DSD: Explain it to me because I'm not getting it

[audiventory]

What You means by "more fast"?

Native PCM DAC like PCM1704UK? Or something else?

 

Sorry, I wanted say about ADC (analog digital converter).

 

Voltage matrix switch to any voltage momentally. For sigma-delta modulation need up to several inerations for achieving desired voltage.

 

For processing (into devices) wide band telecommunication applications better native PCM format due more optimal for calculations than DSD.

 

For wide band application desirable to minimize sample rate for saving of used resources. For it need maximally fast switch to measured voltage.

 

Thus wide band better capture and processing signal as PCM.

 

For audio applications (narrow band 0 ... 20 kHz) DSD DAC simpler native PCM DAC. No need speed for ADC due huge ratio [useful band]/[sample rate] even for DSD64.

 

I'm not expert in ADC/DAC. I red PCM1704UK's datasheet. I'm not sure that it is fully native PCM DAC. Me seems there some combined method:

 

"The PCM1704 employs an innovative architecture which

combines the advantages of traditional DACs (e.g., excellent

full-scale performance, high signal-to-noise ratio, and ease

of use) with superior low-level performance."

 

"By interleaving

the individual bits of each DAC and employing precision

laser-trimming of resistors, a highly accurate match between

the two DACs is achieved."

 

"traditional DACs" - it is native PCM DAC, as I think.

 

Better or worse it than traditional DAC need learn more.

[Miska]

Thus wide band better capture and processing signal as PCM.

 

Not really. The problem is that you cannot have both high word length and high speed at the same time, because you would need faster settling time to smaller error. As you increase bandwidth, settling time becomes issue and you need to relax how much quantization error you can tolerate. Sigma-delta fixes the problem by not requiring high sample precision to achieve high accuracy. Resolution and speed are inversely proportional in electronics.

 

It's exactly the same as in statistics, you get better result by increasing your sample set instead of trying to increase precision within a small sample set.

[EuroDriver]

Not really. The problem is that you cannot have both high word length and high speed at the same time, because you would need faster settling time to smaller error. As you increase bandwidth, settling time becomes issue and you need to relax how much quantization error you can tolerate. Sigma-delta fixes the problem by not requiring high sample precision to achieve high accuracy. Resolution and speed are inversely proportional in electronics.

 

It's exactly the same as in statistics, you get better result by increasing your sample set instead of trying to increase precision within a small sample set.

 

Miska, thanks for this insight, a Eureka moment for me :-)

[audiventory]

Not really. The problem is that you cannot have both high word length and high speed at the same time, because you would need faster settling time to smaller error. As you increase bandwidth, settling time becomes issue and you need to relax how much quantization error you can tolerate. Sigma-delta fixes the problem by not requiring high sample precision to achieve high accuracy. Resolution and speed are inversely proportional in electronics.

 

It's exactly the same as in statistics, you get better result by increasing your sample set instead of trying to increase precision within a small sample set.

 

Possibility applying DSD instead PCM in practical solutions depend on value of processed band and dynamical range. I said about full range of applying signal processing, not only audio.

 

1. For sigma delta modulation need more reserve of band.

Here very rough calculations:

 

For well known us DSD64 (that almost same by dynamical range with PCM16 in band 0 ... 22 kHz)

effectivity using of band is:

 

Edsd=(1-([sample rate]-[useful band])/[sample rate])*100%=(1-(44100x64-22050)/(44100x64))*100%=1%

 

For 44/PCM16 in band 0 ... 22 kHz reserve is:

 

Epcm=(1-([sample rate]/2)/[sample rate])*100%=50%

 

Thus PCM has effectivity by using sample rate 50%, DSD has effectivity 1% for almost same dynamical range.

DSD64 has better abilities, of course (5-20 dB, may be possible more. Depend on realisation).

 

2. If we increase sample rate, we must increase calculation power due 1 bit modulation processed not simply than multibit modulation.

 

 

Even audio applications (with very narrow useful band) demand significant resources for DSP of DSD signal only due high sample rate.

 

Resume:

Comparing with DSD, PCM processing (for same dynamical range):

1. Possible for more wide band (relatively sample rate value).

2. Less demands in calculation power.

[4est]

The Teac ud 301 is $300 at Crutchfields.

