[Currawong]

The Audio-gd DACs are a good match with the usually fairly bright Stax Lambdas and others, especially if you're more into jazz and classical and appreciate natural-sounding instruments.

 

[FauDrei]

Quote:

Do you know where can I find informations about DSP-1 Dip switches ? I have a Dac-19 DSP too, and i would try NOS mode.

 
Turn off pin 7 for NOS.
 
IMO, the default settings are pretty good, possibly the best. I preferred playing with attenuation (ambiance, focus) and dithering (details) settings. NOS was dull compared to 8x OS.
 
YMMV. Try for yourself.
 
NOTE: before switching DSP-1 dip swithces - turn your DAC off.

 

[Newk Yuler]

This is the same image that Currawong originally posted for the Reference 1 and what Kingwa is emailing for anyone who asks about the settings.  Vague at best and somewhat useless without further details.  Can anyone give proper explanations to the functions?
 
Normally all switches are in the off position (which is an active configuration for some settings) except the one that puts the DSP in line, which is switch 7.  According to Kingwa, if I interpreted him correctly, switching number 7 to off will bypass the DSP.
 
I'm going to stick my neck out and show my ignorance for the sake of someone correcting me and enlightening us all.  I know some of you are a lot smarter than I am.  (Hurt me! 

)
 
Switch 1.  PLL.  Active by default and normally best active.  Signal buffering and jitter reduction.
 
Switches 2 and 3.  Stopband Attenuation.  This is a bandpass filter with selectable attenuation.  Default on the highest possible setting.  Anyone know the actual spec?
 
Switches 4 and 5.  Oversampling.  This is pretty straight forward.  Default is 8x which is probably the best.  Very curious about the 1x setting, though.  It appears to provide no oversampling while allowing the benefits of running the signal through the DSP instead of a complete bypass (switch 7).
 
Switch 6.  Data dithering.  Active by default.  Filtering quantization distortions.  This might be one for experimenting depending on how you feed the DAC19DSP.
 
http://en.wikipedia.org/wiki/Dither#Digital_audio
 
Switch 7.  NOS.  DSP bypass to the PCM1704UK.  Actually eliminates DSP benefits instead of creating a simple NOS state.
 
Switch 9.  Output format.  PCM (Ref 1) and TDA1541 (Ref 2).  What is "Ref 1" and "Ref 2"?  PCM makes sense but what is TDA1541?  Is that a reference to the Philips DAC chip?
 
Switches 11 and 12.  Output bit.  Preset to the spec of the PCM1704UK.
 
Now I'm looking at this written out it really appears the switches are already at optimal settings for normal use.  Experimenting with the data dithering might be interesting.  Switching oversampling to 1x would allow the benefits of all the other active settings while delivering NOS to the PCM1704UK.  Unless I'm not really understanding 1x oversampling.
 
Some of you have said switching the DSP to NOS makes the DAC19DSP sound dull.  Can you elaborate on that?  The DSP is eliminating jitter from the input signal and switching to bypass would be allowing more jitter to pass to the PCM1704UK.  If you aren't using a sure low jitter source (such as a Hiface by SPDIF) the extra jitter should be much more apparent with the DSP out of line.  You also have to consider the jitter inherent to Kingwa's SPDIF circuit.  Higher jitter causes the sound field to collapse and focus/definition to disappear.  On the other hand, if you have a sure low jitter source, the affects shouldn't be as apparent.  The USB input on the DAC19 is a weak adaptive interface and not the best way to feed it, likely producing a high jitter input signal and poor performance with the DSP bypassed.

 

[FauDrei]

You pretty much got it right. I'd just add a couple of comments:
 
Switches 2 and 3.  Stopband Attenuation: how much "music energy" will slip out over stopband frequency. -130dB means almost nothing "gets away". OTOH, -90dB and -50dB are "letting some energy out" and in the process altering the balance of perceived audible focus and ambiance - lesser attenuations tend to have somewhat more "relaxed" sound. Worth experimenting.
 
