Pro Audio AD/DA convertor

[BillEl]

Has anyone on the HiFi stereo side of things done any extensive testing using these as a DAC?

 

Seems like they might have the most accurate conversions.

 

 

Thanks, Bill

 

[PeterSt]

You will receive a 100 responses with "Yes, true", but your first is a "No, not".

 

You don't have many options in that area. Notice that we talk about the playback side of things, and not the recording part.

 

While for me it is a fact that there's nothing much like "accuracy" in these devices there's also some logic in it : musicians don't need it.

 

So, after this the 100 posts with the "Yes, true" will appear. Me against 100. haha

 

Go listen to a Zanden and the like and compare. And even a Zanden is far from accurate, but then you'll know what to listen for.

Stuff like the a Zanden will cost you a car, so know what you are asking for ...

 

All the best,

Peter

 

[PeterSt]

Before someone else will say it, my example of the Zanden is a complete stupid one because it is so expensive and not even good (IMO). I took it as en example though, because of its known (very) poor measurement performance (throughout all the aspects). Actually it is a good case of proof how accurate one may think a random other DAC is, by looking at measurements.

 

Actually I didn't want to dive into this subject (of measurements) but thought to be funny to refer to a DAC which measures the most poor, but still will sound better than a well known pro DA/AD. And this is because the measurements of a DAC like the Zanden are honest (for harmonic distortion I mean), while the measurements for any pro gear DAC are not.

This is the difference between non oversampling DACs (like the Zanden) and oversampling DACs (nearly all others).

 

In the end a more honest or better reference would be (I think) the Trinity DAC. But I don't think I saw any measurements from this anywhere.

Audio Note will also be a better example.

 

All 'n all, "accuracy" is very dangerous in this (DAC) realm, because it can't be measured properly for OS DACs. But you can easily hear it.

 

Sorry for a confusing post,

Peter

 

[souptin]

Musicians may not 'need' accuracy, in the sense that they have learned to hear and understand what is happening musically under less than ideal circumstances: Venues with poor acoustics, audience noise, etc. I would question whether this also applies to other musical professionals though: Surely the recording and mastering engineers need to hear the recording with all its flaws as well as the underlying performance?

 

I actually find accuracy a difficult thing to judge. With multi-tracked recordings, logically you might think that the most accurate way to listen would be on a similar set up to the mixing room (ie professional kit), but one of the steps in the mixing process is to listen to the mix on other equipment, for example a car stereo, mp3 player, ghetto blaster etc.

 

Oh, and by the way, sign me up to the 100's saying that professional gear is worth investigating for home playback. I don't say that it is by definition better, just don't rule it out.

 

[PeterSt]

OF COURSE you are correct that the recording engineers want to hear back as accurately as possible what they are creating. But the gear just is the pro gear (with everything around it necessary for the job) and no "Audio Notes" have been built in their tables. One exception (I just don't know of any others) : the DXD mastering making use of that one and only DAC I know doing 24/384 for the Pro world, used in Pyramix systems. Sorry I forgot the name, but it's hardly a commercial DAC for our world, and is used by e.g. 2L and IIRC even has its origin there. It may be from Switzerland (it's an expandable thing, and can be bought in 192 version also).

 

You can well say : if you master in DXD you will be using that DAC (and Pyramix). If I am wrong on this, I'll stand corrected !

 

So, that is the exception I can think of, but it doesn't mean we should go out and buy that DAC (way too expensive again).

 

Oh, before I forget : there is one other "secret" DAC (maybe that's the one from Switzerland) also made for DXD mixing, but or the development stopped, or it's never finishing.

 

I actually find accuracy a difficult thing to judge

 

You are completely right, and this is exactly why I put in that strange post (which was too strange to my own likings). Again this is about the measuring stuff we shouldn't dive into, but for me this is about the "knowledge" I have by now and which nobody else can have because I created that stupid DAC which tells me all (and which does not exist in any similar form). So, as a crazy audiophool right now I only want to share share and share whereever I can. Here is such a place, where it comes down to "accuracy" as such.

