Digital, but not by the numbers

[abraxalito]

Originally Posted by Trevor White

2.8MHz square waves with fast rise times which means at least a 28MHz bandwidth to maintain signal integrity !!

I've already pointed out that fast rise times aren't part of the design philosophy - quite the opposite. But you're so fond of your strawmen you can't resist

Your rats nest circuit design topology will be ringing all over the shop not to mention crosstalk and false trigger !! It will certainly not be by the numbers !!

If false triggering were happening then I'd hear it as glitches in the audio. But I don't - strange eh?

[abraxalito]

So now the thread has gone a little quieter here's something about D/A desgn in digital systems which I've learned from my couple of years of tinkering with DACs.

DF96 raised the point that to make my DAC design work I'd need good mixed signal design skills. That's true of getting any DAC to sound good, not just the design I've posted on my blog. The way I approach mixed signal design is I take everything as analog, so its no longer mixed any more. Digital being just a narrow subset of analog.

There is of course another way to do it - treat everything as digital. This seems to me closer to the way that DAC chips are designed nowadays. The idea with a S-D chip design is to make as much of the DAC digital as possible. Such designs are invariably built on digital (CMOS) processes. After all, analog by comparison is expensive - check out the price of the PCM1704 against the PCM1794 which is supposed to be its replacement, according to TI.

So in treating all digital as analog this leads towards the building of digital circuits more like analog ones, rather than as Trevor has been suggesting, with ground planes and 'signal integrity'. Groundplanes suit high speed digital design well, but don't fit with my philosophy for a couple of reasons. Low inductance means high rates of change of currents meaning higher noise. And also I do like to consider the current loops in my designs, using a groundplane means I'm never quite sure where the current is flowing. There's a third reason - groundplanes really can't be so easily implemented by the DIYer, they rather call for multilayer PCBs. A ground fill not being quite the same as a groundplane even though many people don't distinguish between these two terms. To me a groundplane has to be unbroken with tracks, though of course it has holes for vias.

PSRR is a consideration when designing analog circuits but digital circuits have PSRR too. PSRR comes in two flavours - for the two rails and in the case of analog its usually the case that the negative one sucks worse than the positive. In the case of CMOS digital circuits, its 6dB for the positive supply and 0dB for the negative. So digital is similar in this respect - in both, the negative PSRR sucks. Clean grounds therefore are twice as important as clean positive rails meaning that my digital circuits have asymmetry in design - the positive rail has filtering - a series inductor or ferrite bead typically - but the negative does not.

[marce]

EMC could indeed be a problem but as its not being put in the market, there's no need for regulatory compliance. As regards your other complaints, they're not an issue for the sound or the rudimentary measurements I've so far made.

I am sorry Abraxalito, but I have to STRONGLY disagree with the statement, it is not just you bit other DIYers that belive this is a valid staement. Wrong, signal integrity and EMC are two sides of the same coin. It is a niave view of design, EMC will affect your design whether you choose to ignore it or not in your design, I see numerous posts regarding noise, RF pick up etc and yet we have statements like this.
Sorry to go on but if you want to design somthing that is better than other things you cannot ignore Electromagnetic Compatability Engineering, and I would suggest you read the book of the same title, as Henry Ott says EMC is 20% external and 80% internal, ie its your own equipement that will suffer.
Marc

[marce]

Just curious, what rise time is being used.
And how do you define what a high speed design is?
And how do you determine the frequencys of the square wave you have to worry about?

[Julf]

Originally Posted by abraxalito

The way I approach mixed signal design is I take everything as analog, so its no longer mixed any more.

Let me make a prediction - if you succeed to make your DAC work, it will be a true "audiophile" DAC. It's measured performance will be really bad, but True Audiophiles will love it's sound - and describe it as "analog", and probably "round", "warm", "musical" and "without any digital glare". Because it is "designed as analog, not digital".

