A/D Converter Shootout (with samples)

[dysenterygary]

wow, and now 3 is taking the lead! When I first listened that was my least favorite of all for some reason, but then I kinda started liking it. I'm still staying w/5 as my favorite though.

[Ben F]

Should we start taking bets?

[ISedlacek]

Very admirable intention, but as Mr.Adebar very rightly said, it is very difficult to judge anything from it, since we are basically listening to the sound of Mytek DA converter, which we don't know unless we have it. So the results entirely depend on its sound qualities and if it (just for example) sounds "thin and bass light - or thick and bass heavy" or whatever" , the AD converter that sounds like that (comparing to the original) would be actually the best and most transparent one, but we would have a tendency to say it sounds far from the original and that's why it is not that good or transparent ... So I am afraid this test will not tell us much about the real quality and transparency of each of the AD converters (unless we use the same DA converter as during the test). I used to use Mytek converters long time ago and to my experience they did not sound as much open and natural as the ones I used later on ...

[mischa janisch]

Originally Posted by ISedlacek

Very admirable intention, but as Mr.Adebar very rightly said, it is very difficult to judge anything from it, since we are basically listening to the sound of Mytek DA converter, which we don't know unless we have it. So the results entirely depend on its sound qualities and if it (just for example) sounds "thin and bass light - or thick and bass heavy" or whatever" , the AD converter that sounds like that (comparing to the original) would be actually the best and most transparent one, but we would have a tendency to say it sounds far from the original and that's why it is not that good or transparent ... So I am afraid this test will not tell us much about the real quality and transparency of each of the AD converters (unless we use the same DA converter as during the test). I used to use Mytek converters long time ago and to my experience they did not sound as much open and natural as the ones I used later on ...

Only Robin could shed some light on how they hold up to what the DA delivered... But judging from his comments #3 should be the winner with #4 comming in second (my interpretation of his comments - post #31)

[HomeProducer]

I compared everyone with the original

1. darker
2. like the original (a little less punch? I am not sure)
3. fuller sound than original (more bass) thats why you like it
4. cannot hear any difference
5. cannot hear any difference
6. guitar arpeggio sounds slightly flatter

the differences are small. My ears can fool me easily.

[Spartacus]

Originally Posted by Kees de Visser

Hi Robin,
thanks for the interesting test.
It seems that the peaks in samples 1 and 4 got chopped off. Can you please verify that ?
If true, what could be the cause ? AFAICT there is no digital clipping so perhaps there was some analog mismatch between DAC and ADC ?

and it looks also that there is a balance issue in the low-meds with converter 6, seems like the sound is moving to one channel more than the other...

my favourite is 3 for the moment
1 and 5 seems to be the more coloured

[inlinenl]

the person who fits all names to the right converter , should win some price ...
some special avatar-comment or so by the mods ....
or free chicks ..
whatever ...

now Robin, tell me before the weekend ...

[24-96 Mastering]

Originally Posted by ISedlacek

Very admirable intention, but as Mr.Adebar very rightly said, it is very difficult to judge anything from it, since we are basically listening to the sound of Mytek DA converter, which we don't know unless we have it. So the results entirely depend on its sound qualities and if it (just for example) sounds "thin and bass light - or thick and bass heavy" or whatever" , the AD converter that sounds like that (comparing to the original) would be actually the best and most transparent one, but we would have a tendency to say it sounds far from the original and that's why it is not that good or transparent ... So I am afraid this test will not tell us much about the real quality and transparency of each of the AD converters (unless we use the same DA converter as during the test). I used to use Mytek converters long time ago and to my experience they did not sound as much open and natural as the ones I used later on ...

Right, that's the limitation of this test. There's a D/A in the mix with all A/D samples. To put this in perspective, it must be said though that this is always the case whenever one is comparing A/D converters. There's always a difference of one D/A conversion between the original and sources to compare to, whatever the test setup is. That's unfortunate but inevitable and no one has ever chosen converters in any other way.

If one thinks that additional D/A step has a significant influence on one's preference, one should repeat the same test with a different D/A and see if the choices made remain true. Unfortunately I didn't think of doing that, so I have no additional samples to offer.
Personally, I'm guessing that the tendencies the samples show would remain similar with a different D/A, but I must admit I have no grounds on which to base that.

