Are Measurements of Schiit Yggdrasil DAC Inconsistent?

[garbulky]

I saw the results and he said it was 7/7 or 8/8 or some such thing. Fascinating thing is that when asked at our local AES meetings how audible these distortions are at such low levels he said that one had to play low level signals and then boost them by tens of dBs to be audible. Perhaps that is how he conducted his tests.

I am happy to test him and am confident he won't be able to achieve any results like he is stating.

That sounds like a high number he got right. Are you impyling that he is lying or incorrectly performing a DBT? I'm unaware of the DBT results so I don't know what he was testing for. It seems unlikely he would be playing low levels and boosting them tens of dbs to be audible. I don't see a reason to do so unless testing whether somebody can hear those distortions when doing it in such a manner. (I don't see what relevance it has though to real world usage though. You would be purposefully exagerrating the noise floor which you can use to make any dac sound bad.

 

[amirm]

That sounds like a high number he got right. Are you accusing him of lying or incorrectly performing a DBT?

I found his post and this is what he says:

So it wasn't the Yggdrasil but still, it is hard to fathom him getting 8/8 right in 7 trials. Shame the display in the ABX test above doesn't demonstrate one of those successes. Or the music used and what was heard that led to the detection.

His references are quite suspect. Harman's how to listen is all about identifying frequency response errors. Nothing remotely there will help you identify issues here unless the two products have significant frequency response errors. In which case you don't need training because they tend to be quite audible to everyone.

Likewise, Dave Moulton's training from what I have read is going to help one to hear the types of distortions we find in DACs.

From this bit, it seems to be hand waving and throwing terms out there that he knows listeners don't know and hoping to score points.

 

[amirm]

BTW, note the part about equal cable lengths for AC and signal. If he can tell those differences, he will forever change the nature of what AES and audio science believes. All of this leads to me to question how genuine his claims are in this regard.

 

[rebbiputzmaker]

This thread is about those measurements, why are we having a debate about neuroscience ?

Because garlic bread tastes good.

 

[flipflop]

Harman's how to listen is all about identifying frequency response errors.

You're wrong. It's also about identifying channel mismatch (left/right & front/rear), distortion, reverberation, and humming.

 

[amirm]

You're wrong. It's also about identifying channel mismatch (left/right & front/rear), distortion, reverberation, and humming.

That must be a later addition as I took the training at the time it was released. And also met with Sean Olive where he presented the tests to us and it was exactly as I stated.

Regardless, those additions do not in any way help you figure out fidelity differences between DACs. DAC differences is a test of small differences. Speaker testing is a test of large. Two different things that require very different training.

 

[garbulky]

Well I still don't know what he was testing for. Was it preference? Was it identifying if there is a difference? whatever it is, he was able to differentiate it 8/8 or 7/7. It's a small sample but the odds of him getting it correct are hard to put down to random guessing. But also that was the Gungnir MB. Not the Yggy. There's significant design differences between the two. And we don't have a Gungnir MB measurement.

 

[Randem Tox]

Regardless, those additions do not in any way help you figure out fidelity differences between DACs.

I believe the reference to those tests is in the context of being able to correctly identify which DAC was which/that a difference exists and NOT in making a determination as to which was higher fidelity. Those are two very different things.

 

[Grave]

You do not need magic ears and cables do not matter. He is obviously lying.

 

[amirm]

I believe the reference to those tests is in the context of being able to correctly identify which DAC was which/that a difference exists and NOT in making a determination as to which was higher fidelity. Those are two very different things.

My reference to "fidelity" was in regards to detecting differences. Not preference.

He on the other hand, seems to be presenting the results of ABX difference test as proof of fidelity differences in the way you mention.

 

[Randem Tox]

cables do not matter.

Oh, sure they do.

Not in the sense that choosing silk or cotton or air over PTFE for your dielectric, or silver vs. copper will result in audible changes in a headphone cable or interconnect or AC cable, but cables are specced the way they are for a reason.

Take a digital connection, like USB ... and take audio out of the equation even, so there's no need/room for subjectivity in evaluating the result.

Get the conductor lengths too out of whack, or the characteristic impedance of the data transmission line too far from 90 ohms, and you'll wind up with a cable that either can't get anywhere near it's rated speed or flat out doesn't work. And this effect is prominent enough that with some USB cables, bending (which will effect it's characteristic impedance) it will reduce its transmission speed.