 

And so, I hope I can do it once again;

If you look into e.g. the Sterophile review of that Zanden I mentioned at first, you can see how really bad it measures. This is just nothing more or less than NOS/Filterless (but with analogue "end" filtering). How come that so many prefer NOS ? Now notice that in these times this is becoming "hard" because the hires world depicts that the DAC should be 24/192, and in NOS this doesn't exist (although we're still in doubt of the Pacific Microsonics Model Two which is out of production). So, eager to find out what actually is going on, I created one myself BUT I shovled the lot such that my THD figures are at -105dB througout the whole frequency band up to 1/2fs (in my case the 1/2fs is 192KHz). The Zanden is "capable" of -30dB ...

 

The story is much much longer, but to summarize it all, just listen. Today I claim indistinguishable playback from live performance (which can also be in a studio). Quite a claim, right ? Well, it happens even with 16/44.1.

There is no way the (a.o. RME) gear I have myself comes even close, nor does the other stuff which floats by regularly.

 

I mean to say that "accuracy" is an easy thing to listen for, once you can listen to "the most accurate" and thus have the reference. If you *now* combine this with measuring, it becomes clear that measuring itself is wrong. Again, this thread should not be for that, but when my stuff measures 0.004% THD throughout while e.g. an ESS Sabre based system would measure 0.00002% (maybe there's a zero too much in there) which I can easily let "pure distort" on some music tracks, there must be something going on (hence wrong).

Add to this that I can let drop my 0.004% to 0.02% by means of stuff like inserting a volume control, and the sound suddenly is completely dead and wrong.

I want to say : what is measured on an NOS DAC is reality, while what is measured on an OS DAC is not at all. This is about the inherent means of measuring, and periodic waves which are no music. Blahblahblah.

 

I can let it boil down by saying that each Pro DAC will be an oversampling one because 24/192 NOS does not exist (but think about the PMII) and that just kills all. Not because I like to say it or like to have an opninion, but because I prooved it.

 

But remember, all is relative; If you listen to a random DAC and you find it accurately sounding, so be it. But this is why I ask you to listen to something like a Zanden, Audio Note, Trinity (the latter *is* a multi bit DAC !), and then compare with all those so much better measuring no matter what gear. You are fooled !!

 

Djeez, if only Julian Dunn would still be around, I would call him and ask him politely to make something better for us.

 

Regards,

Peter

 

[PeterSt]

Btw, if you combine my before post with this one : http://www.computeraudiophile.com/content/Upsampling-Just-Illusion#comment-53043

and my response to that, it might even become logic (Chris, as mud ? :~). So, think in the "bit perfect" realm, try to see that OS never can do that, while I just did that with NOS.

 

IOW, if we nag about "bit perfectness" why not extend that beyond our players ?

And it really helps ...

 

Man, I'm full of it. Sorry ... :-)

 

Peter

 

[Audio_ELF]

I think the general "mania" for bit-perfect is because Windows (XP especially) desn't have a very good sample rate converter. This is improved in Windows Vista/7 and quite good in Apple OS X (Core Audio). It's also a case of making sure that the software isn't applying EQ, etc. without the user being aware.

 

Beyond that it's up to the end user to decide what processing he/she wants to occur. As commented in another thread, sometimes a little EQ can actually improve the sound from a particular track (or all tracks).

 

Eloise

 

[PeterSt]

Okay ...

 

And how is this related to this thread ?

 

I'm sorry if I encouraged you by the title of my previous post, but I really don't see how yours relates to this all.

 

No big deal of course (not even a small one), but I don't think there should be another "bit perfect" debate in the PC area.

 

But as said, sorry if I encouraged for it. :-)

Peter

 

(PS: I now used a title which should be ok. Haha)

 

PPS: Please don't compare the tiny whiney :-) dithering bits (like from XP) with the huge changes the complete stream undergoes as the result from filtering.

