Wavelength Cosecant v3 USB in Stereophile

[manisandher]

Hi,

 

Does anybody know why John Atkinson did not give an absolute value for the jitter performance of this device?

 

All he wrote was, "This is the best jitter performance I have measured from a DAC fed USB audio data." But as we know, this isn't really saying much...

 

Could it be that because the device measured so poorly elsewhere, he didn't want to 'rub salt into the wounds'? This is pure speculation on my part, but I have never read a Stereophile review of a DAC in which an absolute figure for jitter hasn't been given (even though the usefulness of such figures is often disputed).

 

Mani.

 

[PeterSt]

Just a guess : The analyser doesn't have USB digital output so it can't have any direct feedback.

As the review said : the judgement of jitter performance has been done through a player (well, a Mini and a MacBook).

 

In this case jitter can be observed only by relative comparison (and knowing how jitter in a given situation (like here the 11025 and the 229Hz toggled LSB) looks like.

 

On the other hand I'd say that you can always express it in dB and derive the time figures from there; if you look at the first picture (from the Mini) I think it is fair to say that

 

a. jitter expresses at -95dB in this picture;

b. compare to all the other pictures with dB on the y-axis the jitter pictures have been shifted down 10dB;

c. ... so we can't see the peak of the 11025.

 

In theory the peak can be at -60dB, and thus jitter would be at -35dB relative. That is (the most) poor.

But of course the peak will be at -0dB and it just never fits the picture, which makes the jitter -95dB ("poor" but inaudible IMO).

 

If the truth is somewhere in the middle, that middle would be a peak at -30dB and jitter of -65dB.

 

That's when a Mini is used; with the MacBook it would be -85dB.

 

What am I doing ? oh well, playing around a bit. But I guess I don't like reviews with "small errors" nobody will notice.

 

Regards,

Peter

 

[manisandher]

Compared to the Cambridge Audio Azur DacMagic's USB input, the Cosecant's USB input definitely performs better. However, the former's 'conventional' input performs significantly better than the Cosecant's USB input.

 

Peter, how do you derive the jitter time figures from the dB figures?

 

Mani.

 

PS. I retract my earlier comment that I've never seen Atkinson omit an absolute ps value for jitter - he didn't for the DacMagic either.

 

[PeterSt]

Since I'm not good at my maths today, I think you can get it from here : http://www.computeraudiophile.com/content/Some-additional-thoughts-about-measuring-jitter#comment-14010

 

I haven't been doing many jitter measurements myself, but I think I recall from the analyser that 1ns would be 70dB down.

 

Peter

 

PS: I am rather confident jitter isn't much bothering us. Other things in the DAC are way more important.

 

[manisandher]

Peter, you may well be right.

 

I suppose the real question I'm asking is, "Is the USB jitter performance of the Cosecant 'good enough' for us not to be concerned with?" Which is different from asking whether it's the best-performing USB DAC.

 

Reading some of JA's other measurements, it looks like a value of say 250pS is considered 'OK' (I'm not sure about 'good enough', because there seems to be some debate as to what this minimum figure actually should be). So, is it below this? I'm not sure how JA derives his time values from his dB values, but surely he could easily have quoted a figure and compared this to 'conventional' ways of piping the data. Was there a reason why he didn't?

 

Mani.

 

[Audio_ELF]

It seams like sometimes we get caught up in the micro measurements of things like DACs - what is it's jitter level, which has the better output stage, etc. - when what's most important for most people is "How does it sound at the output?" If something sounds bad, then maybe you start to look at why it sounds bad (high jitter, cheep op-amps used in output stage, etc.) but if it sounds good and does the job YOU need, then thats what's really important... isn't it?

 

Just like an amplifier with low distortion figures can sound worse than one with high distortion figures, a DAC with high jitter can sound better than one with practically zero jitter. We're back again to the importance of listening to a device in your own setup - and then going to have to bemoan the death of audio / HiFi shops and the inability to get Wavelength products to test in the UK, etc. Makes life difficult I know.

 

Eloise

 

[manisandher]

I sympathise with the gist of your post - ultimately if you like the sound of something, then go for it. (But personally, I like to know there's some objective underpinning to my subjective preferences... in audio at least.)

 

But I felt there was something 'strange' about the lack of comment on the Cosecant's jitter performance... in relation to other 'conventional' (i.e. non-USB) devices. After all, 'Streamlength Asynchronous USB' is the Cosecant's USP, so why not tell us how this USP measures up against other competing methods... not just other USB methods.

 

Mani.

 

[Audio_ELF]

I'm not sure of how they go about measuring the devices, but maybe the point is that you can't (fairly) compare the USB on the Wavelength to SPDIF (or AES) on another device. The Wavelength DAC cannot be used with a normal CD-spinner transport ... that much is obvious. You can use an SPDIF DAC with a computer, but how would you take into account what jitter comes from the computer to SPDIF interface (weather thats a USB, Firewire or inbuilt interface).

