Tonepad's Filter/Sample&Hold troubles (not only).

[jrc4558]

Good day, gentlemen!
Yesterday I finished what appears to be my new favourite envelope filter. Its the tonepad's FSH-1. I built it will all the mods and it really sounds great with good tracking an nice timbre.

I didn't have a 1M liner potentiometer for Sample&Hold speed control, so I used a 500kB, doubled the size of .33µF capacitor (now ~.68µF) and decreased the 180K resistor to 100k. (Those are the current values in the feedback loop of the S&H oscillator).
I already found some info on the so-called "noise" transistor. I am assuming that that's the one with it's base lifted. I will try and select the one that will give the most pronounce effect, but just in your opinion, what could be the one t hat will do the job best? What parameter makes it good for this application? (right now I'm using a 2N3904).

I also have a problem with clicks from the S&H circuit bleeding into the audio path of the envelope filter. (I understand that the filter is same for both, but I can hear this bleed in both Filter and S&H positions).  It is audible only when the filter swings towards high frequency (for obvious reasons), and varies with the turn of the speed knob (it actually is in synch with the clicks of the S&H, so I assume it is in fact THEM)
I power this effect from the adaptor, well regulated and generally very quiet. I connect my negative lead directly to ground too.

What could be the reason for this noise? I have been looking at the schematic and my build (its on PCB, developed correctly, etc...) and found no mistakes so far...

[Ry]

Do a search about the ticking.  I have a few posts where I describe how I fixed it.  The problem (as I found it, FP disagrees) is in where the noise oscillator is located in the layout.  I moved that part of the circuit to a perf board and BAM, problem solved and it sounds fantastic.

Ry

[Mark Hammer]

The noise is apparently not from the number of the transistor (i.e., type/model) but from the specific unit.  In just about every circuit I've ever seen that used a transistor as a noise source in that manner, the accompanying text always encourages the builder to socket and try out a bunch, in search of the noisiest one.

[jrc4558]

Thank you gentlemen! I have some 2N2222A, as these seem to be so popular for the function.
Is there any theory writing on what makes the S&H section tick? I found some Sample and hold circuits,  but still didn't figure out how they work...
In any case, I'll try some several transistors tonite and report with findings.

[Mark Hammer]

Okay, go to http://www.tonepad.com/getFile.asp?id=96 for reference purposes and look in the lower right-hand corner.

The noise source (Q1) simply provides a random signal.  Think of the transistor as being like a barrel full of random numbers between 1 and 1,000,000.  IC3a provides fixed boost (because the noise/random variations aren't huge in amplitude) and the 10k trimmer fine-adjusts the range of random voltage changes (it could productively be a panel-mount control in some instances). 

The LFO/clock formed by IC3b isn't a whole lot different than the single op-amp LFO used in the MXR Phase 90 and similar circuits.  It provides a regularly occurring pulse that is high enough to turn on the gate of Q5.  When Q5 tujrns into what is a virtual straight wire (or tiny resistance), that allows the .05uf cap to charge up to whatever the random voltage coming off IC3a is at that moment.  When the pulse from IC3b goes low again and Q5 is turned off, the charge in that cap is essentially "frozen".  It can't drain off backwards through Q5, and Q6 presents a high enough input impedance that it can't drain off through there either.  So what happens is that Q6 ends up "seeing" whatever the present voltage is on that cap, until such time as the cap drains off naturally, or the LFO tells Q5 to let another brief sample pass through.  Note that when Q5 goes low-resistance, that is also a kind of opportunity for the .05uf cap to discharge via the 10k trimpot.  That's important because otherwise, unless the LFO was going super-slow, the cap would never discharge and random voltages would simply keep being added to what is already there.

And that, my friend, is how the noise is "sampled" and the resulting voltage "held".

The cap used for that .05 "holder" should be a low leakage type.  I forget which type that is, but I'm sure someone will chime in. (tantalum?)

S&H units often have a portamento/glide control, which can vary the extent to which the unit instantaneously changes from the one sampled voltage to another, versus gradually changes.  The FSH-1 lacks such a control, but incorporates a tiny bit of portamento via the .15uf cap you can see in the dead center of the schematic, and its accompanying decay-resistor provided by the 2M7 unit in parallel.  Variations in that combo would result in different rates of change between successive samples.

Alternatively, what one can do to kind of "de-jitter" the random changes is to stick a cap in parallel with the 1M feedback resistor of IC3a.  Remember that the transistor generates broadband noise, which is amenable to filtering.  Stick a .01uf cap in parallel with the 1M resistor, and the random noise is filtered starting above around 16hz.  This won't smooth the transition from sample N to N+1, but it will change the likelihood that sample N+1 will be radically different from sample N.  You can use whatever kind of filter cap value you want to change the overall pattern of random sample changes.