 

[Purite Audio]

Quote,'Today I claim indistinguishable playback from live performance (which can also be in a studio). Quite a claim,

You are heartily plugging your new dac Peter, when will it be ready, I am looking forward to 'live' music in my room.

Keith.

 

 

[Audio_ELF]

Well I was just responding to your title thread and comments on bit-perfectness... No more shall be said!!

 

To the original poster: yes some people do use pro-DAC/ADC in their stereo hifi, you'll often find people using Metric Halo, Weiss DAC2 and Apogee, not to mention Benchmark and Lavry which are essentiall pro-audio DACs.

 

Eloise

 

[PeterSt]

Keith, please cut that out if you still can (edit). I am NOT doing that at all.

 

Grrrr ... guys, let me please say the stuff I want to say.

 

I have one DAC and one DAC only, and it will NOT be for sale.

... until further notice.

 

It is just the proof of things.

 

I better quit posting here.

 

bye

 

[BillEl]

Wow, I love the lively discussion.

 

As a musician it's true the first concern is good sound as opposed to accurate sound. The accurate part only comes in when you've achieved a good sound and it is before a necessary conversion.

 

Most of the time I find that conversion is AD as it involves capturing the good sound. Once it's in there the DA is more about monitoring but inevitably you mix for a wide variety of outputs hoping the majority of listeners hear your good sound.

 

Even with accurate sound you only get one mix for everyone and that could be listened to with head phones, a computer or a stereo all with differing degrees of quality.

 

I was hoping something out there, outside of all the marketing and hype that HiFi equipment seems to be victim to, would be more accurate. Pro audio gear seems more focused on the results and good gear is just good gear but it seems like the search for accurate might be so subjective accurate might not be relevant after all.

 

Thanks again for all the thought.

 

 

 

 

[Audio_ELF]

"I was hoping something out there, outside of all the marketing and hype that HiFi equipment seems to be victim to, would be more accurate. Pro audio gear seems more focused on the results and good gear is just good gear but it seems like the search for accurate might be so subjective accurate might not be relevant after all."

 

As I commented, a lot of people are using pro-audio DACs.

 

You see regular mentions of Metric Halo and Prism Sound Orpheus along with occasional mention of Apogee - all at higher end pro-audio devices. That's aside from Anterlope and Weiss who are pro-audio companies who make consumer devices and offerings from Benchmark and Weiss. I'm sure there's others too.

 

Eloise

 

[BillEl]

I did see your previous post, thanks for the info. My comment was more toward the subjective issues of accuracy mentioned above as opposed to what pro audio gear people are using. That issue is relative to any piece of gear pro audio included.

 

Forgot to add Lavry. ; )

 

 

 

[Audio_ELF]

It's only one opionion, but I believe that Barry Diament (www.barrydiamentaudio.com) - an experienced Mastering Engineer - feels the Metric Halo ULN-8 is as "close to the live mike feed as he's every heard" (his words as I remember them).

 

Eloise

 

[PeterSt]

Hey Bill,

 

I think your few more posts by now, actually may make the subject even more interesting than I already know it is. But *you* could ask yourself what you mean or want with "accurate".

 

IMO no such thing exists like "representing an instrument accurately" because the only reference you have is the real instrument, and it is represented so "inaccurately" through any system, that it makes no sense to see this as a target.

And mind you, this time I am NOT speaking about systems which may do that, so this time forget about my previous posts. Instead, let's try to see what it means that we are very far from the accurate representation of instruments through loudspeakers.

 

It is my experience that it is an all or nothing situation, implied by the pure technical merits under the hood. And these differ so vastly per system, that your observation of "accuracy" becomes a totally subjective matter.

 

a. It is not right in the first place;

b. So it will be up to you how close to reality it sounds;

c. Because of it being wrong in the first place, the degree of realism depends on the instrument (etc.);

d. When you tweak the one (in the end technical aspect) for the better, the next collapses.