 

I do agree that some objective figures are useful, especially to backup claims which don't always logically add up. There are jitter graphs in the Stereophile report ... but I guess thats not what you're wanting? I'm not sure there is anything suspicious about the lack of absolute figures, just for some reason they were either unavailable, or missed in the report. To be honest - I'm actually unclear about what is different in the report on the Wavelength compared with other DAC reports I've read through quickly.

 

[Addition] Sorry I understand now ... in the test for the Wavelength he says [..]This is the best jitter performance I have measured from a DAC fed USB audio data.[..]

[Wavelength]

Guys,

 

The Miller Audio Test set use to be JA's number machine. The problem was that this test set used a 16 bit ADC board from National Instrument and is only capable of medium scale testing. When you get down really low like a lot of the newer pieces are then there is no way for the Miller's SN ratio to compete with the unit under test.

 

The AP and Prism tests sets do not give numerical output of jitter. A bigger problem is that unlike SPDIF using the JTEST does not really give you the spuriae that typical SPDIF dacs have. Data related jitter just would not happen like it does with SPDIF. As the 1/192 1 bit oscillation of the JTEST would be lost in the SN range of the dac chip. This is why Prism says that the best avenue for testing USB devices is using a 1/4 Fs -10dB signal and also looking at the distortion with different oscillators in their arsenal.

 

When you are testing an SPDIF dac you can with these test sets (AP and Prism) inject jitter of differing types and see what kind of rejection they show. But also these tests set have ultra low jitter output so that when testing SPDIF dacs they can show the true nature of the rejection of jitter.

 

I test jitter at the Word Clock input to the dac chip using a Wavecrest DTS system. This is really the best way to evaluate jitter and is kind of like an empirical Meitner LIM device as it is capable of measuring clocks to better than 500fS. I also measure the jitter @ Master Clock which is more important than Word Clock jitter for most contemporary dac chips.

 

But this would mean that if a Magazine were to use this device to evaluate product that they would have to open the products and figure out were all these signals were and then it would be really an unfair test.

 

I know John really well and have worked with him over the last couple of years on a varied number of projects. Do note... testing products like this takes a ton of time. We worked on the iPod jitter testing together and man... it took nearly a week of time to do this.

 

Testing is a good thing for all involved, but in the end we all know that this is merely one of the factors for why something sounds the way it does. It's always best to try it out in your system before buying a new purchase.

 

Thanks

Gordon

 

[manisandher]

I'm either going blind or insane...

 

Does he say, "The Miller Analyzer indicated some 250 picoseconds peak–peak of jitter, though this is close to the resolution limit of that test gear."? For the Cosecant... in the online version?

 

But that's exacty what I would have liked to see. For example, for the Benchmark DAC1, JA wrote:

 

"The measured jitter level was now just 119ps p–p, which is both superbly low and at the limit of the Miller Analyzer's resolution."

 

Or perhaps I should stop being so paranoid and simply see it as a non-important omission...

 

Mani.

 

[Audio_ELF]

Sorry something happened to my addition to my post and it lost some of my words - I obviously need to buy a higher grade Cat5 cable to connect to my router (Joke!) ... I meant to say ...

 

Sorry I understand now ... in the test for the Wavelength he says [..]This is the best jitter performance I have measured from a DAC fed USB audio data.[..] However in a test for the Meridian 808.2i CD player it's stated that [..]The Miller Analyzer indicated some 250 picoseconds peak–peak of jitter, though this is close to the resolution limit of that test gear.

 

I'm assuming you think there should be a similar comment, with a ps figure, for the Wavelength?

 

Again sorry for the confusion...

Eloise

 

P.S. Like you I've also seen Gordon's post but thought I should still correct my earlier post anyway. Sometimes feel like bashing head against the wall.

 

[jtwrace]

As a happy owner of this DAC, I think it measures GREAT to my ears.

 

[Wavelength]

Mani,

 

Do note the measurements for the DAC1 USB differ from that of their SPDIF inputs.

 

But really this is just sillyness and numbers. When it get's to this level and you don't listen to it then it's just a waste of time. For your system either one of these products could sound better. Only you will know.

 

Thanks

Gordon

 

[PeterSt]

I never believed in measuring much, which maybe springs from thinking NOS is better than OS in general (but it needs *good* NOS to start with) and NOS measures so bad. However ...

 

Diving into the matters lately, I found it to be a GREAT help to measure, and in the end just using that as a tool to improve the performance of the DAC. Without this tool it just wouldn't have been possible, but I can tell you ... better figures do sound better ...

 

(I am not talking about jitter, just the general figures like harmonic distortion types)

 

I can also tell you that often things are tradeoffs, and the one figure "making the sound" being better, another can be worse - but inaudible to your findings. So it is the end result which counts (and this is in your ears).

Sadly this will not say that any random reviewer is capable of finding those tradeoffs, or as a matter of fact he/she won't, because the ready DAC can't be toyed with. Besides that, it would take 6 months to do it all.

I know ...

