Jump to content


Photo
* * * - - 4 votes

Mutibit vs Delta Sigma (old vs new)


  • This topic is locked This topic is locked
306 replies to this topic

#151 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 01 April 2012 - 10:18 AM

I dont agree. I have had great results with most of the S-D chips from AD and TI/BB. The trick is to use the right power delivery to them.


Sounds plausible. So far I haven't experimented with supplies any more exotic than cheap series and shunt ICs. Then can
you explain why, given that I use the same power delivery arrangements in both S-D and multibit, that the multibit has the
jump factor and the S-D does not? What might it be about multibit that makes it immune to PSU regulation and decoupling?

#152 zenelectro

zenelectro

    Member

  • Members
  • 355 posts
  • Location: NSW

Posted 01 April 2012 - 10:28 AM

I dont agree. I have had great results with most of the S-D chips from AD and TI/BB. The trick is to use the right power delivery to them. It's similar a lot of op-amps. If you dont do the right power delivery (regulation and decoupling caps etc.), then they dont perform. Most designers dont get this right, and as a result, the audio is flat and uninvolving, not dynamic. No jump factor. I modded a LOT of DACs in the 10 years before I started designing my own, so I know what effect this has on otherwise mediocre DACs.

The ENTIRE power subsystem, not just the power supply, must be optimized to achieve jump factor. What you must understand is that even the best voltage regulators and power supplies dont respond to load changes the same at all frequencies. It is the understanding of this that allows you to "tune" the regulation speed and decoupling cap locations and sizes so that the response to load changes (current) is uniform across the entire audio frequency range. The use of proper connectors, PC board layout practices and cable design also plays a major role in this.

Steve N.


WRT S-D versus multibit these days it's just not that easy to separate them. All S-D DAC's now -are- multibit and some have quite a few
actual physical bit levels before factoring in the PWM and noise shaping. Sabre is a good example.

Absolutely agree that SD DAC's impose a much more difficult load on the power supplies. Not to mention that all this noise gets spread
to other parts of the DAC.

Further to this SD DAC's usually are working at much higher clocking rates. Most people run the Sabre at 100MHz. Compared to the lazy old
TDA1541 running at 2.8Mhz and it's internal current mode logic we have two entirely different beasts.

The other curved ball is that we also have to consider the types of distortion profile of the DAC itself. My hunch is that multibit have a
much more benign distortion profile that is more musical and pleasing. This is confirmed by the fact that people actually 'tune' the sound
of the 1541 by adjusting the size of the I-V resistor. Larger = more musical and richer, smaller = more accurate. Somewhere in between
is a good balance. Needless to say, the larger the I-V resistor, the more OP swing of the DAC = more distortion.

WRT opamps, it's funny, if you look at the data sheets, opamps are pretty much infallable! :D In reality they are a long way from it.
If you haven't read it yet, I recommend the work by Samuel Groner on opamp measurements, very interesting and shows up their very
real weaknesses.

WRT regulators, Ultimately it's pretty much a juggling act between open loop gain versus frequency extension of the regulator.

For a long time, people got obsessed with super low output impedance (jung super reg etc)
and forgot about flat OP impedance versus frequency and good transient response.

Going the next step, if you take that philosophy and apply it to the opamps themselves, you'll throw them in the bin and design
either 1) an open loop SS circuit or 2) a good tube based circuit.

IME, once you do a decent implementation of either, it's hard to go back to using opamps.

But there are plenty of highly regarded designs out there that use them, so as they say YMMV.

#153 Telecine

Telecine

    Member

  • Members
  • 4,824 posts
  • Location: QLD
  • Country: Australia

Posted 01 April 2012 - 10:32 AM

Ah i didn't phrase my question well enough - apologies. I know what AV meant but I could not figure out why that world would be different from the general audio world where DACs are concerned. The 'Shark' would be ADI's 'Sharc' floating point DSP chip - they also make great DACs (or rather, they did make great DACs). Nowadays they've almost totally gone over to S-D with the exception of the AD1865 (I think).


The Lampi does use a S-D chip.

Edited by Telecine, 01 April 2012 - 10:36 AM.

Learn from the mistakes of others - you'll not live long enough to make them all yourself!

#154 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 01 April 2012 - 10:51 AM

WRT S-D versus multibit these days it's just not that easy to separate them. All S-D DAC's now -are- multibit and some have quite a few
actual physical bit levels before factoring in the PWM and noise shaping. Sabre is a good example.