 

Although the one device (but let's keep it to DACs) will have more anomalies al together than the other, and thus the one may sound better to you than the other, both will be plain wrong as a fact. It is easy : if this weren't so, you'd perceiving the real instruments. And, we are 100% sure it can't be done, right ?

 

Let's say I measure a DAC, and because of the way it is technically setup, I see false harmonics in the audio band. Keep in mind, we're talking about the impossibility to have it right, which THUS means false harmonics will be there. It is really that simple.

This is a situation without preamplifier, and it means I'm looking directly at the DAC output. But now I'll hook up a preamp, and look again. I see that my false harmonics are way down now, or even gone. Great !

No, not great at all, because whatever frequency I noticed before - now being disappeared, will thus disappear because of the preamp, which really doesn't know the difference between a fundamental tone, a harmonic of it, or a false harmonic because of inherent harmonic distortion. Thus, it filters as hell, and while it filters out the wrong stuff, it also filters out the good stuff.

 

What I am saying here, is at using *any* means of volume control in the signal path (even a single very good resistor) shows behaviour similar to this.

 

So, at using any device or setup with a volume control, I can 100% guarantee you or anyone, that you will end up in that leage of subjectiveness.

 

Because this is technical impeeded stuff, you may try to tell me what is better : those false harmonics staying there, or the real harmonics being filtered out. It is one or the other you know, and you can never have both. Oh, theoretically you can, by not having the false harmonics in the first place, but remember, this was not our base, just because in the real world it isn't done, and for OS DACs it can't even be measured.

 

Well, stuck. Anything with (btw analogue !) volume control is ruled out now. It doesn't matter what brand it is, but as soon as a VC is involved it is the most easy to guarantee it won't be accurate. And not to forget : "accurate" inherently means 100%. No less. Also : E.g. 80% is completely out of control because of the mixed bag of anomalies, the one not leting sing a cymbal, the other killing a bass, the next naking a violin sound like a flute.

 

Which brings me to the OS DACs themselves;

Because it can't be measured what damage they do, it can only be reasoned out which is a long threaded thing to do. I have done it a couple of times, and still it is hard to accept, because way earlier science (math) told it is okay. Me against science. However, what I at least *can* point out easily for anyone knowledged about these matters, is that the same algorithms are used for creating a wave which ought to be the good one, as the analysing equipment uses to measure the perfectness of it. So, if I would say that 1+2=4 and made a wave of that, and next I measure the correctness of that wave, I will hold it against my 1+2 must be 4, and if that is the case, all is right.

This is really how HD (harmonic distortion) in DACs are measured, but it is *me* who observes that 1+2 is not 4 at all. It is 3.

 

(NOTE : This is only a virtual presentation of the story, because analysers don't anticipate on any math at all, but just want to see nice sines. Those sines though are created from (filter) math which will only work on a repetitive stream of the same "wave", while in music this doen't happen.)

 

This is about the filtering part of an OS DAC, which means is plain wrong, and every insider knows it, but won't admit it;

It is plain wrong, and it is easy to see too, because any filter rings (it echoes, post or pre and post) and *everybody* will agree that ringing is no good at all. Sadly this is what each DAC with (remember from the earlier post) necessary filter does, and per definition this is not accurate.

That our analysers present it as 100% accurate must be the biggest fault in digital audio. But they do.

 

(NOTE : Again, the analysers are accurate allright, but DO present the means to measure HD. Therefore it's actually pilot error to do it in the first place, but ... with the lack of anything else, what to do ...)

 

Right. So each filter rings, ringing can't be accurate, so all DACs with filter are ruled out now. Bye bye.

 

Bill, can you see what this is heading for ?

I'm just going to quit here with the example, because the story was told already in this very thread. It is the most easy to rule out DACs because they can't be accurate. 100% of the Pro DACs belong in that leage. Not because the Pro world wants to make inacurate DACs, but because they perceivedly need 24/192 and the whole Pro world is one big gang where you can't behave differently from the 1000 others around (or otherwise things get expensive -> Pro needs thousands of these devices, while consumers need tens, relatively speaking).

