Thoughts on a bunch of DACs (and why delta-sigma kinda sucks, just to get you to think about stuff)

[fzman]

   
Recommendations are generally assumed positive as a genre. That's why it's a "Recommendation" and not an "External Review of Candidate's Potential." 
 
Reviews are not assumed positive as a genre, except - apparently - in high end audio. 

No analogy is perfect-- only high end audio components can be perfect (Oh, and Messiah's too, I suppose) 

 
My point was regarding whether the content is "honest" and who is solicited to write them.  As to your main point-does it make sense to use the phrase 'negative recommendation'?
 
In addition, it does seem like many people uses reviews as a form of buying recommendations- not just as fact-sources.  Your point is a good one, and I think we should go back to dacs-- pm is ok though

 

[prot]

Not to veer to far from the discussion of dacs - How many people have sought letters of recommendation, or used as "job references" people who would give really honest reviews of their abilities and performance- and not just say good things about them........   Thought so.  Bad reviews do not get written - if reviews are written by people who are blatantly lying or who write stuff because they have been bribed-- different story.

Buy things from vendors who allow returns - judge for yourself - problem largely solved.

Or not. The try&return option surely helps but how many components can you test? I cant even test one per month. So no, that option does not solve much.

I do not agree with purrin all the time ... and I'm a fan of chocolate (although not in the chocolatey-sound sense) ... but I really appreciate his "that sounds like ****" rants. It's his opinion and should be same as proeminent as other's "sounds amazing" opinions. He even calls my dac (minimax) "cheap china ****". So what, I still like it and I think it's one of the best price/perf deals.

Generally, I do appreciate ppl who state their opinios directly and openly. The bullsh*t consensus that ppl (cough marketers) usually try to build along various components makes me puke. It's not true and it'll never be because there are millions of ears and tastes. And I dont need anyone to filter those for me and only show the nice stuff. That is a false, artificially sweetend picture and personally I dont want it. I guess it's the kind of pocture that Carlin had in mind when he spoke about the "*****fication of the western man".
I'm a grown up. I can take an honest opinion. Give it to me and keep your sweeteners.
/rant

 

[arnaud]

   
Indeed, it is rather amazing! Yes, based on my own experience with so many DACs, examining their designs, I can almost get a sense of how something will sound.**
 
I'm more than inclined to believe that the TOTALDuc will have great tonal, textural, liquid qualities with no delta-sigma treble hash. However, I have always, always, been suspicious of the resolving capabilities of R2R DACs. The fact is, delta-sigma designs do on average (for audio applications) have better accuracy than R2R designs, especially the discrete or magical black-box solutions. Cheap, small, and resolving were always the advantages of delta-sigma and why delta-sigma was permeated 99.999% DACs sold today.
 
...
I mean, why would I want to spend $10K and then down convert to 13/14-bits every CDs that I have?
...
 
P.S.
Another thing regarding the blathering from Schiit: I like to trust but verify. This is why I went out to get Theta Gen V (I suspected same / similar DSP) and audition a four dozen other DACs, a dozen of them in them in the mult-thousand dollar range. I wanted to find out if Mike Moffatt and Jason Stoddard were full of **** of the accuracy thing, missiles, close-form filters, some of their opinions of certain chips of yesteryear, etc.

 
- Thank you for taking the time to write such detailed explanation (you really are one of the most dedicated out there, there's no question about it!)
- However, I think the comment from @Clemmaster below is much more meaningful than simply looking at bit depth.
- I personally don't have kemar head ears and certainly can't hear a fly fart in the middle of a fortissimo which I think is the argument we're having here.
- For instance, your parallel between using a traditional ladder dac and converting a library to 13-14 bits is flawed: Decimating a whole song to 14 bit is basically throwing away a large part of the ladder array (so you'll loose all low level information). That is not what is happening in practice (the dac can effectively quantise down to 24 bits).
- You insist on the high bits precision error relative to the low bits but does that really matter when, by definition, high bits are active during peak times and there's only so much difference between loudest and quietest sound human can hear (level masking)?
- From my experience so far (i.e comparing the D1 resolving power to all other D/S dacs I have had / heard) tells me that, perhaps, this bit depth thing is a non-issue. 
 

