SynthModules

Soon, a new version of the gc-sah-sr-qua.bas will appear - Original
code written by Woody Wall and revised by Dave Brown.  This program
converts the PSIM into a Serge Analog Shift Register - actually
several of them, allowing for up to 16 stages, which would take 6
Serge triple ASRs to replicate.  Thanks so much to Dave Brown for his
time this weekend in this effort - I very much appreciate your incites
on the PSIM.  

After some experimentation, some realizations are evident to me.

1).  Dave's analysis of the PSIM's accuracy is spot on.  ALL PSIM
programs that are designed to have accurate input/output ratio
throughput should reflect Dave's research.

As Dave describes:

For input, the range is 1024 steps over 10 volts, or 9.76 mV 
increments.

For output, the range is 4096 steps over 10.666 volts, or 2.60 mV 
increments.

Thus to pass an input to output, the scale factor is (10/1024)X
(4096/10.666) which is 3.75.

Thus to scale the input to output, it needs to be multiplied by 
3.75, (not 4).  I typically multiply by 15/4 to keep the math 
integer based.

Note that 10.666 volts was chosen  to calibrate to semitone 

[end Dave's quote]

2).  For the purposes of a shift register application such as
gc-sah-sr-qua.bas, IMHO, quantization is unnecessarary.  

First off, many of the patterns that one may want to "arabesque-asize"
with a Shift Register will be already quantized - this makes the
issues of "arguments" between the PSIM's interpretation of what the
input is possible - I found that, even with Dave's more accurate
input/out ratios installed that my PSIM was constantly arguing with my
Sequantizer about what pitch it should e playing.

Secondly, I found that the unquantized shift register is not only more
accurate, but the wider range of the unquantized version is much
better use when utilizing the Shift Register to controll non-pitched
modules.  It is also much more accommodating to patterns with wider
pitch ranges (such as playing a sequence that spans more than two
octaves).  In this regard, the unquantized version performs brilliantly.

Thirdly, the unquantized version operates identically the Serge Analog
Shift Register - which, for me, is the architypical Shift Register in
my life....

Thanks, Dave!

Gary

A couple of other comments on the PSIM.

If you plan to quantize, quantize on the output, not the input.  Since 
the voltage reference was chosen to be a multiple of a semitone (e.g. 
83 mV), you can quantize by simply zeroing out the last 5 bits of the 
output value.

Also, if you plan to quantize, any noise or errors can cause you to 
quantize to the next lower semitone.  I always add a half a semitone 
offset before I quantize, which in this case is by adding 16 to the 
output, then zeroing the last 5 bits.  This forces any error to 
quantize to the nearest semitone, instead of the next lower semitone.

The PSIM is quite accurate if you account for the correct scaling with 
the different reference voltages for input and output.  It is also 
much more versatile with MIDI and / or an LCD display.  John Loffink 
has expansion boards available that can add MIDI and LCD functionality 
quite easily to a PSIM.  See the ComputerVoltageSource Yahoo list for 
details.

Dave


--- In SynthModules@yahoogroups.com, "Gary Chang" <gchang@...> wrote:
>
> Soon, a new version of the gc-sah-sr-qua.bas will appear - Original
> code written by Woody Wall and revised by Dave Brown.  This program
> converts the PSIM into a Serge Analog Shift Register - actually
> several of them, allowing for up to 16 stages, which would take 6
> Serge triple ASRs to replicate.  Thanks so much to Dave Brown for 
his
> time this weekend in this effort - I very much appreciate your 
incites
> on the PSIM.  
> 
> After some experimentation, some realizations are evident to me.
> 
> 1).  Dave's analysis of the PSIM's accuracy is spot on.  ALL PSIM
> programs that are designed to have accurate input/output ratio
> throughput should reflect Dave's research.
> 
> As Dave describes:
> 
> For input, the range is 1024 steps over 10 volts, or 9.76 mV 
> increments.
> 
> For output, the range is 4096 steps over 10.666 volts, or 2.60 mV 
> increments.
> 
> Thus to pass an input to output, the scale factor is (10/1024)X
> (4096/10.666) which is 3.75.
> 
> Thus to scale the input to output, it needs to be multiplied by 
> 3.75, (not 4).  I typically multiply by 15/4 to keep the math 
> integer based.
> 
> Note that 10.666 volts was chosen  to calibrate to semitone 
> 
> [end Dave's quote]
> 
> 2).  For the purposes of a shift register application such as
> gc-sah-sr-qua.bas, IMHO, quantization is unnecessarary.  
> 
> First off, many of the patterns that one may want to "arabesque-
asize"
> with a Shift Register will be already quantized - this makes the
> issues of "arguments" between the PSIM's interpretation of what the
> input is possible - I found that, even with Dave's more accurate
> input/out ratios installed that my PSIM was constantly arguing with 
my
> Sequantizer about what pitch it should e playing.
> 
> Secondly, I found that the unquantized shift register is not only 
more
> accurate, but the wider range of the unquantized version is much
> better use when utilizing the Shift Register to controll non-pitched
> modules.  It is also much more accommodating to patterns with wider
> pitch ranges (such as playing a sequence that spans more than two
> octaves).  In this regard, the unquantized version performs 
brilliantly.
> 
> Thirdly, the unquantized version operates identically the Serge 
Analog
> Shift Register - which, for me, is the architypical Shift Register 
in

> my life....
> 
> Thanks, Dave!
> 
> Gary
>