[sdiy] too quiet...

[Harry Bissell]

Hmmm... I've been simulating this for years, and have even built units which the simulations
said would NOT work (they lied). I have a fairly robust model for making a guitar string 'sound'
and could even (in theory) input a real .wav file.

otoh, what isn't modeled well is the noise floor. That is where the simulation gives little value.
Sure it works at 100mV, and at 50mV... but at what point does the noise make the signal unusable ?

I have two options I'm using to convert from a filtered guitar fundamental to square (pulse) wave.
One is a variant of the method used in the Gentle Electric P/V converter, only my method uses two
opamps as comparators rather than the original method. The gist of this one is to use two peak
tracking detectors, and compare the input signal against some percentage of the peak (maybe 90%), then
use this to drive a schmitt trigger latch.  You set it high with (hopefully) the biggest peak in one
polarity, reset with the biggest peak of the opposite polarity. The original works well as long as the input
signal is strong, and stops working (rather cleanly btw) as the signal drops.  My variation tries to keep
working right down into the soup...

The other scheme is to use a single peak detector and a sample/hold (of sorts). I got this from Electronics
magazine design idea (cir. 1980 or so) and this method is the basis for the Calman Gold patent (New England Digital).
I've used this one in my present guitar synth. 

Opamp count is about the same (maybe 5 - 6 opamps). 

The freaky thing is that the simulation outputs for my improved Gentle Electric method and the 1980's method are
virtually identical in output. The simulation traces lie almost on top of each other.

The peak detecting method has the disadvantage of having some phase jitter as the input signal has multiple peaks
on one side (or other) of the waveform.  The up-side is that this phase shift distortion is not as obnoxiuos to deal with
as most other methods, which tend to 'octave-hop'.  

The methods can be used over a fairly wide range of input frequencies (unlike the Roland analog method, which can only deal with
a single string).

H^) harry


----- Original Message -----
From: Magnus Danielson <magnus at rubidium.dyndns.org>
To: synth-diy at dropmix.xs4all.nl
Sent: Thu, 21 Jul 2011 11:08:28 -0400 (EDT)
Subject: Re: [sdiy] too quiet...

On 07/21/2011 02:26 PM, Harry Bissell wrote:
> I haven't seen any messages for two days, so I wonder if everyone is
> out of town (or the list is down).

Everybody is charging their capacitors, and the cycling takes many turns 
of alcohol, pole-cleaning, pondring, lie out in the sun to dry up etc. 
essentially all the fundamental aspects of Yak-shaving.

So yes, the list is down for maintenance and what you see are ghost 
emails haunting you.

> I'm working on Guitar Synthesis using the Electro-Harmonix Deluxe Octave Multiplexer
> tracking filter with some enhancements (true opamp integrators in the VCF, changing the
> filter tracking point on the cutoff slope, and two methods of processing the resulting
> wave to a square.
>
> It seems (in simulation) to be working very well...

It always work in simulation, this is what simulation companies make 
sure, that the customer quickly think they have achieved something.

Cheers,
Magnus
_______________________________________________
Synth-diy mailing list
Synth-diy at dropmix.xs4all.nl
http://dropmix.xs4all.nl/mailman/listinfo/synth-diy

-- 
Harry Bissell & Nora Abdullah 4eva