[Eb7+9]

The noise source (Q1) simply provides a random signal. 

I looked at an original VCF pedal a while back, if you probe with a scope you'll notice that "noise" circuit produces random FM of a squarish pulse - it's not the usual type of white/pink noise source as with Collector-open circuits ... having tried both on the sampler circuit I can say it has a slightly different effect on the sampler output - this I attribute partly to the exponentially decaying clock pulse not being narrow enough from top to bottom ... a sharper sampling pulse and better noise source/amplifier gives better (more random) performance IMO ...

the other places where I've seen this Base-open arrangement is in the fuzz-feedthrough suppression of the Acoustic360 and in the VCO's of the EMS synthesizers - in those cases you have to choose devices that have specific quality also ... non-oscillatory in the 360 case of course ... would be nice to come up with digital-random where outputs are fixed to harmonically related filter positions, maybe Zack has thought of this already ...

~JC

[Mark Hammer]

Hi JC,

From the sounds of it, the FSH-1 might be improved in peformance by subbing some sort of variable mark-space ratio CMOS for the simpler oscillator it currently has.  The on-period of even something with a 3% duty cycle might be longer than what the current one is, but when it's off, its OFF.

[soggybag]

I built one of these also. From what I remember of the thread a while back. Ry seemed to have gotten rid of the clock noise by moving IC 3 off the board.

I didn't like this solution as it would make mine too big to fit my box.

I keep thinking that that it might be possible to to fix the problem by cutting a trace or two and using a jumper, maybe shielded.

I don't quite get how the noise is getting into the system. Is it the clock pulse effect a neighboring trace? Or is the signal passing through the LFO side of IC 3 to the sample and hold side as the LFO signal spikes?

Speaking of short sharp signals, I found a schematic for a variable duty ocsilator made from a 555 timer. I built it on the breadboard and it worked well. I could dig up the link if anyone is interested.

[jrc4558]

Got 12 pcs of 2N2222A's. Wil try them all tonight.
Thanks for your help, gentlemen!

[Paul Perry (Frostwave)]

would be nice to come up with digital-random where outputs are fixed to harmonically related filter positions, has thought of this already ~JC

If you send a random voltage to a LED bargraph driver, then you can put different resistors on each of the outputs, to give different output levels.. choose the resistors right, and you get your musical intervals, I guess. kind of an arbitrary quantizer.

[Ry]

I built one of these also. From what I remember of the thread a while back. Ry seemed to have gotten rid of the clock noise by moving IC 3 off the board.

I didn't like this solution as it would make mine too big to fit my box.

I keep thinking that that it might be possible to to fix the problem by cutting a trace or two and using a jumper, maybe shielded.

I don't quite get how the noise is getting into the system. Is it the clock pulse effect a neighboring trace? Or is the signal passing through the LFO side of IC 3 to the sample and hold side as the LFO signal spikes?

I did indeed move IC3 and the noise transistor to a daughter board.  The problem that I found was that the traces running near IC3 picked up too much of the oscillator noise.  I tried cutting as many of them as I could on the board and re-routing them as shielded cables.  I tried creating a ground plane from a piece of un-etched PC board and running all of the grounds to it.  Neither of these fixes cut down on the bleedthrough in an appreciable manner.  After many hours of peering at the layout, I realized that the traces on the layout were picking up the oscillations.  The oscillator is placed in a fairly central location in the layout, so it has all kinds of chances to bleed into the signal path (which was what was happening).  So I moved the ocillator and noise transistor to a perf board.  I ran power and ground from the original PCB and the output of the daughter board runs to the switching transistors.  There is absolutely no bleed through now and the pedal sounds great.  I built one from the GGG layout (the original circuit) as well and I think that the Tonepad version is much better sounding. 

FP will disagree with me about the source of the bleedthrough, but after all the time I spent debugging it, I'm quite certain that it's the position of the noise source in the layout.

Ry

[jrc4558]

Success! 2N2222A turned this pedal into a really weird thing! Me like.
Noise was tamed somewhat by suspending the oscillator IC and decoupling it more with 47 ohm-10µF capacitors.

Thank you for all the advices!