  I'm not sure your ear can resolve a signal at -90dBFS when there's another signal 70dB higher in level at the same time, due to the combination of masking and ATH. In a sense, being able to resolve at both extreme of the DAC's scale, at the same time, is probably not necessary.
What if the totalDAC didn't use a linear scale? It would be able to both 1) have a great dynamic range (140dB, as reported by someone; couldn't find that on the site, at first glance), 2) have an excellent low level resolution (as shown on the website: the -100dB sinewave looks extremely clean), but not at the same time.

 
 

   
I think you will be very surprised at how much information is contained in redbook content. When I listened to the Yggdrasil, it had been after about two years of bugging Marv and demoing DACs in the 15-20k or less range. The Yggdrasil pulled information from familiar recordings that I had NEVER heard before the sense of realism was ridiculous.
 
I'm curious, what DACs have you compared to the TotalDAC? What gear and with what recordings?

 
- I already get surprised each and every day so I can't wait for the Iggy's revolution .
- You're asking the wrong guy about extensive gear comparisons, I have nowhere near enough time on my hands to ever consider doing that.
- I have been looking into a DAC upgrade for a couple of years though and here are the guys I had opportunity to listen to with my own amp / phone (I listen to other dacs at shows but it's meaningless without my recordings / downstream gear as reference): Yamamoto YDA-01 (still own it) , Mark Levinson 30.5, Ren. Labs Invicta, Eximus DP1.
 
cheers,
arnaud

 

[arnaud]

   
Unsuitable for a custom title, unfortunately. 

 
 
Iggy vs. Vivaldi stack would be interesting. 

 
That's it, I am reporting you to a mod!
 
Oh wait...

 

[purrin]

   
- For instance, your parallel between using a traditional ladder dac and converting a library to 13-14 bits is flawed: Decimating a whole song to 14 bit is basically throwing away a large part of the ladder array (so you'll loose all low level information). That is not what is happening in practice (the dac can effectively quantise down to 24 bits).
- You insist on the high bits precision error relative to the low bits but does that really matter when, by definition, high bits are active during peak times and there's only so much difference between loudest and quietest sound human can hear (level masking)?

 
Agreed. The proper method is to introduce LSB error to screw things up so the ENOB gets reduced. I've actually written a WAV file converter to do this according to a transfer function / INL plot.

 

[Sorrodje]

   
 

• Muddy Waters Folk Singer UDCD593 (GAIN system. R2R A-D converter designed by Pass+Moffat using milspec accuracy A-D chips with oven to control temps.)

 
Absolutely great recording. Thks for that.

 

[purrin]

   
Absolutely great recording. Thks for that.

 
If you like music along those lines, another one using the same Pass/Mofatt GAIN A-D.:
 
UDCD659 B.B. King - Lucille.
 
Now pick up stuff before GAIN (no what what A-D converter MFSL was using), still sounds pretty good, but not as resolving. Not as much plankton.
Then pick up stuff after GAIN - the one-bit A-D GAIN2. Sounds different... I know the first GAIN2 release, Tom Petty is like "shhhh shhh shhh shhh shhh."
 
Yeah, I actually went out and got a bunch of the old MFSL stuff, just to see if the GAIN thing / Mofatt were full of ****. As I like to say, "trust but verify" - much better than be skeptical and pontificate.

 

[purrin]

 
- You insist on the high bits precision error relative to the low bits but does that really matter when, by definition, high bits are active during peak times and there's only so much difference between loudest and quietest sound human can hear (level masking)?

 
You are either presenting a straw man argument or not grasping the concept of relative accuracy. The issue is what happens when a DAC is not monotonic to 16 bits? What if small tiny increases in volume according to the bitcode is translated into small tiny decreases instead? Or what if the bitcode calls for an increase of 0.021V for the next sample, but instead get increase 0.017 volts or perhaps even a decrease of 0.012 volts? 
 
Again, think of this as "jitter" on the Y-axis. This has nothing to do with high level signals masking low level signals (since you made a reference to Clem's post).
 

 
- From my experience so far (i.e comparing the D1 resolving power to all other D/S dacs I have had / heard) tells me that, perhaps, this bit depth thing is a non-issue. 

 
What satisfies you may not satisfy me. What I notice, you may not. You have to remember I'm coming from the Berkeley, Bricasti, Wadia (yuck), MSB, PSA, Aurulac, Invicta (yuck), DACs.

 

[wnmnkh]

  So does this mean the Yggy portion of the R2R ladder DAC is 14-bits?
 
