I have owned an Audio-gd PCM1704 DAC for a few years now and have put some of the best DAC's I could get my hands on against it.  The Audio-gd always wins.  I tried his ref1 and ref 7.1 but always felt something missing compared to the older models.  I believe the issue to be jitter from the FPGA integration.  I found a DAC from Rockna called an RD-30,  it uses the Xilinx Spartan6xling FPGA but the DAC designer worked Xilinx's engineer to get the most out of the chip including a claimed synchronous pll with the CCHD-575 clock.  It also has an asynch usb input where the CCHD-575 is switched to master clock.   If the claims are true this could be a real nice alternative to the Audio-GD products.  They also come an I/V stage with current steering NFB free topology.  I went a head and ordered one when I learned the oversampling filters can be chosen from 4 different types including an apodizing filter.
 
  Rockna has a good reputation in Europe with multi-thousand dollar DAC's and this RD30 although requires a minor amount of assembly looks on paper to be a real nice alternative to the Audio-gd PCM1704 DAC which are truley nice but I am hoping that this RD30 will be a final answer to a true hirez PCM1704 DAC with less jitter than the DSP-1 and a better analog stage than the ACSS modules.
 
  Regardless its nice to finally have a modern alternative,  hopefully the DAC will get here before Christmas and I can post some comparisons to the Audio-gd PCM1704 DAC's if folks are interested.

The Rockna DAC sounds very interesting. Is there any page with detailed information about it? I searched around and found a few sites selling it, but with very little information.

I found it via a thread at diyaudio there is al ink to the web-page.  I have high hopes it will dethrone my Audio-gd standby which is unfortunatey limited to 16/44.1.  If this RD30 doesn't do it I just going to wait a few years till HDMI DAC are common-place,  but by then the PCM1704 will probaby be extinct so I figured its xmas might as well take a chance.

Quote:

catfish said:

  I have owned an Audio-gd PCM1704 DAC for a few years now and have put some of the best DAC's I could get my hands on against it.  The Audio-gd always wins.  I tried his ref1 and ref 7.1 but always felt something missing compared to the older models.  I believe the issue to be jitter from the FPGA integration.  

Click to expand...

 
Probably not jitter as I find its effects to be inaudible on multibit. FPGAs though are well known to be noisy - nowadays they're made in very fine geometry CMOS and hence ultra high edge speeds. So probably its introducing extra digital hash into the equation via the power supply.

Quote:

sapientiam said:

 
Probably not jitter as I find its effects to be inaudible on multibit. FPGAs though are well known to be noisy - nowadays they're made in very fine geometry CMOS and hence ultra high edge speeds. So probably its introducing extra digital hash into the equation via the power supply.

Click to expand...

 
Yes groundbounce and VHF noise can muck things up this is one of my fears in ordering this thing with the FPGA,  but the key is how the PCB is layed out.  I just am never going back to CDP's so everything is on the computer,  there is no technology I can find that really isolates high speed "edges" if you use a computer without using SPDIF.  Even so called adum "isolators" are just a bandaide,  so my latest thought is go with a designer who knows how to route high speed ground bounce issues and keep them from mucking things up.
 
 But I do have my fingures crossed as in general I don't like the sound of any high-speed digital device be them FPGA's, S-D Modulators, etc. 
 
With my Audio-gd  (its an old model weighs as much as a 7.1 but pre DSP-1). I found a huge improvement when I added an Art Legato as the transport,  so I still think Jitter is important even with multi-bit DAC's,  I mean peope go to huge efforts to keep jitter from mucking up multibit TDA1541 DAC's.
 
I am also of the camp that thinks oversamping/filtering is very important to good accurate sound reproduction by a DAC.  There is a great article by Steve Nugent at audio stream http://www.audiostream.com/content/qa-steve-nugent-empirical-audio . 
 
Finally the analog stage is critical,  ultimately all Multi-bit i-out DAC's convert to voltage output via resistance.  Wether one uses an opamp, a current conveyor, a resistor, or a transformer, ACSS,  there is always a resistance to produce a v-out (its ohms law.)   My Audio-GD uses a resistor then post amplification/diamond buffer,  this sounds great and measures very well. 
 
This new RD30 uses a different approach,  it amplifies the current first then dumps the current thru a resistor.  There is no NFB so I am concerned that there will be too much 2H distortion,  but I have high hopes ...

