Passlabs D1 (PCM63) vs. Threshold DAC-2 (Ultra Analog D20400a) - who have heard both?

Attachments

  • Threshold DAC-2 front.jpg
    Threshold DAC-2 front.jpg
    30.8 KB · Views: 983
  • Threshold DAC-2 silver.jpg
    Threshold DAC-2 silver.jpg
    98.3 KB · Views: 965
  • Threshold DAC2 rear.jpg
    Threshold DAC2 rear.jpg
    6.8 KB · Views: 1,013
  • Threshold DAC-2 PCB D20400a.jpg
    Threshold DAC-2 PCB D20400a.jpg
    72 KB · Views: 1,270
  • d1-1b-pass-labs.jpg
    d1-1b-pass-labs.jpg
    155.4 KB · Views: 1,610
  • pass-labs-d1-front.jpg
    pass-labs-d1-front.jpg
    96.3 KB · Views: 610
  • pass-labs-d1-dac rear.jpg
    pass-labs-d1-dac rear.jpg
    63.3 KB · Views: 527
  • pass-labs-d1-dac-r2r-burr-brown-pcm63k.jpg
    pass-labs-d1-dac-r2r-burr-brown-pcm63k.jpg
    59.3 KB · Views: 502
Last edited:
  • Like
Reactions: 1 user
There is also a DIY forum here about DACs. It seems the "DAC nerds" are into what they call "NOS DACs" (No Over Sampling DACs) and they build towers where you can have like 8 or 16 DACs in parallel. I have yet to understand why it is better than a conventional DAC and why "over sampling" or "up sampling" is not good. It is also very important to differentiate between these two terms.....I have found out.
 
I didn't audition both DACs so far, but about the basic DAC designs I can tell the following:

- Measurements of the PCM63 chip seem to reveal a design flaw, which heavily compromises Differential Non Linearity (DNL, as a measure of resolution) and noise for positive codes (0.2%..100%FS), while codes near zero and below are much (about 10dB) better. This design flaw can only be addressed by using two or more matched PCM63 per channel differentially (not only parallel, which Pass fortunately did right), but nevertheless this chip is pretty fishy in that regard.

PCM1702/1704 (As a potential upgrade to the D1) could be better, but so far were not tested regarding DNL. Bad DNL in a DAC produces nasty ("digital" sounding) distortions. A good DNL is IMO necessary for a good sound, but nevertheless doesn't guarantee it (as many Ʃ/Δ-DACs show).

- The D1 got rave reviews (Still seeking that German review, where they compared it to a dCS converter) and is still highly regarded. I think this is mainly due to its Crystall PLL for 44.1 kHz (Being tolerant to mediocre, jittery sources) and the ingenious (at its time) I/V converter. Many other designers copied OPA2604 I/V stage from the BB reference designs and this is indeed a bad idea to implement a PCM63 right. Nowadays there are better I/V implementations possible, without using open loop circuits with high Cgd MOSFETs.

- The Pass I/V stage uses power MOSFETs, which generate lots of high frequency distortions, due to unnecessarily high, non linear Cgd (Miller) Capacitance. Change glitch distortion against Cgd distortion of the I/V ...

- The D20400A is voltage output by design and thus doesn't need an external I/V stage or the like.

- The D20400A contains a deglitcher circuit, which no other high resolution (Except the Lavry DA924) DAC ever provided. The code dependent glitches of R2R DAC designs provide huge data dependent distortion/jitter, which only can be eliminated by "reclocking" the raw DAC output with a sample- and hold circuit (deglitcher) in the analog domian. If you don't use a deglitcher after a R2R-DAC (PCM63 included), every crystal PLL in front of it is almost in vain.

- The Threshold DAC is a maximum Implementation of the UltraAnalog "Chipset" using two D20400A differentially, AES21 input receiver (also with secondary PLL), and also the ingenious PMD100-Filter, which hopefully is implemented right (analog gain switching for non HDCD).

- I auditioned the Classé DAC-1, which is a very similar design to the Threshold (both products most likely designed by UltraAnalog Inc. directly). This DAC is among the best DACs (probably the best) I ever heard. I doubt that even the best PCM63 implementation (Pass D1 ?) could match it regarding clarity and sound stage, even if you would try to use 16 hand picked PCM64P-K per channel to get rid of the DNL problem.
 
  • Like
Reactions: 1 user
The distortion on a D1 was .001% at FS - 1 bit not exactly high but not as good as today’s chips.

With technically correct digital implementation we found the analog I to V most important followed by the digital filter. Things may be different today I don’t really know. I preferred the PCM 63 sound over the 1704 but they were close unless you had 24-96 data. I know the chip designers had a preference for the 63’s internal clock circuit IIRC.
 
The distortion on a D1 was .001% at FS - 1 bit not exactly high but not as good as today’s chips.

With technically correct digital implementation we found the analog I to V most important followed by the digital filter. Things may be different today I don’t really know. I preferred the PCM 63 sound over the 1704 but they were close unless you had 24-96 data. I know the chip designers had a preference for the 63’s internal clock circuit IIRC.

Mr Wayne, did the D1 feature a digital filter and oversampling, or was it NOS and filterless?

Did the I/V stage offer roll-off like a lowpass filter?