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

[Bluespower]

The amplitude errors appear as noise - not as inaccuracies in the signal. This noise is shaped, giving greater than 20 bits of resolution in the audio band.

to you there is not any advantage of multibit ? I think 24 bit 96 hz pcm in nos'mutibit is the most accurate cause you don't do anything to the bits and you can filter more easily. And my'ears also confirm that and no tests can confirm'that cause test are only to show good aspects of delta sigma dacs.

 

[andreasmaaan]

to you there is not any advantage of multibit ? I think 24 bit 96 hz pcm in nos'mutibit is the most accurate.

No, I’m happy with 16/44.1 PCM, although I think an argument can be made for 24/88.2 or 24/96.

Objectively speaking, NOS is not the most accurate. The amplitude response isn’t flat to 20KHz (I think even with a 96KHz sample rate, although I haven’t investigated this), and the lack of an anti-imaging filter introduces a (relatively) high level of intermodulation distortion in the audio band. Both of these could be audible depending on the DAC.

I think both R2R and DS DACs can be accurate enough to introduce zero audible distortion. I prefer DS because it can be done far more cheaply. An accurate R2R DAC requires very tightly matched resistors, which costs a lot of money, and even then, performance isn’t quite as good as a better DS DAC.

I’m not an expert though

 

[Bluespower]

Well thank you all for sharing your point of view. To me tests that are usually done are too'simple and they only show'the good aspect 'of ds dacs and not the bad aspects. Then this is not objectif to my mind. It s marketting to'pretend 32 bits on dac and a low noise floor doesnt mean that every level of the tones in real music is accurate.

No, I’m happy with 16/44.1 PCM, although I think an argument can be made for 24/88.2 or 24/96.

Objectively speaking, NOS is not the most accurate. The amplitude response isn’t flat to 20KHz (I think even with a 96KHz sample rate, although I haven’t investigated this), and the lack of an anti-imaging filter introduces a (relatively) high level of intermodulation distortion in the audio band. Both of these could be audible depending on the DAC.

I think both R2R and DS DACs can be accurate enough to introduce zero audible distortion. I prefer DS because it can be done far more cheaply. An accurate R2R DAC requires very tightly matched resistors, which costs a lot of money, and even then, performance isn’t quite as good as a better DS DAC.

I’m not an expert though

But you should be fair and admit that ds has not the same,precision. It makes more approximations in levels. And no test shows that and it seems nobody want to accept that here. Ds is more accurate in frequency. R2r is more accurate in level. I think most recent cds have too much compression to hear it. Also NOS doesn,t suffer from intersample distortion which might be worse that imd. Well let's stop if nobody want to think on it and just say i'm wrong. Anyway thank'for taking time to answer. Tests are selected to give good results on delta sigma dacs. And filter arz selected to give good results on these test and often people,prefer filters that have less good'results, that mean some test should show why some,peaple'prefere filter with less good results. What's the point then of'making test that don't show what people,like? Good result to,make money but not to have better quality.

 

[andreasmaaan]

But you should be fair and admit that ds has not the same,precision. It makes more approximations in levels.

The thing I don’t think you’re appreciating is that these inaccuracies are pushed out of the audio band / below the noise floor by noise shaping.

I don’t have the math ability to explain it, but noise shaping means that above the noise floor, a perfect DS DACs is 100% precise.

Of course no DAC is perfect in reality, but in practice the best DS DACs are are more precise than the best R2R DACs because the architecture means that a lower level of precision in the components can produce a higher level of precision in the output.

Tests are selected to give good results on delta sigma dacs.

Why then do R2R DAC manufacturers or proponents not provide their own alternative tests that show up the alleged disadvantages of DS DACs?

Surely at least one R2R manufacturer or enthusiast would by now have done the tests that show this? The reason they haven’t is that they can’t

 

[March Audio]

Well thank you all for sharing your point of view. To me tests that are usually done are too'simple and they only show'the good aspect 'of ds dacs and not the bad aspects. Then this is not objectif to my mind. It s marketting to'pretend 32 bits on dac and a low noise floor doesnt mean that every level of the tones in real music is accurate.
But you should be fair and admit that ds has not the same,precision. It makes more approximations in levels. And no test shows that and it seems nobody want to accept that here. Ds is more accurate in frequency. R2r is more accurate in level. I think most recent cds have too much compression to hear it. Also NOS doesn,t suffer from intersample distortion which might be worse that imd. Well let's stop if nobody want to think on it and just say i'm wrong. Anyway thank'for taking time to answer. Tests are selected to give good results on delta sigma dacs. And filter arz selected to give good results on these test and often people,prefer filters that have less good'results, that mean some test should show why some,peaple'prefere filter with less good results. What's the point then of'making test that don't show what people,like? Good result to,make money but not to have better quality.

I'm afraid this isn't true. Level linearity is often better in DS. Amir measures level linearity as part of his tests and good DS designs are essentially perfect down to his measurement limit of - 120dB

 

[Bluespower]

I'm afraid this isn't true. Level linearity is often better in DS. Amir measures level linearity as part of his tests and good DS designs are essentially perfect down to his measurement limit of - 120dB

It's doesn't proove anything to test levels one after the other. It's not the same as Testing one complex wave with multi tone with'multi,level harmonics

 

[Bluespower]

What are you expecting to see?