Which S-D chips these days use PWM? I haven't seen any myself. Sabre doesn't, to my knowledge. I
think the new NAD being talked about here might use PWM.

When I say multibit I mean a DAC which hasn't used noise shaping and truncation. So even the venerable
TDA1540 counts as multibit (despite being just 14bits) whereas Sabre (which uses 6) does not. So the distinction
does remain clear to me at least.

Further to this SD DAC's usually are working at much higher clocking rates. Most people run the Sabre at 100MHz. Compared to the lazy old
TDA1541 running at 2.8Mhz and it's internal current mode logic we have two entirely different beasts.


Agreed - Sabre almost guarantees the designer will need a groundplane. With TDA1541 this isn't essential. So
how do high speed decoupling requirements translate down to audio? If, as SteveN implies its about keeping the
supply series impedance flat across the audio range then how is this affected by the impedance presented to the
DAC chip at RF?

The other curved ball is that we also have to consider the types of distortion profile of the DAC itself. My hunch is that multibit have a
much more benign distortion profile that is more musical and pleasing. This is confirmed by the fact that people actually 'tune' the sound
of the 1541 by adjusting the size of the I-V resistor. Larger = more musical and richer, smaller = more accurate.


Ah, we differ on whether 'musical' and 'accurate' are polar opposites. To me they're definitely not. Transparent to me means accurate
and transparent to my ears is also musical. Having a larger I/V resistor will tend to increase the low order distortion (2nd mainly).
Its not true (from the DACs I've examined) that multibit in general has a more benign THD profile - take a look at the recent review by
Colloms of the Metrum DAC as an example - I've never seen an S-D looking quite as aggressive as that.

WRT opamps, it's funny, if you look at the data sheets, opamps are pretty much infallable! :D In reality they are a long way from it.
If you haven't read it yet, I recommend the work by Samuel Groner on opamp measurements, very interesting and shows up their very
real weaknesses.


+1 for Samuel's paper. Very informative. But in my experience not that relevant to sound quality which isn't really strongly correlated
with those measurements.

#155 Arg

Arg

    The Argentinian

  • Members
  • 949 posts
  • Location: SA
  • Country: Australia

Posted 01 April 2012 - 12:23 PM

.... No jump factor. ...


Is 'jump factor' the new PRAT? A vague and imprecise term that seems misleadingly precise, until you realise it's just an excuse for subjectivist reviewers to say something unmeasurable, so it gives them high ground to stand in judgement of any equipment without being disprovable.

Let's say one person says it has 'jump factor' and another says it has no 'jump factor'. Well, does it have it or not?

#156 zenelectro

zenelectro

    Member

  • Members
  • 355 posts
  • Location: NSW

Posted 01 April 2012 - 01:07 PM

Which S-D chips these days use PWM? I haven't seen any myself. Sabre doesn't, to my knowledge. I
think the new NAD being talked about here might use PWM.

Apologies, I should have wrote PDM. Late night - and a few beers :)

When I say multibit I mean a DAC which hasn't used noise shaping and truncation. So even the venerable
TDA1540 counts as multibit (despite being just 14bits) whereas Sabre (which uses 6) does not. So the distinction
does remain clear to me at least.

Yes I understand where you are coming from and there are many with your perspective on this.
http://www.esstech.com/PDF/sabrewp.pdf
However from the white paper and from discussions at DIA, with Dustin Formin one of the designers, they have put
a huge amount of effort in overcoming traditional problems with the modulators. For example noise floor modulation
etc.
The biggest frustration with Sabre for me is they dumped an ASRC in the chip aswell. I would have loved to see that as
a separate chip and keep the DAC as simple as possible. Apparently they did consider this.

Agreed - Sabre almost guarantees the designer will need a groundplane. With TDA1541 this isn't essential. So
how do high speed decoupling requirements translate down to audio?

This is the million dollar question. Objectively or subjectively??
I know what works and sounds good subjectively, I know what is the correct way to do the design from an RF design perspective,
Somewhere in between is the answer. This is the problem of audio.

If, as SteveN implies its about keeping the supply series impedance flat across the audio range then how is this affected by the impedance presented to the
DAC chip at RF?

