[sdiy] MC1495 multiplier chip discontinued completely

[jhaible]

Good overview, MC.
A few additions:

> The Fet sign changer is simple, but only useable for "effect" sounds.
> I.e.: very agressive modulation, basically the same as a square wave
> driven real multiplication device.
> So it's use is very limited. Those "soft" sounds I have mentioned
> are not available at all.

But there is a FET based RM that uses the FET as a VCR, and not just as
a switch. As used by Korg. I have copied this circuit to preserve the
PS-3x00 series own character. It's part of this drawing:
http://www.oldcrows.net/~jhaible/polykorg/jh_3200_signal_3of4.gif
Look for the part between "direct out" and "AM out". The LDRs just perform
a wet/dry mix; the multiplication is done with a 2SK30AFET.
This is *not* clean (the modulator wave is quite distorted. The PS-synths
don't even have a sine wave modulator, so a distortaed triangle might be as
good as an undistorted triangle ...), but it's certainly not a crude square
wave
modulation, either.


> The 13600/13700 do have the problem that they are only two quadrant.
> You can overcome this with using two vca devices, each for two quadrants
> will make four quadrants. The problem is then, like in B type output
stages,
> the crossover distortion. This kind of distortion will not go away
> if the signal gets smaller, so no good prospect.

Come on: forget about the second VCA device, and you're forced to go for
"class A" (stretching the terminology to the mixing part, not ´peaking
of the OTAs interior), so you kill two birds with one stone.
Here's the circuit:
http://www.oldcrows.net/~jhaible/tonline_stuff/jh1_ring.gif
It is not clean either, and not temperatire compensated, but you can
overcome both problems if you use a linearised VCA instead of the
3080.


> Perhaps there is a possibility to have a tricky bias scheme,
> similar to A/B or A operation of output stages. However, a lot of mirrors
> are involved, much more circuitry than the normal gain cell
> has, I do not like that idea.

The most funny thing in terms of RM design is that there is a circuit
which deliberately uses class B (a real gap, though ever so tiny) instead
of class AB or whatever. That's from ARP (as used in the 2500
or 2600, maybe in both, I don't remember), and it has a reputation
of being especially *clean* !!

Why? Because clean can mean two different things: No higher sidebands
from nonlinearity and/or no carrier bleedthru when you stop playing.
Obviously the latter can be achieved even without perfect trimming
if the *signal* (not the modulator) is fed into the input that suffers
from the dead band.


> The problems of symmetry also applies to the last one, RC4200.
> They are still around at Farnell, but I need to try them.

You need a lot of resistors, and a lot of trimming, if you want to make this
a decent 4Q multiplier. And be careful about the impedance match of
signal inputs and the other inputs used for trimming. These are bipolar
devices, and you can ruin all the trimming with drift of offset _currents_.
If you really want the trimming to be good, you need 3 resistors and a
capacitor in addition to the trimpot, and this all x3.
(Two resistors to divide down the trimpot voltage, the 3rd resistor
to increase the impedance of the divider, as seen from the IC's pin,
and a capacitor to make the pin a GND for high frequency.
(I know that the MS-50 and also the circuit in Barry Klein's book
uses a simpler approach, but you loose some of the chip's specs
that way.)

JH.