Just got a Yggdrasil!

[Jud]

6 minutes ago, Samuel T Cogley said:

 

The problem with A+ is you can't configure it to only output 20 bits, right?  Or is there a way to do it that I'm missing?

 

 

But Yggy should round off the extra bits, right?

[semente]

4 minutes ago, Jud said:

 

But Yggy should round off the extra bits, right?

 

Wouldn't you rather have it truncate them for that all time audiophile favourite blackest of blacks background? 

[Silly goose]

Blows my mind people can be so materialistic that they resort to extreme fanboyism. Come on audio gear is aluminum, copper and plastic. Not flesh and blood. 

[Samuel T Cogley]

1 hour ago, Jud said:

 

But Yggy should round off the extra bits, right?

 

Now who's trolling? 

[wgscott]

5 hours ago, manisandher said:

 

Haha... There's a part of me that's going to miss him. But why have I got a certain Vera Lynn song on my mind right now? (Perhaps only the Brits will understand what I mean.)

 

Mani.

 

It was featured at the end of Dr. Strangelove.

 

 

 

[Ralf11]

Wow - does this win the CA most dramatic thread award?

 

also, can we use the term "ygg or yggy" as a curse?

 

as in "oh yeh, well Ygg You!"

[Speed Racer]

The Yggdrasil ALWAYS upsamples. And it ALWAYS rounds down to 20 bits just before the data is sent to the DAC chips.

[Jud]

49 minutes ago, Samuel T Cogley said:

 

Now who's trolling? 

 

I was actually completely serious, but I understand how it could have been taken as poking fun.

[Jud]

27 minutes ago, Speed Racer said:

The Yggdrasil ALWAYS upsamples. And it ALWAYS rounds down to 20 bits just before the data is sent to the DAC chips.

 

It does, but if you send it 176.4 it only does 2x upsampling; if you send it 44.1 it does 8x.  Does it sound any different?  Heck if I know.  Many DACs do 8x in three 2x rounds, but I don't know if Yggy is set up that way.

[mansr]

8 hours ago, Jud said:

Mike's thinking is interesting to me: it essentially boils down to a philosophy that PCM if done well will serve the music better than  a format conversion via delta-sigma modulation.  Entirely aside from how well one thinks Yggy does PCM, the philosophy of avoiding an extra format conversion is intriguing.  Of course you'll also find fervent disagreement here (regarding whether staying in PCM or delta-sigma modulation sounds better).

It's an intriguing philosophy, but one that doesn't pan out in reality. Sigma-delta was invented because direct conversion wasn't good enough.

[Speed Racer]

30 minutes ago, mansr said:

It's an intriguing philosophy, but one that doesn't pan out in reality. Sigma-delta was invented because direct conversion wasn't good enough.

 

Sources?

[mansr]

2 minutes ago, Speed Racer said:

Sources?

Spec sheets clearly show that sigma-delta converters are more accurate.

[Jud]

27 minutes ago, mansr said:

Spec sheets clearly show that sigma-delta converters are more accurate.

 

The specs are better (currently at least), but I think cost had a lot to do with it also.

[Speed Racer]

31 minutes ago, mansr said:

Spec sheets clearly show that sigma-delta converters are more accurate.

 

That doesn't mean they were "invented because direct conversion wasn't good enough". Sources, please.

[GUTB]

14 hours ago, Jud said:

 

What I read in JA's testing was his *assumption* regarding truncation (prior to Jason Stoddard's clarification on Head-Fi that rounding is used), and no statement from him about false manufacturer claims.  Do you know of any such statement by John Atkinson?

 

[Silly goose]

59 minutes ago, Speed Racer said:

 

That doesn't mean they were "invented because direct conversion wasn't good enough". Sources, please.

Why did we go from the steam engine to diesel for trains? Steam engines are great for nostalgia purposes, but seems like that evil old man technology caught up with things. 