I don't know that Steve was implying this. But they are separate issues, power supplies for opamps and in the Sabres case supplies for
analog OP stage, digital stages etc. Personally I would treat them differently.

Ah, we differ on whether 'musical' and 'accurate' are polar opposites. To me they're definitely not. Transparent to me means accurate
and transparent to my ears is also musical.

Yes, if you look throujgh this forum and all others you will find years of 'colorful discussion' on this subject.

I used to have exactly your view for many years. However these days, after many years of building stuff for recording and playback
I have changed my views to a certain degree. Your view applies to a perfect world with pefect studios and perfect playback.
I haven't found that perfect world yet. In my engineering heart I think -exactly- like you do. Unfortunately the reality of life's
experience has been different and this conflict has many times made me even want to give up audio design altogether.

Having a larger I/V resistor will tend to increase the low order distortion (2nd mainly).
Its not true (from the DACs I've examined) that multibit in general has a more benign THD profile - take a look at the recent review by
Colloms of the Metrum DAC as an example - I've never seen an S-D looking quite as aggressive as that.

Haven't seen the review or measurements. In fact does anyone know what DAC they are using?
I was referring to R2R dacs like PCM63/1704 etc that always have increased low order distrtion with OP voltage swing.
Subjectively they sound richer / less precise with more voltage swing.

+1 for Samuel's paper. Very informative. But in my experience not that relevant to sound quality which isn't really strongly correlated
with those measurements.


IMO, read between the lines so to speak and look at common mode distortions and OP loading distortions. These accoun for a
lot of the sound of an opamp.

#157 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 01 April 2012 - 04:45 PM

Is 'jump factor' the new PRAT?


I'd suggest yeah its a very close sibling if not the same animal. But I'm not a reviewer so what do I know?

A vague and imprecise term that seems misleadingly precise, until you realise it's just an excuse for subjectivist reviewers to say something unmeasurable


I'd wanna correct this inaccuracy - we don't currently have measurements which correlate well with it. That's not the same thing as its unmeasurable.
I'm naive but I still hope we'll find some noise modulation measurements which pin it down somewhat.

, so it gives them high ground to stand in judgement of any equipment without being disprovable.


Yep you raise a good point - if its not falsifiable, then its not science but religion. We all know where that leads....

Let's say one person says it has 'jump factor' and another says it has no 'jump factor'. Well, does it have it or not?


Listen to it yourself and make up your own mind would be my suggestion. To me these subjective terms are rather akin
to phrases like 'sheer driveability' for cars - just because that's not measured does it mean 'driveability' doesn't exist? I
doubt whether any petrolhead would suggest that would they?

#158 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 01 April 2012 - 05:01 PM

Yes I understand where you are coming from and there are many with your perspective on this.
http://www.esstech.com/PDF/sabrewp.pdf
However from the white paper and from discussions at DIA, with Dustin Formin one of the designers, they have put
a huge amount of effort in overcoming traditional problems with the modulators. For example noise floor modulation
etc.


I think I've already mentioned that credit should be given to ESS for even flagging it up as an issue. I have yet to
see the same degree of transparency from TI and ADI. But as I'm an engineer I'm keenly aware of the difference
between a problem mitigated and a problem dissolved. They're most likely the best, but they're the best of a bad
bunch. How much consolation do you wish to draw from that?

The biggest frustration with Sabre for me is they dumped an ASRC in the chip aswell. I would have loved to see that as
a separate chip and keep the DAC as simple as possible. Apparently they did consider this.


That's the part I can't understand. But Bruno Putzeys seems jolly happy with their explanation and he's one of the gurus
of ASRC.

I used to have exactly your view for many years. However these days, after many years of building stuff for recording and playback
I have changed my views to a certain degree. Your view applies to a perfect world with pefect studios and perfect playback.


You really think so?

I haven't found that perfect world yet. In my engineering heart I think -exactly- like you do. Unfortunately the reality of life's
experience has been different and this conflict has many times made me even want to give up audio design altogether.


I wasn't relating my thinking though, I was relating my experience. So you've mis-read me. None of my experience in audio
engineering has turned out to be 'unfortunate' - rather the mistakes have led me to deeper insights into what's going on.

Haven't seen the review or measurements. In fact does anyone know what DAC they are using?


Here's a thread I've been active in over on DIYA which turns up quite a few nuggets about the Metrum:

http://www.diyaudio....chips-used.html

IMO, read between the lines so to speak and look at common mode distortions and OP loading distortions. These accoun for a
lot of the sound of an opamp.


