1 bit Dac?

[antcollinet]

https://sjeng.org/ftp/SACD.pdf

TL;DR impossible to dither, so intrinsecly distorted, even if at very low level.

Every day is a school day. Thanks for that.

I really need to improve my maths capability so I can follow articles like that deeper than at a superficial level.

 

[Sokel]

https://sjeng.org/ftp/SACD.pdf

TL;DR impossible to dither, so intrinsecly distorted, even if at very low level.

It's a 2001 paper so some gear were not at their prime yet, but it seems to me that the below (blanket) statements didn't age well:

We know where this stuff is also applicable, right?

 

[BeeKay]

Every day is a school day. Thanks for that.

I really need to improve my maths capability so I can follow articles like that deeper than at a superficial level.

Would this genAI wrap-up be mostly correct?

Title: Why 1-Bit Sigma-Delta Conversion is Unsuitable for High-Quality Applications


Authors: Stanley P. Lipshitz and John Vanderkooy


Main Thesis:
1-bit sigma-delta converters (such as those used in DSD/SACD formats) are fundamentally unsuitable for high-quality audio because they cannot be fully linearized. Proper dithering causes constant overload, which leads to unavoidable distortion, noise modulation, instability, and limit cycles.

Key Points:

• Dithering, a crucial method to eliminate quantization artifacts, is incompatible with 1-bit systems, since full dither already exceeds the overload threshold.
• Multi-bit sigma-delta converters (≥2 bits) can be fully linearized with proper dither, resulting in theoretically perfect performance.
• Simulations show that even with optimal design (e.g., Lip7ZP), 1-bit converters require severely limited dithering to avoid instability—thus nonlinearities persist and cannot be removed.
• The use of 1-bit systems in formats like DSD/SACD is criticized as technically regressive and damaging to audio quality, especially with multiple processing stages (editing, mastering).
• The authors present alternative designs (e.g., 8-bit PCM with noise shaping) that achieve superior audio quality at half the data rate of DSD—and with no nonlinear artifacts.

Key Takeaway:

1-bit sigma-delta converters are too simplistic for high-end digital audio. They inherently introduce audio-degrading artifacts that multi-bit, properly-dithered PCM systems avoid. For highest fidelity, multi-bit architectures with noise shaping are clearly superior.

 

[wwenze]

DS DACs usually are 4-5 bit because we are already pushing the limits of both sampling rate and linearity at high bit depth so why not take a step back in both directions and share the load? On one extreme you have a non-oversampling 16-bit or 24-bit DAC and on the other a 1-bit DAC, and for years the better ultimate performance as well as cost-performance has always been the thing in-between.

 

[piaseczek]

I am thinking about 1.5 bit dac

Here you go:

MU2 – Grimm Audio

 

[antcollinet]

Would this genAI wrap-up be mostly correct?

Yes - but at an even higher level of superficiality than even my limited maths achieves.

 

[Sokel]

Would this genAI wrap-up be mostly correct?

The downside is that based on this paper's logic we have to scrap class D amps too.

 

[Scytales]

It's a 2001 paper so some gear were not at their prime yet, but it seems to me that the below (blanket) statements didn't age well:

View attachment 443956


View attachment 443957

We know where this stuff is also applicable, right?

Those excerpts you have opportunely selected illustrate the main issue with this 2001 article, which still regularly pops up whenever the phrase "1 bit" is employed on audio related Internet webpages. The non-technical lexical field about feedback the authors use in these excerpts says it all. The title is enlightening all by itself : "Why 1-Bit Sigma-Delta Conversion is Unsuitable for High-Quality Applications". There is obviously an hidden agenda behind the technical discussion.

This article was part of the "Format War" between SA-CD and DVD-A at the very end of the '90 and the early '00s. At that time, both camps, on the one side Sony and Philips (and some others : dCS, Sharp) and supporters and at the other side Meridian, Matsushita/Technics and others and supporters fought with each others through AES papers in order to convince the potential professional clientele to choose their system proposal rather than the proposal of their competitors. Fair enough.

The response of the DSD camp at the very same 2001 AES convention was not much different that the above infamous Lipshitz and Vanderkooy paper: http://tech.juaneda.com/en/articles/dsd.pdf

So, the statements in this kind of paper must be taken with a pinch of salt and, at the very least, been properly contextualized.

And the world has moved on since 2001.

 

[Sokel]

Those excerpts you have opportunely selected illustrate the main issue with this 2001 article, which still regularly pops up whenever the phrase "1 bit" is employed on audio related Internet webpages. The non-technical lexical field about feedback the authors use in these excerpts says it all. The title is enlightening all by itself : "Why 1-Bit Sigma-Delta Conversion is Unsuitable for High-Quality Applications". There is obviously an hidden agenda behind the technical discussion.

This article was part of the "Format War" between SA-CD and DVD-A at the very end of the '90 and the early '00s. At that time, both camps, on the one side Sony and Philips (and some others : dCS, Sharp) and supporters and at the other side Meridian, Matsushita/Technics and others and supporters fought with each others through AES papers in order to convince the potential professional clientele to choose their system proposal rather than the proposal of their competitors. Fair enough.

The response of the DSD camp at the very same 2001 AES convention was not much different that the above infamous Lipshitz and Vanderkooy paper: http://tech.juaneda.com/en/articles/dsd.pdf

So, the statements in those kind of paper must be taken with a pinch of salt and, at the very least, been properly contextualized.

And the world has moved on since 2001.

Another exceptional post, thanks for that and the references.

 

[Hayabusa]

Here you go:

MU2 – Grimm Audio

Exactly! having 3 levels makes sense.
(btw 3 levels is exactly log(3)/log(2) bits ~1.5849625 bits)

 

[jsrtheta]

I’m still waiting for a half-bit DAC, hopefully soon.

How about a 2 bit dac?

 

[gvl]

How about a 2 bit dac?

Perhaps it’s too subtle, but the joke was about 1/2 bit being nonsense as one bit is the smallest unit of digital data.

 

[obsessive9119]

The first Discman I bought in the late 90s had a 1 bit DAC. Didn’t know what that meant then… still don’t now.

 

[jsrtheta]

Perhaps it’s too subtle, but the joke was about 1/2 bit being nonsense as one bit is the smallest unit of digital data.

Yeah, I got that.

 

[solderdude]

The first Discman I bought in the late 90s had a 1 bit DAC. Didn’t know what that meant then… still don’t now

Had something like the CXD2565M DAC chip in it with an output similar to that of DSD (but was PCM to DSD conversion) and required steep filtering to get rid of the substantial HF noise it generated.
This is why current converters are multibit DS. Makes the post filtering much more simple to do.
Another option is to make it run at considerable higher clock-rates which allows filtering to be simpler.

Other brands used that chip too. Technics for instance while developing MASH (4-bit DS) but they did not mention it as they could not explain to buyers why the bit race went from 16 to 20 in those days (and increased oversampling) and why 1 bit would be beneficial or not.
Later on Technics introduced MASH.

Mostly all to get around existing and pending patents and to convince potential buyers that 'their' solution 'sounds' best.

 

[Count Dacula]

My 10 year old Sony 5 CDP is 1-bit. I strained to hear if it was different than the older DACs and units I had, and the Sony sounds crisp.

 

[solderdude]

CDP-XA5ES ? this one is 1995 so ... 30 years old.

 

[yys310]

The downside is that based on this paper's logic we have to scrap class D amps too.

You do dither on class D amps?

 

[Sokel]

You do dither on class D amps?

You don't, but you do filter, heavily, at the same principle.

 

[Leejb1970]

Ibasso DX340 is a 1 bit dap. Superb player.