Well for delta sigma'it's much more difficult to render micro details and high details at the same,time. The test only shows that it can render them separatly. No records have song with levels at -10db or -20db so it shows nothing. But on music you can have harmonics at -40db while the global,level is at -2db.

 

[March Audio]

Well for delta sigma'it's much more difficult to render micro details and high details at the same,time. The test only shows that it can render them separatly.

on what basis do you make that claim? I can demonstrate it doing high level and low level signals simultaneously quite happily.

BTW I have a Soekris R2R dac sat in the room right now. Shall we do a comparison?

 

[Bluespower]

on what basis do you make that claim? I can demonstrate it doing high level and low level signals simultaneously quite happily.

Ok for a demonstration that's exactly what i m looking for. A test that show that ds can do that properly. And maybe experimenting at wich difference in level between the tones there will be some loss. That is what i ask from the beginning

 

[March Audio]

OK, I know what the result will be but it's only fair to demonstrate.

I will endeavor to do this over the weekend. As a heads up, the R2R will struggle due to the high harmonic distortion which is directly related to IMD. Low level details will get obscured by spuria

 

[Bluespower]

Some'article that explain that dsd and pure'delta sigma'converters loose resolution at high frequencies. That'is a drawback of them.

https://support.auralic.com/hc/en-us/articles/206806457-What-s-the-Difference-Between-PCM-and-DSD-

 

[DonH56]

Sounds like marketing FUD to me. All data converters lose resolution at high frequencies. For audio I don't see delta-sigma designs particularly worse than any others. At RF it is hard to sample rapidly enough to provide sufficient in-band noise-shaping for broadband delta-sigma designs, and you need high-Q resonators for narrowband (bandpass) delta-sigma designs. Conventional designs have other issues at high frequencies but you'd have to tell me what "high" means. I have designed and worked with data converters sampling from audio to many GHz and they all have their challenges.

 

[Bluespower]

A quote:
Instead DSD leads to constant high levels of noise at these frequencies. The dynamic range of DSD decreases quickly at frequencies over 20 kHz due to the use of strong noise shaping techniques which push the noise out of the audio band resulting in a rising noise floor just above 20 kHz. The dynamic range of PCM, on the other hand, is the same at all frequencies.

That means that good resolution below 20khz is articial from strong noise shaping techniques. Then if noise rise above 20khz it's like a non oversampling dac. But filters in dsd are usually way above 20khz. I guess some people hear noise shaping effect and that also this hf stuff can interact with amp and speakers and resonnate at hearable levels.

 

[andreasmaaan]

That means that good resolution below 20khz is articial from strong noise shaping techniques.

Trying to distinguish between “natural” and “artificial” makes no sense here. The question is: how accurately does the device reproduce the signal within the audio band? Whether you get there through noise shaping or some other method makes no difference if the end result is the same. You don’t hear the process, you hear the outcome.

 

[Bluespower]

Trying to distinguish between “natural” and “artificial” makes no sense here. The question is: how accurately does the device reproduce the signal within the audio band? Whether you get there through noise shaping or some other method makes no difference if the end result is the same. You don’t hear the process, you hear the outcome.

For example on tv if you watch a vhs on old tube cathodic tv it renders more natural than on a upscaling full hd tv. The full hd has sharpness filter and upscaling that shows lots of detail but you can feel something not natural on the picture. Do you think the same can apply whith noise shaping in audio?

 

[andreasmaaan]

For example on tv if you watch a vhs on old tube cathodic tv it renders more natural than on a upscaling full hd tv. The full hd has sharpness filter and upscaling that shows lots of detail but you can feel something not natural on the picture. Do you think the same can apply whith noise shaping in audio?

No, they can’t be compared, it just doesn’t work the same way it does with TV (although I know little about the latter).

 

[Bluespower]

I wanted to add that with dacs that don't have digital volume you cannot avoid intersample distortion from spdif. (Except some dacs that have headroom). Only nos dac doesn't have intersample distortion. Then wich is better ? No intersample distortion or no imd distortion? On usb it's obvious the difference when you set -3db in foobar on record loudly mastered.

 

[vitalii427]

No, they can’t be compared, it just doesn’t work the same way it does with TV (although I know little about the latter).

Actually it does pretty the same to me. I have few CRT TVs and Pro Video Monitors for retro gaming. And it is lot like LP and tube amps with lots of distortions and it quite pleasing but not High Fidelity.

 

[Veri]

I wanted to add that with dacs that don't have digital volume you cannot avoid intersample distortion from spdif.

why's that?

 

[Bluespower]

Actually it does pretty the same to me. I have few CRT TVs and Pro Video Monitors for retro gaming. And it is lot like LP and tube amps with lots of distortions and it quite pleasing but not High Fidelity.

Some people prefer good sounding low fi than bad sounding hifi