Sure, but there's the thorny issue of RF rejection which for me is a biggie and he's not even looking anywhere near that. Its my
hypothesis for a reason why JFET opamps do better in many I/V applications for example.

#159 zenelectro

zenelectro

    Member

  • Members
  • 355 posts
  • Location: NSW

Posted 01 April 2012 - 08:32 PM

I think I've already mentioned that credit should be given to ESS for even flagging it up as an issue. I have yet to
see the same degree of transparency from TI and ADI. But as I'm an engineer I'm keenly aware of the difference
between a problem mitigated and a problem dissolved. They're most likely the best, but they're the best of a bad
bunch. How much consolation do you wish to draw from that?


Consolation - I don't know - I'll let you know when I see some good evidence. Maybe you can give us some better insight through
quantifiable measurements that the problem with Sabre is significant. I'd be interested to see them - we can all learn something.

There's a certain irony in this - I am defending Sabre, where as I had my money (quietly) on the Arda 1401 which promised great performance.
The wait just got too long and I've been informed it's on the backburner.

I'm not totally sold on all of the instrumentation DACs that I have seen, no doubt there are others.

That's the part I can't understand. But Bruno Putzeys seems jolly happy with their explanation and he's one of the gurus
of ASRC.


There are mixed opinions and some claim significant jitter. But it's hard to argue with Putzeys - very deep knowledge.

I wasn't relating my thinking though, I was relating my experience. So you've mis-read me. None of my experience in audio
engineering has turned out to be 'unfortunate' - rather the mistakes have led me to deeper insights into what's going on.


WRT misreading - sorry unintended. Deep insight is a good thing! At >50yo I have quite a bit of it myself.
But it has led me to understand that we have some serious issues WRT quality of recording in most of our facilities that are becominng
entrenched. It's an involved discussion and not really correct in this thread but I'm thinking you get my drift so to speak. :)

Here's a thread I've been active in over on DIYA which turns up quite a few nuggets about the Metrum:

http://www.diyaudio....chips-used.html


Yes I glanced over that one. What is your DIYA 'call sign'
So, what DAC chip do you think they are using?

When Lampizator mentioned he was using an instrumentation type multibit DAC I had a dig around but couldn't
find anything that seemed to fit the bill.

Sure, but there's the thorny issue of RF rejection which for me is a biggie and he's not even looking anywhere near that. Its my
hypothesis for a reason why JFET opamps do better in many I/V applications for example.


WRT RF, good point, however I'd be surprised if he wasn't well versed - he's employed by Weiss, it should be part and parcel of his job there.

WRT Jfet opa's, that's my experience also. But I don't generally use opamps for I-V applications. It's open loop SS or tubes.

#160 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 01 April 2012 - 08:51 PM

Consolation - I don't know - I'll let you know when I see some good evidence. Maybe you can give us some better insight through
quantifiable measurements that the problem with Sabre is significant.


Well to me the evidence is in the listening. Its for measurements to catch up with what we hear. Which is pretty much what the
CEO of Y G Acoustics said in a recent article I read about him. He made the point that we listen in order to check that we're
measuring the right things. When our ears tell us something's amiss, if our measurements don't show it then we're still making
the wrong measurements and we need to revise them. ESS on their datasheet just show noisefloor changes vs DC level. That's
hardly likely to be representative of real music (which has no DC) but its a step further than the other guys go.

I'd be interested to see them - we can all learn something.


As you're fond of irony maybe you'll enjoy this - I came here to hang out because so many of you guys were posting
up first hand listening reports, descriptions. That's raw data for doing science. Evidence is what our senses tell us,
not what our models do.

There's a certain irony in this - I am defending Sabre, where as I had my money (quietly) on the Arda 1401 which promised great performance.
The wait just got too long and I've been informed it's on the backburner.


Yep and I see a market opportunity in that (virtual) DAC's demise as plenty of people (like yourself) are looking like
they really wanted it.

WRT misreading - sorry unintended. Deep insight is a good thing! At >50yo I have quite a bit of it myself.
But it has led me to understand that we have some serious issues WRT quality of recording in most of our facilities that are becominng
entrenched. It's an involved discussion and not really correct in this thread but I'm thinking you get my drift so to speak. :)


The irony of my own DAC development path is that as I've improved it, I've heard more and more problems with recordings.
So I agree with you - something's going wrong and it'd be really nice to have a hand in putting that to rights. :)

Yes I glanced over that one. What is your DIYA 'call sign'


abraxalito

So, what DAC chip do you think they are using?


