[sdiy] MIDM-CV DACs, Part #3 (last)

[phillip m gallo]

Regarding DACs for VCO's...

Speed of conversion and forms of propagation delay that reflect in "settling
time" are important.

There is a bit of choice here and a lot of room for personal criteria but
you want a DAC that quickly "settles" to your intended voltage (with few
switching transients of low amplitude with a specified accuracy) as your
following OP-AMPs will increase (by adding their own) the resulting settling
time.

"Built out DACs" have some challenge ensuring the propagation delay for each
bit is consistent enough to achieve the shortest "settling time" (the
purpose of those doubled up buffers on the MSBs). If you have a lot of delay
or experience a lot of glitching artifacts you use a track and hold (which
is easier than a sample and hold).   If you "mux" the DAC you have to be
stable soon enough to account for the acquisition time of the many hold
circuits and the "refresh rate".

Linearity

DAC deviation from the expected output response demonstrate in the driven
VCO (if of sufficient magnitude).    The Integral Non-Linearity (INL) (as
Paul indicates) is an important specification impacting the total accuracy
across all codes.   Dynamic Non-Linearity (DNL), also important, indicates
the code to code deviation from the ideal straight line response. As you
might imagine these two characteristics relate.

Simple Keyboard Interfaces/Accuracy

In the case of encoding a 5 8ve Keyboard manual (60 keys) it's easy to see
that 6 Bits (0-63) is enough to binarily encode this keyboard.  These key
codes then need to cause the DAC to emit the "straight line" linear response
expected for 1V/8ve.  (We are ignoring Gate and Trigger).

We know each semi-tone note is spaced .083 V (1V/8ve) apart as Paul
indicated.  Now applying a generally available DAC08H (+/-.1% Full Scale
non-linearity) to produce the 0-5 Volts, we can expect any step to deviate
from ideal by 5V/1000 or .005v.  This lack of accuracy means your first
semi-tone (which you expect to emit as .083V) could emit as .088 or .078V
(each extreme equal to ~6 cents of deviation from expected). This
non-linearity then adds with the Non-Linearity of the driven VCO.

Before deciding on what 6 cents of deviation means to you (ever tune a
piano?), instrument a fully analog keyboard.  How many bit's of accuracy do
they have (especially across time)?  Everybody talks resistors and current
sources but it's the damn Sample and Hold that is the final limit.  Droop
sucks if you don't want it (with a bow to the standard synthesizer reality
that even bad stuff can sound good).

Digitally scanned keyboards with DACs, rigged up no more exotically than
Gene Stopps previous post, perform very well.  MIDI2CV interfaces based upon
similar technology work as well.  Good Hi resolution DACS are readily
available, easy to implement,  and not as expensive as you might imagine.
You can also sum multiple 8 bit DACs and achieve enough accuracy to solely
drive a VCO (encoding FZ, Offsets and Modulators).

regards,
p

---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.389 / Virus Database: 220 - Release Date: 9/16/02