Teac UD-501 $600 - $800

 

(highly regarded by some of my local audiophile friends)

[Hiro]

Resume:

Comparing with DSD, PCM processing (for same dynamical range):

1. Possible for more wide band (relatively sample rate value).

2. Less demands in calculation power.

 

I think that your summary overlooks the aspects of temporal performance. Because interpolation and decimation filters can be avoided with DSD, the DSD signal isn't degraded by rounding, aliasing, and frequency/time-domain errors.

[audiventory]

I think that your summary overlooks the aspects of temporal performance. Because interpolation and decimation filters can be avoided with DSD, the DSD signal isn't degraded by rounding, aliasing, and frequency/time-domain errors.

 

Hi Hiro,

 

Here I said about wide bands - 0.5, 1.0, 10.0, 100.0 MHz and more that used in telecommunications, not audio.

 

This discussion began from comparing speed of DSD and native PCM DAC. And domain of applying last ones.

 

For audio applications not so strong problems

 

If you convert DSD to PCM, degrading of DSD depend on precision of calculations, quality of filters, quality of sigma delta modulator. There are noise, aliases. As usually

 

Hardware decissions strongly (comparing with PC) limited in computing resources.

 

Quality of DSD/PCM (any direction) conversion significantly depend on how successfully developer can manage quantization noise.

 

Best regards,

Yuri

[Jimmypowder]

If you had to guess , how much DSD content is there as a percentage of PCM music?2,5,20 percent ?

Will there be significant increases in the availability of DSD content in the future?

[Jud]

Depends very much on genre. Much greater percentage for classical than others. So let's say 5% overall, and about the same going forward (of course I have no idea if DSD downloads will become more popular, but just throwing out a figure for discussion). I happen to like a lot of classical, so there's much more available to me than I'd ever be able to buy.

 

But even if I didn't care about classical, I'd still want DSD capability. The best mastered versions I've heard of classic albums by artists like the Stones, Who, Dylan, Pink Floyd, Dire Straits, Beach Boys, The Band, Grateful Dead, Rickie Lee Jones, Tom Waits and others are in SACD or DSD format. This seems likely to continue through outlets like Mobile Fidelity, Acoustic Sounds, and CD Japan.

[Bill Lord]

If you convert DSD to PCM, degrading of DSD depend on precision of calculations, quality of filters, quality of sigma delta modulator. There are noise, aliases. As usually

 

Hardware decissions strongly (comparing with PC) limited in computing resources.

 

Quality of DSD/PCM (any direction) conversion significantly depend on how successfully developer can manage quantization noise.

 

Best regards,

Yuri

 

Interesting thread, lots of knowledge and experience.

 

Who/what has the most precise calculations?

 

Who/what has the best filters?

 

Who/what has the best Sigma Delta Modulator?

 

Who/what has the lowest noise?

 

Who/what has the least amount of aliasing?

 

Who/what manages to quantization noise the best?

 

 

It there a software and hardware combination that incorporates all or most winners from each of the above questions?

 

What have read DSD is simpler and because of it's simplicity it has an advantage.

[mordante]

Actually, without being ornery or mean, or even insulting, you are correct. There is a great deal more. But nothing on the planet justifies a $40K DAC, much less a $150K DAC. *Nothing*

 

Now if you have the cash and want to buy one, more power to you. But it is not and will not be better than the $2500-$6000 DAC out there, which represents about as good as a DAC can physically be, given today's technology.

 

And *those* DACs, while better than a $500 DAC, are not "night and day" better - not any where near as impactful on a system as the speakers are. Spending an extra $2K on speakers rather than a DAC can make an enormous difference in a system.

 

And, DSD is a far less expensive way to get the same kind of audio quality, or better, than the best of the PCM only DACs.

 

Maybe you are right maybe you're not. But that is beside the point. You come across like a preacher. I do want to be preached to. So please tone done a bit on your DSD gospel.

 

I guess there is little love for NOS DACs in DSD crowd.

[Miska]

Possibility applying DSD instead PCM in practical solutions depend on value of processed band and dynamical range. I said about full range of applying signal processing, not only audio.

 

You are now ignoring physics related to real world implementation of the A/D and D/A converters and the necessary analog antialiasing and reconstruction filters...

 

For sigma delta modulation need more reserve of band.

 

Naturally, this is on purpose to make the analog side work...

 

Edsd=(1-([sample rate]-[useful band])/[sample rate])*100%=(1-(44100x64-22050)/(44100x64))*100%=1%

 

Now you are ignoring step response performance and looking at frequency domain in a limited way. Useful range of DSD64 is not limited to 22.05 kHz, although it is the focus band. But if you turn to look into step response with analog filters specified for DSD it looks pretty different ...