Yes people do fuss about jitter rather a lot, but oddly for me its never been an issue. Whereas noise is an ever present threat to great sound in my system. I take the view that most of what people talk about as 'jitter' is actually noise. So perhaps your Art Legato improved the sound more through noise absorption or filtration than through hitter reduction? I'm rapidly coming to the conclusion that differences in SQ of DAC chips are in the main down to glitchiness. This explains why the TDA1543 sounds pretty good even though its one of the poorest measuring chips. Its low glitch by virtue of being made out of bipolar devices. The TDA1545 OTOH has slightly better measurements but sounds grey if the output isn't carefully filtered. The faster you run a DAC chip, the greyer the sound gets so I'm no fan of oversampling.

Quote:

sapientiam said:

I'm rapidly coming to the conclusion that differences in SQ of DAC chips are in the main down to glitchiness. This explains why the TDA1543 sounds pretty good even though its one of the poorest measuring chips. Its low glitch by virtue of being made out of bipolar devices. The TDA1545 OTOH has slightly better measurements but sounds grey if the output isn't carefully filtered. The faster you run a DAC chip, the greyer the sound gets so I'm no fan of oversampling.

Click to expand...

 
That's very interesting. How does one measure the glitch level of a DAC chip (so one could compare one chip to another)? Does it show up on conventional harmonic distortion measurements? What do you make of the statement in the PCM1704 datasheet that it has "glitch-free output"? Does this just mean the glitch level is relatively low? How does it compare to the TDA1543?

Quote:

moonunit said:

 
That's very interesting. How does one measure the glitch level of a DAC chip (so one could compare one chip to another)? Does it show up on conventional harmonic distortion measurements? What do you make of the statement in the PCM1704 datasheet that it has "glitch-free output"? Does this just mean the glitch level is relatively low? How does it compare to the TDA1543?

Click to expand...

 
Interesting questions, thanks. So far I haven't come up with a way to measure the glitch - I suspect they're very high speed, spikes. Some of the more modern DAC chips designed for communications (rather than audio) applications have a specification for glitch because it matters there. They measure it in V.nS (volts X nanoseconds) and often have a plot of the shape in the datasheet. These are the chips that I reckon will be the best sounding but as yet I've not built a DAC using them. One reason for this is that the best ones for glitch seem to max out at 14bits so would need a bit of DSP to adapt them to redbook input. Here's the latest generation datasheet : http://www.analog.com/static/imported-files/data_sheets/AD9704_9705_9706_9707.pdf. This has an extremely impressive gltich spec of 5pVs, the lowest I've seen anywhere by a substantial margin. As a comparison the DAC used in the Metrum (also rated very highly for SQ by Martin Colloms) is about 100X worse for glitch. I have a hunch though that even that figure's better than R2R audio DACs.
 
No, the glitch does not show up in conventional THD measurements. One reason for this is that the glitch is extremely high frequency, almost like an impulse so very wideband. THD+N measurements are alwoys low pass filtered, so relatively narrow band. Hence little glitch energy makes it into the audio band to increase the noise. But when the glitch gets into active circuitry it causes IMD leading to greying out of the sound. Only this month I've found a huge benefit in SQ by installing a steep passive filter after the passive I/V before the active stage and the improvement is in the tonality - particularly noticeable on piano. Its taken on a highly alluring richness now. The soundstage depth also shows substantial improvement, which was the biggest surprise.
 
Your question about the 'glitch free output' is extremely relevant. If you read further on down the datasheet you'll see that it doesn't really mean 'glitch free' rather it means they've addressed the R2R problem of the zero crossing glitch. The glitch is still there, just it got moved away from the zero crossing - now since they're using two DACs the glitch will be at 25% and also at 75% of full scale. So the low level performance improves but in return you get twice as many glitches at -6dB output level. Interestingly listening reports seem to correlate well with this as discerning listeners report the dynamics of PCM1704 aren't the best which is precisely what you'd expect if extra noise is being added only at higher levels.
 
I'm only guessing at the TDA1543's glitch level based on listening. Since its bipolar and not R2R I reckon it escapes the glitchiness of the CMOS (and BiCMOS) technology DACs. As does the TDA1541A which is ranked the highest subjectively by many listeners.

Thank you so much for the detailed response. I definitely learned something today.
 
Quote:

sapientiam said:

Only this month I've found a huge benefit in SQ by installing a steep passive filter after the passive I/V before the active stage and the improvement is in the tonality - particularly noticeable on piano. Its taken on a highly alluring richness now. The soundstage depth also shows substantial improvement, which was the biggest surprise.
 

Click to expand...

 
I've noticed that a number of designs use that approach. It's something I was contemplating in my next build as well.

If you're interested in building your own DAC then you might well get something out of this thread too : http://www.whatsbestforum.com/showthread.php?8858-Digital-that-sounds-like-analog

  The glitch deal is an issue.  The problem with audio measurements is the same as the Heisenburge principle you can measure in the frequency domain or the "timing"/phase domain but not both at the same time, so no one has the "answer".  That leaves trial and error (its still an empirical science.)
 
