Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Are there still any high-end 1-bit DAC ICs being made? (Read 5469 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Are there still any high-end 1-bit DAC ICs being made?

I wasn't buying audio equipment when SACD was introduced (1999), so I'm curious what people were told was the advantage of DSD. I can't seem to find anything on the matter in Google Books from around that time. I would think the sales pitch highlighted that the DSD signal more closely resembles what hardware converters (ADC/DAC) work with. That is, if the hardware indeed uses a 1-bit delta-sigma ADC and DAC.

Of course that glosses over the difficulties of editing a signal in the DSD form, but let's for now assume that an analog master is digitized with 1-bit delta-sigma ADC and played back with a 1-bit delta-sigma DAC.

That brings me to the question: are such ICs still being made? Back when SACD was introduced, there were already a few multibit delta-sigma DACs around:

  • The Burr-Brown PCM1710U, PCM1712U and PCM1715U (introduced in 1994) had a 5-level (so a little more than 2-bit) delta sigma
  • The AKM AK5350 (introduced in 1996) had 'Enhanced Dual Bit' delta-sigma DACs
  • The Crystal/Cirrus-Logic CS5396 (introduced in 1997) had a 3-level (1.5-bit) delta sigma converter.

It seems datasheets from present-day DACs don't spoil their internals as much as the older do, except when there is some novelty. The AK4499EX uses 7 bits (128 level), but that isn't secret because that DAC needs a front-end that is on a separate chip, which means the number of bits is in plain sight (the pins required). All datasheets I found that say something about their inner workings seem to mention being multibit.

So, my question is: are DACs that use DSD unaltered still a thing at all? Could someone build a DSD DAC or SACD player using DSD unaltered without having to design their own DAC IC? Or would they have to use obsolete ICs? It seems most DAC IC accept DSD input, but convert it to their internal multibit format anyway.
Music: sounds arranged such that they construct feelings.

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #1 –
I think 1 bit DACs were already obsolete  before SACD launched.  Not sure if anyone still makes them after all these years, but probably could find some old stock of early 1990s ICs if you really wanted to build one.  It would not perform nearly as well as modern hardware though, so no point in doing that. 

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #2 –
For some reason DSD had a revival in the audiophile community in the 2010-2015 area.
I don't know the internals (DAC's used) but Schiitt offered a Loki DSD DAC

If you want it high-end have a look at MSB Technology
TheWellTemperedComputer.com

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #3 –
you can't compare quality if there's no precision. (fingernearthehead.jpeg)

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #4 –
For new chips, have a look on the CS43198 datasheet. It supports DSD direct mode without going through the "DSD processor", digital filters and multibit modulators. Also they have different specs in DSD direct mode and DSD processor mode, and DSD direct mode has slightly inferior specs in comparison. There is also CS43131 which is similar but does not support DSD direct mode.

For AKM, their single chip implementations (AK4490/4493/4497/non-EX 4499) should do similar things as well.

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #5 –
Quote from: bennetng on
supports DSD direct mode without going through the "DSD processor" [...] DSD direct mode has slightly inferior specs in comparison.

Begs the "why include it?!" question.
Faster / more efficient / less power-hungry? Or compatibility with something else?

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #6 –
Quote from: Porcus on
Quote from: bennetng on
supports DSD direct mode without going through the "DSD processor" [...] DSD direct mode has slightly inferior specs in comparison.

Begs the "why include it?!" question.
Faster / more efficient / less power-hungry? Or compatibility with something else?
I suppose one thing is to fulfill the market demands, as some audiophiles prefer minimal digital processing. For example, there are also DACs with discrete circuits advertise true multibit, non-oversampling operation, which is exactly the opposite of DSD.

Another thing regarding DSD direct mode specs of DAC chips is that because DSD direct mode bypasses the DAC chip's built-in modulator, so the performance of the modulator is based on the input DSD data, which for example, could be modulated from another ADC chip, or from a PCM to DSD encoder. Therefore without knowing how the input DSD data is exactly generated, there are some uncertainties about the specs listed on the DAC's datasheet. For example, a DSD encoder can be configured to have high DNR (e.g. above 125dB) below 20kHz with a sharp rise on ultrasonic noise, as well as lower DNR (e.g. below 110dB) below 20kHz with a softer noise shaping curve.

In DSD direct mode, the DAC chip relies on analog filters to cut off ultrasonic noise, if the filters are too weak, even if the DSD encoder has high <20kHz DNR in digital domain, the analog output can still be polluted by intermodulaton distortion and result in rise of noise floor below 20kHz.

