Maestro FSH-1 explaination

[mth5044]

Anybody care to give a little explaination as to what is going on in the circuit?

http://www.tonepad.com/getFile.asp?id=138

It seems there are building blocks in there not common to the normal stompboxes.

Thanks!

[PRR]

Let us start with what building blocks you do recognize. It would waste everybody's time to type-up stuff you already know.

[mth5044]

Good idea!

I understand the power section with the 1044, where the +/- 9 and 8v's are coming from.

The part with IC1(a&b) - LM13600 - based on the CV labels a look at the data sheet, I'm guessing it is some kind of VCA / filter? The resonance takes the effect (filtered) signal and feeds it back into the.. filter. This is the circuit block that effects what frequencies are being filtered. Is there a way to switch the frequencies (high, low, band)?

The input/IC1(c) section I'd imagine is a buffer using the on board buffer of the LM13600.

The filter part on the bottom left takes signal from the input, and uses that somehow to control the voltage of the filters in the LM13600 section. Attack and decay control the come up and down time, respectively, for the voltage to the filter. What does range do? What does the direction mean (up, down, no direction/middle pole)? What is going on with the signal there?

The SH side looks like a convoluted LFO stuck between a FET and a hard place. Not any idea really what is going on there, except it seem to randomly generate a voltage.. possibly?

What is that common bit of circuitry to both modulation sources on the switch?

A decent amount of that I had just figured out while typing Thanks for making me type it out and think for a change. Can anyone fill in the blanks?

Could a normal LFO be substituted in to control the voltage on the filters? I guess normal needs some defining. Perhaps an LDR/LED with the resistance varying the voltage? I guess that bit of circuitry common to both the filter and the S/H is tricking me too. There seems to be a lot of problems with the S/H side of things, but it sounds like it is making random voltages for the filters. Perhaps that problem could be solved with a uC? I can do some coding on picaxes, but as I said I'm not totally sure what is happening in there.

Danke.

[Mark Hammer]

The basic principle of analog "sample and hold" units, whether they are for guitar or anything else, is that a capacitor is allowed to charge up, and then "sealed off" with a FET or other gate-like device on one side, and a high-impedance op-amp (or FET) on the other side.  Having only a path to ground and the other end essentially floating as if unconnected, the cap will hold that charge for as long as its own leakage properties dictate.

When does that cap get charged up? (i.e., when is that sample sampled?)  At a point dictated by a pulse sent to the FET/gate.  The FET/gate is actuated by the pulse and the cap has a brief opportunity to either discharge (if its current voltage is higher than the new one being sampled) or charge up even higher.  After that moment, the cap is once again "sealed off", and we have a new sampled voltage.

Part of what you see in the FSH-1 schematic is a noise source, to provide random voltages, or at least a source that is aperiodic.  Another part of what you see is simply an LFO that provides the pulses which briefly turn the FET/gate on to sample that random source.

The storage cap is like a food server.  The LFO is the person who rings the bell at the window between the kitchen and the counter and yells to the cap "Club sandwich. Pick up!".  The noise source is the person working in the kitchen who keeps cranking out food.

Bucket-brigade delay chips sort of do the same thing, except for a few differences.  First, the source being sampled is not noise but an instrument or mic signal.  Second, the rate at which samples are being taken, and briefly stored, is much higher/faster.  Third, where a S&H has one storage element, whose output is fed to many possible different destinations, a BBD has many storage cells, and passes the sample (or what's left of it, because those little capacitors in the chip leak pretty fast) from one storage cell to the next.  Where the clock/LFO in the S&H unit simply "instructs" the circuit to take another sample, the clock feeding a BBD instructs it both to take another sample AND to feed that last sample to the next cell in line.

The storage caps used in S&H units are selected for their leakage properties; specifically very low leakage.  Because the caps in BBDs have relatively poor leakage properties, there are limits to how slowly you can clock BBDs.  Any slower and the sample may have already leaked out enough that the sample no longer reflects the original input signal.  That is also one of the reasons why repeats in an analog delay tend to decline in audio quality; there is enough leakage of each sample every time it has to stay in a cap for a moment that after 4 repeats, it has been shifted from cell to cell - and leaked a little each time - some 16384 times.

There are a number of other things that happen with slower clocking, but our focus here is on the sampling and holding of an analog signal.

The noise source used has an impact on the degree of variation in how the filter moves around.  If the noise source is heavily amplifed, then most of what is sampled may be near ceiling or floor, with little variation in between, just constant ah-ooh-ah-ooh.  If the noise source is low amplitude and lowpass filtered, there will also be less variation, more like ooh-oh-oh-ooh.  If the sampling rate is fast enough, and the noise source lowpass filtered, then the next sample won't be than much different from that last one, just a bit higher or a bit lower.  And so on.  You can alter the feel of the S&H output, and what it is driving, by monkeying around with the parameters of the noise source that is being sampled.