   I do agree about the PCM1704 dynamics issue,  its a especially a problem with R2R's that use BPO/servos on the output. They are there for integration with opamp I/V's.   The PCM1704 has these servos integrated and is one of the reasons it tends to sound less dynamic that previous generations like the PCM63K where the servo  can be disabled.
 
  There is a trick to slow thePCM1704  BPO/servos to the point they do nothing by using large capacitors on those pins (Audio-GD does this.)
 
  The other thing that affects dynamics is the over sampling filter.  For example an Apodizing filter works to eliminate pre-ringing by decreasing dynamics.  Trade-offs with everything,  Grado's can be a good cure to dynamic loss
 
  But I am in love with orthos,  have a pair of HE-5LE,  and the achilles heal of orthos is again dynamics.   So again I am worried I made the wrong choice,  but ultimately it comes down to the sound.  If this doesn't work I will just focus my energy on finding a TDA1541,  the "king of dynamics" but 16 bits is tough to integrate with a computer if you like to dial knobs (eq, reverb, stereophonic-to-binaural dsp, etc.)  That 32 bit floating math is a b*itch!
 
I found that Rockna is offering the DAC I ordered pre-built,  they call it  the Audiobyte Unimorph.   They seem like a decent company as far as I can tell with their high end discrete R2R DAC called the WaveCrest (which is way out of my price range),  looks like a european MSB DAC.
 
Some of the posts I found on Rockna from years back showed that they would make a great DAC but delivery could be slow,  they promised it will be here by Christmas , knock on wood.

Quote:

sapientiam said:

If you're interested in building your own DAC then you might well get something out of this thread too : http://www.whatsbestforum.com/showthread.php?8858-Digital-that-sounds-like-analog

Click to expand...

 
Thank you! That's a long thread but I read through the whole thing. I'm a bit of a DSP geek, so I'd like to try my hand at writing custom oversampling filters first before I try NOS, but I think you may be on the right track in a lot of ways.
 
One question I have: what is the effect of paralleling DACs on the glitch level? It's possible to improve the traditional SNR of the PCM1704 by ~3dB by paralleling two. Does the glitch level also decrease by roughly the same amount, or does it stay the same or double?

Quote:

catfish said:

   There is a trick to slow the PCM1704  BPO/servos to the point they do nothing by using large capacitors on those pins (Audio-GD does this.)
 

Click to expand...

 
Which pin(s) are you referring to? REF DC, SERVO DC, or BPO DC? Also, roughly how large of a capacitor are we talking about?
 
It's really hard to find out any concrete information about the purpose/use of the REF DC, SERVO DC, or BPO DC pins.

Quote:

moonunit said:

 
Thank you! That's a long thread but I read through the whole thing. I'm a bit of a DSP geek, so I'd like to try my hand at writing custom oversampling filters first before I try NOS, but I think you may be on the right track in a lot of ways.
 
One question I have: what is the effect of paralleling DACs on the glitch level? It's possible to improve the traditional SNR of the PCM1704 by ~3dB by paralleling two. Does the glitch level also decrease by roughly the same amount, or does it stay the same or double?

Click to expand...

 
As you're a DSP geek then here's another of my threads that you might enjoy - even longer than the last one: http://www.diyaudio.com/forums/digital-line-level/195791-open-source-dsp-xos.html. I've played a bit with minimum phase OS filters on my ARM M0 and M3  chips, the sound wasn't as good as NOS which was one result that lead me to my hypothesis about glitching being the major SQ determinant.
 
Yes paralleling DACs cannot reduce the glitch level and in fact may make it worse because of layout issues. If there's power supply bounce different between the two DAC chips then potentially the glitching could be worse when added together. The glitches double in the best case but then so does the signal (current) so no net difference provided the power supply noise is identical on each (hard to achieve - I do this by stacking chips, but then I'm not using 1704). BTW its only my theory about power supply noise making the SQ worse in paralleled DACs, its based on hearing improvements by fixing up the grounding on my Muse TDA1543 * 4 so I could easily be wrong there.

I finally finished building mine from the OEM boards, saved some money and prefer it much over the AudioGD Ref7.1,  this has the Si570 clock and fifo along with an excellent oversampling algorithm.  The I/V stage is fairly unique,  its a discrete no global feedback common base BJT design,  simple and nice sounding.  For those who like the PCM1704 but would like a modern implementation on the digital side this is a good alternative.
 
The finished product can be found here:
 
http://www.audiobyte.ro/unimorph.html