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #7 –
Quote from: Roseval on
I don't know the internals (DAC's used) but Schiitt offered a Loki DSD DAC
The DAC used (AK4396) doesn't seem to do direct DSD like the CS43198 does. It goes through a delta-sigma modulator according to the block diagram.

Quote from: bennetng on
For new chips, have a look on the CS43198 datasheet. It supports DSD direct mode without going through the "DSD processor", digital filters and multibit modulators. Also they have different specs in DSD direct mode and DSD processor mode, and DSD direct mode has slightly inferior specs in comparison.
Thanks!

I wouldn't call the specs 'slighly inferior' though. Sure, these numbers are all excellent, but I wouldn't call the difference slight. For PCM or 'processed' DSD the typical SNR is 130 dB(A), THD+N is typcally -115dB. When switched to direct DSD mode however, these figures drop considerably, SNR to 116dB(A) and THD+N to -105dB. When fed with DSD128 (double speed DSD) the THD+N drops even further to -101dB.

So, that is a difference of 14dB(A) SNR and 10dB THD+N, or even 14dB THD+N for double speed.

Quote from: Porcus on
Begs the "why include it?!" question.
Probably because of marketing. I have a reasonably priced receiver, not high-end by a long shot, and still it has something called 'pure mode' disabling as much processing as possible. I think it sounds way better with all processing on (roomEQ, upmixing etc.) but I guess some people will find 'peace of mind' with something like this.

Quote from: bennetng on
Therefore without knowing how the input DSD data is exactly generated, there are some uncertainties about the specs listed on the DAC's datasheet.
I don't understand, could you elaborate? As I see it, the DAC was able to get a ~14dB boost in performance (both in THD+N and A-weighted SNR) by processing the DSD signal instead of feeding it directly to the output stage. So, I'd say, the way the signal was generated doesn't matter, switching processing on or off makes all the difference.

Quote from: bennetng on
For AKM, their single chip implementations (AK4490/4493/4497/non-EX 4499) should do similar things as well.
Thanks! They indeed seem to have a volume bypass option which lets the DSD signal go directly to the output stage, bypassing filtering and modulation. Sadly there are no analog specs listed with that option enabled.
Music: sounds arranged such that they construct feelings.

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #8 –
Quote from: ktf on
I wouldn't call the specs 'slighly inferior' though. Sure, these numbers are all excellent, but I wouldn't call the difference slight. For PCM or 'processed' DSD the typical SNR is 130 dB(A), THD+N is typcally -115dB. When switched to direct DSD mode however, these figures drop considerably, SNR to 116dB(A) and THD+N to -105dB. When fed with DSD128 (double speed DSD) the THD+N drops even further to -101dB.

So, that is a difference of 14dB(A) SNR and 10dB THD+N, or even 14dB THD+N for double speed.

I don't understand, could you elaborate? As I see it, the DAC was able to get a ~14dB boost in performance (both in THD+N and A-weighted SNR) by processing the DSD signal instead of feeding it directly to the output stage. So, I'd say, the way the signal was generated doesn't matter, switching processing on or off makes all the difference.
OK, I can remove the "slightly" wording. Regarding this specific CS43198 chip I've not seen any 3rd party measurement which explicitly used DSD direct mode, but for the AK4490/4493, specifically, the RME ADI-2 series, differences between DSD direct and multibit modes are not that big as long as the 50kHz filter setting is used when using DSD direct mode.
https://www.audiosciencereview.com/forum/index.php?threads/digital-filter-game.23795/post-934885
https://www.audiosciencereview.com/forum/index.php?threads/octave-music-don-grusin-high-resolution-music-analysis-video.31620/post-1138796

Here are some examples that different DSD encoder settings showing differences in the analog output.
https://archimago.blogspot.com/2021/10/measurements-look-at-dsd-and-using-sox.html


Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #9 –
Found another one. The ROHM Semiconductor BD34301EKV. Very expensive at $100 per chip when ordering 25 pcs at Farnell and Digikey. Might not seem like a lot of money for a DAC, but it is for a single chip. The CS43198 goes for about $15 per chip when ordering 25 at a time.

Anyway, that IC does pretty much no processing, unlike other DACs mentioned here which usually at least have a volume control. For PCM it has an upsampler and a delta-sigma modulator, for DSD it uses the signal directly.