As an extra bonus, it will probably be very sensitive to things like cable capacitance and rise times, so the True Audiophiles will also love it for proving them right in that cables do matter, not to mention all sorts of power conditioners, extra dampers and whatever it might take to deal with the EMC issues...

[abraxalito]

Originally Posted by marce

I am sorry Abraxalito, but I have to STRONGLY disagree with the statement

You quoted more than one statement so please clafify which one? The one about EMC compliance or about measurements?

Wrong, signal integrity and EMC are two sides of the same coin. It is a niave view of design, EMC will affect your design whether you choose to ignore it or not in your design, I see numerous posts regarding noise, RF pick up etc and yet we have statements like this.

Indeed - I do agree that EMC is an issue, so where precisely are you disagreeing with me? I have posted about how its important to run slow rise times for example. I shall go on to talk more about noise issues in isubsequent posts because its an area often overlooked by digital designers judging by the pictures of the internals of DACs I've seen.

Sorry to go on but if you want to design somthing that is better than other things you cannot ignore Electromagnetic Compatability Engineering, and I would suggest you read the book of the same title, as Henry Ott says EMC is 20% external and 80% internal, ie its your own equipement that will suffer.

I'm not particularly concerned about 'being better' than something else, I'm rather concerned about making the best DAC I can make, within the constraints I've set for myself. A rather different problem.

Incidentally, how many products have you shepherded through EMC complaince testing yourself? By which I mean been the final design authority for ensuring the device gets through and on to the market?

[abraxalito]

Originally Posted by Julf

Let me make a prediction - if you succeed to make your DAC work, it will be a true "audiophile" DAC. It's measured performance will be really bad, but True Audiophiles will love it's sound - and describe it as "analog", and probably "round", "warm", "musical" and "without any digital glare". Because it is "designed as analog, not digital".

This sounds like fun as a hypothesis, but it won't pass muster as science because its not falsifiable. To make it falsifiable you need to tighten up your definitions of currently very woolly terms like 'very bad measurements'. How bad in numberical terms would you predict and in which measurements in particular? THD, full scale or low level? FR +/- how many dB ?

As an extra bonus, it will probably be very sensitive to things like cable capacitance and rise times, so the True Audiophiles will also love it for proving them right in that cables do matter, not to mention all sorts of power conditioners, extra dampers and whatever it might take to deal with the EMC issues...

As above, tighten up on 'very sensitive' to make this falsifiable.

[abraxalito]

Originally Posted by marce

Just curious, what rise time is being used.
And how do you define what a high speed design is?
And how do you determine the frequencys of the square wave you have to worry about?

There's no one rise time being used, but I'll probably get around to measuring some waveforms at some point in which case I'll post up the data. I don't define 'high speed design' (the term was Trevor's if I recall) I just know 2.8MHz doesn't meet it - call that a gut feeling if you like

[marce]

EMC

[qusp]

hes not talking about clock speed

I have no idea how to contribute at this point. my view was not, not by the numbers, but despite them. I believe with modern kit AFTER its been verified as working within accepted margins, we have some leeway to make it sound as we like because its good enough now that we can afford to trade some off, perhaps trading some bandwidth for subjective sound, or stripping away some complexity while still staying above audible range. basically making decisions with eyes open, not JUST chasing the numbers for the sake of numbers.

I have no desire to be aligned with charlatans like Altmann, who makes such ridiculous claims about his little '24/192' (which is printed on the chassis) portable player with a 16bit dac chip its not funny. he avoids all technical queries to back up his claims with the usual subterfuge about measurements being meaningless etc. claims like the 24/192 when only the DSP can read them and downsample them to send to the 16bit dac, claims that it can drive any headphone with its 3-4v (MAX) swing and tiny cellphone battery that lasts ages, so cannot have much by way of current either, 32bit digital volume with a 16bit dac... the list goes on. and he wants 700EU for it or something...

so he either doesnt have a clue, or hes taking the p1ss