Most home-made non-oversampling DACs nowadays have sin(x)/x roll-off, which at 44.1 kHz sampling frequency can well be audible.
For what it's worth, I was at a demo last month where someone had made three multibit sigma-delta DACs with PCM1794 chips. One had a normal oversampling filter, one just a zero-order hold and one a zero-order hold and an analogue filter that corrected for the sin(x)/x roll-off. I didn't hear any difference between the first and the third DAC and had the impression that the second had slightly less treble. Mind you, it was not a blind test.
I also demonstrated my own DAC, which has an oversampling filter with very little passband ripple and plenty of headroom for intersample overshoots, a properly (2 LSB peak-peak triangular) dithered sigma-delta modulator with a kind of embedded pulse width modulator and a single-bit two-tap FIRDAC. It was very much liked, also by people who normally only listen to non-oversampling DACs with sin(x)/x roll-off.
One possible explanation is that they liked it because of the attention paid to details, like the passband ripple (which relates to pre-echoes), the headroom and the dither.
Another one is that they were subconsciously impressed by the looks of it: most of them are used to a DAC being an IC on a small board, so they were rather surprised when I came with a wooden enclosure of 63 cm by 33 cm by 20 cm with a big four-layer board with an FPGA module and ten valves and a filter board with big potcores and two signal transformers.
Since the event took place in two rooms, a third explanation is that people who didn't like it fled to the other room. I can't possibly say which explanation is correct.
To get back on topic: I prefer single-bit sigma-delta, provided it has an embedded pulse width modulator and proper dithering, but that's strictly on theoretical grounds.
For what it's worth, I was at a demo last month where someone had made three multibit sigma-delta DACs with PCM1794 chips. One had a normal oversampling filter, one just a zero-order hold and one a zero-order hold and an analogue filter that corrected for the sin(x)/x roll-off. I didn't hear any difference between the first and the third DAC and had the impression that the second had slightly less treble. Mind you, it was not a blind test.
I also demonstrated my own DAC, which has an oversampling filter with very little passband ripple and plenty of headroom for intersample overshoots, a properly (2 LSB peak-peak triangular) dithered sigma-delta modulator with a kind of embedded pulse width modulator and a single-bit two-tap FIRDAC. It was very much liked, also by people who normally only listen to non-oversampling DACs with sin(x)/x roll-off.
One possible explanation is that they liked it because of the attention paid to details, like the passband ripple (which relates to pre-echoes), the headroom and the dither.
Another one is that they were subconsciously impressed by the looks of it: most of them are used to a DAC being an IC on a small board, so they were rather surprised when I came with a wooden enclosure of 63 cm by 33 cm by 20 cm with a big four-layer board with an FPGA module and ten valves and a filter board with big potcores and two signal transformers.
Since the event took place in two rooms, a third explanation is that people who didn't like it fled to the other room. I can't possibly say which explanation is correct.
To get back on topic: I prefer single-bit sigma-delta, provided it has an embedded pulse width modulator and proper dithering, but that's strictly on theoretical grounds.
I always thought it was impossible to properly dither a 1-bit DAC since you need about 3 bits or so? Is there some sort of trick that I missed?
I can't take NOS DACs seriously. Tinkerer stuff. They went out of fashion in the mid-'80s for good reasons. Even the last CD players with traditional multibit DACs had something like 4X or 8X oversampling. If I had to run a NOS DAC, I'd be resorting to computer-based playback with a nice software upsampler (e.g. SoX VHQ) to do 176.4 / 192 kHz at least.
Single-bit sigma delta can be cranky and needs all kinds of trickery to get jitter sensitivity and high frequency noise output down (switched capacitor filters etc.), but with those can deliver pretty good results, even on a lowish power budget.
Personally, I think multibit sigma delta is where it's at. I'm not going to argue with anything that yields a 130 dB dynamic range. Plus, there's chips like the CS4399 that can run in "NOS mode" if you really want to roll your own digital filter in software.
Sigma Delta. It has matured over the years. NOS multibit is old news and chips are becoming rare.
Make that consciously as many audio people listen with their eyes. It really is better to not show the equipment until listening is over. Many tube people will automatically prefer anything that has tubes.
Make that consciously as many audio people listen with their eyes. It really is better to not show the equipment until listening is over. Many tube people will automatically prefer anything that has tubes.
Last edited:
I'm a fan of NOS DACs but they vary too much to generalize about how they sound against S-D DACs. S-D DACs also vary a lot, depending in the main on whether they use on-chip opamps for output stages.
If the comparison is a PWM-based single bit DAC (say STA326) against unfiltered multibit I'd go with the PWM DAC. If the multibit NOS is well filtered and doesn't include opamps then the multibit NOS would likely win.
