Delta-sigma vs “Multibit”: what’s the big deal?

[Frank Dernie]

Yep. But any measurement feticist (including most people attending this forum) will never admit that they may fall victim to the same fallacy. Wereas it is obvious to anybody with a modicum of critical thought.

OK, let me ask a some simple questions.
Do you accept that the only connection between one item if hifi and the next is the cable between them?
Do you accept that the only thing going down the cable is a voltage having magnitude, frequency and phase? (there is nothing else btw)
If so then it is of no consequence whatever how the signal is achieved, from a simple IC to a complex circuit containing exotic components, massive feedback or none, the only thing that matters is that this signal has not been compromised in any audible way.
We can measure a dynamic range of 140dB, measure frequency from DC to megahertz, so way more than any human can hear, so whilst I do indeed believe people hear differences between components which measure the same or with minimal change, the only effect we can't measure is the placebo effect and that is very strong, so strong that a sugar pill can actually cure some diseases if the individual is susceptible, so the only plausible explanation of us not measuring the right thing is that we can not measure the placebo effect. So all sorts of sales pitches can influence what people hear, from low feedback to silver capacitors to shakti stones even though the actual sound is the same.

 

[mocenigo]

OK, let me ask a some simple questions.
Do you accept that the only connection between one item if hifi and the next is the cable between them?

Mmmmm, are you ignoring, for instance, the power ground which is usually common to them? The fact that both devices are connected to power? Ever heard of ground loops?

Do you accept that the only thing going down the cable is a voltage having magnitude, frequency and phase? (there is nothing else btw)

Of course this is right.

If so then it is of no consequence whatever how the signal is achieved, from a simple IC to a complex circuit containing exotic components, massive feedback or none, the only thing that matters is that this signal has not been compromised in any audible way.

Of course this is right.

We can measure a dynamic range of 140dB, measure frequency from DC to megahertz, so way more than any human can hear, so whilst I do indeed believe people hear differences between components which measure the same or with minimal change, the only effect we can't measure is the placebo effect and that is very strong, so strong that a sugar pill can actually cure some diseases if the individual is susceptible, so the only plausible explanation of us not measuring the right thing is that we can not measure the placebo effect.

This is all correct.

So all sorts of sales pitches can influence what people hear, from low feedback to silver capacitors to shakti stones even though the actual sound is the same.

All correct.

However, I never talked about sales pitches.

I never even dared assume that R2R is intrinsically better than DS, even though just the suspicion of thinking it equals to a condemnation in a measurement forum. I only state one thing: are you sure that measuring first and second order steady signals is sufficient to establish whether there is a difference between two items of equipment? I know for a fact that in other areas where we DO measure signal traces such a claim would be so ludicrous that nobody would ever take you seriously if you did this. However, this is normal.

Regarding the R2R vs DS debate: I do not believe that the measurements of the DAC1421 are representative, even though it would not surprise me at all if most modern DS designs actually measure better. After all, the latter, regardless of whether they sound fine or not (and in general they sound so fine that they are mostly indistinguishable and all very good) are designed to excel in these measurements, since they are the easiest to understand. Also, I would never buy a DAC with a bad SN ratio, since I have power amps with high gain (26Db) and sensitive horn based speakers (95+Db/W/m). The most I hear from the compression driver tweeter loaded by the horn is an extremely faint crackle only if I stick my head in the horn. And nothing in the headphones. Absolutely nothing. At this point, other parameters are relevant.

 

[Thomas savage]

What you really need is a Himalayan salt lamp, hifi never sounded so good.

 

[mocenigo]

What you really need is a Himalayan salt lamp, hifi never sounded so good.

???
Misunderstood some irony?
(Even though, if class A and triodes fight depression, let us make everything else illegal!)

 

[Veri]

What you really need is a Himalayan salt lamp, hifi never sounded so good.

And as a bonus you can season with it / eat it, too!

 

[JJB70]

Yep. But any measurement feticist (including most people attending this forum) will never admit that they may fall victim to the same fallacy. Wereas it is obvious to anybody with a modicum of critical thought.

Not true. People do question the point of obsessing over measured performance improvement beyond the point of audible transparency and if you read the listening impressions bit of DAC reviews here you will see even most of the badly designed and implemented ones sound OK.

I do think that becoming fixated on measurements beyond any audible difference is pointless and a form of subjectivism but at least stellar measurements indicate good design if you do want to spend $$$$$s on a premium product. Too many expensive products rely on florid prose and shill reviewers to try and divert attention from mediocre design, build and performance (see clueless DACs for example).

The reality is most competently designed electronics have been audibly transparent or as near to as makes no real difference for decades.

 

[SIY]

Ok, I will switch to triodes! Really, I know that from the point of view of "hifi" they are all measure like crap, but they sound lovelyl

Mine measure quite well, thank you.

 

[mocenigo]

Not true. People do question the point of obsessing over measured performance improvement beyond the point of audible transparency and if you read the listening impressions bit of DAC reviews here you will see even most of the badly designed and implemented ones sound OK.

