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

[jacal01]

 
Also, isn't the Nagra HD-DAC discrete ladder type? Can't see anything about the DAC module from pics as it's encaged but I thought I read something like this.

 
I believe you're right; I'm reading a 6 moons review that states it as using a discrete R2R for PCM conversion. Interestingly, it also says that it separately realizes true single bit DSD conversion via FPGA programming.  Sounds right to me.  A mere $25K.
 
Also, I've learned that the PCM1704 is still readily available for current DAC products, only its price has dramatically increased.  Anyway, it's not square... 

 
But you know, thinking on it, the single middle square chip probably is the DSP processor.  You need 2 or 4 of those oblong DAC chip suckers to make a balanced output, and I'm not seeing any of that.  I vote underneath the USB input converter board.

 

[Currawong]

Now if the largest chip on the DAC board is an FPGA....
 
I know it is not uncommon for the better than average DAC makers to use an off-the-shelf DAC, but none of the built-in filters, instead programming their own. However, Mike has said that the Yggy will keep all the bits going through, so he must have programmed something I reckon.

 

[purrin]

 
But I see what you've been doing; trying to match the square chip profile in the Yggy middle section with the usual and unusual multi-bit R2R chip suspects.  Any luck?   Sourcing a large volume inventory or open stock R2R chip, which are by and large pretty extinct beasts nowadays, for an upcoming production DAC does seem to be the tipping point.

 
Yup. Now you know what I was thinking. No luck though. As I've said, been trying to bait Jason and Mike. They don't budge.
 

 
Also, I've learned that the PCM1704 is still readily available for current DAC products, only its price has dramatically increased.  Anyway, it's not square... 

 
I don't think the Schiit guys are big fans of the PCM1704. I'm sure the constant availability/retirement alerts wouldn't give any manufacturer confidence to base a product off of it. Based on DACs that I've heard (and there are always good, mediocre, and bad implementations), I'm beginning to think the PCM63 was the pinnacle of DAC chips. Another favorite would be the UltraAnalog, with the UltraAnalog being smoother and organic, and the PCM63 being cleaner and more precise. All PCM1704 implementations had at least some murkiness in the bass and a tendency of sounding soft.
 

  Now if the largest chip on the DAC board is an FPGA....
 
I know it is not uncommon for the better than average DAC makers to use an off-the-shelf DAC, but none of the built-in filters, instead programming their own. However, Mike has said that the Yggy will keep all the bits going through, so he must have programmed something I reckon.

 
You look at what Theta (Gen VIII does not count) did in the past which was different from most of their competitors. I was curious about certain special sonic qualities of the Gen V and Mike told me [redacted] I grabbed the Gen V on the hunch there would be some similarities with the Yggy because I wanted an Yggy-like DAC before Q1 2015. It turns out [redacted]

 

[jacal01]

Now if the largest chip on the DAC board is an FPGA....
 
I know it is not uncommon for the better than average DAC makers to use an off-the-shelf DAC, but none of the built-in filters, instead programming their own. However, Mike has said that the Yggy will keep all the bits going through, so he must have programmed something I reckon.

 
Jason already said that they are using programmed DSP and not FPGA for filtering.  But yeah, the DSP chip would be large and square, too.
 
I'm looking at the mil spec DAC8581, altho the datasheet is stamped that it's not recommended for new designs (new replacement chip?) and I'm not sure it's even suitable for audio application.  What's the difference in SQ between 'resistor string' and 'resistor ladder'?  2 aspects of it that is intriguing, tho, is that it calls for an off-chip digital filter, such as the DSP that Schiit is implementing, and that it allows for a piece-wise linear (PWL) approximation to cancel linearity errors from the resistor string, which can also be resourced from DSP program.   Hell, if true bit perfect data retention can be obtained via the closed form equations in the 1917 Western Electric paper that they're programming the 18K+ timing interpolation filter taps algorithm from, then quite maybe they won't need PWL approximation for resistor mismatch error.
 

 
I don't think the Schiit guys are big fans of the PCM1704. I'm sure the constant availability/retirement alerts wouldn't give any manufacturer confidence to base a product off of it. Based on DACs that I've heard (and there are always good, mediocre, and bad implementations), I'm beginning to think the PCM63 was the pinnacle of DAC chips. Another favorite would be the UltraAnalog, with the UltraAnalog being smoother and organic, and the PCM63 being cleaner and more precise. All PCM1704 implementations had at least some murkiness in the bass and a tendency of sounding soft.