DAC ARCHITECTURE
The architecture of the AD5791 consists of two matched DAC
sections. A simplified circuit diagram is shown in Figure 49.
The six MSBs of the 20-bit data-word are decoded to drive 63
switches, E0 to E62. Each of these switches connects one of 63
matched resistors to either the VREFP or VREFN voltage. The
remaining 14 bits of the data-word drive the S0 to S13 switched
of a 14-bit voltage mode R-2R ladder network. To ensure
performance to specification, the reference inputs must be force
sensed with external amplifiers.
 

 
I think this is the reason why two of them are used on each channel for proper 20bit+ performance.

 

[JohnnyCanuck]

  I think this is the reason why two of them are used on each channel for proper 20bit+ performance.

 
Either that or to maintain balanced channels.

 

[Sapientiam]

   
I think this is the reason why two of them are used on each channel for proper 20bit+ performance.

 
The DAC's a low-frequency monster and to get it sounding as good as it does requires an external deglitcher. So the 'secret sauce' such as it is, is in that deglitcher they're using - without that it'll be no more than, say a 12-bit DAC. The spec sheet says its good to 12bits after 1uS - at 16X OS the sample period is 1.4uS.

 

[prot]

The DAC's a low-frequency monster and to get it sounding as good as it does requires an external deglitcher. So the 'secret sauce' such as it is, is in that deglitcher they're using - without that it'll be no more than, say a 12-bit DAC. The spec sheet says its good to 12bits after 1uS - at 16X OS the sample period is 1.4uS.

What's a deglitcher?! And generally, would you care to explain your post a bit? Not all of us have EE degrees

 

[arnaud]

   
You are either presenting a straw man argument or not grasping the concept of relative accuracy. The issue is what happens when a DAC is not monotonic to 16 bits? What if small tiny increases in volume according to the bitcode is translated into small tiny decreases instead? Or what if the bitcode calls for an increase of 0.021V for the next sample, but instead get increase 0.017 volts or perhaps even a decrease of 0.012 volts? 
 
Again, think of this as "jitter" on the Y-axis. This has nothing to do with high level signals masking low level signals (since you made a reference to Clem's post).
 
 
What satisfies you may not satisfy me. What I notice, you may not. You have to remember I'm coming from the Berkeley, Bricasti, Wadia (yuck), MSB, PSA, Aurulac, Invicta (yuck), DACs.

 
- Fair enough on the stuff one notices / pays attention being people dependent (and you seem to have a particularly solid ear)
- About the ladder precision thing, let say I will remain skeptic until I can hear this for myself.
- 14bits effective resolution is 80dB plus of dynamic range.
- That means this "Y-axis jitter" term you're using comes into play for signals that are 80dB below the peak value at any given time block.
- You're thus saying you can hear these "fly farts within the orchestra playing".
- Fine, but there are concepts like level masking that long exist and prove this makes little sense.
 
cheers,
arnaud

 

[SoupRKnowva]

   
The DAC's a low-frequency monster and to get it sounding as good as it does requires an external deglitcher. So the 'secret sauce' such as it is, is in that deglitcher they're using - without that it'll be no more than, say a 12-bit DAC. The spec sheet says its good to 12bits after 1uS - at 16X OS the sample period is 1.4uS.

that might be part of why they aren't doing 16x oversampling, they seem to only be doing 8x oversampling according to what mike said about things being oversampled to 352 or 384khz, though that could also have been a limitation in the computational ability of the DSP they are using...who knows

 

[Sapientiam]

What's a deglitcher?! And generally, would you care to explain your post a bit? Not all of us have EE degrees

 
It does get a bit technical - for more technical background, go see my blog on DIYA which I linked to further up.
 
Here goes for a non-techy explanation. The Yggy DAC's designed to give the best performance with low frequencies - that's the application its intended for, where the output doesn't change at a fast rate. Think of it as a low sample frequency DAC where it only gets its stellar performance when updated relatively slowly (less than 10,000 times per second). Because of this, when the DAC's output changes, its spending a fair bit of time on its journey to the new commanded output. A deglitcher is used to remember the DAC's output and provide a copy of it to the following circuit while the DAC itself is changing its output. Think of the deglitcher as a switch and a memory of the last sample out of the DAC - when the DAC's told to update, the switch is opened and the memory element (normally a capacitor) provides the voltage to the downstream circuit. Then when the DAC's output has fully settled, the memory is updated with the new value.