 

 

This video explains a lot, but not everything:

 

 

[Jud]

1 hour ago, GUTB said:

 

 

Stronger than I'd seen before, but "My measurements suggest..." I would say is not quite conclusive.

[audiventory]

5 hours ago, mansr said:

Spec sheets clearly show that sigma-delta converters are more accurate.

 

Sigma-delta modulator is pure digital device, that have 1 important property noise floor in audible range or operation band. Precision here is noise level. General rule here: higher sample rate - lesser noise. because we have more band reserve for noise pushing out operation band.

 

Electrical precision begins in analog sigma-delta demodulator in DAC.

There need provide 2 levels only. There may be several cascades, as far as I know.

Contrary, in "resistor matrix" DAC need provide several levels. Some time ago at CA we discussed demanded precision of resistors. Unfortunatelly, I can't found it now.

That is difference.

Though precision of implementation depend on ... implementation

[manisandher]

5 hours ago, Jud said:

 

Stronger than I'd seen before, but "My measurements suggest..." I would say is not quite conclusive.

 

Well, here's his comment in full:

 

Quote

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

 

[Highlight mine.]

 

To me, that "suggest" implies (suggests?) that the data is actually conclusive, but that he's being a little 'polite'. Anyway, here's the measurement he's talking about:

 

 

To me, those spikes show the result of truncation, hence increased quantization distortion, but I'm out of my depth here. Can anyone here conclude definitively that this measurement shows truncation vs. rounding?

 

Mani.

[manisandher]

 

15 hours ago, manisandher said:

... why have I got a certain Vera Lynn song on my mind right now? (Perhaps only the Brits will understand what I mean.)

 

9 hours ago, wgscott said:

It was featured at the end of Dr. Strangelove.

 

Haha...

 

Mani.

 

[esldude]

7 hours ago, Silly goose said:

Why did we go from the steam engine to diesel for trains? Steam engines are great for nostalgia purposes, but seems like that evil old man technology caught up with things. 

 

 

This video explains a lot, but not everything:

 

 

Leaves out alot, but I believe I understand your point.

 

I once had a relative who was a huge historical train buff.  He made a living as historian.  I live near the place where all old vintage locomotives get restored and worked upon.  Through him I actually got to ride those old steam locomotives.  They gave him permission to take them out a few times.  An old steam locomotive of the last couple generations of such would pull no less than the same load as 8 modern diesel electrics would pull.  Yet for efficiency, cost and sizing to the load, the consistency of the diesel electrics, the fact you could link them together and have them work as one more than made up for the other factors. 

 

Not a bad comparison at all with sigma-delta DACs and the old multi-bits.  All things considered the non-oversampling multi-bits simply cannot compete.  When they can have a few advantages it comes at very high cost and having to be individually tuned. For peanuts any sigma delta of good design can match and usually exceed that performance. I don't know why the nostalgia for those NOS multi-bits. 

 

I can both admire the design acumen of the guys at Schiit, and feel quite comfortable they don't make the best DACs.  Glitches, no glitches, glitches that are inaudible, makes no difference.  When the competition has no glitches, and offers clearly measurably better performance your product doesn't cut the mustard. Anything else is nostalgia and hyperbole marketing to a niche that values faith over results.

[esldude]

1 hour ago, manisandher said:

 

Well, here's his comment in full:

 

 

[Highlight mine.]

 

To me, that "suggest" implies (suggests?) that the data is actually conclusive, but that he's being a little 'polite'. Anyway, here's the measurement he's talking about:

 

 

To me, those spikes show the result of truncation, hence increased quantization distortion, but I'm out of my depth here. Can anyone here conclude definitively that this measurement shows truncation vs. rounding?

 

Mani.