Well, yes, what I read is that a lot of people say, yeah, they measure differently, but there is no audible difference. But in most cases they write this here because they have seen the measurements first, and maybe they have not listened to them.

Also, what makes an excellent DAC? Just the engineering? If a Soekris and a DS DAC sound indistinguishable, then they are both either excellent or not.

I do think that becoming fixated on measurements beyond any audible difference is pointless and a form of subjectivism but at least stellar measurements indicate good design if you do want to spend $$$$$s on a premium product.

Ok, I would not spend money on something just because it measures in a stellar way, if there is no other difference. But who am I to stop somebody from parting from their money?

Too many expensive products rely on florid prose and shill reviewers to try and divert attention from mediocre design, build and performance (see clueless DACs for example).

Absolutely true. And my BS alarm starts to ring as soon as I read the description of stuff like the Synergistic Research Orange fuses. Because what they write is pseudo-scientific jargon. They do not reveal what they are actually doing, and cover it with buzzwords. If they speak like cons (to somebody with a scientific background) then I am going to assume they are cons.

The reality is most competently designed electronics have been audibly transparent or as near to as makes no real difference for decades.

Not sure about this. Also, what competently designed means. This would open not one but several cans of worms. Too many variables.

 

[mocenigo]

Mine measure quite well, thank you.

Splendid. Then they must sound neutral, and have no curative effect ;-) [I was just making some irony]

 

[SIY]

Splendid. Then they must sound neutral..,

Yes they do. They are designed to make a small signal bigger, no alchemy, just engineering.

 

[sev1]

It's been a bit of a trend in the DIY community (not just over at DIYAudio, but in Asian DIY communities as well) over the last couple of years to take the approach you suggest and use software-based filtering with old multibit chips. It's an approach that makes a lot of sense, if you're playing around with multibit chips. PCM1704 is specced at 768kHz, and most older multibit chips can be driven at 384kHz, sometimes higher.

First Thanks to everyone who replied. I could see this becoming a huge trend in Audio. Software sales have a much shorter supply chain and inventory investment than hardware sales. Will we see the ASR Filter Plug-In's on sale soon???


Edit: I realize this is a bit off topic. So took out the last portion.

 

[JJB70]

Well, yes, what I read is that a lot of people say, yeah, they measure differently, but there is no audible difference. But in most cases they write this here because they have seen the measurements first, and maybe they have not listened to them.

Also, what makes an excellent DAC? Just the engineering? If a Soekris and a DS DAC sound indistinguishable, then they are both either excellent or not.



Ok, I would not spend money on something just because it measures in a stellar way, if there is no other difference. But who am I to stop somebody from parting from their money?



Absolutely true. And my BS alarm starts to ring as soon as I read the description of stuff like the Synergistic Research Orange fuses. Because what they write is pseudo-scientific jargon. They do not reveal what they are actually doing, and cover it with buzzwords. If they speak like cons (to somebody with a scientific background) then I am going to assume they are cons.



Not sure about this. Also, what competently designed means. This would open not one but several cans of worms. Too many variables.

I think it is quite a salutary experience to do a level matched blind test, ideally it'd be double blind but even if you try and level match two components and then do a blind test switching between them it is eye opening. To be clear, I think speakers and headphones often do have very obvious differences which are immediately apparent (although even there, ease of discerning differences doesn't necessarily mean significantly better/worse listening enjoyment) and amplifiers need to be appropriate for the load but I have honestly found that provided amplifiers are suitable for the load then even pretty cheap PA type amps from Crown, Behringer and similar perform well.

I do think it is possible to discern differences between some electronic components, but my honest opinion is if doing so causes brain overload and results in anxiety if you are doing a blind test and want to identify equipment then it basically say's that even discernible differences are actually irrelevant in terms of using audio equipment for its intended purpose of listening to music (or other material such as spoken word).

In terms of what makes an excellent DAC, I'd say it is one which is well made and audibly transparent. I can understand why people like coloured sound and depending on what I'm listening to I am quite partial to adding to the bass but to me the sensible way to approach audio is to design equipment to be audibly transparent and accurate and then use tools such as EQ and DSP to adjust to your preference if you like colour.

I have no issue with people buying anything, as ultimately we all make our choices. Even though I am not into the sort of audio subjective attitudes and florid prose of some audiophiles, pretty much see audio as simply a tool to enjoy music and think you can get audibly transparent electronics for peanuts I do actually like superior quality and build. I've always had a soft spot for Accuphase and classic statement Japanese audio from the 70's and 80's and love Benchmark equipment. However, the reason I'd buy a Benchmart or Accuphase set up is because I see some beautiful in the stellar build quality, design and just because I like it. What I wouldn't be buying it for is superior SQ as you don't need to spend anything like the admission price for such gear if you just want great sound.

