Trouble with biasing 2N3906 in a Spring Reverb

[EAsch]

Hey everyone, thanks in advance for the help, this is only my second project, so please forgive the fact that I am likely making one or more mistakes that are obvious to the more experienced, but woefully over my head.

I am attempting to build G. Forrest Cook's 'Guitar Reverb Effect Version 2'. I have the circuit completed, and am having trouble with biasing it, and it is presently not working. I built the unit to the best of my knowledge, exactly in accordance with the schematic with the only exception being the addition of a true bypass circuit, which appears to work fine. When bypassed, the guitar sounds normal. When the effect is engaged, I get an incredibly soft (barely audible) signal, which abruptly changes to a rather loud, harsh crackling when the low strings are struck particularly hard. The crackling and weak seems to be from the 'dry' signal, as turning down the drive level and reverb level have no effect on it, but turning down the input drive or output level will silence the signal entirely. The only other sound is a light scratching when I adjust the reverb level pot, which I would guess is just a cheap scratchy pot, but at least means that the signal from the reverb recovery stage is passing to the output stage without issue.

I was hoping that this might merely be an issue with the 2N3906's in the input and reverb recovery sections being improperly biased, so I finally got a decent multimeter and started trying to bias them. I will freely admit I was having trouble understanding both the instructions in Mr. Cook's page under the 'alignment' heading, as well as free online resources I was coming across.

My current process is this: I am connecting the negative probe of my multimeter to ground, and the positive to the collector pin of the 2N3906. The transistor is fully connected to the complete circuit, as it would be in operation, and I have unplugged the audio cables from both input and output, to guarantee that there is no input signal. I then adjust the trimpot corresponding the given transistor and check again.

However, the voltage is not changing at all, even with fairly significant changes to the trimpot resistance. The transistor in the input stage settles at 10.33 V, and the one in the reverb recovery stage settles at 10.39 V, regardless of the resistance. I presume that either a) I am measuring the bias voltage incorrectly, b) have constructed the circuit improperly and cannot find where, or most likely c) both.

If anyone could provide any guidance it would be greatly appreciated, particularly on biasing, so at least I can eliminate that as a potential cause of the issues. As I said, apologize for what I'm sure is a fairly obvious mistake, but I am not finding a comprehensible answer online, and I have embraced what I read somewhere that 'one need not understand a circuit to construct it'. My only other project was much simpler (fuzz face clone) and did not involve any biasing, so perhaps I bit off more than I can chew with this spring reverb. Thanks again for your patience and assistance.

[antonis]

Plz, post a schematic (diagram) of your build circuit and some info, like power supply voltage etc.

[R.G.]

Good insight - you have a biasing problem.

Also good advice from Antonis - post a link to the schematic, and to the device pin voltages. See the sticky "Debugging; what to do wnen it doesn't work".

Just based on the statistical history in this and other forums, the problem (or A problem) is likely to be either a misunderstanding of the pinout of the transistor, or a mis-connected. wire.

[EAsch]

Thank you both.

The schematic I'm using is at:
http://solderman.esy.es/Spring/Spring%20Reverb%20By%20Forrest%20Cook.pdf

The supply voltage is 12 V, and the instructions recommend the following with regards to biasing the transistors: "With no
signal input, adjust each pot so that the corresponding collector is at 6V."

With regards to the pin I am connecting the positive probe to, it would be the rightmost pin on the transistor when looking at the flat side of the 'can', and with the pins facing down.

[Rob Strand]

Check the voltage on the trimpot wiper is changing.  If not you may have a pin out issue with the trimpot.

If you have use 1uF polarized electrolytic caps for instead of non-polars for the 1uF NP caps then make sure you have them in the correct orientation ie. +ve to the transistor base.

Other things to check is you have use the correct resistor values. 

[EAsch]

Thanks for the suggestions, I just checked the trimpot wiper, and it indeed changed voltages as it was twisted.

The 1 uf capacitors are the non-polarized type as specified in the schematic so that shouldn't be an issue.

I'll go through and double check my resistor values, I do recall having some trouble differentiating a few colors in the color codes as I was building, but thought I had gotten it figured out by comparing different resistors and eliminating potential values that were not in the set they came in, but I will definitely go through and double check.

Thanks again for the time and suggestions!

