Four quadrant multipliers:Results of my searching

[Magnus Danielson]

From: harry <harrybissell at prodigy.net>
Subject: Re: Four quadrant multipliers:Results of my searching
Date: Sun, 24 Dec 2000 00:44:11 -0500

> Hi Kids:

Hi Harry!

> I like the "matter of taste" comment. The advantage of the "no trim"
> solution is if you are building a commercial lot, you do not have to
> run units through very expensive "calibration" department where
> everything must be done just so by high-priced employees... or done
> poorly or not at all by low priced employees.

Rigth. Every trimmer has a price tag. But then, if you have to
calibrate your machine, the more advanced support setup in instruments
you need (and I guess that most people isn't up to my level of
instrumentation, a thing to recall as important) the more expensive
(time and money-wise) will the trimming be and the less frequently
will the trimming ocur.

For instance, I have just done my Christmas trimming of my OB-8 (OK, a
few trimmings left to be done, but it was running late, I will do them
now). It is truely fashintating to see how much effort Oberheim put
into making much of the trimming automated or microprocessor
assisted. The VCO trimming is simplistic and many offset errors takes
support of a DVM to get done. If they only had put the microprocessor
into use when trimming the filters...

> I have trouble with getting cals on a 4-20 mA current loop output. This has
> a Span and Zero cal... which interact. Sound familiar ???  ;^)

Yes. But one of the "problems" is that most synthesizer designs has
the offset value scaled, so trimming becomes unecessary
problematic. If you instead have the offset parameter AFTER the scale
trimmer, then you start of with a scaled contribution of 0 V and then
adjust the frequency of the lowest tone (C0 for instance) and then
onces that is set, you only have to bother with the v/oct scale. This
was scaling can be done accurately and easy. The reason to have a
offset trimmer before the scaling is to offset the scale on a
keyboard, that can be done without affecting the oscillator core
offset. Most oscillator designs allows for simple insertion of an
unscaled offset trimmer with minimalistic component addition.

The moral is, when one design a curcuit, consider how to calibrate it
by easy methods and with minimal of supporting instruments.

Requireing a DMM migth be acceptable in most cases. Requiring a
Spectrum Analyser might not be that good. But if people have the fancy
tools, don't forget to give sufficient hints on what you are really
trying to acheive.

Another classical fault is to use trimmers reaching over a too large
range, so that the usefull range is only a small portion of where you
can trim. Given a certain precission in your mechanical positioning
abilities, those errors will have greater effect than if you had
proper ranges. Also, mechanical repositioning due to vibration, chocks
and even gravity will be allowed to contribute to larger errors. I
prefer multiturn trimpots over 300 degree pots. Then again, some
prefer having no trimpots at all ;)

> My Rev3 Prophet V  had a reduction from 160 trims (earlier units) to only 80
> !!! I added
> 40 trims back in because I insisted that the Poly-mod section track from
> note to note so
> I could do "bagpipes" and other calibrated offset bends.  It was a bitch to
> calibrate.

Hehe... adding trimers can make trimming much easier, if done well.

> Yo Magnus and JH. Face it. You are just like me. You will buy the laser
> trimmed part because it is tighter tolerance to start with, and you
> can hear, want, deserve, and can afford ? the best. Then you will
> decide that it is not "good enough" yet.... and add the trims
> because it can be BETTER.  So you will have an awesome result and
> few will follow in your footsteps... or even see your footsteps....   ;^)

Yeap. I have to admitt, there is some obscure pleasure in being waaaay
over the required spec ;)

> If you want fun... modify one of Bob Moogs designs. I think he gets perverse
> pleasure out of making every component do THREE things at the same
> time, so you can't change one parameter without screwing up two
> others !!! ARRRGGGHHHH!!! 

That's typical of doing economical designs. That is an artform of its
own and doing it good is in itself worthy of respect in my book.

> PS:
> Hey Magnus: when they come out with a "Divide by Zero" be sure and let me
> know... (or divide near zero, through zero...etc.)   ;^)

Hehehe... yeah, I will then make the annoncement over my 1 TOhm
impedance speaker, will that be OK for you? ;)

Cheers,
Magnus