 

Even audio applications (with very narrow useful band) demand significant resources for DSP of DSD signal only due high sample rate.

 

DSP resources are cheap and very good quality compared to trying to achieve similar in analog domain. Have you ever tried to design a 22.05 kHz analog brickwall filter to reconstruct 44.1k PCM without oversampling? How do you think phase response of such analog filter would look like?

[Miska]

If you had to guess , how much DSD content is there as a percentage of PCM music?2,5,20 percent ?

Will there be significant increases in the availability of DSD content in the future?

 

I would ask the question in a different way. Which format makes best performance out of a DAC? Whatever that format is, we can output any source content format in the format that is optimal for the DAC.

[audiventory]

Interesting thread, lots of knowledge and experience.

 

Who/what has the most precise calculations?

 

Who/what has the best filters?

 

Who/what has the best Sigma Delta Modulator?

 

Who/what has the lowest noise?

 

Who/what has the least amount of aliasing?

 

Who/what manages to quantization noise the best?

 

 

It there a software and hardware combination that incorporates all or most winners from each of the above questions?

 

What have read DSD is simpler and because of it's simplicity it has an advantage.

 

 

> Who/what has the most precise calculations?

 

Only developers know it.

 

 

> Who/what has the best filters?

> Who/what has the best Sigma Delta Modulator?

> Who/what manages to quantization noise the best?

It measured. But also need audio test.

 

> Who/what has the lowest noise?

> Who/what has the least amount of aliasing?

It measured.

 

Archimago (Archimago's Musings) correctly do same tests via PCM->DSD->PCM and RMAA.

[audiventory]

You are now ignoring physics related to real world implementation of the A/D and D/A converters and the necessary analog antialiasing and reconstruction filters...

 

 

 

Naturally, this is on purpose to make the analog side work...

 

 

 

Now you are ignoring step response performance and looking at frequency domain in a limited way. Useful range of DSD64 is not limited to 22.05 kHz, although it is the focus band. But if you turn to look into step response with analog filters specified for DSD it looks pretty different ...

 

 

 

DSP resources are cheap and very good quality compared to trying to achieve similar in analog domain. Have you ever tried to design a 22.05 kHz analog brickwall filter to reconstruct 44.1k PCM without oversampling? How do you think phase response of such analog filter would look like?

 

General spectral energy [useful signal+noise] limited by dynamical range of analog part.

 

As example, we extend useful band (in digit).

 

Need reshape noise (in digit).

 

Reshape noise - more energy of noise due the noise correlated with useful signal (in digit).

 

We can manage noise shape, but we can't manage general value of noise quantization energy.

 

We can limit noise quantization energy if we limit useful signal energy.

 

More energy noise - lower general dynamical range (in analog) due limit of analog part unchanged.

 

We can choose between: more useful band or more dynamic range.

 

I'm not against DSD And I don't like develop 44.1 kHz filters. I like high sample rates! You too?

[Hiro]

I guess there is little love for NOS DACs in DSD crowd.

 

Native DSD DACs playing DSD files operate as NOS DACs.

[Paul R]

Maybe you are right maybe you're not. But that is beside the point. You come across like a preacher. I do want to be preached to. So please tone done a bit on your DSD gospel.

 

I guess there is little love for NOS DACs in DSD crowd.

 

I only preach on Sundays. The point is, however, that at some point, you have to face facts, and those are as close to facts as anyone is going to get. Oh, you might slide the figures a little - perhaps it is a $10K DAC instead of a $6K DAC that marks the point of best technology at this time. But there is no technological advances in a $50K DAC over that - just marketing by some slick fellas trying to make money by taking advantage of other people's honest desire to get the best sound possible. That disgusts me, same as pretty much any other con.

[Miska]

As example, we extend useful band (in digit).

 

My point was that practical limitations of analog implementation need to guide what is done in the digital domain. And the bandwidth is not as simple as having a single digit. Your transition band has substantial impact on your step response...

 

There are many aspects in analog reconstruction and the digital domain approach has huge impact on the net result.

 

I prefer to optimize both frequency and time domain, to get as good step response as possible too. As you can see here:

Squarewaves from DACs - Blogs - Computer Audiophile

Optimizing for both simultaneously is a nice challenge.

[Jimmypowder]

I would ask the question in a different way. Which format makes best performance out of a DAC? Whatever that format is, we can output any source content format in the format that is optimal for the DAC.

