Review and Measurements of Schiit Yggdrasil V2 DAC

[solderdude]

Sorry... WOT following, techies please read bottom part !

The problem with trying to match 'heard' differences with measured differences is that one can always find small to not so small differences well outside the audible range.
I discussed this with Bob (e-mail not in person) when he published his findings (the jitter not being that different and Bob mentioning this is audible).
Didn't agree about the audibility of this level realizing (more like knowing) that the dynamic range (in practice), that I measured for myself' is around 70dB and that test did not even cover masking.
It is pretty easy to tie small measured differences to what is (thought) is 'heard' when you are of the opinion that the hearing is 'better' than is generally thought/assumed/measured under more, or less, relevant circumstances.

That's what I had discussed with Bob a while ago. He draws his conclusions based on own listening tests, experience with acoustic labs and conversing with coworkers he trusts.

I too would have liked him to not vent his 'subjective opinions' as much and just stick to what he does best (measuring) and not 'interpreting'.
Realizing that he is in the highly subjective 'circles' that can hear a flea fart at -150dB while listening at 80dB SPL peak levels, he is often asked for his opinion (being regarded as specialist) may play a part.

Bob and Luis Gilberto are about the only guys I respect and admire at a certain place and both seem to be in a total knot about Amir's measurements and spew foul words which is totally out of character for both men. Passion is in play here, just like with Amir who is being dragged through the mud because he doesn't like poor measuring DACs.
Them DACs being from their personal friends (Schiit) and rallied on by fearless leaders and minions makes them conclude (think, be convinced) that Amir HATES Schiit and that must be the reason for 'borking' measurements on purpose. Thereby completely passing by on the fact it measures poorly. I understand the position of lower levels being 'irrelevant' to SQ as they are below audible limits (my sentiment as well see -70 dB remark).

I don't know Amir at all nor his connections so have no opinion about this but all I see is Amir does not like poorly measuring stuff.
I do hope he never measures my designs (Bob did) as they won't be recommended for sure hurting the business of a good friend (not mine, not into business myself)
One also has to remember that Jason and Mike are personal friends of key people from SBAF and Jude.
Knowing Jason and Mike are really nice folks I can understand 'protective' bashing of 'non SBAfriends' but find their way of 'rebutals' against not like-minded people highly unprofessional/childish. To which they openly admit under the cover that it is just 'audio' and nothing to get in a knot about... untill friends are 'attacked' that is which seems to change their modus operandi.


Enough ranting and now and on to the actual reason of my post.

I have a question for those in the know about FFT's which involves the interpretation of 'skirting' and 'spikes'.
Preamble:
Having played around a bit with FFT's I realize that most of them are heavily averaged in order to get the noise floor down low so only always present harmonics and frequencies that should not be there but are are shown clearly and not drowned in noise.
Question:
This involves the dreaded jitter. (I do believe practical jitter in modern DACs is not an audible issue) but do have some questions regarding the plots and how they should be interpreted.
One has a constant tone (say 10kHz) which, because of jitter jumps in frequency a bit. This follows a Bell curve I presume. Lacking accurate equipment I cannot test myself. What I have seen ( a lot) is skirting. A sign of LF jitter I am told.
What basically happens in FFT's is that when a frequency deviates to a lower (or higher) frequency once in a while and still has the same amplitude it is shown as peak of high amplitude when no time averaging is done. This is giving a very noisy floor in the plot. But when this only occurs briefly once in a while and heavy averaging is going on (to obtain a low noise floor in the plot) that peak is lowered in amplitude each time it isn't there.
When it only happened a few times (and not at the exact same freq) and thousands of samples are averaged that peak will not show while the deviation has been there.
Is this shown as skirting ?

Is it possible to measure a single tone in 'memory scope' mode which would show the actual frequency being measured and the highest and lowest measured frequency in either equal brightness or less bright depending on number of occurrences over time.
Just being curious as to why such plots are never seen.
Having played with this option on my scope debugging jitter in long distance data communication lines at comparable speeds gave me insight in what caused the errors that were observed once every blue moon by 'recording' over long time periods.

 

[amirm]

One has a constant tone (say 10kHz) which, because of jitter jumps in frequency a bit. This follows a Bell curve I presume. Lacking accurate equipment I cannot test myself. What I have seen ( a lot) is skirting. A sign of LF jitter I am told.
What basically happens in FFT's is that when a frequency deviates to a lower (or higher) frequency once in a while and still has the same amplitude it is shown as peak of high amplitude when no time averaging is done. This is giving a very noisy floor in the plot. But when this only occurs briefly once in a while and heavy averaging is going on (to obtain a low noise floor in the plot) that peak is lowered in amplitude each time it isn't there.
When it only happened a few times (and not at the exact same freq) and thousands of samples are averaged that peak will not show while the deviation has been there.
Is this shown as skirting ?

