Schiit Yggdrasil -- 21 bit?


Schiit says that Yggdrasil is a 21 bit DAC. But the DAC chips that they put in the device ( Analog Devices AD5791BRUZ, 2 per channel) are 20 bit with the error of plus-minus 0.5 LSB.

How can the DAC be 21 bit if the chips are 20 bit? Using two chips per channel does reduce the RMS voltage of the noise by  a square root of 2. But how can you get to 21 bit from there?

Can someone please explain.
defiantboomerang
Yeah Robert Harley is with Absolute Sound he's the reviewer that got excited about the Yaggy.
.
At least Stereophile try to backup their reviews with measurements and you can draw some kind of parallel between review and measurements (if you can understand them) most readers can’t, so they need to learn.

Where Absolute Sound is more like "audio poetry" wafting in the wind, but granted of late they are bending and have had "some" independent measurements done on just a little of what they review, would be good to see more of it.

Cheers George
@bacobits1 
@falconquest 
@georgehifi 

Ok, I see now. I couldn’t care less about Harley’s views and writings. Irrelevant stuff.  
Yeah, the publication industry and a good deal of the industry have become so oddly slanted, its become the true impediment to high fidelity.

One example, and this has vastly greater importance than this bit depth discussion, is the  idiotic effusive behavior of the publishing community in regards to high resolution recordings and the primary driver of why high bit depth digital converters are sold for.

Lets take a modern multi-track recording that might be released and adervertised as high resolution, but in truth hardly qualifies. I run analysis of the files and find a single instrument being recorded above the standard RB 16/44, with the remainder showing no a single aspect of a high resolution recording. So out of this entire multi-track production, in instrument or sample was in HiRes. It gets mixed together, with all  the other tracks going through and up sampling and sold as high resolution recording. At best, is RB+ or 16/44+. In every rag, each and every month, every reviewer should be checking and reporting and being critical about the state of recorded music. All this gear and too many piss poor recordings.

Also, today we do have gear that can exceed the modern Red Book recording. But keep in mind that the vast majority of speakers and amplifiers couldn't cope on 20 bit recording at scale. Some can do over 19 bits worth, but it won't be composed doing so. Amps and pre are also under this limitation.

Now as for that Benchmark DAC3 HGC review, those are some excellent bench test results. It seems that ESS had improved upon some of their filtering as well as some other tweaks being added. While I could pick on some level of improvements from lesser DAC devices my own, or even something like this, at this level the differences between other high performance devices that measure properly are hardly audible. An engineer can modify the final output to their preferences, but that's a designer coloration. But my hats to Benchmark, this is an engineering marvel!

Amusing to read various reviews on the DAC with most reviewers stating they couldn't discern a difference between it and other competitors at near or well above its cost. Big old surprise right there!  
It's really striking when audiophiles ask engineers to tell them how good something sounds. I'd love Stereophile to put on their bench a Guarneri 
vs a Stradivarius. My next concert will be incomplete without precise measurements.
I'd love Stereophile to put on their bench a Guarneri
vs a Stradivarius.
Probably could, if they plugged in.

Cheers George  
@mmeysarosh

Good points about performance measurements. I was surprised about the DAC3. I didn’t expect to hear a difference but I believe I could. It wasnt a difference in sound but more akin to better blacks on a TV - the sound popped out of the background in more detail. Not sure I could detect this in blind A and B tests - definitely subtle. I believe the performance in many of the latest DACs is on a similar excellent par. My speakers have THD distortion below -65 dB across the spectrum and also can play cleanly up to about 120 dB SPL which helps to discern such small improvements at low signal levels.


@defiantboomerang


ATC EL150ASL which is the more aesthetically pleasing elliptical version of the much more affordable ATC SCM 150 ASL big box studio monitor. ATC have been successful at selling pro studio main monitors for over 30 years. These are like a microscope on the recording and not something most people would be seeking for home listening.

Client list

http://atcloudspeakers.co.uk/client-list/

my setup

https://systems.audiogon.com/systems/6257

@shadorne 

The ATC's can be highly dynamic, but go ahead and call ATC if you like, they publish the maximum SPL is 117db at 1m, and at typical listening distances will be around 110db or less. Also note, the distortion figures aren't provided but I would wager it has peak distortion levels of around 10%. This is not to say across the audio band, just simply the highest level of distortion at any point across the audio band. The vast majority of speakers produce this at 20-250hz, but many exhibit audible distortion at 95db average levels in the mid or treble ranges. There are very few speakers that can achieve this kind of dynamic range, but bass drivers are typically massive to keep distortion under control with mid and tweeters often being horn loaded or in an array configuration to withstand the load.