TI DAC8581

WRT RF, good point, however I'd be surprised if he wasn't well versed - he's employed by Weiss, it should be part and parcel of his job there.


'Shoulds' are often quite far from reality. If they were really listening I'd say they'd not have chosen TI PCM1792 but would be fans of multibit
like me :P So I think they're measurement driven, not subjectively driven and therefore since RF effects don't show up in our current suite of
measurements, why bother?

#161 kajak12

kajak12

    ENJOY YOUR JOURNEY

  • Members
  • 5,096 posts
  • Location: WA
  • Country: Australia

Posted 01 April 2012 - 11:23 PM

The Lampi does use a S-D chip.

http://www.ebay.com.au/itm/Lampucera-High-End-Tube-DAC-/200733334134?pt=UK_Consumer_VintageAudio_RL&hash=item2ebca39a76

cheap lampi ........... easy to modd

Give a man a gun and he can rob a bank......

Give a man a bank and he can rob the world..........


#162 Telecine

Telecine

    Member

  • Members
  • 4,824 posts
  • Location: QLD
  • Country: Australia

Posted 01 April 2012 - 11:43 PM

http://www.ebay.com....=item2ebca39a76

cheap lampi ........... easy to modd


Sold as a kit through the Lampi store. See here. It is the Lampucera which isn't quite the same as the higher end models.
Learn from the mistakes of others - you'll not live long enough to make them all yourself!

#163 kajak12

kajak12

    ENJOY YOUR JOURNEY

  • Members
  • 5,096 posts
  • Location: WA
  • Country: Australia

Posted 01 April 2012 - 11:55 PM

Sold as a kit through the Lampi store. See here. It is the Lampucera which isn't quite the same as the higher end models.

I am sure somebody would put it together (i would just for fun)

Give a man a gun and he can rob a bank......

Give a man a bank and he can rob the world..........


#164 Telecine

Telecine

    Member

  • Members
  • 4,824 posts
  • Location: QLD
  • Country: Australia

Posted 02 April 2012 - 12:07 AM

I am sure somebody would put it together (i would just for fun)


Very generous of you but I am not in the market for one.
Learn from the mistakes of others - you'll not live long enough to make them all yourself!

#165 empirical

empirical

    Manufacturer/Designer

  • Members
  • 175 posts
  • Location: ACT
  • Country: Australia

Posted 02 April 2012 - 03:48 AM

Sounds plausible. So far I haven't experimented with supplies any more exotic than cheap series and shunt ICs. Then can you explain why, given that I use the same power delivery arrangements in both S-D and multibit, that the multibit has the jump factor and the S-D does not? What might it be about multibit that makes it immune to PSU regulation and decoupling?

I have not experienced this myself, so I'm not convinced that you can generalize. I used PCM1704 in my older Spoiler tubeDAC. However, most older chips dont have the digital filtering of S-D or the oversampling, so they are generally running at lower frequency internally. With CMOS, this means lower power consumption and less dynamic current demands. This could explain the difference. Older chips also tend to have fewer pins and sometimes smaller packages, which improves inductance and on-die power delivery. And BTW, cheap shunt and series regulator IC's will be okay for a $500 DAC maybe, but not any higher pricepoints IME. The regulation and input noise rejection is poor on most of these. Steve N.

Edited by empirical, 02 April 2012 - 04:17 AM.


#166 empirical

empirical

    Manufacturer/Designer

  • Members
  • 175 posts
  • Location: ACT
  • Country: Australia

Posted 02 April 2012 - 03:56 AM

"I don't know that Steve was implying this. But they are separate issues, power supplies for opamps and in the Sabres case supplies for
analog OP stage, digital stages etc. Personally I would treat them differently."

And I do. I like to use a separate regulator for the digital side and the analog side if possible. Speed of regulation is more important for the digital side, to minimize jitter. For the analog side, I have found that speed is also important, but decoupling caps (the optimum sizes, mix and locations) are at least as important. Its really this mix that delivers the full analog dynamics. Same with Op-Amps.

Steve N.