[Eb7+9]

If you send a random voltage to a LED bargraph driver, then you can put different resistors on each of the outputs, to give different output levels.. choose the resistors right, and you get your musical intervals, I guess. kind of an arbitrary quantizer.

great idea Paul - thx (!!)

lately I ended up using the jFET-shunting LFO from my Nyquist circuit - it runs on 9volts and because timing is dominated by the "on" resistance of the FET as it discharges the cap - much lower than the time constant of the Maestro clock - it works well above 20kHz also ... though at lower speeds if the clock pulse is too wide, esp in a direct coupled filter circuit, you'll hear some noise tracking through the filter ... not necessarily bad or anything - I just wanted to see how clean a sampler I could get and this does it ... otoh, a wider pulse will make the sampler react differently to a given noise source - not sure if it's a little more random with a finer pulse but I ended up using a Collector-open noise source with a properly buffered amplifier for a less choppy noise source also ... it's less finicky to tune in the end ...

incidently, I think there's a little known "offset" issue with the Maestro sampler circuit ... notice that the minima on the sampling range, which directly sets how low the filter will go down to, will depedend on both the VGS(DC) bias voltage (providing a vertical offset from the sampled voltage at its Gate) and how the sampler altogether tracks the noise signal limits ... for this reason there's a good reason to try out different jFET's in the follower position of the sampler so as to better match how close to the audio cut-off point you can bring the filter to when increasing the range control ... with some devices you end up below the audio range at full noise signal range (providing no audio output at full range - ok, that can be kewl) but with another you might not get close enough to the lower limits of the audible filter range to get those huge whopping low ones  ... otherwise, you can modify the current follower interfacing the filter to include variable offset and scale controls (to make the filter hang in a zone for example) and forget about jFET specs ...

here's a trick I discovered for helping get rid of clock injection ... instead of using the customary filter resistors and caps to isolate the clock ckt from the audio portion try using signal diodes like 1n4148's in place of resistors - one diode (anode to cathode) going from analog +9 to digital +8.3 (call it, say) and one going from digital +0.7 (again) to analog gnd ... the diodes provide higher dynamic resistance [... albeit varying/non-linear, but so what ?! ...] to current without stealing proportional amounts of voltage since by nature diodes are voltage limited ... in my 9v-opto-mutronIII and Maestro-EMS combos there is absolutely NO feedthrough after using this ...

~JC

[jrc4558]

Interesting. I never considered using diodes for power supply isolation...  Thanks for the idea.

[taku319]

Hi,guys.
I built tone pad's FSH-1 and having a problem.
My FSH-1 have a little bit of ticking, and now I'm trying to solve this.
Here is sound sample
http://sund1.sakura.ne.jp/uploader/source/up14909.mp3
Ry said the Q2 voltages are
D 8.43
S -0.01
G -0.04

but, mine is
D 0.01
S -8.31
G 0.81

There are very big diferrence, isn't it?
And what I want to ask is GGG and Tonepad's transistor/IC voltage are different, or not.
Please tell me someone.

[soggybag]

I'm very curious to try out the diode isolation idea to solve the ticking in the FSH-1. This could be a very simple and elegant solution, that would require no modification of the PCB.

Looking at the FSH-1 pcb at tonepad, http://tonepad.com/getFileInfo.asp?id=96, it looks like the +9v and -9v power supplies feed the LFO and sample and hold sections. While the +8v and -8v supply feeds the filter. Between the two are two 470 ohm resistors.

I will give this a try, let me ask a couple questions to clarify.

1) It seems Eb+7 is saying the 8v supply feeds the LFO. In this case it is reversed, 9v feeds the LFO. I'm guessing that this should not matter.

2) If I understand correctly these two 470 ohm resistors would be replaced with diodes. Cathode closer to the more positive side? On the negitive supply the diode would be reversed?

3) It sounds like the two 220uf caps on the 9v side could be removed? Is it required these be removed if the diodes are added?

[soggybag]

I am about to give the diode idea a test, my soldering iron is heating up right now. I'm wondering if  removing the caps matter? It would be easier to remove the 470 ohm resitors by themselves.

[soggybag]

I replaced the two 470 ohm resistors with 1N4148 type diodes. Same result I can hear the sample and hold filter when there is no input.

I did leave the both pairs of 220uf caps in place. Eb+7 suggested taking these out. I might try that also...

[ovnifx]

Hey soggybag, are you certain it was those two 470's he was referring to?

[soggybag]

Thats a good question! I'm trying to understand why these diodes would solve a problem with the LFO leaking into the audio path?

First thought would it sucks a lot of juice when it spikes. Which pulls the voltage down on the audio side. The diodes could stop this back flow since they only conduct one way?

Looking at the schematic and think about what Eb+7 said, providing I understand it, it looks like the power supply feeds the LFO first from the 9V supply, and the Filter is fed second form the 8V supply. This sounds opposite to what Eb+7 is describing.

To reiterate the problem, in the sample and hold mode the sound of the LFO can be heard and is being filtered randomly. The sound is just like analog synth sample and hold, though quiter than the sound of the instrument.

In the filter mode a faint ticking can be heard when not playing. When playing the ticking is a little louder and effected by the filter.