Switches 4 and 5.  Oversampling: it's about upper frequencies (treble). Lower the OS - greater the treble rolloff and lower treble definition (details). Some would argue less OS or no OS at all sound more natural. IMO "more natural" in majority of the cases is euphemism for "filtering in my DAC sucks so no filtering sounds better" (please do not shoot - it is just IMO). A-GD DACs tend to do OS and filtering right and having full OS (8x) means you are having detailed, crystal clear, not at all harsh treble. Thus my "dull NOS" comment - you are missing so called "treble sparkle" (or treble at all). Again, just IMO.
 
Switch 6.  Data dithering: if you feed DSP-1 with 16 bit data and dithering is on, DSP-1 will "expand" 16 bits to 24 bits. This would be done not just by adding 8 zeroed bits, but 16 bit value will be relatively expanded to full 24 bit range. Dithering will be used to lower the relative error of such expansion. It should have no effect if DAC is fed with 24 bit data. Another setting worth experimenting with. Dithering slightly refines and "rounds" the sound, in certain cases no dithering can be perceived as more detailed. Try and decide for yourself.
 
Everything over dip 7 is DAC dependent and sets up DSP-1 module to work with it's host DAC.
 
FYI, Ref-1 is Reference One - first A-GD dac that had DSP-1 module. Ref-2 is Reference Two - discontinued twin brother of Ref-1 which used Philips TDA1541 DAC chips instead of PCM1704.
 
Hope this clarified DSP-1 a bit more.

 

[slim.a]

Quote:

You pretty much got it right. I'd just add a couple of comments:
 
Switches 2 and 3.  Stopband Attenuation: how much "music energy" will slip out over stopband frequency. -130dB means almost nothing "gets away". OTOH, -90dB and -50dB are "letting some energy out" and in the process altering the balance of perceived audible focus and ambiance - lesser attenuations tend to have somewhat more "relaxed" sound. Worth experimenting.
 
Switches 4 and 5.  Oversampling: it's about upper frequencies (treble). Lower the OS - greater the treble rolloff and lower treble definition (details). Some would argue less OS or no OS at all sound more natural. IMO "more natural" in majority of the cases is euphemism for "filtering in my DAC sucks so no filtering sounds better" (please do not shoot - it is just IMO). A-GD DACs tend to do OS and filtering right and having full OS (8x) means you are having detailed, crystal clear, not at all harsh treble. Thus my "dull NOS" comment - you are missing so called "treble sparkle" (or treble at all). Again, just IMO.
 
Switch 6.  Data dithering: if you feed DSP-1 with 16 bit data and dithering is on, DSP-1 will "expand" 16 bits to 24 bits. This would be done not just by adding 8 zeroed bits, but 16 bit value will be relatively expanded to full 24 bit range. Dithering will be used to lower the relative error of such expansion. It should have no effect if DAC is fed with 24 bit data. Another setting worth experimenting with. Dithering slightly refines and "rounds" the sound, in certain cases no dithering can be perceived as more detailed. Try and decide for yourself.
 
Everything over dip 7 is DAC dependent and sets up DSP-1 module to work with it's host DAC.
 
FYI, Ref-1 is Reference One - first A-GD dac that had DSP-1 module. Ref-2 is Reference Two - discontinued twin brother of Ref-1 which used Philips TDA1541 DAC chips instead of PCM1704.
 
Hope this clarified DSP-1 a bit more.

Hi FauDrei,
 
It is exactly what I heard trying different settings on the DSP-1 of the DAC19.
 
Regarding the dithering, I think it is done post-processing. I remember reading somewhere that the DSP-1 uses 32bits for processing (oversampling, passband...). Since the PCM1704uk accepts a maximum of 24bit word length the data (post filter) has to be converted from 32bits to 24bits. It can be done either by means of trucation or by dithering.
In theory, it shouldn't matter since 24bits (144 db) is well below the noise floor of the PCM1704uk(120db). But that is only theory ... as it seems that the dithering does indeed affect the sound (details) and we heard the same thing.
 
After having tried different settings, I personally returned back to the stock settings. But I suspect that with different headphones, I might prefer different passband settings.

 

[slim.a]

Quote:

burson opamp arrived today.

what a surprise: no muffled, but very dynamic, full-bodied, warm and mellow sounding music flowing through the rca connection. now the acss connection almost feels anaemic ... well, let's see, no, listen again tomorrow.