The ATC design will produce a notable amount of distortion in the bass if driven, but being a doped paper driver and our lack of sensitivity in that audible range makes it somewhat negligible. Doped paper, while not really capable or producing a rigid diaphragm, is quite good with internal dampening and generally makes the system work well. Its lightweight makes cone movements easier to achieve with a given motor assembly. Those soft domes used in the large dome mid and treble drivers also follow the same ethos. Light and well damped, which is pretty reliable approach.
@defiantboomerang

The speakers are active - 6 poweramps one for each transducer so I don’t need a power amp for this model.

@mmeysarosh

I believe ATC claim that a single ATC 300 cabinet will deliver an astonishing 121 dB SPL at less than .3% THD (with 10dB of headroom). The 300 cabinet is the same build as mine except for two 15" woofers per speaker instead of one (300 litre volume cabinet instead of 150 litre). So 117 dB SPL for one of my speakers at 1 meter sounds about right.

... please direct further questions on my system thread as I don’t think it appropriate to continue to discuss ATC speakers further on a thread that is specifically about Schiit’s excellent Yggdrasil DAC. I prefer not to hijack this thread.
Ahh that would be Robert Harley of TAS, you can read here:
http://www.theabsolutesound.com/articles/schiit-audio-yggdrasil-dac/

Build of a Yggi here
https://www.youtube.com/watch?v=R3YZmaIRb6Q

Video review here:
https://www.youtube.com/watch?v=AFvUyK1oQRQ

Really interesting discussion with Mike Moffat and Jason Stoddard:
https://vimeo.com/194658219

I love the Yggi, its amazing for the price.

Brett

@shadorne Nice ATCs!  I've never seen those, but being familiar with ATC, I can guess they sound pretty amazing from an analog front end!
@shadorne 

The specification in itself is honestly rather vague, which is another gripe the entire industry as measurements must be standardized to permit consumers to compare. Far too many brands take liberties with specifications. Sure, it could be set out to lessen consumer confusion when providing only what is deemed necessary, but when a specification is incomplete such as the one listed on ATC's own website, its best to leave it out instead. 
Back to Yggdrasil, my worst fears as posted above are true.

I found this statement in Stereophile

"When you have 24-bit data but 20-bit DACs, you need to dither those data to match the DAC. Otherwise, simply chopping off the 4 LSBs, called "truncation," reintroduces quantizing distortion. Schiit’s Jason Stoddard has subsequently said that the Yggdrasil "rounds" 24-bit data but my measurements suggest that the LSBs of 24-bit data are simply truncated.

John Atkinson
Editor, Stereophile"

This is a disaster. Every studio mastering engineer will carefully dither high resolution 24 bit data when converting to 16 bit lower (bit depth). This is in order to reduce quantization noise! This is not optional - This is industry best practice! If JA is to be believed Schiit do NOT understand digital signal processing at all - they just truncate the data!!! They may be geniuses at headphone amplifiers and analog circuits but this product is a failure if you are interested in high fidelity.

Be warned - this DAC should only be used with files that are 20 bit or less. If you use a PC for playback to this DAC then I would recommend doing a proper dithered conversion of 24 bit source files to 20 bit on the PC prior to sending to this DAC as it will sound much better. Programs like Audacity can do this conversion for you.
@shadorne and @defiantboomerang

But JA also states "Now you could say that you prefer the sound of truncation against that of redithering and I have no argument with that opinion."

Not to mention that JA’s measurement "suggest" that the LSBs of 24-bit data are simply truncated. So he is acknowledging he is not absolutely certain with his assertion/measurement understanding.

JA is super duper diplomatic or euphemistic as always.

Here is a five minute youtube that explains why dither is so important in digital audio.

https://m.youtube.com/watch?v=zWpWIQw7HWU
JA is wrong, Schiit uses rounding, not truncation.
JA believes that MQA (lossy, proprietary, licensed platform)
is " the birth of a new world".
The "extra" 1 bit in Schiit balanced multibit Dacs is deducted from the 2 phases +6dB.
If you watch the video you will see that rounding is practically the same as truncation and does similar damage.

Sorry but Schiit don’t understand what they are doing. This is common when good designers with a strong background in analog design start making digital components.
After a failed appeal to authority now  comes a referral to youtube and arguing ad hominem. Scores of other logical fallacies still at hand.
Rounding and truncation are similar, but with rounding being a broader function over truncation in this use. The idea behind rounding in the case of the Yiggy was that either method would increase quant noise, rounding would result a potentially less quant error and they decided it was low enough in level not to address it in by using a dithering filter. Adding dither would have added cost and complexity to the design and did so knowingly it would not measure in an ideal manner. In the test, it had shown that the rounded method used produced the exact result as truncating. No surprise in that part as rounding can produce the exact same figure. Rounding will in some cases, produce a slightly different result over truncation and have an very minor difference in quant error in comparison to truncation.