It could. If Jitter is periodic, then it shows up in FFT spectrum as vertical spikes. If it is random but broadband, it raises the noise floor. If it is random but low frequency then it broadens the skirt.

Is it possible to measure a single tone in 'memory scope' mode which would show the actual frequency being measured and the highest and lowest measured frequency in either equal brightness or less bright depending on number of occurrences over time.
Just being curious as to why such plots are never seen.

You can get a plot of frequency distribution in time domain. This is indeed what digital scopes show. You would need very good scopes though to measure to high accuracy that we are able to do in FFTs.

We don't go there in audio much because it is hard to analyze its psychoacoustics. When plotted in time domain, it becomes much easier to see the effect.

In other domains like communications, we only worry about how bad the error is so the kind of graph you are speaking of is much more common.

 

[solderdude]

Thanks for your reply.

Does this mean that frequencies actually shift (due to jitter) as far as the bottom of the skirt ?

Will the accuracy of the measurement equipment be of influence on the width of the skirt ?

All digital scopes I know aren't nearly accurate enough in their timebase as well as trigger point. small amounts of hum will also shift trigger points.
This rules out scopes for me.
Therefore the question is can the AP do this ?
If it can why is this not a standard measurement in this case ?
Does one measure jitter of the source (AP most likely) or the DAC itself or the AP trigger capabilities and how to tell them apart in that case ?

 

[amirm]

Does this mean that frequencies actually shift (due to jitter) as far as the bottom of the skirt ?

Yes. That is why I say they are "low frequency random jitter" because I can see the excursions left and right there.

 

[amirm]

Will the accuracy of the measurement equipment be of influence on the width of the skirt ?

It can of course. If the measurement noise floor is too high, then you won't see the skirt. The noise floor would be above it. With longer length FFTs we can overcome this to some extent.

 

[amirm]

All digital scopes I know aren't nearly accurate enough in their timebase as well as trigger point. small amounts of hum will also shift trigger points.
This rules out scopes for me.
Therefore the question is can the AP do this ?

Correct. You need a high-end scope with very accurate timebase to do time domain analysis. DonH56 is our expert in this domain.

With AP and other analyzers we take advantage of their very high dynamic range (24 bit ADCs) and FFT transform to show very, very low levels of jitter. This by definition forces us to show such in frequency domain.

AP can also make measurements in time domain but it is not as precise there. It can create a histogram also but I have only played with it once.

 

[amirm]

Does one measure jitter of the source (AP most likely) or the DAC itself or the AP trigger capabilities and how to tell them apart in that case ?

As long as the noise floor is not too high, then the analyzer is not an impediment here.

 

[trl]

Sorry... WOT following, techies please read bottom part !
[...]

I don't know Amir at all nor his connections so have no opinion about this but all I see is Amir does not like poorly measuring stuff.
I do hope he never measures my designs (Bob did) as they won't be recommended for sure hurting the business of a good friend (not mine, not into business myself)
[...]

Hello Frans,

That would be a nice challenge, although C.H.A.M.P. or Polaris have been already measured in the past by so many that I'm 125% sure that on Amir's desk will measure perfectly as well, definitely more than 18dB of "Amir bits".

Best,
Raul.

 

[solderdude]

Anybody know how much deviation in frequency (percentage) is audible when the change is sudden ?
Remember reading it somewhere but can't find it.

Realizing a piano tuner or musician may be more sensitive than your average HiFi buff although the first one mostly makes use of specific techniques involving 'floating' of 2 tones where one of them is reference.

P.S. Thanks Raul, totally forgot about C.H.Amp ... old age I guess but is not a real product in that sense.

 

[DonH56]

Jitter threads:

Part 1: https://audiosciencereview.com/forum/index.php?threads/digital-audio-jitter-fundamentals.1922/
Part 2: https://audiosciencereview.com/foru...igital-audio-jitter-fundamentals-part-2.1926/

There are precursors on data converters, and other threads on jitter, but these two will provide a decent overview, I hope.

 

[solderdude]

Thanks for the links John,

These threads didn't answer the question about the audibility of a frequency changing suddenly and what is considered audible by the most discerning folks.
I know I read it somewhere but could not find it.
There is a threshold where people can detect a difference in frequency in a single tone and or relative to a second tone. Research has been done about this subject.
My thought was to compare this minimum detectable difference and compare that to the percentage of skirting bandwidth.
Just for the fun of it.
I have not heard 'jitter' in any of the products I own and have tried.
Just curious about the numbers really.