For plain PCM the datasheets states a typical SNR of 130 dB(A) and a THD+N of -115dB. Double-rate DSD gets SNR of 125 dB(A) and THD+N of -113dB. All are excellent figures, but here also PCM does better than DSD.

Anyway, looking at the specs here I can't help but think that it seems DACs have improved quite a bit the last 10 years, but it seems ADCs haven't really. Even when not allowing for any headroom on the recording, there is no ADC that can match the performance of these DACs.
Music: sounds arranged such that they construct feelings.

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #10 –
Quote from: ktf on
I wasn't buying audio equipment when SACD was introduced (1999), so I'm curious what people were told was the advantage of DSD. I can't seem to find anything on the matter in Google Books from around that time. I would think the sales pitch highlighted that the DSD signal more closely resembles what hardware converters (ADC/DAC) work with. That is, if the hardware indeed uses a 1-bit delta-sigma ADC and DAC.
You cannot find all the marketing lies by reading reasonably accurate technical documents. One of the often claimed benefits of DSD is high sample rate, and false illustrations of impulse response, while neglecting the fact that it only has 1 bit with more than 97% of bandwidth covered up with noise.
https://www.merging.com/highlights/high-resolution


Also plain silly illustrations by drawing flat lines on FFT plots to show how "limited" 24-bit PCM is...


Yet the reality is something like this:
https://pcmdsd.com/Software/PCM-DSD_Converter_en.html

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #11 –
Quote from: bennetng on
You cannot find all the marketing lies by reading reasonably accurate technical documents. One of the often claimed benefits of DSD is high sample rate, and false illustrations of impulse response, while neglecting the fact that it only has 1 bit with more than 97% of bandwidth covered up with noise.
https://www.merging.com/highlights/high-resolution
[....]
That is pretty ridiculous indeed. They compare analog as it is being generated to storage formats. The proper comparison would be to compare to vinyl or tape. This response can probably be stored on videotape, to store it on audio tape you'll probably need specially hacked gear.

Also, conveniently left out all the noise for DSD from that plot.

Quote
Yet the reality is something like this
It also leaves out that there is no analog-to-digital gear that can match the performance of these formats, except 16-bit audio. So, this only applies to synthesized audio I guess  :))
Music: sounds arranged such that they construct feelings.

Re: Are there still any high-end 1-bit DAC ICs being made?

Reply #12 –
High performance DACs these days can have 30dB better dynamic range than 16-bit, so the "Sony DSD Direct" and "Korg AudioGate 4" plots above for example, will bottleneck these DACs, also some unwanted spikes on the wiggling noise floor.

However, ADCs on those old and DSD capable recorders could actually be using similar modulators, so even without considering analog pollution, these modulators themselves have no more than 19-20 bits of dynamic range even in digital domain, and probably only 18 bits of analog performance.

The meaning of AES17 dynamic range:
https://www.ap.com/technical-library/more-about-signal-to-noise-ratio-and-dynamic-range/

BTW, here are some reviews of a DAC with 4x CS43198
https://www.l7audiolab.com/f/measurements-of-topping-d30pro-dac/
https://www.audiosciencereview.com/forum/index.php?threads/topping-d30pro-review-balanced-dac.20259/

This one uses 2x CS43198
https://www.l7audiolab.com/f/millon-audio-cs43198-dac/

One CS43131 chip:
https://www.audiosciencereview.com/forum/index.php?threads/review-and-measurements-of-sound-blasterx-g6.7016/

ROHM BD34301EKV
https://www.l7audiolab.com/f/smsl-d300-dac/
https://www.audiosciencereview.com/forum/index.php?threads/smsl-d300-review-balanced-dac.28919/

Sadly, there is no DSD direct tests.

The CS chips are relatively cheap and often used in portable devices. For desktop products, quite a number of manufacturers use several chips in parallel to improve measurement results.

Still, all the products above are cheaper than for example, the ones using ES9038/9039Pro or AK4191+AK4499EX from the same manufacturers. ESS is known to not support DSD direct, I've seen some products, despite using the same ESS chip, have very similar PCM vs DSD performance, yet some products are like 10dB noisier when using DSD, and the DSD test signals being used are identical in the tests.

[edit]PCM-DSD_Converter in previous post is FOSS product, and allows manual input of FIR data on upsampling and IIR data for noise shaping, the FIR coefficients can be easily generated, but I don't exactly understand how to set the IIR data, I tried to use Octave to generate the data but the data format seems incompatible.