In fact, high-end "audiophile" USB cables are notorious for being horrible when it comes to maintaining full bandwidth transmission of USB data (most noticeable when you use them with an external hard drive and do a large transfer ... an easy way to determine if they're fundamentally "to spec").

And try sending a 4K HDR signal down a 3+ meter cheapy HDMI cable that doesn't meet the spec for the HDMI standard you're trying to support (despite having all the same pins and connections) and see what happens. The specs for those cables change because they VERY MUCH matter.

And those are both examples in the digital domain, where you have ECC and ER/rTx.

There's also a reason why, say, studio and event cabling tends towards using star-quad internal wiring rather than simple twisted-pair or linear arrangements. Run that simple twisted pair cable over a stage PSU or a lighting pack supply/control and you'll be amazed at how much crap it picks up. Changing the cable geometry to star-quad, which is essentially making twisted-pair behave the way it should in an ideal world, is the difference between constant noise and interference, and none.

 

[Grave]

All not broken cables sound the same.

 

[Randem Tox]

All not broken cables sound the same.

That's a much more reasonable statement.

It's not the same thing as "cables do not matter".

And I only bring this up, because when you're arguing about signal artifacts > -90 dB from the fundamental, precision matters. Precision in measurements, analysis, and in the expression of that analysis.

Absolute statements, of an over-broadly general nature, are, by definition, imprecise and do not help establish credibility or defend a position that is, rightly or wrongly, under attack.

 

[Jimster480]

So how much do you handicap sigma-delta DACs to equivalency with the superior sounding inferior measuring multi-bits?

Why do the multitude of measurements atomicbob does if he then ignores what they are telling him?

Because hes paid in marketing dollars obviously.
People think that because hes providing graphs, that what he is saying is legit. And it backs up the marketing they are reading. This is psychology rather than sound science.

I think Schiit do a great job!
Let me clarify. They are a business and their success is measured by sales/profit. It think the idea was too concentrate all of their efforts into marketing knowing full well that sound quality is subjective and people are likely to follow the crowd and believe what they are told to be true. They even chose the company name "Schiit" to make their point! They may have even had a side bet regarding the success/failure of this project. So, well done Schiit!

Yes they have done a great job scamming everyone with inferior products (some of which are worse than your onboard sound card) by lieing to you about the "quality" you will experience when you downgrade your audio experience using one of their devices.

 

[Jimster480]

I'd really like to dump the jotunheim and move on. I'm searching for separate dac+amp for say $600-$800 total with xlr pre out (for my lsr 305:s), xlr headphone out for the amp, and preferably xlr out from dac to amp. Headphones I have at the moment are the HD 650's and I plan on purchasing HD 800:s in the future (s version ). Could anyone recommend a suitable combo? I've searched everywhere and am yet to find one

The DX7 does all of this and costs less than your budget. So why not just buy that consdiering the performance is nothing less than stellar?

 

[Jimster480]

@ garbulki and rebbiputzmaker

What makes you think that an ADC can make a (near) perfect recording (the ones you listen to) with all of the subtleties and details and when played back on 'musical gear' sounds wonderful but measurement gear, using even more accurate ADC's and analytic software, cannot capture and analyze it ?

You guys are aware that a stereo signal is just 2 voltages varying over time... right ?
variables: voltage and time. Which can both be measured over 1000x more accurate than any human can do.
The whole measurement suites cover this range way, way, waaaayyy beyond our rather limited hearing capabilities.

Are you perhaps suggesting there is something 'travelling along' a 'hidden signal', something illusive that can influence the sound and passes through certain gear (components, cables) better than other and is of yet unknown and thus not measurable ?
No measurement buff or lab has ever (accidentally) run into unexplained signals of behavior over all these years, yet the mighty 'measuring/analytical' ears can ?
Only ears and brains are capable of detecting this illusive signal ?
A signal that also can be converted to soundwaves by (substantially flawed) transducers being either electrostatic or magnetic based ?
The air can conduct these soundwaves + unknown signal and brain/body/ears can detect that what is not yet measurable ?