 
I'm beginning to think that as long as they get the digital filtering right, it doesn't much matter which multi-bit R2R chip they use.  Quite of bit of latitude in the implementation.  I may be having an epiphany here.
 
MOAR EDIT:  OK.  If the 24-bit PCM1704 isn't in play because it sounds like "ass" compared to the PCM63, then my vote is the 16-bit PCM56, which is still in production, even if not recommended for new design.  Note that the analog output signal filter requirements are greatly simplified through the use of 2-channel digital oversampling, up to 4x per DAC chip (page 8).  'Course I could completely be talking out of my own ass if not hat.  The PCM56U is fairly square, too. 

 

[jacal01]

OK, best guess:
 
TI has trio of 16-bit monotonic DAC ICs that are designed to replace the traditional monolithic multi-bit resistor ladder DAC chips, which suffer the disadvantages of conversion error from output signal sample-and-hold-deglitcher circuit imperfections and higher order analog filter phase response non-linearity, largely reduced to low order linear phase by employing digital filter oversampling, as alluded in my PCM56 post above, and high-swing output buffers, as well as voltage drift in multiple chip implementations due to different temperature coefficients.  This monotonic DAC chip family is as follows:
 
DAC8580 – Uses ‘resistor string’ architecture to largely eliminate glitching error and analog filter and output buffer requirements.  It includes an onboard programmable digital data interpolation filter capable of input data oversampling up to 16x, which can be bypassed, switched on, or switched on/off on the fly.  The external DSP interpolation filter might then be used to in conjunction with the onboard programmable filter in the bypass mode.
 
DAC8581 – Is the same in design as the DAC8580 above with similar resistor string architecture, but without the onboard programmable digital interpolation filter.  Since the Yggy will employ external DSP programmed digital interpolation filtering, anyway, this may be the appropriate DAC chip to use between the two, replacing the suggested external microcontroller digital filter PWL linearity approximation, as I had previously proposed.
 
However, both of these chips are not recommended for new product design, trending toward obsolescence.  They are both proposed to be replaced by the
 
DAC8811 – This 16-bit monotonic DAC does use resistor ladder architecture, and with a THD of -105 dB.  An external op amp is to be used in conjunction to impose output linearity.  The external DSP programmed input data interpolation filter should significantly enhance the output waveform fidelity, but is not discussed as a requirement or capability in this DAC implementation circuit design.   This DAC chip production is actively supported at present, may well offer some additional D/A signal conversion advantages for audio application, and so probably is most likely choice of the 3 options for predicting the Yggy DAC topology.
 
And the best part is that they are all square!   

 

[hans030390]

My guess is it isn't either of the first two of those three chips, though there's a good chance another DAC manufacturer does. That third looks interesting regardless if anyone is using it yet.

 

[BirdManOfCT]

  OK, best guess:
 
TI has trio of 16-bit monotonic DAC ICs that are designed to replace the traditional monolithic multi-bit resistor ladder DAC chips, which suffer the disadvantages of conversion error from output signal sample-and-hold-deglitcher circuit imperfections and higher order analog filter phase response non-linearity, largely reduced to low order linear phase by employing digital filter oversampling, as alluded in my PCM56 post above, and high-swing output buffers, as well as voltage drift in multiple chip implementations due to different temperature coefficients.  This monotonic DAC chip family is as follows:
 
DAC8580 – Uses ‘resistor string’ architecture to largely eliminate glitching error and analog filter and output buffer requirements.  It includes an onboard programmable digital data interpolation filter capable of input data oversampling up to 16x, which can be bypassed, switched on, or switched on/off on the fly.  The external DSP interpolation filter might then be used to in conjunction with the onboard programmable filter in the bypass mode.
 
DAC8581 – Is the same in design as the DAC8580 above with similar resistor string architecture, but without the onboard programmable digital interpolation filter.  Since the Yggy will employ external DSP programmed digital interpolation filtering, anyway, this may be the appropriate DAC chip to use between the two, replacing the suggested external microcontroller digital filter PWL linearity approximation, as I had previously proposed.
 