[mischa janisch]

Originally Posted by 24-96 Mastering

Right, that's the limitation of this test. There's a D/A in the mix with all A/D samples. To put this in perspective, it must be said though that this is always the case whenever one is comparing A/D converters. There's always a difference of one D/A conversion between the original and sources to compare to, whatever the test setup is. That's unfortunate but inevitable and no one has ever chosen converters in any other way.

It's not the conversion that bothers some - it's that unless you listen trough a Mytek 192 it's a hit and miss... I think that the different listening DACs that people are using yield different results in judging which AD is truer to the source.

For example: Your DA may be a bit shy in the upper mids (I'm making this up), compared to my DA which is a little more agressive. When comparing samples of the ADs to the original I will probably pick a more agressive AD as truer to the source than you would since my DA will make the original more forward sounding - you have been playing back through a softer sounding source, hence the AD has to make up for this softness for me to sound closer... Of course your observation would be the more valid one in this case.

[24-96 Mastering]

Originally Posted by mischa janisch

It's not the conversion that bothers some - it's that unless you listen trough a Mytek 192 it's a hit and miss... I think that the different listening DACs that people are using yield different results in judging which AD is truer to the source.

For example: Your DA may be a bit shy in the upper mids (I'm making this up), compared to my DA which is a little more agressive. When comparing samples of the ADs to the original I will probably pick a more agressive AD as truer to the source than you would since my DA will make the original more forward sounding - you have been playing back through a softer sounding source, hence the AD has to make up for this softness for me to sound closer... Of course your observation would be the more valid one in this case.

I don't understand the idea that because I have the same D/A in my monitoring chain that was used in preparation of the samples, I'm better able to judge its effects. It's one additional Mytek D/A conversion in any case that is the same for everyone listening. For me too, whatever I'm monitoring on.

Put differently: When we both listen to those samples, it's ONLY the monitoring D/A in your chain that differs, otherwise we're listening to the same thing. We're both comparing

SOURCE -> Mytek D/A -> A/D #X -> our individual monitoring chain that we always listen to
with
SOURCE -> that same monitoring chain that we always listen to

That different monitoring chain features a different monitoring controller too, and a different room and different ears. I don't understand what's special about the D/A, or rather about the correlation between monitoring D/A and the D/A used in preparing the samples.


PS: I could monitor the samples using my Crane Song D/A and see if I make the same observations... now there's an idea

[kjg]

thank you for posting, Robin. very interesting.
i voted 4. although it does seem to be a touch mid forward, it was my overall favorite. 3 & 6 came in shared second place.

really curious what is what. would indeed also be interesting to see how your observations might change with a different monitoring DA.

thanks again,
klaas-jan

[Shy]

Originally Posted by Boro

I am not saying it's in the original master, as most probably it's due to the several conversions (WAV-FLAC-WAV with all inherited artifacts)

I don't know what you're talking about. FLAC is a lossless format and when decoded it results in a bit-identical copy of the original PCM data.

[24-96 Mastering]

Originally Posted by Boro

I agree with Ivo that we'd better had some input material of more "uncompromising" quality, though for a quick test this can do...

My first listening impressions were somewhat disappointing in terms of overall dynamics handling, although I am far from being a recognized gold-eared expert... Then I obviously had to find out why all different converters sounded so unequal at peaks while the overall loudness was approximately the same... probably manipulated after the conversion.

Then I checked the peaks and WOW: most of them performed extremely heavy peak limiting, in some cases of obvious clipping character... That might be due to improper input level, but I believe that such high-end manufacturers would have designed their analogue input stage(s) to have the necessary headroom to cope with any input signal... Well, some of them might have clipped due to weak/inferior power supply, or many other reasons but I absolutely doubt the designers were at fault, at least at that super pro level!

Wow, that's some serious detective work there, although I'm not sure such an overly dramatic presentation is in order. There is certainly no heavy peak limiting going on.

As said, file 1 and 4 are clipped at SOME points, the most of which is 9 consecutive samples at one single point in time. I'll go out on a limb here and say that with this material, one those peaks in question, that will not be audible.