No,
But personally I think that the only possibilities here exist in how a DAC actually plays back music.
Let me explain what I mean by this, because its very common in the world of computers to over-optimize specific performance or "cheat" in certain respects.
So when you are playing music, there are many tones playing at the same time. These tones are obviously the music itself which is compromised of possibly 50 different tones at the same time, vs 1-2 tones at once. Therefore DAC vendors can indeed optimize single and dual-tone performance far beyond that of say 50 tone performance and without that multi-tone performance measured it leaves room for error.

In the past we have seen companies like Nvidia (and they have done it many times) optimize their graphics hardware (and sometimes drivers) specifically for tests that were done by industry recognized testers. In order to achieve artificial performance increases. The same has been done with Intel and specific testing software aswell as in their architectures where they optimized specific types of workloads not commonly found in everyday tasks but only found in these specific benchmarking tasks so that they could claim large generational performance increases; which in turn would be shown by the industry's reviewers who weren't smart enough to see through this. While some smaller vendors did alot of real world testing, only to find out that there was a 1% gain at best and sometimes there were negative "gains" at the same clockspeed.

This is the only place that I see fault in the world of audio testing, where there is room for improvement. Because THD, SNR, IMD measurements are all given a specific tone and not even the same standard tone per DAC vendor.

 

[Blumlein 88]

No,
But personally I think that the only possibilities here exist in how a DAC actually plays back music.
Let me explain what I mean by this, because its very common in the world of computers to over-optimize specific performance or "cheat" in certain respects.
So when you are playing music, there are many tones playing at the same time. These tones are obviously the music itself which is compromised of possibly 50 different tones at the same time, vs 1-2 tones at once. Therefore DAC vendors can indeed optimize single and dual-tone performance far beyond that of say 50 tone performance and without that multi-tone performance measured it leaves room for error.

In the past we have seen companies like Nvidia (and they have done it many times) optimize their graphics hardware (and sometimes drivers) specifically for tests that were done by industry recognized testers. In order to achieve artificial performance increases. The same has been done with Intel and specific testing software aswell as in their architectures where they optimized specific types of workloads not commonly found in everyday tasks but only found in these specific benchmarking tasks so that they could claim large generational performance increases; which in turn would be shown by the industry's reviewers who weren't smart enough to see through this. While some smaller vendors did alot of real world testing, only to find out that there was a 1% gain at best and sometimes there were negative "gains" at the same clockspeed.

This is the only place that I see fault in the world of audio testing, where there is room for improvement. Because THD, SNR, IMD measurements are all given a specific tone and not even the same standard tone per DAC vendor.

Multi-tone testing is not a big problem. Two tones is enough. If a DAC handles two tones fine it will handle many more fine, because each tone has to be lower in level as you add in more of them. Two tones will show non-linearities at higher levels than results in larger numbers of tones.

 

[Jimster480]

Multi-tone testing is not a big problem. Two tones is enough. If a DAC handles two tones fine it will handle many more fine, because each tone has to be lower in level as you add in more of them. Two tones will show non-linearities at higher levels than results in larger numbers of tones.

Well I would like to test a multi test tone anyway to see what it would be like.
You know something with like 5 tones or more just to see.

I spoke to amir about this before and he said that he was actually working on a way to test some sort of music via the analyzer for comparison.

 

[Blumlein 88]

Well I would like to test a multi test tone anyway to see what it would be like.
You know something with like 5 tones or more just to see.

I spoke to amir about this before and he said that he was actually working on a way to test some sort of music via the analyzer for comparison.

I don't know if I still have the results. Did some testing with up to 8 tones. It simply wasn't worth doing.

I've also put in filtered portions of high level noise leaving an octave empty here and there to see if anything shows up in the empty portions. Again not worth doing. Yes some poor performers show something, but you already knew that from the conventional testing.

 

[Jimster480]

I don't know if I still have the results. Did some testing with up to 8 tones. It simply wasn't worth doing.

I've also put in filtered portions of high level noise leaving an octave empty here and there to see if anything shows up in the empty portions. Again not worth doing. Yes some poor performers show something, but you already knew that from the conventional testing.

I see,
Well if anything doing these multi tone tests would help I think to turn the skeptics into believers. As I have friends who tell me that "analyzer's aren't your ears and music isn't just one tone".