idea.

http://audio.rightmark.org

[electronic_pulse]

sounds great. do it!!

[kraku]

The software seems to be PC only. I have a Mac. Could someone create the test file using that software for me?

I could also just do the little bit naughty thing and take a short audio clip from some well engineered acoustic recording and bounce that 10 times. I already have an excellent candidate on my mind...

[kraku]

Wow! Not the results I expected from this test. Here are the test results using RME Fireface 400:

AD/DA loop back test results for RME Fireface 400

The original signal is normalized and I also normalized the signal between each bounce. I used the inputs channels which didn't have mic pres on them. As you can hear the stereo image drifts pretty dramatically. This is why I included both 5 and 10 bounce versions. If you listen to those clips in mono, you'll clearly hear how the high frequencies get lots of sizzle.

Could someone try the same thing using a different audio interface and the original sample file for cross comparison purposes. I'd be really interested in hearing some really high end interfaces and how their results compare to my mundane FF400 (also the new Steinberg interface is of interest to me here)

EDIT:
The patchbay between the I/O caused this. I redid the test without the patchbay and updated the results. I also left the old faulty results in the same location for educational purposes on the dangers of using patchbays.

[Fred_Abstract]

i think a techno drum loop with a bassline are perfect for these test, transients, imaging, low end, high end .beeing "extreme" i find this material very good at testing converters and to exhibit their sound.

[jbjoubaud]

wow indeed !

I have a uln 8. I'll try do the test tomorrow with the same sample.

[jbjoubaud]

here's the file after 10 ADDA conversion for the metric halo uln 8.

I had to adjust the levels every time, and I adjusted the last file with sonalksis free G for to get it to match the first file a close as possible (only used the trim button to add .90 dB)

10.2.wav

[jbjoubaud]

I just compared your file with mine and boy am I glad I switched to metric halo !

[electronic_pulse]

Originally Posted by jbjoubaud

I just compared your file with mine and boy am I glad I switched to metric halo !

Hmmm.. I would think sooo!!! The metric halo cost 4X the price of the RME FF400.

[Silas Holmes]

Wow!

[kraku]

Originally Posted by jbjoubaud

here's the file after 10 ADDA conversion for the metric halo uln 8.

I had to adjust the levels every time, and I adjusted the last file with sonalksis free G for to get it to match the fist file a close as possible

Attachment 229365

If the Free G is just a simple fader which doesn't process the signal in any other way, then all I can say is: WOW!

It seems that ULN 8 is way better audio interface than the FF400!

The stereo image doesn't drift much at all and and signal still has it's fullness to it. Nice! I like very much!

[YOHAMI]

very good idea

[kraku]

Originally Posted by Fred_Abstract

i think a techno drum loop with a bassline are perfect for these test, transients, imaging, low end, high end .beeing "extreme" i find this material very good at testing converters and to exhibit their sound.

I would prefer to use the original reference file, since it contains lots of natural sound sources which people are accustomed in hearing every day and they have incredible amount of nuances in those sounds. Also the dynamics are excessive. All these have lots of potential in revealing any flaws with the AD/DA conversion the given interface has.

[jbjoubaud]

Yes the Free G is basicelly a fader, it's got pan pot (which I didn't tuch), mute and phase buttons. Only used it to add 0.90 dB

[jbjoubaud]

Originally Posted by electronic_pulse

Hmmm.. I would think sooo!!! The metric halo cost 4X the price of the RME FF400.

Price doesn't always mean everything.
I compared a $200 power amp (Nikko) to a $1200 Bryston and the difference wasn't day and night (at all.)
I'm just glad it is in this case.

[kraku]

Brothomstates kindly provided Focusrite Saffire LE results, which can be found in the same URL as before.

[kraku]

I'm beginning to think that the patchbay which was between the I/O of my audio interface has somehow degraded the stereo image... I'm going to redo my FF400 test tomorrow and post its results again.

[GoldMember]

Originally Posted by kraku

All the test results people post in this thread will be mirrored in this URL, including the original reference audio file:

AD/DA 10x Loopback Test Results

Current audio interface results in the above URL:

Focusrite Saffire LE
Metric Halo ULN 8
RME Fireface 400 (re-testing tomorrow - I suspect the patchbay between the I/O somehow affected the stereo image)

Originally Posted by jbjoubaud

here's the file after 10 ADDA conversion for the metric halo uln 8.

I had to adjust the levels every time, and I adjusted the last file with sonalksis free G for to get it to match the first file a close as possible (only used the trim button to add .90 dB)

Attachment 229365

are you shure the ULN and saffire are true 10x times ?
they sound like 1 time.

a -1dB in the lows = -10dB with the 10x time loop.

[GoldMember]

Originally Posted by kraku

The software seems to be PC only. I have a Mac. Could someone create the test file using that software ?
...