If the comparison is a PWM-based single bit DAC (say STA326) against unfiltered multibit I'd go with the PWM DAC. If the multibit NOS is well filtered and doesn't include opamps then the multibit NOS would likely win.
That's the embedded pulse width modulator. The quantizer has three bits, which are then converted to a single bit stream by a pulse width modulator algorithm. I also randomly rotate the PWM pattern to get rid of its distortion.
Personally, I don't care whether the dynamic range is 80 dB, 130 dB or 200 dB. In all cases, I can't hear the noise at normal volumes.
Last edited:
Thanks to all your venerable opinions (and, please, continue to post). 
I don't want to resurrect the never-ending ancient diatribe between NOS multibit and DeltaSigma DAC, but I'm just asking this because at the end of my live tests I am really (but really!) undecided between two past glories: a DeltaSigma (CS4303) or NOS MultiBit (PCM56).
Beyond the technical peculirities of the single chips (the first RTZ, with pure digital differential outputs that gives full freedom to the designer to realize the entire analogue stage; the second with its pure current output and "zero crossing" noise trimmer which really minimise distortion at -80dB), both sound terrific to me (and better than modern commercial machines
) but totally different - given the same source, receiver and power supply.
It would be pretty easy for me, now finally, to mention pro and cons of the sound signature of the two DAC topologies, but if I just had to define the difference in perceived sound with a metaphor, I would say that if you enjoy "listening to music", you should get an NOS multibit; if you like more "musical details", then go for a delta-sigma.
If you allow me another comparison with other similar audio realities and perceptions, at first place I would put the diatribe if "Zero" Global Negative Feedback tube amplifiers sound better. At the end, if you think, Delta Sigma has a feedback loop too inside the modulator that (even if it is digital and not analog) it reduces distortion and extend response but potentially introduces time delays and artificials as well.
And with the passing time and with the hearing being lowered but much more trained, I find that I am getting more sensitive now to phase distortions than harmonic distortions and roll-off response.
I don't want to resurrect the never-ending ancient diatribe between NOS multibit and DeltaSigma DAC, but I'm just asking this because at the end of my live tests I am really (but really!) undecided between two past glories: a DeltaSigma (CS4303) or NOS MultiBit (PCM56).
Beyond the technical peculirities of the single chips (the first RTZ, with pure digital differential outputs that gives full freedom to the designer to realize the entire analogue stage; the second with its pure current output and "zero crossing" noise trimmer which really minimise distortion at -80dB), both sound terrific to me (and better than modern commercial machines
) but totally different - given the same source, receiver and power supply.It would be pretty easy for me, now finally, to mention pro and cons of the sound signature of the two DAC topologies, but if I just had to define the difference in perceived sound with a metaphor, I would say that if you enjoy "listening to music", you should get an NOS multibit; if you like more "musical details", then go for a delta-sigma.
If you allow me another comparison with other similar audio realities and perceptions, at first place I would put the diatribe if "Zero" Global Negative Feedback tube amplifiers sound better. At the end, if you think, Delta Sigma has a feedback loop too inside the modulator that (even if it is digital and not analog) it reduces distortion and extend response but potentially introduces time delays and artificials as well.
And with the passing time and with the hearing being lowered but much more trained, I find that I am getting more sensitive now to phase distortions than harmonic distortions and roll-off response.
Last edited:
It is possible to make a multi-bit sigma delta dac very musical and very detailed at the same time, but it is usually fairly costly to do. Usually, we are talking well over $1,000 for one stereo dac, probably closer to $2,000. Or, you can make your own (not quite as good, but pretty close) for maybe around $250, but a lot of work in that case.
Making a multibit NOS DAC musical and detailed isn't as costly as the figure Mark quotes for his ESS DAC - my lingDAC's BOM cost is in the region of $50 or so. Partly this is because the multibit DAC I use isn't so critical of its power supply, the chip has at least 30dB of PSRR. Having a current output also makes I/V stage design easier and filtering can be added fairly simply too. There's still work to do in selecting inductors and capacitors for the filter, to get sufficient passband flatness though, given that inductors come in at 20% tolerance.
There are better chips out there for a very long time now (especially on the Delta Sigma front).
Last edited:
1) switched capacitor filter
2) on board CMOS opamp
3) not RTZ output
Cannot compete with CS4303 design flexibility in my opinion...
The dac chip is certainly important but I would say the biggest contributor to the way a DAC sounds can be found in the power supply and, above all, in the analog output stage. The analog output stage is what determines the tonal colour or neutrality of the overall sound.
