[sdiy] Deceptively simple question: 8 bit DACs

[phillip m gallo]

Andre's analysis is exactly right for 8 bit conversion with 1/2 LSB error,
however 8 bit converters are commonly found to work quite nicely
polyphonically or otherwise.  Why ???

The .25 LSB error noted requires the dynamic use of the LSB and as MIDI is 7
bits you have the option of holding this LSB static.  Doing this it's
contribution to error can now be "trimmed" out as it is a fixed "offset".
The .25 LSB error noted is "robbed" of the LSB's dynamic error and is that
which is attributable to the remaining 7 bits.  On top of this the 1/2LSB
error is not a guaranteed spec but a maximum spec.  So if the error is less
to begin with the non-conformance is less.

Additionally some 8 bit MIDI Converters hold their range to 6 bits providing
a range of 64 Note values (just greater than 5 8ves) which allows very good
accuracy and a good matching for synths with range switching.

I am only superficially aware of the PAIA converter but know that is uses
the DAC08.  As a mature workhorse this DAC is often "dis'd" by those haven't
used it critically but it is in fact a great converter (albeit of low
integration), very fast settling (well suited for muxing as in the PAIA
converter) and available inexpensively with tight parametric quality.  The
DAC08-H being available with Full Scale Linearity of .1% (where .39% would
be 1 LSB).

Where 8 bit converters fall down is that they do not have the bit depth to
correct errors that exist in a VCO whose linearity is not likely to be as
small (across 10 8ves) as .1% of full scale, to provide these corrections
DACs of 13 bits or greater are required.  Additionally, there is no
"freeboard" to allow the addition of modulation effects or the
implementation of a digitally generated glide.  The former is usually
accommodated by companion DACs and the latter is usually compensated for by
analog glide circuitry after the converter.

If i where to get squeamish about the PAIA converter it would be the mux
circuitry as this would have to be done carefully to achieve the accuracy of
a DAC08-H converter.

An 8 bit MIDI DAC i've constructed using the equally mature AD558 results in
errors no greater than 2 cents which occur in the 7 and 8 octaves of it's
range.  This error is due to accumulating DNL as the error reduces in
octaves 0 thru 6 and octaves 9 thru 10.

regards,
p


-----Original Message-----
From: owner-synth-diy at dropmix.xs4all.nl
[mailto:owner-synth-diy at dropmix.xs4all.nl]On Behalf Of Andre Majorel
Sent: Sunday, August 28, 2005 1:08 AM
To: synth-diy at dropmix.xs4all.nl
Subject: Re: [sdiy] Deceptively simple question: Panels


On 2005-08-27 13:25 -0400, Aaron Bader wrote:
> On Fri, 26 Aug 2005 19:14:10 -0500
> "J. Larry Hendry" <jlarryh at iquest.net> wrote:
>
> > In this case, you do get what you pay for.  I own two of the PAiA
> > MIDI 2 CV8s.  They are 8 bit resolution.  I'd say plenty good for
> > controlling a mono synth or perhaps two voice.  If you get into
> > polyphonic operation, you might want something a little better. I
> > have a Kenton pro-2000 which I really do like.  I am waiting
> > patiently for the MOTM MIDI2CV also.
>
> This confuses me a little. How does an 8 bit dac resolution degrade
> polyphonic operation? Or is it more because you ONLY get cv/gate when
> playing 4 voices with this particular converter?

Precision is more critical when playing chords. I suppose that the
sort of cheap 8-bit DAC you're likely to find in a PAiA product is
not precise enough.

If your 8-bit DAC is precise to +/-1/2 LSB and you use it so that
a semitone is 2 LSB (128 possible notes -- just right for MIDI),
you have a +/-25 cents error on each note !

--
André Majorel <amajorel at teaser.fr>
http://www.teaser.fr/~amajorel/


--
No virus found in this incoming message.
Checked by AVG Anti-Virus.
Version: 7.0.344 / Virus Database: 267.10.16/83 - Release Date: 8/26/05