In large, this won't present much of an issue to most of recorded music. To the ones that it may have, will be quite low in level.
Mike Moffat knows a thing or two about DACs  since his theta digital dac in 1988  and MoFi Gain system. Introducing dither to Yggdrasil is a software change. He chose not to do so because with the 20 bit chips there is no significant difference between rounding and dither. Last but not least  let's not forget that there are hundreds of milions of digital tracks in 16 bit in distribution and only thousands in higher resolution. 
Shadorne’s technical points and the statements in the video he linked to all make sense to me, at least theoretically, as does the response by Mmeysarosh.

FWIW, though, I took a close look at Figures 6 and 7 in JA’s measurements of the Yggdrasil, which show the responses to undithered 16 bit and 24 bit signals at very low levels of 90.31 db below full scale. The bottom line is that the response to 24 bits looks vastly better than the response to 16 bits, and looks quite good aside from glitches occurring at the zero-crossings. And for that matter a comparison of the 24 bit response shown in Figure 7 with the corresponding Figure 12 for the comparably priced Mytek Brooklyn DAC (and its 32 bit converter!) shows the Yggy looking at least as good and probably better, aside from the zero-crossing glitches. Even though JA described that output of the Mytek as being "a well-formed sinewave," and its overall measured performance as being "superb."

And regarding the zero-crossing glitches, those likely reflect what we’ve previously presumed to be the switchover between the two DAC chips that are used on each channel, that would occur at the zero-crossing, and they would therefore be unrelated to truncation error or rounding error. And given that the glitches are more than 100 db below full scale and are very short in duration they would seem likely to be audibly insignificant.

FWIW. Regards,
-- Al

The video from Nigel Redmon is all clear and right on truncating /rounding, especially with the examples of going from 24 bits to 16. To get his whole message it’s worth going to his site and read about dithering. When and where to use it to make recordings sound better. Surprisingly he advises against dithering at many stages of production.
@almarg



As always I agree with pretty much all your comments.

I would add that the 24 bit will always look much better because at the same -90dbfs you have more bits to characterize the signal. The 16 bit looks square because there are only 2 bits to describe the signal. Even with a 20 bit DAC there are 6 bits available to characterize the sinewave at 24bit (so it looks smooth).

The zero crossing errors are the most alarming in those particular measurements.

"Audibly insignificant" depends so much on the frequencies. Low level noise distributed evenly and randomly across the spectrum is harmless. Rounding or Truncation can lead to high frequency noise of distinct tones which is harmful even at low levels which is why studios go to lengths to dither digital when reducing bit depth.

JA notes the jitter performance is less than desirable also. There has been much importance given to jitter even though it is very very low level noise - and this is because non-random jitter creates distinct tones that are not harmonically related to the music. 

What is actually audible as distortion is very much related to tone and frequency rather than purely a signal level.
@almarg 

I'm very appreciative that you would chime in on this thread. I don't want to speak or write on the OPs behalf, but he too should be.

All -

I'll up-the-anti. I'd like to suggest for the moment that you forget entirely about measurements. I don't care if its voodoo. The Yggy sounds spectacular. Please read my last post in the New Yggdrasil - First (and second) Impressions thread.
I would hazard a guess that for several reasons (only 20 bit, R2R and no upsampling), by necessity the filtering in this DAC will be close to Nyquist and therefore quite aggressive or sharp (like Non upsampling CD players). I suspect it is this smoother sound from filtered highs and the fantastic analog section that makes this DAC so musical and desirable to some folks over competing ESS 9018 and 9028 style DACs.

Measurements only validate accuracy or high fidelity capability of a device. The musicality aspects are all in the ear of the beholder.
@shadorne 

You are correct that measurements only validate accuracy or high fidelity capability of a device. The musicality aspects are all in the ear of the beholder.

In my case, there is something else noteworthy.

I already have a "virtual" Benchmark DAC 3. I call it an "Oppo UDP-205". It, has a MEASURABLY BETTER AND MORE ADVANCED DAC chip than the Benchmark.

Some folks might now write back and cry foul; reminding me that the DAC chip itself represents only 20 percent of the story, and that the implementation of said chip, analog stage and grace of God comprises the remaining 80.

I will then write back and point out that because the Oppo uses a Sabre ESS9038 instead of an ESS9028, that my 20 percent should really equate to 30 percent.

And away we go.

@gdhal 
 
The Oppo UDP 205 is indeed outstanding and I expect that it sounds a bit brighter and thinner than the Yggy. It may well outperform the DAC3 as you claim. Perhaps the Oppo UDP 205 will be another Class A or A+ Stereophile recommended player. Per JA above, we can easily surmise why the Yggy is not.
@gdhal 

I'll up-the-anti. I'd like to suggest for the moment that you forget entirely about measurements. I don't care if its voodoo. The Yggy sounds spectacular. Please read my last post in the New Yggdrasil - First (and second) Impressions thread.