 

Best,

Peter

 

[manisandher]

Hi Gordon,

 

Thanks for your detailed explanation earlier. I suspected that measuring this stuff is difficult , but now I know that measuring this stuff is difficult (fS measurements?)!

 

In 'Why Fix Jitter?' on your 'Concept' page , you make a clear inverse correlation between jitter and sonic performance. This being the case, I think it's reasonable for me to ask, "How much better (i.e. lower jitter) is 'Streamlength Asynchronous USB' than 'regular Adaptive mode USB'... and also than the other 'conventional' modes (and perhaps even firewire)?" I still maintain that it would have been useful if JA could have given us an idea of this.

 

But seeing as he didn't, perhaps you could?

 

Mani.

 

PS. I totally concur that testing is merely one of the factors for why something sounds the way it does.

 

[manisandher]

Peter,

 

You said, "I am rather confident jitter isn't much bothering us. Other things in the DAC are way more important."

 

Do you have a 'maximum level' for jitter, below which it isn't 'much bothering us'? What other factors are more important then? How well do they correlate with sonic performance? How easily are they measured?

 

Sorry, not meant to be an interrogation - I'm just very curious...

 

Mani.

 

[PeterSt]

Hi Mani,

 

Of course I am asking for questions like yours by stating that other factors than jitter are more important, but I better hadn't said it, because it is a complex of things which hardly can be explained by their elements.

 

As you will have noticed from the (still basic) measurements, my DAC crosses quite some boundaries when it comes down to THD+N, like worst case figures of 98dB down for the whole frequency range which is "unheard" for an NOS. Now, I could start telling you that this is related to a 4 layer board design, a special setup of the 8 chips, an output stage unheard of and blahblahblah, but it comes down to the insight (hence good obervation) of which does what.

 

A DAC chip (and all active elements) expresses distortions, and those distortions are stated in datasheets to begin with. However, this is about minimum, typical, maximum, and is often expressed for one frequency like 1KHz and for one or two levels like -0dBFS and -20dBFS.

One of the "benefits" in me, is that I not only can do some programming where needed, but that I also wrote a software player, and I thus know a little what should be going on over at the DAC side, one of the dimensions being the digital volume in the player. Of course, people may think the digital volume is a convenient thing, or it may be a good thing to leave out the pre-amp, but in the end it controls the quality of the DAC. So, remember about these min, typ, max I just mentioned, because by this little means you can twist the DAC in your direction (to a certain extend).

 

Here we stop, because a discussion about digital volume takes ages alone, so I hope this is enough example of what can be done inside a DAC design itself which, however, is also related to strategic decisions as software filtering.

 

The reason why I said that jitter isn't bothering us is because the above implied things have such a huge impact on the sound, that either jitter will be masked by it anyway, or otherwise it just can't have its influence. I mean, go to Sterophile and grab the figures of the Zanden; People may think it sounds good, but with that huge amounts of THD+N it just can't. You are tricked. Things get emphasized which shouldn't, and although one may like it, it is not natural sound. This is so enormeously bad, that I'd get rid of that first before thinking about jitter.

So the real answer is : I don't know because I don't think I have ever experienced it. For me this is like with cables : only at the very last I will be working on cables, and I am thinking this for over 30 years now. They may matter, but other things influence far more, first.

 

If you don't mind, this is all such a pile of "strategic thinking" and making basic decisions, that it can't be captured in a few sentences anymore. I already hear you thinking "what about the influence of XXHighEnd then ?". So yes, that is part of the whole game too, and only when that influence has gone the DAC has a chance to be good. I hope you can understand.

When I decide for "filterless DAC", in fact I decide for my own filtering. Along with that goes NOS, otherwise this isn't possible.

If I say "my own filtering", what I really say is that I will be in control of pre- and post ringing and the phase change how I want it, which is all just a very different means of rounding the too square higher frequencies which may be have been sines in the original. OS does this too, but completely out of control and it only destroys.

 

When these basics are sufficiently under my control, I may start a small project on jitter.

Or maybe I then just sit back and listen.

 

Peter

 

[Wavelength]

Mani,

 

In our tests with other TAS1020 Adaptive implementations the Async protocol would result in a 35x increase in performance (jitter tested at the Word Clock output of the TAS1020, Wavecrest DTS Calibrated last 12.19.09). In off the shelf controllers like the Cmedia and TI/BB PCM27/PCM29 series it is actually much higher.

 

But do note that everyone using these controllers does have some sort of jitter reduction circuitry. So again and I am not beating a dead horse but really everyone should listen to this stuff.

 

Sitting around and asking questions like jitter is really kind of a waste of time. There are so many factions that decide how something sounds that this is just one of the variables you should be looking (listening) at.

 

Thanks

Gordon

 

[manisandher]

Hi Gordan,

 

I'll stop 'wasting time' then. Thanks for your explanations though...

 

Mani.

 

[lxgreen]

for those of you considering the wavelength cosecant, are you comparing this to the new Ayre USB DAC? i guess the big difference is tube vs ss but since the Ayre has just come out, is there anything about it that would make it a better choice? It is also cheaper.