School Me on FSH-1 Noise Generation

[Ripthorn]

I have wanted the sample/hold sound for some time, but have heard horror stories about the build. I was looking at the schematic and I am trying to understand the noise generation section. Specifically, the noise that gets output to the VCF section. What is the voltage range? Is it +/-9V? +/-8V? 0-9V? I'm just trying to understand this section at present, but it isn't making sense. Someone school me on it!

[Mark Hammer]

It is pretty much standard practice amongst modular analog synth builders to install a socket and try out different transistors for optimal/"best" noise.  Even within the same numerical transistor type (e.g., 2N3904)  individual units will vary with respect to noise.

So it is very difficult to suggest or declare a specific noise range from the transistor itself.  Having said that, you will note the 10k trimmer, which is there to adjust for differing noise levels.  You will also note the presence of a 1M feedback resistor in that op-amp stage.  If the noise produced by the transistor is insufficient, you can always increase the value of that resistor.  The trimmer is then used to achieve a range of bleep-bloop-blurp that is pleasing to your ears.

Traditionally, the cap used to hold the sampled noise voltage should be a very low leakage type, preferably a good quality plastic type.  The idea is that, once sampled, the noise voltage is "held" in the cap, with nowhere to go, due to the FETs on either side.

I found that an interesting sound could be achieved by undermining the degree of holding with a bleed-off resistor; what I referred to as a "droop" effect.  Resistances in the >500k range are required for this.  You can hear a poor demo of it around 2:05 here: https://www.youtube.com/watch?v=YbWtreDE7tM  I get the droop and stock S&H functions mixed up.  The stock arrangement is more of a stairstep, while the droop is a sort of reverse glissando.

[PRR]

> the noise .... What is the voltage range? Is it +/-9V? +/-8V? 0-9V?

It is random. For a specific part, random over time (what we want). And as Mark says, different for different parts.

Because it is cap-coupled, we can assume it averages-out to symmetrical (+/-) peaks.

IIRC, such a thing with a broken emitter and an opamp booster makes maybe 100mV average. Peaks are statistically unlimited, but you may wait forever to find a peak much over 1V.

Obviously the hiss voltage can be changed with the amplifier.

There are other ways to make noise. Both halves of TL072 may, IMHO, be more predictable in level and 1/f corner. There used to be a pseudo-random chip that repeated every 2 seconds. R.G. and others have recently been thinking about doing that on a one-buck CPU with repeats longer than the universe. 

[Ripthorn]

I am tinkering with replacing the noise section with a microcontroller that would allow for pattern, multiplier, and tap tempo functions. I have several things working, but I don't want to produce too large of a voltage. I can drive 0 to 5v directly geron the uC, but would need an opamp stage to get negative voltages.

[Nitefly182]

It helps to go into transistor selection with an open mind as that transistor can heavily influence how your specific build works. Its a good way to get a more interesting version of the effect than what all the commercial versions provide. My experience was that you can get a much more electronic sound that latches onto a specific frequency for a few beats before moving on. The completely 'random' burble can get boring after a while.

[skyled]

The FSH-1 is an easy build when it is built properly. Most of the layouts have the noise transistor set up wrong.

Ciaran Haslett28 September 2016 at 10:26

Few notes...

Ticks like a ticky thing ticking! Even when bypassed before a high gain pedal. Going to try our usual tricks and see what happens.

The floating base has to be wrong. I've read thru every FSB post (Nocentelli's included) and can confirm that there must have been a schematic error thats been continuously copied from one version to the next. Tying the Base to -9V (the collector socket in this layout and float the collector) gave instant results! People here and on Madbean have constantly reported that tuning this thing was a nightmare....but with the collector now floating it's a breeze.

The Top trimmer now seems to set the range of CV that will be triggered (i.e. how low and high the filter will open) and the right most sets where within that range it will open (kinda like a depth range). Instant results with every BJT I tried. The gain does have an effect but like I said, you can redial the operating range in quite easily.

Don't use multi turn trimmers....it'll break your heart!

All in, a great sounding effect. Cheers lads
Reply
Replies

    Rory Crane31 July 2018 at 22:06

    I just wanted to add more support to the noise transistor being set up wrong in the official build. Float the collector, not the base.

    Here's a little article on the subject:
    http://aquataur.hilpold.net/aquataur/musicstuff/fsh-1.html
    MAO30 September 2018 at 21:59

    I also tied the base to -9v and floated the collector, and it's much easier to dial in. With more range than I remembered vs the original pcb.

    Thanks for the tip Ciaran and Rory!

Taken from https://tagboardeffects.blogspot.com/2016/04/madbean-sharkfin-maestro-fsh-1-filter.html

As for making a custom noise section, I can't help you there.