CS4303 is infinitely more flexible than the other DAC with respect to its digital differential RTZ output stage: the complete freedom to entirely design ad-hoc the active analogue stage suited to personal tonal preferences and adopting the most performing and preferred electronic components, is the main reason why I'm nearly in favour of this Delta-sigma sound. Because, we cannot deny the fact, a NOS multibit DAC sounds more relaxed and emotional in every way ...
Last edited:
One of the greater mistakes in audio is to judge chips by judging the used technology. I know as I made the mistake way too often Many audio enthusiasts listen with their eyes. Never dismiss a certain DAC chip because of a certain topology. I dismissed ES9023 at first but we all know where that lead to Besides that implementation is key. "Tonal preferences" means colouration which a DAC should not do IMHO.
Is this really a fact or is this hype? It is true that older R2R chips are often well treated in very nice NOS designs with extremely well-thought out power supplies (sometimes costing more than a complete Delta Sigma device). Delta Sigma chips are often used in recent reasonable priced devices. Direct comparison is often a bit uneven. I recently got feedback on one of my Subbu V3 DACs that it really was not as good as another tested device but the difference was very small. It turned out to be a comparison with a 2500 Euro device that was used with USB input on a different source than the Subbu was connected too via SPDIF... I have seen and heard this "NOS multibit dogma" debunked on a few occasions....
Many audio enthusiasts listen with their eyes. Never dismiss a certain DAC chip because of a certain topology. I dismissed ES9023 at first but we all know where that lead to Besides that implementation is key. "Tonal preferences" means colouration which a DAC should not do IMHO.Is this really a fact or is this hype? It is true that older R2R chips are often well treated in very nice NOS designs with extremely well-thought out power supplies (sometimes costing more than a complete Delta Sigma device). Delta Sigma chips are often used in recent reasonable priced devices. Direct comparison is often a bit uneven. I recently got feedback on one of my Subbu V3 DACs that it really was not as good as another tested device but the difference was very small. It turned out to be a comparison with a 2500 Euro device that was used with USB input on a different source than the Subbu was connected too via SPDIF... I have seen and heard this "NOS multibit dogma" debunked on a few occasions....
Last edited:
I'm pleased to say that I have finally decided and
among Delta-Sigma (CS4303, PCM1710 and AK4452)
and NOS multibit (TDA1543, PCM56) listening tests
the winner is.....
....
....
....
....
....
....
....
....
....
....
....
.... NOS PCM56 (3 in parallel) !!!!
Why? In one sentence, it is the one which transmitted me more the emotional message of the music! (that is not an hype ! )
Measurements (Jitter, THD, frequency response) are not the best ones, only better than TDA1543 which is the worse. Absolutely great ONKYO AK4452 DAC measurements, but this is on-board and not connected via SPDIF.
Anyway old NOS PCM56: THD@0dB is less than 0.003%, THD@-80dB is less than 8%, response @20kHz -1.5dB. I'm expecting to improve a little bit by tuning the circuitry. But not necessary for the listening!
Bass response is, on the other hand, stunning, extended full and clear, the best among all the above DACs.
I hope with PLL reclock to further improve these my positive feelings.
Cheers
among Delta-Sigma (CS4303, PCM1710 and AK4452)
and NOS multibit (TDA1543, PCM56) listening tests
the winner is.....
....
....
....
....
....
....
....
....
....
....
....
.... NOS PCM56 (3 in parallel) !!!!
Why? In one sentence, it is the one which transmitted me more the emotional message of the music! (that is not an hype !
)Measurements (Jitter, THD, frequency response) are not the best ones, only better than TDA1543 which is the worse. Absolutely great ONKYO AK4452 DAC measurements, but this is on-board and not connected via SPDIF.
Anyway old NOS PCM56: THD@0dB is less than 0.003%, THD@-80dB is less than 8%, response @20kHz -1.5dB. I'm expecting to improve a little bit by tuning the circuitry. But not necessary for the listening!
Bass response is, on the other hand, stunning, extended full and clear, the best among all the above DACs.
I hope with PLL reclock to further improve these my positive feelings.
Cheers
Especially for you I dug up 2 new/unused AD1860 which are the 18 bit PCM56 pin compatible replacements. Let me know by PM if you want to have them.
BTW PCM1710 is not a very good DAC chip. Do you also work with more modern DAC chips like those from ESS? They are among the best in Delta Sigma. One can choose to stay in the past for nostalgic reasons but there have been many improvements in technology (not all that succesfull but still)
BTW PCM1710 is not a very good DAC chip. Do you also work with more modern DAC chips like those from ESS? They are among the best in Delta Sigma. One can choose to stay in the past for nostalgic reasons but there have been many improvements in technology (not all that succesfull but still
)
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.