#167 empirical

empirical

    Manufacturer/Designer

  • Members
  • 175 posts
  • Location: ACT
  • Country: Australia

Posted 02 April 2012 - 04:15 AM

SD8581 in the Metrum? How do they use 4 chips if it has a voltage output? Hard to believe that a voltage output DAC would sound this good.

I thought someone said it had 8 pins like a TDA1543.

Steve N.

#168 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 02 April 2012 - 10:10 AM

I have not experienced this myself, so I'm not convinced that you can generalize.


What you're saying here isn't clear to me. You haven't experienced jump factor with multibit? What was I generalizing about?

I used PCM1704 in my older Spoiler tubeDAC. However, most older chips dont have the digital filtering of S-D or the oversampling, so they are generally running at lower frequency internally.


The way I read what you were saying it was the speed of decoupling at audio frequencies which was important - that's the transient
demands of music ISTM. Now you're saying that its the speed of decoupling at the bit rate (>1MHz) ? To me those are two different
things - regulators can handle up to 20kHz effectively, but they have nothing left to give >1MHz which is why we use local decoupling.

With CMOS, this means lower power consumption and less dynamic current demands.


Have you compared decoupling a bipolar (say TDA1543) DAC compared to a CMOS one (like TDA1545) ? Yes CMOS has lower
power consumption but that comes with much higher dynamic current demands. Internally CMOS has myriads of little crowbars sitting
across the supply lines. They're triggered on every clock cycle.

#169 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 02 April 2012 - 10:15 AM

How do they use 4 chips if it has a voltage output? Hard to believe that a voltage output DAC would sound this good.


They just put sharing resistors between each - similar in how chip amps (also voltage out) can be paralleled. Except
here the resistors are individually in the 100's of ohms. Yes I agree - with CMOS opamps how does it get to sound so
great? That's a puzzle.

I thought someone said it had 8 pins like a TDA1543.


That person wasn't paying attention. There are internal pics on the 6moons site.

#170 empirical

empirical

    Manufacturer/Designer

  • Members
  • 175 posts
  • Location: ACT
  • Country: Australia

Posted 03 April 2012 - 04:24 AM

What you're saying here isn't clear to me. You haven't experienced jump factor with multibit? What was I generalizing about?


I have experienced jump factor with all D/A chips. The generalization I heard is that all newer chips have dynamics problems.

The way I read what you were saying it was the speed of decoupling at audio frequencies which was important - that's the transient
demands of music ISTM. Now you're saying that its the speed of decoupling at the bit rate (>1MHz) ? To me those are two different
things - regulators can handle up to 20kHz effectively, but they have nothing left to give >1MHz which is why we use local decoupling.


I do use fast reacting regulators for logic, and I can use slower for analog.

Have you compared decoupling a bipolar (say TDA1543) DAC compared to a CMOS one (like TDA1545) ? Yes CMOS has lower
power consumption but that comes with much higher dynamic current demands. Internally CMOS has myriads of little crowbars sitting
across the supply lines. They're triggered on every clock cycle.


I know this. I have not compared the D/As you mention.

Steve N.

#171 zenelectro

zenelectro

    Member

  • Members
  • 355 posts
  • Location: NSW

Posted 03 April 2012 - 10:50 AM

The way I read what you were saying it was the speed of decoupling at audio frequencies which was important - that's the transient
demands of music ISTM. Now you're saying that its the speed of decoupling at the bit rate (>1MHz) ? To me those are two different
things - regulators can handle up to 20kHz effectively, but they have nothing left to give >1MHz which is why we use local decoupling.


WRT regulator performance, for an IC regulator, yes at 1MHz PSRR and OP impedance are shot to pieces.
However a carefully designed discrete / hybrid reg can have very good performance right up to 1MHz.

As I suggested earlier, generally, you are trading off OLG with bandwidth. With reasonably low OLG, the regs BW
can far exceed 1MHz. It depends on the design and components used.

Z

#172 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 03 April 2012 - 12:20 PM

WRT regulator performance, for an IC regulator, yes at 1MHz PSRR and OP impedance are shot to pieces.
However a carefully designed discrete / hybrid reg can have very good performance right up to 1MHz.


Yep. I would only resort to a discrete design if all else fails... Lots of work to do to get it right. Its almost like
a whole amp design. With my lazy streak I definitely prefer to steer around this.

