Sample/Hold questions.

[digi2t]

I've been interested in incorporating the sample/hold angle into my Infinitphase for a while now, so I've been breadboarding for a couple of days now. Mike, over at Prophesysound, was cool enough to point me towards what they've used in the original units, so that's what I've been futzing with. Everything shown below is running off the +/-15v of the Infinitphase, and on the breadboard, I'm using another LED to see the output of the s/h.

The s/h section is LF398 based, lifted from Rene Schmitz page;



The noise source, Mike recommended is this one;



So far, I've managed to get it working, but not as it's drawn. This has me wondering if I'm doing something wrong. In the s/h section the (I'm assuming) rate LED, just above the BC548, would blink, but not in an on/off fashion. It was more of a bright/less bright type of deal. I decided to take the lead above the 2K2 resistor from the +15v, and switched it to ground, and now I have a truly blinking LED. Why does the diagram show connection to +15v?

Another thing, the 10K resistor, just after the 470p cap in the s/h. Using it where it is drawn, kills the s/h output (output LED stops flashing). For fun, I moved it to before the cap. The s/h seems to work, but judging by the output LED, the best result seems to be without the resistor installed at all. Why?

Have I done something wrong here, or am I on the right track? I mean, it seems to work right now, but not as drawn.

Insofar as the noise section goes, it seems to work fine, but requires a handpicked transistor to get random goodness. So far my best candidate has been a 2SA1319, but the cool thing is that the feedback trimmer, just above the first opamp, allows me adjust the gain for different transistors. This kind of opens things up a bit as to transistor choice, but as with the Maestro, this part is simply plugging in every transistor in my arsenal, one by one, and looking at how the LED dances.

[slacker]

It sounds like you're doing something wrong around the BC548.  The LED can't flash with the 2k2 connected to ground because the emmiter is also connected to ground so there's no where for any current to come from. My guess would be that you've connected the emmiter to -15V instead of ground. Maybe the same thing with the 10k?

[digi2t]

Dang!...

The power supply of the Infinitphase is actually +15/0/-15. I'm using the 0 as ground, rather than the case ground. Could this be the problem?

I'll recheck my power connections. Bizarre though, that the LED is flashing normally. 

[bluebunny]

The LED can't flash with the 2k2 connected to ground because the emmiter is also connected to ground

The 2K2 is connected to V+, not ground??

[slacker]

The 2K2 is connected to V+, not ground??

It's supposed to be, Dino said he connected it to ground though.

[Mark Hammer]

If you can find a way to add variable drain/decay rate to the sampled control voltage, see if you can do it.  Generally accomplished by means of a medium-to-high resistance in parallel with whatever is used as your buffered storage cap.  I did this to my Maestro FSH, and its a heap of fun.  Sounds like a space raygun: pyew, pyew, pyewww.

[digi2t]

Friggin' hell!!! I have the LED flipped around.   I guess I needed you guess to grill me a bit to open my eyes. Thanks!  

In any case the LED flashes the way I would want it to flash now, that's to say completely on/ completely off. The other way it only slightly dims in intensity as it cycles.

This is what is on the board right now (with flipped LED).



The 10K resistor is out, it's a real killjoy. The noise source works as advertised, and I'm getting really nice randomness on my output LED. Video anyone? I can shoot a quick clip if anyone is interested.

EDIT: Maybe the 2K2 needs to be adjusted for the LED to flash properly the other way? Gain of the transistor at play? Tried it, didn't work. My modified drawing works.

[duck_arse]

I guess I needed you guess to grill me a bit to open my eyes.

grilled digit? what would (the other) DA say about that?

yr diagram still shows the 2k2 to ground, tho. how does the led forward bias?

[edit :] and isn't that nand following the 470p effectively an un-terminated/un-pulled input?

[slacker]

What he said, that shouldn't work, the 4093 shouldn't be doing anything with it wired up like that either. Something in what you've built doesn't match what you've drawn, or something funky is happening that's putting some Voltage on the inputs of the the 4093 and the collector of the BC548. What voltages do you measure on the transistor pins?

[digi2t]

OK OK OK.....

I'm a dumb-ass. I had, for some unbeknownst reason, stuck a jumper in between pin 4 and 5 on the 4093. 

Everything works as it should, and all is right in the universe, no thanks to me.

Thank you for bearing with me folks.

Next step, integrate this into my Infinitphase.