If a person wants to get fancy, it is also possible for the gating pulse itself to be delivered in a less predictable fashion.  Imagine you had a 2nd noise source, hooked up to a comparator.  Should the noise signal momentarily exceed the reference voltage of the comparator, a pulse gets emitted and the S&H now snags another sample.  You won't know in advance when that will happen.  Or, the gating pulse could be delivered according to a pattern, via a PIC or sequencer, etc.  However you do it, the basic principle is the same: a random voltage source is sampled, and that sampled voltage used to produce novel, unpredictable changes in somethng else.

Old mono analog synthesizer keyboards worked the exact same way.  A voltage would be produced by pressing a given key and dividing down a source voltage, and that voltage sampled and held, and used to drive an oscillator, and maybe some other thngs as well.  Many old synths would use the keyboard voltage to feed the onboard filter as well, so that the filtering was always adjusted according to the key being pressed at the time.

One of the features implemented on many of those synths was the transition between one sampled keyboard press and the next.  This was normally referred to as "glissando", the degree to which one moved immediately and directly to the next note or swooped up or down to it.  Basically, one is simply adding some lag to the output of the sampled voltage.  The same thing could be done to the FSH-1, such that it isn't quite as stepwise, but behaves more like a wah that can't really make up its mind, or a wah with a drunk operator.

make sense?

[jdub]

Mark, your explanations are awesome.  I always feel a bit smarter after reading them 

[mth5044]

Mark! Your explanation is fantastic. I wish data sheets were written that manner  Thank you for the very detailed explanation on the SH works. Smoothing out the random steps to glissando sounds pretty interesting. I'll have to do some searching to see what kind of circuit is responsible for filtering off those edges.

I was checking out the LM13600 data sheet, and they have a similar set up what is found in the FSH, with the bandbass tapped off from after transistor between the two stages, but also have a lowpass out at the end. See page 13

www.komponenten.es.aau.dk/fileadmin/komponenten/Data_Sheet/Linear/LM13600.pdf

Looks like it could be possible to replicate the fet section found at the end of the first amplifier at the end of the second and have a seperate tap our for lowpass filter. Is there an option for highpass?

Another question, what control voltage is the chip looking for? +/- supply (8v)? Probably not. Maybe +/- 5v? Couldn't seem to find much info on that in the data sheet.

Thanks again.

[PRR]

> I had just figured out while typing

Yes, that was a trick to expand your conscious understanding as well as mark the borders you want to cross.

Mark has obviously spent too much time working with Sample/Hold.

The filter is "two integrators with feedback".

The S/H and the 2-integrator ideas are not well explained in modern datasheets because they are very basic concepts (in analog computing) which were sorta well explained in 1950s texts and 1960s application notes.

The OTA chip does not have a control Voltage, it has control Current, zero to ~~1mA. Conventionally we have a voltage, so we use a resistor to turn that voltage into a current to feed the OTA. When you have more than one OTA, you really should use more than one resistor.

"The filter part on the bottom left" is an Envelope Follower. A signal rectifier and a filter to smooth the 440Hz jiggles and just follow the loud/soft of the signal. There is no perfect Envelope Follower, especially for plucked strings; also you may want "imperfect" following for musical effect. So two knobs to diddle.

> What does the direction mean (up, down, no direction/middle pole)?

This should be explained in the Owners Manual? Using customary assumptions of polarity, the filter picth rises with volume. Or you could invert that, pitch falls with volume. In either case you want limits on how far pitch goes. Zero voltage can be one limit (there is a constant bleed current so the OTA does not go to zero frequency), but you need another limit so there is a voltage divider.

> Could a normal LFO be substituted in to control the voltage on the filters?

You can adapt ANYthing. Having an early fully-modular synth makes it easier; my ARP had separate noise-source and S/H and LFO and sequencer, all scaled to compatible voltages. Sometimes the scaling (to make 20mV VCOs and VCAs take 0V-10V control) was more stuff than the actual guts. Sometimes if you limit how many things it will do, you can save a lot of parts-- this board is scaled for the two sources (Envelope Follower and Sampled Noise), not to any "standard interface".

What the common-stuff does, I didn't study. Maybe with new clues, you will figure it out for us?

[Keppy]

I just built a variation of the FSH-1 that uses an LFO to drive the filter. I found that the filter circuitry loaded down the LFO somewhat, so I buffered the LFO out into the 22k resistors. YMMV.

[mth5044]

PRR - thank you for the further explanation. I had seen the bit about current control in the datasheet, but the 1mA seemed a little strange so I wasn't sure.

Keppy - do you have a schematic of the variation you built or what the LFO looked like/voltage swing? Good to hear that it works!

[Keppy]

LFO is a slowed-down CE-2 triangular LFO. I ran it through a transistor buffer an then through a 100k(?) resistor to get CV on the FSH-1 schem. The great thing about the FSH-1 being current- rather than voltage-sensitive is you can set the upper range just by changing that series resistor.

[Mark Hammer]

LFO is a slowed-down CE-2 triangular LFO. I ran it through a transistor buffer an then through a 100k(?) resistor to get CV on the FSH-1 schem. The great thing about the FSH-1 being current- rather than voltage-sensitive is you can set the upper range just by changing that series resistor.

Excellent idea!

[Keppy]

Thanks Mark! That means a lot, coming from you. If I finish it in time for the page 1000 competition, I'll get some clips posted.