楽しみ。

Anymore impressions on the Burson opamps? (I am very tempted to order a pair to try).

By the way, for those who are comparing RCA vs ACSS, keep in mind that the RCA is 3db higher than ACSS in the C2.

 

[IPodPJ]

Quote:

In theory, it shouldn't matter since 24bits (144 db) is well below the noise floor of the PCM1704uk(120db). But that is only theory ... as it seems that the dithering does indeed affect the sound (details) and we heard the same thing.

Several people have mentioned over the course of the Audio-gd threads that using 8x PCM1704U-K (a dual differential configuration) will lower the noise floor, bringing it closer to that theoretical limit.  I don't know if this is true or not, as Texas Instruments never answered my query.

 

[slim.a]

Quote:

Several people have mentioned over the course of the Audio-gd threads that using 8x PCM1704U-K (a dual differential configuration) will lower the noise floor, bringing it closer to that theoretical limit.  I don't know if this is true or not, as Texas Instruments never answered my query.

It appears that stacking multiple PCM1704uk does indeed increase the SNR beyond the official specs of a single PCM1704uk. The Ref-7 is specified to have a SNR of 126db, which is 6 db above the SNR of the PCM1704uk under perfect conditions.

From what I understand, by using multiple pcm1704uk in the appropirate configurations, the noise floor of the PCB remains constant while the current output of the DAC rises. According to Wadia which uses a similar method (see here), doubling the number of dac chips improves the SNR by 3db, and using 4 dac chips increases the SNR by 6db (1bit). Since the Ref-7 uses 4 chips per channel: 120db+6db=126db.
If I had to guess, I would say that the advantage of this method is not so much as to reduce purely the SNR but perhaps to increase the low level linearity of the dac. But this is just speculation. I also doubt that there are many ADC and recordings chains out there with a SNR better than 120db.
 
Back when I measured the dac19dsp+c2 with my emu0404 usb, I got a SNR of 106db. It means that it was at least 10db better than the maximum resolution of CDs (which is 96db)... The emu 0404 usb under the same test has 113 db SNR but it sounds less detailed and the low level details are missing.
That is to say that there might be other reasons (besides lowering the noise floor) that Audio-gd and high-end companies such as Wadia use multiple PCM1704.

 

[IPodPJ]

Quote:

That is to say that there might be other reasons (besides lowering the noise floor) that Audio-gd and high-end companies such as Wadia use multiple PCM1704.

Yeah, because R2R ladder DACs sound good.  They have a very natural presentation.  MSB is at the forefront of DAC design and has the highest spec DAC in existence, the Diamond DAC IV.  They make their own ladder DACs.  But admission to this party does not come cheap at $26,000.  I would absolutely love to borrow one.
 
http://www.msbtech.com/products/dac4diamond.php?Page=platinumHome
 
With the Diamond power base it achieves an insanely low noise floor of -158dB.  The DAC is 32 bit/384kHz.
 
http://www.hd-multimedia.com/msb/msb.html

 

[Black Stuart]

Not knowing the make-up of the Audio-gd dacs but hearing how most if not all users are very happy with them, I'm not surprised when I learn that they have stacked Dac chips.
 
On a visit back to the island I stayed with a mate from the now defunct WAD forum and he had constructed two DACs using the above, with one he used 4 x chips per channel and the other 8 x , which I thought was excessive.
 
After listening to both I realised he was'nt mad and I have to say with both the DACs, I had never heard digital produce such a liquid and natural sound - this was 8 years ago.
 
Was the second DAC so much better, no, but it was just that more natural and afterwards it made me question the assumption that vinyl was untouchable compared to digital.

 

[Pricklely Peete]

Quote:

Yeah, because R2R ladder DACs sound good.  They have a very natural presentation.  MSB is at the forefront of DAC design and has the highest spec DAC in existence, the Diamond DAC IV.  They make their own ladder DACs.  But admission to this party does not come cheap at $26,000.  I would absolutely love to borrow one.
 
http://www.msbtech.com/products/dac4diamond.php?Page=platinumHome
 
With the Diamond power base it achieves an insanely low noise floor of -158dB.  The DAC is 32 bit/384kHz.
 
http://www.hd-multimedia.com/msb/msb.html

Yeah MSB stuff is impressive to say the least. They certainly zeroed in on what is important when they dropped their lower end line (Link series) to concentrate solely on producing what they consider the best (which it likely is) but the price tag...ugh...