 

So, no Weiss, no Benchmark, no Lavry, no nothing there is close to accurate. They have been ruled out by their technical merits. Nothing to do.

 

What's left are the NOS DACs. They have a chance.

Funnily (or sadly), my story above started with inherent HD which is exactly what *will* be measured from NOS filterless DACs. Nothing changed the inherently wrong sine, and the HD coming from that will nicely show up. And this time, unlike OS, it is true what you see. The 1+2=4 rule (filter) just wasn't applied. It is true what you see, because it doesn't ring, with the notice that you can't see ringing from FFT measurements.

Again : Filtering does it's job with the explicit means of ringing, while ringing can't be measured as part of HD. The HD figure will be wrong and the least thing one can say is that all has to be far from accurate.

And : No Filtering does no filtering job, and thus doesn't ring. The HD which shows up is honest though and accuracy can be derived from it, as a relative means to devices which measure differently to this respect.

The Zanden again is the example, but no matter what, what you see is true. Don't hold this against any OS measurement, because that doesn't say a thing.

 

 

 

 

See how easy it is for judgement, where it comes down to accuracy and the first part of my story (the volume control);

What you see is a 1000Hz test signal (almost completely at the left) and red is without the (fixed resistor relay) VC, while blue is with. The volume is full open, test tone is at -0dBFS. Notice that blue overlaps red.

 

Important to note is that even the red is influenced by the VC, just because it's in the system, and its inputs are connected to the outputs of the not VC controlled parts.

You see this in the uplevelled noise up to just over 25KHz.

 

What it's about is the difference for HD between red and blue. The red spikes are inherent (but could just have been real harmonics) but they are filtered out (partly, see the 5-7KHz range or completely, e.g. 15-30KHz) in the blue (VC).

 

What we also see is the many "self spikes" in blue (except for the 2000Hz and 4000Hz harmonics they are not overlapping red, trust me).

 

Of course, by looking at this we could say that it's all inaudible in the first place. Yea, if that only were true. So, notice that this is one test frequency only, while an infinit number of them exist, each carrying their own false harmonics, which for the VC are not even derived from something. Also the level of them changes per frequency.

But I can tell you, with the VC in all music is completely dead. It's also less accurate, easy to hear in the bass.

 

So what did I just say ? "it's also less accurate". Wrong. It is inaccurate as hell, if we only compare it with the red readout, and assume that the best possible for the situation. There's no comparison, and this is only because of a stupid resistor (and relay). Never think this is inaudible.

 

Now think of how dangerous it all is;

Keep in mind, what you see here is from NOS, while OS wouldn't show you any single peak to begin with (but sadly it depends on 1+2 to be 4). But what if I'd left out the red graph ? In other words, what if I presented this all with the VC in it, and THUS can/will show the blue graph only. Just look at it ... it will look BETTER. Way better, just because you won't see those higher (red) spikes in the audio band plus you'd accept -124dB of noise (also SNR here) as good enough, the better SNR of the red graph not being there (remember, the red would show -134dB as you see it from off 30KHz).

But you just learned that this is not more accurate, because red is the accurate one. So that so called better blue graph (apparently) contains a filter, uhm - filtering in the audio band.

 

Peter

 

PS: I never, never found that a better measuring situation didn't sound better as well, as long as you're measuring the right thing.

 

 

[manisandher]

Peter, as you well know, I've been using your zero pre-/post-ringing upsampling scheme (Arc Prediction) with my DAC (which I believe is non-oversampling at 4fs) for a while now. I've made it clear on a number of occasions that I think what you're doing is game-changing. But there is something that still doesn't make sense to me.

 

How is it that 'regular' oversampling can sound so good? I mean, if I set my DAC to 16/44.1, it then uses its internal 8x oversampling filter. And it sounds really very, very good. If I then switch the DAC to 24/176.4 and use your AP upsampling in the software, the sound is different, perhaps better in some ways... but certainly not night-and-day better.