With respect, bro -- this thread started as a simple technical question and we have more or less arrived at an answer. Let's try to keep it technical.

FYI -- I have asked the manufacturer AGAIN about the 21-20 bit issue. Let's hope that this time they will answer.

I am very happy that you are satisfied with your Yggy. I don't share you enthusiasm and will never buy anything made by Schiit. But despite our differing shopping preferences, we can still have a reasoned technical debate about the 21 vs 20 bit question -- as long as we stay technical.
@defiantboomerang

Apologies.

I sincerely thought a wide degree of latitude would be acceptable, because your post (thread title) is specific to Schiit Yggdrasil, and it has been compared within this thread by others - directly and/or indirectly - to the Benchmark DAC 3. So, by extension, I thought - again, apologies because perhaps my thinking is incorrect - that since the DAC 3 uses a Sabre ESS 9028 chip, it would be reasonable to submit the Oppo UDP-205 into the discussion, because it too uses an Sabre ESS chip, albeit a 9038.

Please let us know what you ultimately decide to "shop" for.
My best guess to the 21 bit question is the utilization of the DAC chips in a balanced pair per channel configuration. Not that in any way changes the chips ability to process the data, but possibly lowers its noise floor to a degree in which they decided to market it with that level of performance, but not purely technical aspect of capability.

From what I have read about the AD5791 in use, its an extremely linear performing R2R solution accomplished by using thin film resistor matching, then placed in pairs to provide its performance in a compact and simple to implement package. It also exhibits little drift in variable temperature ranges. Other R2R products that have been shown in the market are utilizing FPGA processors running code that corrects the linearity issues of ladder dac in use. So as long as the linearity drift remains within a range, they can identify the error and correct for it digitally before it reaches the next stage. Obviously a lot larger and may not be as stable as the AD5791 in stability. 
@gdhal 

you make a very good case for the OPPO. Looks like a machine with tons of potential. 
Udishamir,

"When connected with the Lynx sound card, then everything "opens up" started to hear new elements (especially in electronic music, soundtracks)."

Hello, I was wondering if the Lynx card made any difference in the sound of Yggdrasil that you may have noticed or was it more so noticeable only on Soekris? 
Just curious being an yggy owner... 

I just found this post/quote from Mike Moffat and it appears to directly answer the OPs question.

https://www.head-fi.org/threads/schiit-yggdrasil-impressions-thread.766347/page-123#post-12342937

"Two twenty bit DACs (one per phase) double the resolution for a balanced signal. Double the resolution only adds 6db (one more bit) for a total of 21 bits. In this case, 20+20 equal 21."



Just as a side note the AD5791 is close to $100 and the ESS’s around 1/10th of that.
Looks like the AD5791 took over where the PCM1704 left off.
PS Loved the video interview with Moffat and Stoddard.
Cheers George
"Two twenty bit DACs (one per phase) double the resolution for a balanced signal. Double the resolution only adds 6db (one more bit) for a total of 21 bits. In this case, 20+20 equal 21."
Actually, the first sentence isn’t correct. Or at least it is somewhat misleading, depending on what comparison is being implied by the use of the words "for a balanced signal."

As Shadorne correctly indicated earlier, simply using one DAC chip to create one phase (i.e., polarity) of the analog signal and another DAC chip to create the opposite polarity will not increase the resolution beyond what is provided by each of the chips. While it will double the full scale (maximum) output voltage compared to single-ended operation of one DAC chip, it will also double the voltage corresponding to the least significant of the 20 bits. So the number of discrete steps (i.e., possible values) within that doubled voltage range will still be 2^20 (two raised to the 20th power), meaning that resolution will still be 20 bits.

As I described earlier, though, if one DAC chip is used to handle the positive half of the output voltage that is ultimately generated, and the other DAC chip is used to handle the negative half, the resolution would indeed become 21 bits.

Regards,
-- Al

Sorry, my error, Correction:  Robert Harley, editor in chief of the Absolute Sound magazine. He also wrote The Complete Guide to High-End Audio.

https://www.amazon.com/Complete-Guide-High-End-Audio/dp/0978649362
+1 Almarg - I agree that the statement in the link is rather ambiguous. It isnt clear if it really is greater resolution from the greater bit depth of an additional bit or just a higher signal per bit but still 20 bits...

Summing 20 bits + 20 bits is still 20 bit levels as it is simply double the voltage value per bit compared to a single 20 but DAC chip.
2^20 = 1048576
2* (2^20) = 2* 1048576
2097152 = 2097152
2^21 = 2097152