Digital that sounds like analog

DonH50 · Dec 5, 2012

Last I checked, fairly recently (past few months), TI/BB and ADI both had conventional DACs in their audio line, or at least DAC at audio rates. Most if not all of the highly-integrated chips, the ones most used in commercial products, are DS, however, so very few may be using the older designs. I did not check to see what companies like Sabre or Wolfson are using. Crystal (now part of Cirrus, I think?) makes DS and those are widely used in computer sound cards and such.

DS DACs still glitch but the impact is much, much less due to their higher speed and fact that only one level is involved so no change of signal modulation. However, ISI can be an issue, leading to all sorts of other problems. Modern multi-bit DS designs could conceivably suffer from the same sorts of issues other DAC architectures have with glitches, however.

Disclaimer: I have not thought much about this at all; please consider my comments off-the-cuff, IMO, etc. - Don

p.s. On what is the biggest problem with digital today, I am not really sure. I think a lot of it is perception and the fact that people like the sound of e.g. LPs even though they are technically far inferior. Perhaps more effort should be expended on figuring out why, i.e. what it is about that sound that people like, and less on seeking perfection. The perfection is pretty much there by most measurements; what remains is to find out why they do not sound perfect to people. Had I plenty of time and money to play with, that is where I would start. Actually, opus11 is making strides in that direction...

opus111 · Dec 5, 2012

DonH50 said:
Last I checked, fairly recently (past few months), TI/BB and ADI both had conventional DACs in their audio line, or at least DAC at audio rates.

They do but they're disappearing rapidly. From TI, the only audio conventional (multibit) one left in production is the PCM1704 and I don't think that one is staying primarily because of demand in audio. ADI used to have AD1865 up until a few months ago, that's gone obsolete now. ADI recommend a D-S part as replacement for this.

Most if not all of the highly-integrated chips, the ones most used in commercial products, are DS, however, so very few may be using the older designs. I did not check to see what companies like Sabre or Wolfson are using. Crystal (now part of Cirrus, I think?) makes DS and those are widely used in computer sound cards and such.

When I was referring to the older designs I meant multibit ones, not DS. Both Sabre (ESS) and Wolfson only make DS, as does Cirrus. When they were Crystal though, and this is going back to the 80s, Cirrus made some really nice multibit ADCs. I still have a couple of their 1MHz, 12bit parts in my box which I'm tempted to press into audio duty together with some DSP

DS DACs still glitch but the impact is much, much less due to their higher speed and fact that only one level is involved so no change of signal modulation. However, ISI can be an issue, leading to all sorts of other problems. Modern multi-bit DS designs could conceivably suffer from the same sorts of issues other DAC architectures have with glitches, however.

The multibit SDs don't have R2R architecture so that's one big way they have lower glitch. They glitch in rather different ways in that they're pure CMOS so are crowbarring the supply on every clock edge - the glitches are on the power supply when then feeds through to the output.

p.s. On what is the biggest problem with digital today, I am not really sure. I think a lot of it is perception and the fact that people like the sound of e.g. LPs even though they are technically far inferior.

I'm not one of those who likes the sound of LP - I don't have any, sold off years ago. When CD came out I didn't notice the noise modulation imperfections I notice now, but I did love the improvement it gave me on piano over LP. I found LP just couldn't do the dynamics of piano well enough. LP lovers bear in mind my TT wasn't an expensive one (Rega Planar 3). The best piano LP I ever owned in terms of SQ was a Nimbus direct-to-disk of Bernard Roberts playing the Diabelli Variations, however my TT/arm/cartridge could not track this without gross distortion. So I really appreciated CD's freedom from hash on piano.

My take is that vinyl lovers love it for its significantly lower noise modulation than digital. For which we still do not have standardized measurements. But absence of measurements is no indication than an effect doesn't exist

opus111 · Dec 5, 2012

What does noise modulation look like on an FFT?

A quick summary of where we are...

I've bleated on about noise modulation in digital and mentioned that its also an issue with analog. Here are a couple of FFT plots to demonstrate how it looks on a tape machine. The first result is a truly excellent one for an analog tape machine, a Studer. The second is decidedly so-so in comparison, this is an Otari. The stimulus here is an approximate 1kHz single sine.

opus111 · Dec 5, 2012

Tape machines are an interesting example of where apparently rather bad measurements still sound good so its interesting to dig somewwhat deeper into what's going on - what really matters in the measurements, in relation to SQ?