But as written before, I need to have a look where in the process this happened and confirm that - if that error is solved - the difference is indeed inaudible.

use an input file of real high quality, instead of "lossless" FLAC source with obvious low-end inadequacies!

Hmmm...

[Shy]

Originally Posted by Boro

But it sounds different - that's why it's NEVER used in pro audio setups!!!
Otherwise we'd be able to save a lot of media space using it! Well, there are too many internet based discussions about it and ALL audiophiles, let alone pros, agree that it's only useful as a better substitute for mp3 and similar compressed formats.

The industry would never go for such surrogates when we all can rely on proven standards.

You have no idea what you're talking about. Bit-identical means bit-identical - exactly the same 1s and 0s, exactly the same PCM data. There is nothing to "prove", it's a fact that can easily be verified.

[mischa janisch]

Originally Posted by 24-96 Mastering

I don't at all understand the idea that because I have the same D/A in my monitoring chain that was used in preparation of the samples, I'm better able to judge its effects. It's one additional Mytek D/A conversion in any case that is the same for everyone listening. For me too, whatever we're monitoring on.

Put differently: When we both listen to those samples, it's ONLY the monitoring D/A in your chain that differs, otherwise we're listening to the same thing. We're both comparing

SOURCE -> Mytek D/A -> A/D #X -> our individual monitoring chain that we always listen to

with

SOURCE -> that same monitoring chain that we always listen to

That different monitoring chain features a different monitoring controller too, and a different room and different ears.

I'm sorry - my thinking was flawed here, but there is still the other problem left: Without knowing the Source DA it is still impossible to judge the AD objectively. It is only possible to say which AD suits your needs the best - we are judging the quality of your analog loop with 6 different converters (As others, including yourself have already posted before).

Hypothetically, to overemphasise the problem:

Let's say your DA has a built in high shelf (-2db@10k)
AD#1 is flat - recording will result in a file that is (-2db@10k)
AD#2 has a built in hi shelf (-2db@10k) - recording will result in a file that is (-4db@10k)
AD#3 has a built in hi shelf (+2db@10k) - recording will result in a file that is true to the source

AD#3 would be the winner since it will produce a result which is the closest to the source file (and yes, everyone will be able to hear this ), even though AD#1 is truer to the (hypothetically flawed) source.

Don't get me wrong - I'm not saying that your DA is actually flawed, it's just an example to show why the result would be different with a Lavry or Prism DA...

[24-96 Mastering]

Originally Posted by Boro

Attachment 141963

Are you sure your pictures are accurate? When I look in my (sample accurate) display, I see no clipped samples at that point in time.

[24-96 Mastering]

Originally Posted by mischa janisch

Don't get me wrong - I'm not saying that your DA is actually flawed, it's just an example to show why the result would be different with a Lavry or Prism DA...

Quite possibly a different D/A might lead to a different result. No doubt a valid concern, and a variable which unfortunately can only be eliminated by using multiple D/As in the test setup. If I still had the converters here, I would be running the samples with a 2nd D/A to create a control group.

Still, I believe that the 'character' of the A/D comes through, to a large extent. But as I said, that can't be proven at this point and this test is not carried out in a thoroughly scientific manner, so I want to advise not to base any significant decisions on it. To be sure, everyone MUST stage their own test to make sure all variables (that one considers to be relevant) are controlled.

[beanface]

Mr Schmidt, when are the results going to be published?

[24-96 Mastering]

Originally Posted by Boro

You're not looking at the exact same peak, just the previous smaller one! Look at 01:11.669

My bad. I was looking at the snare, not the tom. I can confirm that it is indeed clipped for a length of 5 samples.

[24-96 Mastering]

Originally Posted by beanface

Mr Schmidt, when are the results going to be published?

See post # 46

[Spartacus]

so, now I'm confused,
the chain was clipping while during this test ?
looks like all the samples are clipping...

[24-96 Mastering]

Originally Posted by Boro

Yeah, unfortunately No 4 IS CLIPPING 3 dB!!! And all of them at least 1 dB which is clearly audible!!!

1 dB clip (3 samples or so?) on that tom would be pretty hard on that spot with blind ABX identification... dare I say impossible?