However, both of these chips are not recommended for new product design, trending toward obsolescence.  They are both proposed to be replaced by the
 
DAC8811 – This 16-bit monotonic DAC does use resistor ladder architecture, and with a THD of -105 dB.  An external op amp is to be used in conjunction to impose output linearity.  The external DSP programmed input data interpolation filter should significantly enhance the output waveform fidelity, but is not discussed as a requirement or capability in this DAC implementation circuit design.   This DAC chip production is actively supported at present, may well offer some additional D/A signal conversion advantages for audio application, and so probably is most likely choice of the 3 options for predicting the Yggy DAC topology.
 
And the best part is that they are all square!   

Good write-up. Thanks!

 

[twinkle]

  But you know, thinking on it, the single middle square chip probably is the DSP processor. 

I'd think so too.         Edit: there's a similar looking square chip on the USB board.
 

 You need 2 or 4 of those oblong DAC chip suckers to make a balanced output, and I'm not seeing any of that.  I vote underneath the USB input converter board.

What about the left side of the center board that bears the square chip, isn't there some twins or quatuors "oblong" chips there?
 
By the way you can stop looking 'cause I found Yggy's board schematics:

Spoiler: Warning: Spoiler!

http://xkcd.com/730/

 

[estreeter]

And to think purrin thought this thread had run it's course and wanted it closed - from memory, I think that was about 5 or 6 pages back  
 
I still think the obsession with silicon is short-sighted but I guess it fills in time during the long wait for Yggy. 

 

[darkless]

1.  Whoa.  A real live discrete resistor ladder DAC circuit layout?! GOK which vintage DAC it's from if so. And there's no clue as to how long the TI HC754 data flip-flop chip has been around. 

 
The picture shows the R2R ladder arrangements inside of a Totaldac d1. The laser trimmed Vishay VAR Series "Naked" Z-Foil Resistors are a dead giveaway. I own a Totaldac d3-single, so I should know. It's not vintage but an actual production DAC and it definitely belongs in the Class E tier.

 

[twinkle]

  And to think purrin thought this thread had run it's course and wanted it closed - from memory, I think that was about 5 or 6 pages back  
 
I still think the obsession with silicon is short-sighted but I guess it fills in time during the long wait for Yggy. 

 
Pretty good detective work on Jacal01's part though, maybe too good? careful now

, don't wanna give it away...
 
You know you might be right though: that analog implementation looks pretty serious, even sans tubes.

 

[NoPants]

If schitt doesn't care about the bit depth then they can use the PCM56 which is still in production. 16 bits is good enough for most of us. A shame it's $20 a chip.
 
I'm interested in how his filter works since he keeps doting on it. If it really does achieve closed-form digital filtering then it should be on everyone's radar. I have a few reservations about how the SPDIF input is handled though. 
 
I also use an R2R dac as my reference (spectral SDR),  a shame that it probably needs to be recapped sooner rather than later
That being said, I also use a  non-DSD Ayre QB9 which could be considered the antithesis of these vintage DACs. I would be wary of writing off sigma-delta when it has a strong implementation behind it.
 
Also stuck with a Lavry DA11 which I don't find to be all that great

 

[Tachikoma]

I'm guessing...
http://www.analog.com/en/digital-to-analog-converters/da-converters/ad5791/products/product.html
 
Not an audio DAC, but is designed for military/aerospace.

 

[purrin]

 
I'm interested in how his filter works since he keeps doting on it. If it really does achieve closed-form digital filtering then it should be on everyone's radar.

 
The digital filter is very intriguing to me. What matters to me is not necessarily the technicalities or algorithms, but how this secret sauce filter sounds like. How does it translate to what hear? Does the implementation of it in the Yggy have any relation to Mike Moffat's tales from yesteryear?
 
Techno babble is meaningless unless it actually does something.
DSD cough cough.

 

[smitty1110]

   
The digital filter is very intriguing to me. What matters to me is not necessarily the technicalities or algorithms, but how this secret sauce filter sounds like. How does it translate to what hear? Does the implementation of it in the Yggy have any relation to Mike Moffat's tales from yesteryear?
 
Techno babble is meaningless unless it actually does something.
DSD cough cough.

That's the worst part about RMAF is that they teased the Yggy, but not in an environment where you couldn't do proper critical listening. Oh well, fingers crossed for another beta program, and I'll be all over it if Schiit decides to have one.
 
Also, +1 for the DSD cheap-shot.