original and ingenious circuits ... (was: in How the heck does the MS-50 filter WORK?)

[Magnus Danielson]

>>>>> "LC" == List, Christopher <Chris.List at sc.siemens.com> writes:

Hi People!

Thanks Juergen for forking of a interesting thread. You are right,
there is something special about these odd solutions. I just love to
find these myself, and even how bizzare they may get they may still be
fully adequate for it's application at the time it was created.

 LC> Actually, one of my favorites was one you told me about! The ARP (I
 LC> beleive that's who it was) cv quantizer. I had always been thinking that
 LC> the way to quantize was to run the input through a ADC then to a DAC
 LC> with the low bits disconnected - this gets very expensive very quickly
 LC> if you want multiple channels. The ARP method, OTOH, uses a fixed
 LC> frequency stepped voltage oscillator (that could be made from - among
 LC> other things - a counter and a DAC) to drive a comparator + S&H. As soon
 LC> as the input voltage crosses the stepped voltage, you sample and hold
 LC> the stepped voltage at the output. You can have as many channels as you
 LC> want, just by adding more comparator + S&H elements connected to the
 LC> same stepped voltage! That's cool!

This is a interesting twist on things. I like it. For a one channel
application you could do with a quick counter (or better, an SAR) and
stop at the level and keep this until next sample period. If one use a
slow enougth lowpass filter at the output will the approximation
period hardly be noticed and you get a simpler S/H cursuit.

This made me think of a box that I think is about the most bizzare box
that I have seen, and I don't that all people really know all the
twist on this one... it is not all analog, but anyway, it's the
Eventide Harmonizer H949 from 1979. It will do a little set of
programs and has some digital and some analog program parameter
changes. There is many things to be noted about this box, but here is
a few.

In order to pitch the sound it needs to be sampled and played up, for
this will the sound first pass a 7-pole Butterworth lowpass filter and
then be compressed through a dBx compressor card. It is then being
sampled at a rate of about 50 kHz. In order to A/D convert they
kick-start a rampwave that is being compared with the sampled
value. At the same time that they started the ramp a counter started
counting frenetically. When the ramp reaches the sampled voltage is
the counters value stored into memory (which is very early DRAM). Now,
a fast excercis with numbers will tell you that to get a few usefull
bits there can be a terrible number of counts, and this is also how it
works. To get the 11 bits will the counter need to count 2048. Since
2048 counts every 50 kHz cylce or so we need the counter to tick in
about 102,4 MHz. This frequency is generated by a LC-oscillator and a
couple of transistors. The frequency is trimmed by adjusting the core
of the coil. There is both ECL and STTL running at this frequency,
since it will also control most of the other digital clock as well.
You can feel the LSB of the counter since this will run at 50 MHz and
head up the LSB DRAM from it's only input pin!
The highspeed clock is also derived down to a base clock for a 16 bit
2901 bit-slice ALU. This ALU section is together with many other
functions controled by TTL PROMs with a width of 48 bits (VLIW!) which
runs in a looped sequence. Some of the program buttons will modify the
bitstream from the PROMs prior to reaching active logic.
The ALU will not touch data itself, it will only be able to read at
various points in memory. Actually, the ALU will only perform address
calculations. The data read by the ALU is feed to CMOS FIFO banks
prior to reaching the D/As. The output of the D/As is feed to 7-pole
Butterworth filters and dBx expander cards. There is two outputs, main
and delay. The output signals can then be mixed and EQed before beeing
sent back into the input path, this allowing some wierd stuff.
On the control side one should notice not only the manual dial, but
also that the backside contains both frequency tracking and CV inputs
(yes, you can hook it up to a modular!). Both the main and the delay
outputs can be assigned a fixed delay with buttons (you punch in the
delay digitally on 6 buttons per output). On the positive side is that
both input and outputs are fully balanced on XLRs.
I even have a extra algorithm card which I still doesn't really know
how it works.
The analog control side of this box contains 2 SSB generators, which
then are used to control the ALU (obviously).

There is an interesting button on the input labeled REPEAT, which
actually causes the memory to stop updateing, you have just sampled
what was in the memory and are able to play it on you CV kerboard!

The box was marketed with a said dynamic of 96 dB.

This box uses as said a fine suite of ECL, TTL, STTL, LSTTL,
4000-CMOS, 74Cxx-CMOS, analog, radio to combine into a really odd
machine.

To me this is among the more oddly once that I ever encountered in the
audio fx department. It will fills it's 2HE rackrange very nicely.
It is a bit cumbersome to get up stable, but I think it is time to do
something to mine... there is even a track record on mine so I think I
know a set of albums where it apears, of course there is some
interesting effects done!

I still lack a users manual for it, but I should have the schematics
somewhere around (I looked but did not find).

The LFO in the EH SmallStone is a nice one as well, it is based around
one CA3094. Hell, I like the rest of the SmallStone as well, I just
lack a few features...

Cheers,
Magnus