 
Quote:

Not knowing the make-up of the Audio-gd dacs but hearing how most if not all users are very happy with them, I'm not surprised when I learn that they have stacked Dac chips.
 
On a visit back to the island I stayed with a mate from the now defunct WAD forum and he had constructed two DACs using the above, with one he used 4 x chips per channel and the other 8 x , which I thought was excessive.
 
After listening to both I realised he was'nt mad and I have to say with both the DACs, I had never heard digital produce such a liquid and natural sound - this was 8 years ago.
 
Was the second DAC so much better, no, but it was just that more natural and afterwards it made me question the assumption that vinyl was untouchable compared to digital.

 
It's interesting to note that while multiple chips improves the SNR the other side benefit to such designs is raising the theoretical bit resolution which directly effects realism to a large extent. The RE1 and Re7 according to Kingwa has a 25 bit theoretical resolution...even if it's actually 22 bit that's better than most of the stuff out there (other than MSB and the like). I had forgotten about that tidbit of info ...
 
I agree with you though, digital (redbook) can compete with great analog as long as the source material is well recorded/mastered.
 
Peete.
 

 

[IPodPJ]

Quote:

Yeah MSB stuff is impressive to say the least. They certainly zeroed in on what is important when they dropped their lower end line (Link series) to concentrate solely on producing what they consider the best (which it likely is) but the price tag...ugh...
 

At least we're in agreement about that.   I heard one of the earlier Platinum DAC models, I believe the model III awhile back.  It sounded superb.  I didn't have the Ref7 to compare but the MSB sounded extremely natural as does the Ref 7.  But since then they have created the Signature model IV, the Platinum model IV, and the Diamond model IV which could take it up to a whole new level.  It would be interesting to see how they would compare face to face with the Ref 7.  The MSB stuff uses 4 custom DACs, not 8.  I imagine though if they felt 8 of the 1704UK sounded better they would use them.
 
By the way, Chesky records uses a Platinum III DAC.
 
MSB's legacy:
 

1. demonstrated first AC-3 output for LaserDisc to Dolby
2. designed and built first THX approved LaserDisc Player (LJR II)
3. first out board AC-3 demodulator
4. first digital output on a DSS receiver
5. first out board DTS processor (Millennium 2.4.6)
6. first mass marketed 24 bit 96K DAC (LINK DAC)
7. first discrete sign magnitude ladder DAC (Platinum DAC)
8. first discrete asynchronous upsampler (Platinum Plus)
9. first 80 bit DSP based digital filter (Platinum DAC III)
10. first iPod based digital music server.

 
MSB is in California.  Problem is they are 330 miles away from me so I can't just pop over and have a listen.
 
The fact that we can even discuss the Ref 7/1 and MSB together speaks volume for the performance of Audio-gd gear, at a fraction of the price.

 

[pategen]

I'm closing in on the 350 hour mark with my burn-in process. It's rather amazing what proper burn-in can do to this gear. Now, there's not a hint of the extra brightness I reported earlier. There's a little bit more bass as well, and more importantly the bass has some serious punch now. I haven't heard AKGs produce anything near the bass drum impact that they are with this combo. It's really stunning.
 
I'm starting to think I may be done with my headphone gear for a while. What reservations I previously had about the synergy between K601s and DAC-19DSP/C-2 are now pretty much gone. They just sound GREAT, even if the Audio-gds might seem a bit overkill for these phones. Me so happy!

 

[cheaphifi]

Many thanks to Newk Yuler, FauDrei and slim.a for your answers about DSP-1 settings, i think the default settings are optimal and i'm very satisfied with the delicate sound blowed by the Dac-19 DSP and the Swans M200 MkIII
 

 

[fromatoz]

still the same impression: with the bursons the c2 is almost bursting ...
much more body and soul in the music.
i love it.