I think you will see such spikes with truncation or rounding.  The exact level of them would vary a bit.  Think about it. If you have 20 bits and a 24 bit signal what is the difference in rounding to 20 bits versus truncating?  Maybe some modulation of the last two bits vs everything chopped off at 20 bits.  Both are going to result in those spikes without dither.  In one case everything below 20 bits becomes a zero.  Quantization spikes result.  In the other it becomes a zero or one LSB above zero, and both are chopping off the lowest levels and causing those spaced spikes.  One is 20 bits and maybe one is 21 bit or 20 1/2 bits. DSP basics, 24 bit signals at 20 bit without dither is a broken design.  It might work better than expected, it might even not be audible or barely audible at that level.  It isn't the characteristic of an excellent DAC design and behaviour.  On the other hand, use it with redbook sources of only 16 bit and everything would be fine. We know such behaviour compromises Spurious Free Dynamic range. We already know it is Schiit.

[Silly goose]

12 minutes ago, esldude said:

Leaves out alot, but I believe I understand your point.

 

I once had a relative who was a huge historical train buff.  He made a living as historian.  I live near the place where all old vintage locomotives get restored and worked upon.  Through him I actually got to ride those old steam locomotives.  They gave him permission to take them out a few times.  An old steam locomotive of the last couple generations of such would pull no less than the same load as 8 modern diesel electrics would pull.  Yet for efficiency, cost and sizing to the load, the consistency of the diesel electrics, the fact you could link them together and have them work as one more than made up for the other factors. 

 

Not a bad comparison at all with sigma-delta DACs and the old multi-bits.  All things considered the non-oversampling multi-bits simply cannot compete.  When they can have a few advantages it comes at very high cost and having to be individually tuned. For peanuts any sigma delta of good design can match and usually exceed that performance. I don't know why the nostalgia for those NOS multi-bits. 

 

I can both admire the design acumen of the guys at Schiit, and feel quite comfortable they don't make the best DACs.  Glitches, no glitches, glitches that are inaudible, makes no difference.  When the competition has no glitches, and offers clearly measurably better performance your product doesn't cut the mustard. Anything else is nostalgia and hyperbole marketing to a niche that values faith over results.

Yes nostalgia vs innovation can be a deadly internal battle. 

[manisandher]

23 minutes ago, esldude said:

I think you will see such spikes with truncation or rounding.  The exact level of them would vary a bit.  Think about it. If you have 20 bits and a 24 bit signal what is the difference in rounding to 20 bits versus truncating?  Maybe some modulation of the last two bits vs everything chopped off at 20 bits.  Both are going to result in those spikes without dither.  In one case everything below 20 bits becomes a zero.  Quantization spikes result.  In the other it becomes a zero or one LSB above zero, and both are chopping off the lowest levels and causing those spaced spikes.  One is 20 bits and maybe one is 21 bit or 20 1/2 bits. DSP basics, 24 bit signals at 20 bit without dither is a broken design.  It might work better than expected, it might even not be audible or barely audible at that level.  It isn't the characteristic of an excellent DAC design and behaviour.  On the other hand, use it with redbook sources of only 16 bit and everything would be fine. We know such behaviour compromises Spurious Free Dynamic range. We already know it is Schiit.

 

Perhaps I should have posted this also - JA's commentry in the review itself:

 

 

... 

while the noise floor lies at or below –160dBFS, a regular series of distortion components can be seen, in which the third, fifth, seventh, or ninth harmonics are highest in level. This will be due to the Yggdrasil's use of 20-bit D/A converters; the bottom four bits with 24-bit data will be truncated.

[Highlight mine.]

 

Mani.

[Speed Racer]

49 minutes ago, esldude said:

I can both admire the design acumen of the guys at Schiit, and feel quite comfortable they don't make the best DACs.  Glitches, no glitches, glitches that are inaudible, makes no difference.  When the competition has no glitches, and offers clearly measurably better performance your product doesn't cut the mustard. Anything else is nostalgia and hyperbole marketing to a niche that values faith over results.

 

Wow, this has turned into a complete schiit on Schiit thread.

 

I haven't listened to a lot of DACs, but I like my Schiit Yggdrasil more than one friend's Auralic Vega or another friend's Chord Hugo.