Both machines show substantial (by digital standards) THD - the Studer's 3rd harmonic distortion is nearly 0.2%, the Otari's 2nd is nudging 3%. But i didn't post these to talk about THD, that's old hat. Rather the 'skirts' around the base of the tones here show noise modulation artifacts. Its not only the input tone that attracts noise modulation but also the distortion tones added by the machine itself. There's also the issue of LF noise modulation that doesn't seem to be related to any input tone at all, the Otari being by far the worst offender.

The difference I believe is in relation to where the noise modulation occurs, frequency-wise. Apart from the LF, the noise is being added close-in to the tone, hence is very likely masked by our brain/ear perceptual processing. With digital though I have the hypothesis that the nose modulation is much broader band and hence not subject to masking.

JackD201 · Dec 6, 2012

I'd just like to thank Don and Opus for the clarifications. Thanks again guys.

DonH50 · Dec 6, 2012

Thanks Jack.

opus11 -- Hmmm... Sampling will fold everything into baseband, but that happens in the ADC. The DAC's output is broadband. It may be that you discover the limitation is on the source (ADC, recording) side, alas.

opus111 · Dec 6, 2012

Well I'd not see that occurence as an 'alas' - rather an opportunity to sell the studio a better ADC

With the current level of transparency of the DAC its not too hard to hear noise modulation on some recordings. Chesky Records in particular have an ADC which adds artifacts, particularly noticeable over the voice of the announcer on 'The Ultimate Demonstration Disc'. Which is rather a sweet irony given how they market themselves as an 'audiophile' label

LL21 · Dec 6, 2012

opus111 said:
Tape machines are an interesting example of where apparently rather bad measurements still sound good so its interesting to dig somewwhat deeper into what's going on - what really matters in the measurements, in relation to SQ?

Both machines show substantial (by digital standards) THD - the Studer's 3rd harmonic distortion is nearly 0.2%, the Otari's 2nd is nudging 3%. But i didn't post these to talk about THD, that's old hat. Rather the 'skirts' around the base of the tones here show noise modulation artifacts. Its not only the input tone that attracts noise modulation but also the distortion tones added by the machine itself. There's also the issue of LF noise modulation that doesn't seem to be related to any input tone at all, the Otari being by far the worst offender.

The difference I believe is in relation to where the noise modulation occurs, frequency-wise. Apart from the LF, the noise is being added close-in to the tone, hence is very likely masked by our brain/ear perceptual processing. With digital though I have the hypothesis that the nose modulation is much broader band and hence not subject to masking.

Quite cool...seems like you have really started to dissect the problem...i need to re-read this, but thanks for continuing to share an ongoing narrative of your dissection!

LL21 · Dec 6, 2012

opus111 said:
They do but they're disappearing rapidly. From TI, the only audio conventional (multibit) one left in production is the PCM1704 and I don't think that one is staying primarily because of demand in audio. ADI used to have AD1865 up until a few months ago, that's gone obsolete now. ADI recommend a D-S part as replacement for this.

i think i already asked you this...but what do you think of the AD1865? i can see your positive views on the TDA1541 Double Crown. i was curious because i have now read 3 xtremely positive reviews of the Audio Note DAC Fifth Element ($100K!!!!!) which uses this chip. Martin Colloms suggested CD might have been 'hi-res' all along based on this DAC's performance...and its just taken 30 years to get it off the CD.

JackD201 · Dec 6, 2012

The most "analog sounding" all solid state DACS I've heard and own (lacking the strained clipping type of artifacts) have output voltages higher than the norm: 3v and up, some as high as 6v.

Could there be a correlation?

LL21 · Dec 6, 2012

JackD201 said:
The most "analog sounding" all solid state DACS I've heard and own (lacking the strained clipping type of artifacts) have output voltages higher than the norm: 3v and up, some as high as 6v.

Could there be a correlation?

interesting...Zanden is 1v output though [i think] Roy Gregory felt sure it was higher for some reason.

JackD201 · Dec 6, 2012

Actually I had the 5000 in mind when I expressly stated "all solid state DACs". The Zanden is very analog sounding but I suspect that the tube section along with the analog filters is doing the smoothing out of the nasties.

opus111 · Dec 6, 2012

lloydelee21 said:
i think i already asked you this...but what do you think of the AD1865?

I haven't heard any DAC with the AD1865 inside, but hey its multibit so probably it sounds just great. Its gone obsolete now so perhaps its headed for even more fame and glory now its 'dead'

i can see your positive views on the TDA1541 Double Crown.