 

DSD may have the best performance but if there is a very minuscule amount of listening material ,what good is the format ?

 

To me this is a problem . And it's expensive . Maybe Merdian's format will be better?

 

Give me your thoughts .

[Miska]

DSD may have the best performance but if there is a very minuscule amount of listening material ,what good is the format ?

 

You can convert all the content to that format. I'm listening everything as DSD128 minimum.

 

To me this is a problem . And it's expensive . Maybe Merdian's format will be better?

 

Why would source file format and what goes to the DAC need to have anything to do with each other?

[bmoura]

DSD may have the best performance but if there is a very minuscule amount of listening material ,what good is the format ?

 

To me this is a problem . And it's expensive . Maybe Merdian's format will be better?

 

Give me your thoughts.

 

Find HD Music, which lists DSD Downloads from many (but not all) of the DSD Music sites shows over 1,300 Stereo DSD Downloads and over 700 Multichannel DSD Downloads to date. A nice start.

 

As for Meridian's MQA which encodes music to fit in a "CD Quality" space or less for streaming sites and then decodes it at the listener's end, it's certainly not in the same league as DSD when it comes to sound quality based on my listening to decoded MQA sample music tracks.

 

Then again, Meridian is primarily promoting MQA encoding and decoding as more of a streaming solution that reduces bandwidth requirements than a higher quality music format. It may catch on in the streaming music world - even though you will need to buy a new DAC with MQA circuitry to hear MQA decoded.

[Jimmypowder]

You can convert all the content to that format. I'm listening everything as DSD128 minimum.

 

 

 

Why would source file format and what goes to the DAC need to have anything to do with each other?

 

So upsampling a red book Pam file to DSD , somehow improves the sound quality ? Or increases the signal noise?

 

I have no doubt some dacs sound better upsampling pcm to DSD ,my ifi micro idsd for example .

 

But upsampling on my Benchmark 2 hgc to DSD does not sound better through that dac. It sounds worse.

 

What programs converts PCm to DSD ? You mean upsampling right?

[Miska]

So upsampling a red book Pam file to DSD , somehow improves the sound quality ?

 

Upsampling to DSD or to higher PCM rate.

 

But upsampling on my Benchmark 2 hgc to DSD does not sound better through that dac. It sounds worse.

 

Unfortunately it is limited to DSD64. It could still benefit from upsampling to highest PCM rate it supports.

 

The Sabre (same chip family used on Benchmark DAC2) based DACs I have, Mytek Stereo192-DSD, exaSound e28 and Resonessence Labs HERUS sound better to me with DSD upsampling. exaSound and HERUS can also benefit from 352.8k rates (and 384k on exaSound).

[Paul R]

Upsampling to DSD or to higher PCM rate.

 

 

 

Unfortunately it is limited to DSD64. It could still benefit from upsampling to highest PCM rate it supports.

 

The Sabre (same chip family used on Benchmark DAC2) based DACs I have, Mytek Stereo192-DSD, exaSound e28 and Resonessence Labs HERUS sound better to me with DSD upsampling. exaSound and HERUS can also benefit from 352.8k rates (and 384k on exaSound).

 

Benchmark is a bit of an odd duck there, as they already up-sample everything inbound to the DAC to some high rate, but last I looked that rate was less than 192K and way less than DSD data rates. That's probably changed a bit on the Benchmark 2, but given Benchmarks insistence that DSD is a poor format compared to PCM, well - who knows? Here is what Benchmark has to say about DSD in the Benchmark 2 -

 

All digital inputs on the DAC2 D support native DSD conversion. DoP 1.1 is automatically detected on all digital inputs, and the system seamlessly switches between native PCM and native DSD conversion.

 

All of which amounts to the Benchmark soundingly ruthless clear on PCM, and middling on DS64. I think that was more the engineers in Syracuse being stubborn because marketing forced them to include DSD and they just.did.not.want.to.do.that. Of course, that is guesswork on my part, but I would not be surprised to have it confirmed by someone more in the know than I am.

[Jud]

So upsampling a red book Pam file to DSD , somehow improves the sound quality ? Or increases the signal noise?

 

I have no doubt some dacs sound better upsampling pcm to DSD ,my ifi micro idsd for example .

 

But upsampling on my Benchmark 2 hgc to DSD does not sound better through that dac. It sounds worse.

 

What programs converts PCm to DSD ? You mean upsampling right?

 

What software are you using to do the conversion?