#173 zenelectro

zenelectro

    Member

  • Members
  • 355 posts
  • Location: NSW

Posted 03 April 2012 - 06:04 PM

Yep. I would only resort to a discrete design if all else fails... Lots of work to do to get it right. Its almost like
a whole amp design. With my lazy streak I definitely prefer to steer around this.


So on one hand you are looking for magical noise floor modulation to explain SD DAC's supposed limitations,
but on the other hand you won't go to the length of designing a simple discrete reg. IME it's really not
that hard to get it right, fire up spice sim some designs to get a starting point, and go for it.

You could also try some of the aftermarket types or many on DIYA.

Edited by zenelectro, 03 April 2012 - 06:05 PM.


#174 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 03 April 2012 - 06:59 PM

So on one hand you are looking for magical noise floor modulation to explain SD DAC's supposed limitations,
but on the other hand you won't go to the length of designing a simple discrete reg.


I think perhaps you misunderstood me. I was just talking in general that I would prefer using off the shelf
IC regulators in various configurations and only if those couldn't give me the results (i.e. sound) I wanted
would I revert to discrete. What's magical about noise modulation incidentally? I don't go looking for it, quite
the contrary I prefer to avoid it by designing with multibit.

IME it's really not
that hard to get it right, fire up spice sim some designs to get a starting point, and go for it.


Indeed but when IC regs won't handle it, the nut's going to be a tough one to crack. I designed
various incarnations of power supplies when I was a schoolboy, before desktop computers even
existed so I'm sure its not beyond me to do a simple one.

#175 wis97non

wis97non

    Member

  • Members
  • 1,040 posts
  • Location: Other

Posted 06 April 2012 - 08:39 AM

Why all the focus on the type of DAC chip and almost nothing on the receiver chip that feeds the DAC?

That chip may actually be more important to the sonic signature, no?

#176 wis97non

wis97non

    Member

  • Members
  • 1,040 posts
  • Location: Other

Posted 06 April 2012 - 08:39 AM

Why all the focus on the type of DAC chip and almost nothing on the receiver chip that feeds the DAC?

That chip may actually be just as important to the sonic signature, no?



#177 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 06 April 2012 - 09:16 AM

Why all the focus on the type of DAC chip and almost nothing on the receiver chip that feeds the DAC?

That chip may actually be more important to the sonic signature, no?


It might be, yes. In my experience of developing a DAC design, I originally assumed this. But so far I have
focussed all my attention on other aspects because the (annoying) artifacts I was hearing turned out to be
DAC-chip, layout and output stage derived. That's not to say that the receiver chip makes no difference,
rather its to say it hasn't so far been the major difference in my experience.

Some receiver chips are better than others where measurements are concerned - for example technically
the WM8805 is better than the CS8414. It also has a lot more features for the money. Whether I can hear
this difference in measurements in practice is yet to be determined.

#178 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 06 April 2012 - 10:24 AM

For all you doubters out there who don't believe in noise modulation - here are a couple of FFTs from the AD1955 datasheet which show it clearly. What to look out for is the difference in the noise floor between when a signal is present (first plot) and when its not (second one). The difference is >10dB in the region where our ears are most sensitive (around 4kHz) and this is an averaged measurement so short term differences may well be greater.

Attached Files



#179 techspurt

techspurt

    100+ Post Club

  • Members
  • 241 posts
  • Location: Other
  • Country: China

Posted 06 April 2012 - 11:58 AM

Here's a plot from the AK4399 datasheet - another manufacturer's S-D part. The noise modulation can be inferred from the full scale
THD+N measurement of -105dB when set up against the dynamic range (unweighted) figure of 120dB. The FFT plot verifies that the
full-scale THD+N figure given is indeed dominated by the noise as the 3rd harmonic distortion shows about -109dB. (Vertical scale I
cropped - its 10dB per division).

Attached Files


Edited by techspurt, 06 April 2012 - 12:00 PM.


#180 kajak12

kajak12

    ENJOY YOUR JOURNEY

  • Members
  • 5,096 posts
  • Location: WA
  • Country: Australia

Posted 06 April 2012 - 12:02 PM

Why all the focus on the type of DAC chip and almost nothing on the receiver chip that feeds the DAC?

That chip may actually be more important to the sonic signature, no?

I2s is the only way to go.................

Give a man a gun and he can rob a bank......

Give a man a bank and he can rob the world..........