This chip is an interesting case where the measurements and the sound do coincide - the TDA1541A is the best measuring and best sounding 16bit DAC available. I'd like to know what the double crown brings to the party, I know its selected but not totally clear what measurements they did to select them coz the datasheet mentions DNL which is a static measurement.

I put the TDA1541A's superiority down to two aspects of its design. First its current segment, so doesn't have the glitching problems of the R2R parts, second its internal circuitry is bipolar hence none of the CMOS (or BiCMOS) glitching of the power supplies.

i was curious because i have now read 3 xtremely positive reviews of the Audio Note DAC Fifth Element ($100K!!!!!) which uses this chip.

At that price its hard for me to imagine that placebo didn't play a part in the perceived sound.

Martin Colloms suggested CD might have been 'hi-res' all along based on this DAC's performance...and its just taken 30 years to get it off the CD.

I do agree with Mr Colloms on this point - its not about the CD format being flawed at all, rather its the implementations to date which have sucked. Which I plan to remedy, but at a slightly more realistic price point

LL21 · Dec 6, 2012

opus111 said:
I haven't heard any DAC with the AD1865 inside, but hey its multibit so probably it sounds just great. Its gone obsolete now so perhaps its headed for even more fame and glory now its 'dead'

This chip is an interesting case where the measurements and the sound do coincide - the TDA1541A is the best measuring and best sounding 16bit DAC available. I'd like to know what the double crown brings to the party, I know its selected but not totally clear what measurements they did to select them coz the datasheet mentions DNL which is a static measurement.

I put the TDA1541A's superiority down to two aspects of its design. First its current segment, so doesn't have the glitching problems of the R2R parts, second its internal circuitry is bipolar hence none of the CMOS (or BiCMOS) glitching of the power supplies.

At that price its hard for me to imagine that placebo didn't play a part in the perceived sound.

I do agree with Mr Colloms on this point - its not about the CD format being flawed at all, rather its the implementations to date which have sucked. Which I plan to remedy, but at a slightly more realistic price point

Thanks, Opus111! Always like to read your posts...and cannot wait to hear your DAC someday! Until then, happy with my Zanden...

Bruce B · Dec 6, 2012

opus111 said:
Well I'd not see that occurence as an 'alas' - rather an opportunity to sell the studio a better ADC With the current level of transparency of the DAC its not too hard to hear noise modulation on some recordings. Chesky Records in particular have an ADC which adds artifacts, particularly noticeable over the voice of the announcer on 'The Ultimate Demonstration Disc'. Which is rather a sweet irony given how they market themselves as an 'audiophile' label

Well that's dCS for you. Just like some of the new stuff that Sterling is doing for them. They are using 15yr old dCS converters. I had them in here for a year and never liked them.

opus111 · Dec 6, 2012

I'm not claiming that its performance will beat your Zanden Lloyd, as the TDA1541A is a seriously tough act to follow and I'm using CMOS DAC chips with all kinds of glitching. But I do know my DAC will be at least 20dB cheaper.

opus111 · Dec 6, 2012

Bruce B said:
Well that's dCS for you. Just like some of the new stuff that Sterling is doing for them. They are using 15yr old dCS converters. I had them in here for a year and never liked them.

They use dCS now? They started out with Bob Adams' design which was originally dbx and made a big song and dance about how its low level linearity performance was better than multibit in those days (it was too, no kidding). That was around 1990 or so - when did they change?

Kindhornman · Dec 6, 2012

Somehow I remember Chesky using Meridian and Cello as their main components. But that was so long ago perhaps everything has changed. I am now curious enough to open up my Sony ES CDP to see what the chips are. I always thought that it was multi-bit, I will let you know. It is one of the better commercial CDP's that I have heard compared to most consumer units. It is an X33ES model.

opus111 · Dec 6, 2012

Bob Stuart at Meridian has long used S-D type DACs so no surprise Chesky can't hear the noise modulation that their ADC introduces.

A quick bit of Googling reveals your player has the CXD2552 DAC from Sony which is a 1bit device so you're outta luck

Looks like you should have bought the CDP-337ESD instead

Kindhornman · Dec 6, 2012

I guess there is no DAC rolling on that Sony then. But it does sound better than my Denon by far.

Digital that sounds like analog

Member Sponsor & WBF Technical Expert

Banned

Banned

Banned

WBF Founding Member

Member Sponsor & WBF Technical Expert

Banned

Well-Known Member

Well-Known Member

WBF Founding Member

Well-Known Member

WBF Founding Member

Banned

Well-Known Member

WBF Founding Member, Pro Audio Production Member

Banned

Banned

New Member

Banned

New Member

Similar threads