To me competently designed is something designed using sound scientific/engineering principles which performs satisfactorily. Some of it is prosaic stuff like component layout, heat management, product safety etc, stuff which may seem to be a given but which often isn't. To be honest, as an engineer myself I also find it hard to really see anything commendable about adding unnecessary cost and complexity, notwithstanding the fact that I do like high end materials and build quality. Over here we say one of the qualities of an engineer is to do for little cost what most people could do for a lot of money. For example, if Apple can design and build an excellent DAC for $9 I see little reason for products a product like the Mola Mola one to cost $10,000, let along badly designed stuff costing even more.. Premium design and build is one thing, but even with a liking for nice things I really see little to celebrate about stuff which is just extremely expensive.

 

[Esotechnik]

We need to get our technical terms straight in these posts.

Texas ADC chip
1990: PCM1750, 18 bit SAR, 4.5µs min conversion time including s/h
or 222 kHz max capture frequency for sine signals

1993, delta-sigma 'for audio': "PCM1760 and DF1760 combine for a low-cost, high-performance dual 20-bit, 48kHz"
4-bit, 4th order, 64X oversampling ADC
Output Data Delay fs=48kHz 1.5ms
IMHO this is only 667Hz for max amplitude signals!

Another D-S audio ADC is as well slow:
https://www.analog.com/media/ru/training-seminars/tutorials/MT-023.pdf

"the group delay through the AD1871 digital filter is 910 µs (sampling at 48 kSPS) and 460 µs (sampling at 96 kSPS)—this represents the time it takes for a step function input to propagate through one-half the number of taps in the digital filter. The total settling time is therefore approximately twice the group delay time."

1/(2*910µs)=549 Hz

If the recordings from slow DS ADCs do not keep up with the signal, then the sound becomes dull and requires exciter&compressor effects for restore speed.
With low dU/dt, "rotten" recordings, the audible difference between the R-2R and D-S DACs can be negligible.

 

[solderdude]

dU/dt is not the same as latency/group delay.
One can still have good dU/dt yet have a long group delay.

Latency can be a problem in certain cases.

 

[DonH56]

Texas ADC chip
1990: PCM1750, 18 bit SAR, 4.5µs min conversion time including s/h
or 222 kHz max capture frequency for sine signals

1993, delta-sigma 'for audio': "PCM1760 and DF1760 combine for a low-cost, high-performance dual 20-bit, 48kHz"
4-bit, 4th order, 64X oversampling ADC
Output Data Delay fs=48kHz 1.5ms
IMHO this is only 667Hz for max amplitude signals!

Another D-S audio ADC is as well slow:
https://www.analog.com/media/ru/training-seminars/tutorials/MT-023.pdf

"the group delay through the AD1871 digital filter is 910 µs (sampling at 48 kSPS) and 460 µs (sampling at 96 kSPS)—this represents the time it takes for a step function input to propagate through one-half the number of taps in the digital filter. The total settling time is therefore approximately twice the group delay time."

1/(2*910µs)=549 Hz

If the recordings from slow DS ADCs do not keep up with the signal, then the sound becomes dull and requires exciter&compressor effects for restore speed.
With low dU/dt, "rotten" recordings, the audible difference between the R-2R and D-S DACs can be negligible.

Walt is talking about the latency through the digital filter and how it impacts acquisition in a multiplexed system. While he uses the word "settling" he is not talking about the output settling in terms of conventional step response but the time it takes for the filter to fill (or half of it) so the output settles before you can switch (multiplex) to acquire a new input. The context of that section is very important. Input a step and measure the output rise and settling; you'll see a delay, then fast rise and settling.

Latency through the filter is very important in multiplexed systems and whenever time-alignment is required, such as multitrack recording (for audio or instrumentation that depends upon time-alignment of multiple signals).

All the arguments presented to date indicate a lot of reading of papers but fundamental misunderstanding of the technology. Going to back to a technical introduction might help clarify things. A couple of books are a compilation with good intro by Candy and Temes, Oversampling Delta-Sigma Data Converters : Theory, Design, and Simulation, and a more recent (but still fairly old) "continuation" with more practical implications of the theory in Delta-Sigma Data Converters: Theory, Design, and Simulation by Norsworthy, Schreier, and Temes. There may be updates (mine are pretty old -- Dr. Temes was one of my professors many, many years ago) and plenty of other texts but these began it all for me and the fundamentals have not changed.

 

[Esotechnik]

Walt is talking about the latency through the digital filter

Conversion time is critical too.
Internal clock 3-25 MHz in modern D-S DACs is very small.
<0.18um sub-GHz chips is very expensive for audio.

 

[Hayabusa]

Conversion time is critical too.
Internal clock 3-25 MHz in modern D-S DACs is very small.
<0.18um sub-GHz chips is very expensive for audio.
View attachment 45721

conversion time is several orders of magnitude smaller than the filter latency

 

[restorer-john]

...to shakti stones even though the actual sound is the same...

So true Frank. I "upgraded" the crystal in my old soundcard and it made no difference...

 

[DonH56]

So true Frank. I "upgraded" the crystal in my old soundcard and it made no difference...

View attachment 45737

You must experiment with crystals of a different colour... Maybe rose?

 

[Veri]

So true Frank. I "upgraded" the crystal in my old soundcard and it made no difference...

View attachment 45737

Did you properly bath it in himalayan salt, first?