 

Any ideas?

 

(All listening done using software VC and DAC connected directly to power amp.)

 

Mani.

 

[PeterSt]

Hi Mani,

 

One thing I certainly can think of (although it would be a bit weak for an explanation) would be that -if indeed your DAC doesn't "oversample" as such- (like 64 times at least), there is no reason why it wouldn't do similar as what I do in the software. It's a matter of creating it and not believe in the common filtering to begin with.

 

Another (or even additional) explanation could be (but this is a more long shot) that the 8 times your DAC applies, is just that tad id needs to get it right. Because you know, in software you can apply 4x only (because your DAC doesn't allow the higher input). And, at this "drum recording" (see thread elsewhere), I could clearly hear the difference between 4x and 8x, the 4x adding distortion to the cow bell.

 

Lastly and generally, your DAC is just a very good one, so why wouldn't it be very good at this part too ? In the end (I think) the real judgements can only be made when we'd have the two together. I mean, if then my AP sounds worse than your DAC's internal upsampling (careful : filtering), my AP just isn't better. Still dangerous, because now I assume my DAC to be better to begin with. But how to do that when it uses AP in the first place.

 

A shorter answer would have been "I don't know". And actually I don't of course;

We know how difficult it is to talk about audio, and up to some extend it can be done very well (like the XXHighEnd tweaks and results), but at some stage it stops. Add to it, that (as how I think) we are talking about stuff up to the very limit of things, where even one resistor destroys all.

But with open minds we can always try ...

 

Peter

 

PS: One more thing to add;

Your DAC will behave quite differently at 176.4 vs. 352.8. That is, generally that may be expected. Its accuracy will be different (not necessarily worse at higher rates !) the current it draws will be different, and what about the jitter at the input side which is a result of 44.1 in the one case, and 176.4 in the other.

 

[Miska]

Way better, just because you won't see those higher (red) spikes in the audio band plus you'd accept -124dB of noise (also SNR here) as good enough, the better SNR of the red graph not being there (remember, the red would show -134dB as you see it from off 30KHz).

 

With digital volume, there wouldn't be any additional thermal noise added by resistors, noise floor would be static regardless of volume control position.

 

The SNR of the figure is way less than 100 dB, since the first harmonic is already evident at -80 dB and there are other high spikes.

 

[PeterSt]

Miska,

 

The SNR of the figure is way less than 100 dB

 

I assume you mean "way worse", right ?

 

since the first harmonic is already evident at -80 dB and there are other high spikes.

 

But what has this to do with SNR ? A harmonic (though false) is not noise ... It's harmonic distortion. That's why there's the THD+N standard. Total Harmonic Distortion plus noise ...

 

Unless SNR includes the "noise" of harmonic distortion ...

In that case I misinformed by calling wrong numbers. But you could always see it. :-)

 

Thanks,

Peter

 

 

 

 

[PeterSt]

Unless SNR includes the "noise" of harmonic distortion ...

 

Well, maybe I never realized that SNR officially is something like the difference between the wanted signal, and unwanted signal. Still I can't really dig up (Google) how the relation with HD is;

 

Funnily enough, my analyser tells me (mind you, from an older screenshot I saved, and it may differ somewhat) that the SNR is 119dB. So, I looked up how the measurement is done, and this starts with a signal of -60dBFS. Why ?

 

I can only guess (btw lacking the official education, but that is clear for a longer time, haha).

Anyway, assuming HD not to be higher than 60dB and inherent noise levels to be not much better than -120dB, it would (almost explitly ?) avoid HD. This is because HD will drop along with the signal strength and it will drop under the inherent noise level, and thus can't be measured (read : be incorporated in the resulting figure).

 

In the end, all I can tell is that my analyser says 119dB (remember, not in the consistent situation with the earlier picture, but I can see from the particuar measurement session that the first harmonic (1000Hz test signal) is at -93dBFS, which is the normal situation for the setup in question (along with 0.0023 THD+N might you need it).

 

?

Peter