But anyway, let's cut this short and get back on track: If that clipped tom is bothering you as much as the low end loss of FLAC, please just listen to another section of the song...

[24-96 Mastering]

Originally Posted by Boro

The heavy clipping is not the main issue, especially with this particular source material! It would be a nightmare with contemporary rock or dance music!

Are you for real?

Anyway, I advise you to listen to the sections that don't heavily clip instead of posting about the tom hit that does.

Why didn't you leave the the way the corresponding converters had put them out?

In order to align them and calibrate level as close as humanly possible (by objective means) in order to make objective comparison possible. Trust me when I say this this will make a very much bigger difference than a clipped tom.

[Waltz Mastering]

Originally Posted by Spartacus

so, now I'm confused,

Any clipping would only expand the test so that you can hear how that particular AD sounds when it is clipped.

[funka]

There's always a difference of one D/A conversion between the original and sources to compare to, whatever the test setup is. That's unfortunate but inevitable and no one has ever chosen converters in any other way.

That's why we tried to make A/D tests with an analog source, tape.
But with this test, the problem is to be able to listen to the original tape sound, without being at the place where the tests where done...

Big A/D converter test session

Soon maybe we will add some other converters in this test, Lavry Gold...will post on the other thread when done...Sorry for hijacking.

[24-96 Mastering]

Originally Posted by 24-96 Mastering

There's always a difference of one D/A conversion between the original and sources to compare to, whatever the test setup is. That's unfortunate but inevitable and no one has ever chosen converters in any other way.

Originally Posted by funka

That's why we tried to make A/D tests with an analog source, tape.

An analog setup doesn't solve the principal problem that there's one more stage of D/A in the test signals compared to the original source. You can't listen to an A/D without D/Aing it first. Whether your source is analog or digital, that problem remains the same.

In the absense of an ideal D/A converter, it seems the only way to take the variable out of the mix to some extent is to repeat the test using different D/A converters.

[soundgeezer]

I just tried something crazy:

I put the free plug attacker from spl on the master bus and while listening
to a small looped part of the original file I raised the attack knob till the red overload light was blinking.

Then I checked the other files to see what happened.
The loop was like 3 seconds long and on 1 kickdrum the over light turned red
when listening to the original track.
The only 2 other tracks, that made the over light turn red were the tracks
from converter 2 and 3 with the same setting of the plug.

Maybe all other converters don't translate the transient as well as #2 and #3 ?

It might be completely crazy to try this at home but check it out yourself and see what results and conclusion you might get.

Just my 2 ct as usual

[Waltz Mastering]

Originally Posted by Boro

OK, let's set up an "ideal" set of input signals (at least two programs of different dynamic and corresponding spectral contents, each 2 minutes minimum) and define the test conditions! I mean:
- level to the ADC input;
- identical "filter" settings (phase-linear, steep, or whatever they offer identical!);
- NO manipulations of any kind over the input files! after the "agreement"!;
- perfect (as possible) power supply for every single unit;
- optimal interconnects, corresponding to manufacturers' requirements! NO optical connections, of course;
- double checked in-situ settings, i.e. test made at two (at least) different locations.

Please, add what you find appropriate to make it as objective as possible!

When will you be posting the files ; )

[Waltz Mastering]

Originally Posted by Boro

I am not performing the test, just offering a fair conditions! Are you capable to make a really objective test happen? Do you have at hand that bunch of oh-so-expensive DAC/ADC converters?
Well, if someone has it all I'd be more than happy to help with the preparation for a serious and revealing test!

It was a joke; ) Yes, I have done a shoot out between Mytek 8x192, Lavry Black, Blue, Gold, Apogee 16, and Digi192 in my room a while back, but appreciate that Robin and Mr Velvet have gone to the trouble of posting what they have done.

...and like Jaakko and others have said it's how any one component integrates into YOUR rig/room, but certainly some conclusions can be drawn.

[inlinenl]

who cares if it all was 100000000000,9999999999999999 % straight/clear test ...
just have fun, listen, make up your own mind, listen once more .. tickle your brain ...
life is there to enjoy ...

really enjoyed what robin did ...