RMAA623.rar

[kraku]

Originally Posted by GoldMember

are you shure the ULN and saffire are true 10x times ?
they sound like 1 time.

a -1dB in the lows = -10dB with the 10x time loop.

There is no way for me to check that. I just have to rely on people not cheating on this test. I do agree that they sound suspiciously good.


@jbjoubaud:
Can you confirm that you really played and recorded the file 10 times so that the signal degraded during each recording round?

[kraku]

Now that I removed the patchbay from the equation and redid the FF400 10x bounce, I didn't have to match the levels anymore after each bounce. This new RME FF400 test result also doesn't have the stereo image drift in it and the sound doesn't degrade much at all (no sizzle this time). I guess the patchbay caused most of the problems, maybe also the normalizing between bounces.

The new FF400 results are now in the same URL.

I also renamed and left the faulty FF400 results in that directory, just as a warning to people that patchbays can mess up your signal

[Majestikc]

At first I thought this was a really good idea, but now I don't really see the point....

Let's just say for arguments sake that your converter only captures 90% of the incoming sound, if you're only recording that sound once, as we all do, then it's not that big a deal, but obviously then if you run that same sound round and round through the converter 10 times it's going to loose 10% every time etc etc etc............the only thing is nobody does this, it's normally just once in and out.

So the main test would still be to record just once using different converters and then compare them.

I'll still be doing this test anyway just to see HOW much degradation there is.

[kraku]

Originally Posted by Majestikc

At first I thought this was a really good idea, but now I don't really see the point....

Let's just say for arguments sake that your converter only captures 90% of the incoming sound, if you're only recording that sound once, as we all do, then it's not that big a deal, but obviously then if you run that same sound round and round through the converter 10 times it's going to loose 10% every time etc etc etc............the only thing is nobody does this, it's normally just once in and out.

So the main test would still be to record just once using different converters and then compare them.

I'll still be doing this test anyway just to see HOW much degradation there is.

There are two good reasons to this test:

1. You'll get good comparable results how different converters perform to each other.

2. You'll get a good idea how much you should worry about bouncing and processing your tracks with HW gear. If the signal doesn't degrade much at all, then you don't have to worry much about doing all the HW processing in one go.

[GoldMember]

this test is designed to see converters with a 10x magnifier.
the most obvious flaw its frequency responce.
thats why its interesting to loop the RMAA files.

frequency matching can be called cheating...
but allows to see harmonic distortions, stereo image, noise floor, etc...

my 2x Roland MMP-2 with frequency responce matched.
using a drawmer m-clock.
6x meter Proel DieHard xlr cables.
psaudio power plant premier.

[GoldMember]

Originally Posted by Majestikc

At first I thought this was a really good idea, but now I don't really see the point....

Let's just say for arguments sake that your converter only captures 90% of the incoming sound, if you're only recording that sound once, as we all do, then it's not that big a deal, but obviously then if you run that same sound round and round through the converter 10 times it's going to loose 10% every time etc etc etc............the only thing is nobody does this, it's normally just once in and out.

So the main test would still be to record just once using different converters and then compare them.

I'll still be doing this test anyway just to see HOW much degradation there is.

At first I thought this was a really bad idea,
then it got very interesting..
now im ok.

converters at 44.1 16bit capture much less than analog.

lossing money its not acceptable..
lossing audio is?

youtube and .mp3 at 128kbps its unsoportable... takes out all the joy of listening music,.
i have records from 1979 that sound amazing. LOL.

[kraku]

So far I can say that the results (at least to me) are stunning!

[jbjoubaud]

Originally Posted by kraku

There is no way for me to check that. I just have to rely on people not cheating on this test. I do agree that they sound suspiciously good.


@jbjoubaud:
Can you confirm that you really played and recorded the file 10 times so that the signal degraded during each recording round?

Yes, I definitly did and the file I sent you definitly is the one that went 10 times in the convertors.
You have to understand that the uln8 is one of the best convertors out there.
I'm pretty sure that the saffire did not go 10 times in the adda though. I think the word would have circulated if it was that good

[jbjoubaud]

By the way I did a null test with the original file every time to match it in volume as close as possible.
And the more I went in and out the convertors the more I was loosing a little bit a high end and vocals (not so much low end; and low mid losses)

[GoldMember]

Originally Posted by kraku

There is no way for me to check that. I just have to rely on people not cheating on this test. I do agree that they sound suspiciously good.

this is a test noone can cheat... just 1 pass.
Same converter...
$1700usd Clock Jitter: Saw wave 5KHz @ 44.1kHz.

Standard/Common RME Clock Jitter: Saw wave 5KHz @ 44.1kHz.

original file...
http://www.gearslutz.com/board/attac...s-dac-test.wav