[sdiy] Static ADC and DAC recommendations?

[cheater cheater]

Hi guys,
Thanks for the great replies. Some great info there. Especially liked
the bits on high-precision DMM's and on analog computers :)

I've read up on successive approximation some time ago, and just
refreshed myself to make sure I'm not missing something. It's
different from the chain-of-comparators; in fact I don't see a way to
go from one to the other. Successive approximation is a feedback chain
with a DAC (or a number of digital switches) and a single comparator
at the end; this is read by logic (possibly an MCU) which adjusts the
DAC/switches accordingly. Here, we're talking about a single bit being
represented by a simple comparator.

Dual-slope is also different: it involves a voltage reference and a
timing circuit (usually based off a clock and counter) and some
arithmetic logic.

On Thu, Jan 17, 2013 at 11:40 PM, Neil Johnson
<neil.johnson97 at ntlworld.com> wrote:
> Plus the fact that you're having to implement precise gains at every stage
> otherwise performance suffers, especially at the LSB end which is where it
> matters most

when has insufficient THD+N stopped anyone from using a Tube Screamer?
:) Seriously, guys, don't concentrate on performance. Quotes like the
above or this one...

On Thu, Jan 17, 2013 at 11:40 PM, Phillip Harbison <alvitar at xavax.com> wrote:
> An ADC capable of 200K samples
> per second should be good enough for audio.

...are great, thanks for the insight, but that's not really the point.
And I know this is unusual, since normally when someone's designing an
ADC or DAC the only thing that matters is quality, but this is not the
case here :) And the direction the conversation went is not surprising
since everyone I asked was in this exact same mindset :)

The idea isn't to build a circuit that will be good enough to
transcribe audio in pristine condition. The idea is to take a flawed
approach - comparator-based, low component count, no clocking, no
sample holding - and see what disadvantages it brings and how they can
be used creatively. So what if I can't make a perfect 2x multiplier -
it's probably better this way. So what if I can't exactly match the
comparators. So what if it's only 6 bit? Let's not think about how to
build yet another boring box.. :) A circuit which will create
reproduction in every way superior to the approach I am trying to go
for can fit on silicon the size of a pin head and can be found in
billions of $0.99 portable music players people throw away without
thinking. No reason to recreate them, and if I wanted I couldn't even
begin to compete, so I'd just buy an existing high-performance IC.

I idly wonder if you could do this (compare and multiply-by-2) with a
555 per bit (plus some other stuff).... :) But that's only a fun
thought, nothing more.

On Thu, Jan 17, 2013 at 11:40 PM, Phillip Harbison <alvitar at xavax.com> wrote:
> cheater cheater wrote:
>>
>> Wait a sec guys - I thought the usual approach was folding
>> converters. That is, you take the input signal, scale it to
>> e.g. 0-1V, and see if it's the upper or lower half [...]
>
>
> That essentially describes one type of ADC, the successive
> approximation variety. The other type is a flash converter,
> where you have enough comparators on the chip to do this all
> in parallel. The problem is flash converters obviously get
> exponentially complex as you increase the accuracy. You'll
> need twice as many comparators to go from 4 bits to 5 bits,
> so anything beyond about 8 bits is prohibitive, and I've
> only seen it used in extreme applications like video ADCs
> where 8 bits per color is usually sufficient.
>
> Another problem is even with a flash ADC, you'll need some
> settling time so there's still the issue of delay time even
> if you don't consider it clocked. I don't think a continuous
> output is possible or at least not realistic. I second the
> suggestion of using a sample rate high enough that you don't
> notice it is not continuous. An ADC capable of 200K samples
> per second should be good enough for audio.
>
> If you really need continuous signal processing, that is why
> we still have analog computers. In my younger days I had the
> opportunity to visit the McMorrow Labs, a research center on
> Redstone Arsenal (Huntsville, AL) that does simulations to
> test missile guidance systems. I was majoring in computer
> engineering so this was quite a thrill. They had a CDC-7600
> which was the fast supercomputer in those days (the Cray-1
> was not yet shipping). Connected to the 7600 was a CDC-6600
> which interfaced to a room of analog computers about the size
> of a hockey rink. I would guess there were about 200 to 300
> analog computers each about the size to two 19" x 84" racks.
> I asked why they needed all these analog computers, which I
> considered to be fossils, when they had the fastest computer
> in the world downstairs. My guide explained that solving any
> one of these complex differential equations, each of which
> was handled by an analog computer, would bring that 7600 to
> its knees. The analog computers were the right tool for the
> job. The 7600 ran a Fortran program that read descriptions
> of differential equations to patch layouts for the analog
> computers. OK, it did a few other things. :)
>
> "If the only tool you have is a hammer, every problem looks
> like a nail." -- me
>
> --
> Phil Harbison
>
>
>
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at dropmix.xs4all.nl
> http://dropmix.xs4all.nl/mailman/listinfo/synth-diy