[digi2t]

Hey... I wonder how this would sound in an Uglyface? You know... sample/hold the Vactrol, instead of the normal LFO?

[bluebunny]

The 2K2 is connected to V+, not ground??

It's supposed to be, Dino said he connected it to ground though.

Oops!  Missed that bit, Ian!      As you were...

[digi2t]

Alrighty, I've got a decent s/h going, got a vero layout for it done. I tested it with the Infinitphase, and if I tap in right after the CV pots, I can use a switch to switch between CV control, or s/h.

With that said, I was wondering if I could get this to work at +/-9v, because I think it would be cool to pop this into an Uglyface. I still have it on the breadboard, so I used two batteries for the bipolar supply. I think my problem is the noise source, since it's originally designed to run at 15v, which is fine for the Infinitphase. I'm tinkering with it right now, and I'm fine with using a 1044 (or similar) for the +/- 9v, but can someone help me out with the adjustment to the lower voltage? This is the total circuit I'm looking at right now;

[digi2t]

OK, I think I found something that works fairly well. Simple 386 white noise generator to replace the 15v noise source;



So I grafted into service, but found that the 386 "wide open" was over powering the LF398's input. I replaced the 100K trimmer with a 1M, and moved the voltage source over, so I could tone down the voltage feeding the 386. Once I find the point where the 386 will JUST trigger the LM398 with the 1M, I could then use the 10K to get the chosen transistor to "dance" just right. It looks like this now;



Finding the right transistor is a matter of trying a bunch until you find the one that dances best. For now, I have a PN2222 in there, and it dances quite nicely. I tried a bunch of different transistors, but so far, it's the best one. Now I need to see if the LFO in the Uglyface will work as the trigger. If it does, it will be interesting to hear what sample/hold on an Uglyface will sound like.

[JebemMajke]

THis looks cool.

Do you have any sound samples?

[digi2t]

THis looks cool.

Do you have any sound samples?

Not yet. First priority is the Infinitphase, going to try to get that done this week. Sound sample will follow. Uglyface angle is more curiosity than anything else. We'll see what happens with that, as I get smarter.

[anotherjim]

I think the original noise circuit should work with +/-9v. C33 and Q5 ground would be changed to -9v. C32 blocks the different DC level from the opamp input. I don't think the extra 3v would cause any trouble. Noise ought to be better than the single +9v solution since you'd have 18volt, although you could have 18v on the transistor (ground change to -9v), but you'd need a decoupling cap to pin 3.

[digi2t]

I think the original noise circuit should work with +/-9v. C33 and Q5 ground would be changed to -9v. C32 blocks the different DC level from the opamp input. I don't think the extra 3v would cause any trouble. Noise ought to be better than the single +9v solution since you'd have 18volt, although you could have 18v on the transistor (ground change to -9v), but you'd need a decoupling cap to pin 3.

How big of a cap on pin 3? Do you think -9 on c33 and q5 would cause any problems? (I know, I know... see what the breadboard says  )

If the first version works, that would be interesting though. If I use a quad, I could use the third op amp to inverse the voltage from the 398, making it friendly to the Uglyface.

I tried the s/h on the Uglyface yesterday, and I must admit, it's addictive. It sounds really cool.

[anotherjim]

I'd try a non polarized cap first, 100nF could be good. The 386 input self biases via 50k to ground, and the noise transistor will be someplace around ground with +/- supply, so there's no guessing the polarity for a polarized cap.

[Mark Hammer]

I think the original noise circuit should work with +/-9v. C33 and Q5 ground would be changed to -9v. C32 blocks the different DC level from the opamp input. I don't think the extra 3v would cause any trouble. Noise ought to be better than the single +9v solution since you'd have 18volt, although you could have 18v on the transistor (ground change to -9v), but you'd need a decoupling cap to pin 3.

How big of a cap on pin 3? Do you think -9 on c33 and q5 would cause any problems? (I know, I know... see what the breadboard says  )

If the first version works, that would be interesting though. If I use a quad, I could use the third op amp to inverse the voltage from the 398, making it friendly to the Uglyface.

I tried the s/h on the Uglyface yesterday, and I must admit, it's addictive. It sounds really cool.

Now you have to do a "droop" mod.  Put a medium-value resistor in parallel with that 1nf cap to ground on the 398 chip to let the sampled voltage drain off at a modest rate.  Start with maybe 47k and experiment from there.  The ideal droop rate depends on the sample clock speed.  So, slower clock/step rates can tolerate slower droop rates.