Was hoping someone knew about this perception aspect or knows where to find the paper I saw on it.

 

[Frank Dernie]

I have also wondered about the level of jitter needed to be audible.
I note that a particular brand of very expensive CD players and DACs consistently show very high jitter and wonder if they fall into the category of "they sound different and are very expensive so they must be better" class of high end kit?
https://www.stereophile.com/content/metronome-cd8-s-integrated-cd-player-measurements
HiFi News have tested others of their products and their most expensive DAC, £37,000, had jitter of 1205 psec, about an order of magnitude more than most, and two orders of magnitude worse than good ones like the (still expensive) Chord Dave they tested.

 

[Soniclife]

These threads didn't answer the question about the audibility of a frequency changing suddenly and what is considered audible by the most discerning folks.

As a starter Wow and Flutter.

 

[solderdude]

a bit off forum but:
Sorry for the incorrect name Luis emm... Gilberto but thanks for your reply and agree.
Luis is also a great guy b.t.w.

b.t.w. there is no 'shit list' just people who's opinions I don't agree with in many (or some) cases.
Have no ill feelings at all, a bit of frowning now and then.

 

[solderdude]

As a starter Wow and Flutter.

from Wiki but no ref. to research alas:

Wow and flutter are particularly audible on music with oboe, string, guitar, flute, brass, or piano solo playing. .... flutter can alter the sound of the music differently, making it sound ‘cracked’ or ‘ugly’. There is an interesting reason for this. A recorded 1 kHz tone with a small amount of flutter (around 0.1%) can sound fine in a ‘dead’ listening room, but in a reverberant room constant fluctuations will often be clearly heard.[citation needed] These are the result of the current tone ‘beating’ with its echo, which since it originated slightly earlier, has a slightly different pitch. What is heard is quite pronounced amplitude variation, which the ear is very sensitive to. This probably explains why piano notes sound ‘cracked’. Because they start loud and then gradually tail off, piano notes leave an echo that can be as loud as the dying note that it beats with, resulting in a level that varies from complete cancellation to double-amplitude at a rate of a few Hz: instead of a smoothly dying note we hear a heavily modulated one. Oboe notes may be particularly affected because of their harmonic structure. Another way that flutter manifests is as a truncation of reverb tails. This may be due to the persistence of memory with regard to spatial location based on early reflections and comparison of Doppler effects over time. The auditory system may become distracted by pitch shifts in the reverberation of a signal that should be of fixed and solid pitch


Interestingly this seems to be more audible with speakers due to reflections. A conclusion that could be drawn (if it turns out to be true) is that with headphone listening (scrape) flutter (and thus also jitter perhaps) would be less audible ?
and further below:
Anyone with a good musical ear can detect a pitch change of around 1%

Also not backed up with references. 1% variation means skirting should be within 1% of the test frequency so when using 11kHz so within 10890Hz and 11110 Hz which appears to be the case here.

 

[Wombat]

Noticeable wow and flutter reminds me of the feeling of 'wooziness' wrt balance. e.g. sea-leg syndrome(physical balance confusion).

 

[solderdude]

Wow .... indeed

 

[svart-hvitt]

I have also wondered about the level of jitter needed to be audible.
I note that a particular brand of very expensive CD players and DACs consistently show very high jitter and wonder if they fall into the category of "they sound different and are very expensive so they must be better" class of high end kit?
https://www.stereophile.com/content/metronome-cd8-s-integrated-cd-player-measurements
HiFi News have tested others of their products and their most expensive DAC, £37,000, had jitter of 1205 psec, about an order of magnitude more than most, and two orders of magnitude worse than good ones like the (still expensive) Chord Dave they tested.

OT, but couldn’t resist, sorry...

JA’s first sentence in the conclusion goes like this:

«The Metronome CD8 S is a beautiful-looking audio componeny».

I suggest @amirm follows the example and writes more about the looks.

 

[andreasmaaan]

That 1% figure is pretty arbitrary IMO, as there is a huge variation among individuals when it comes to pitch acuity. This ranges from complete inability to perceive pitch (very rare and actually considered a pathology) to ability to discriminate to a couple of Hz right up into the high frequencies. Most of us are somewhere in the middle but even in this middle band the range is wide.

People with absolute pitch (rare in the West but more common in countries with tonal languages) often liken their perception of pitch to colour perception. If the pitch is off, it's as obvious to them as looking at grass that is blue would be to us.

 

[derp1n]

Schiit's testing of this unit are linked from head-fi now.