[Rob Strand]

The weird thing is the voltage isn't changing at all when you change the trimpot.

It might be worth checking which voltages *do* actually change with the trimpot.
Maybe measure the base, collector and emitters when you set vary the pot.
If you can report the voltages in min, middle and max positions it might help
workout what is wrong.

I'm assuming the problem is exactly the same on both 2N3906's.
It has to be some sort of systematic error for two ckts to have the same weird behaviour.

[antonis]

Something is happening with 100k bias resistor(s)..

[duck_arse]

that circuit diagram suggests, to me at least, that the trimpots are multi-turn type (20T). what type did you use?

and can I request your voltages at VF1, VF2 and VF3, please?

[R.G.]

Just a repeat note to the original poster: this is all going to take a very long time and be very uncertain of success at us guessing what is happening unless and until you post the voltages on all pins of the transistors in question.

That's not just me being uncooperative. Experience has shown that with all of the pin voltages, one can make very good guesses about what is actually wrong. Otherwise, we're groping in the dark. There is a near-infinity of possibilities to guess, but actual information about the circuit cuts that down FAST.

[EAsch]

Thank you for all the help everyone. Sorry, I wasn't gathering that you needed the voltages from all the transistor pins. No worries, I don't perceive that as being difficult at all, you are providing assistance with my project to me at no charge and asking for the technical information you need to assist me, no worries at all, like I said, I appreciate the help, and actually, attemping to get those specs, I identified the 'problem'!

Good news! I got it working! I am super excited, and it sounds (to my ear) great! Having a little trouble with the clipping detectors is the only remaining issue, but honestly, that might just be bad or out of spec diodes as I scavenged them off an old effect that I gutted for the enclosure I'm using. I have a pack of new diodes that I'll try swapping out at some point.

I was using the multiturn pots, I actually used 25-turn instead of 20-turn, but they were the 20K ohm specified. I think the issue was that there is a very precise sweet spot where the resistance of the trimpot shifts the voltage at the collector between its minimum value (0 V), and maximum value (the 10.33 V and 10.39 V I was reporting). For the transistor in the reverb recovery stage, this sweet spot was found near the maximum resistance of the corresponding trimpot (within 5 turns of it). I actually discovered it when trying to get voltages for all the pins of the transistor at minimum, middle and maximum values of the corresponding pot. So even though I had twisted the pot 10 or 15 times, I hadn't yet twisted it far enough to enter the 'sweet spot' and have any effect on the collector voltage. But when I measured it with full resistance, it had dropped to 0 V, so I quickly checked, and indeed within about 5 turns of maximum resistance, the trimpot does change the voltage at the collector but turn the resistance any lower than that, and the collector will stick at the 10.39 V I was originally reporting and not increase further, presumably maintained by the interaction of supply voltage and the other resistors in the circuit.

Thanks again for everyone's time and help! I am really excited to have it working!

[Rob Strand]

I think the issue was that there is a very precise sweet spot where the resistance of the trimpot shifts the voltage at the collector between its minimum value (0 V), and maximum value (the 10.33 V and 10.39 V I was reporting).

The whole reason it needs a 20 turn pot is because it is a fine adjustment.  It's not a good practice.  This issue could be totally removed by putting resistor in series with the pots so they only operate over a sensible part of the range. 

IMHO the collectors supposed to be at 6V.   The top clipping detector relies on that.   The lower one is independent of the adjustment.

[PRR]

> the issue was that there is a very precise sweet spot

Yes. It is a "bad design". Mr Cook is a very elegant writer. This electronic design falls short of elegance.

Both those stages could be reduced to this with, IMHO, little change in performance and NO tricky biasing.


Oh wait- he also references the Clipping Detector to the DC in the amplifier. That's not usually done. There actually is good logic in this. I dunno if it will detect "slight" clipping or only gross clipping. (We should rarely care.)

[Rob Strand]

Yes. It is a "bad design". Mr Cook is a very elegant writer. This electronic design falls short of elegance.

There also an issue with pickup loading.  At first you see 100k pots at the input  ... but ... it's worse than that - the 27 ohm resistor only multiplies up by say hfe=200 which presents a 5.4k ohm load (on a good day).

Oh wait- he also references the Clipping Detector to the DC in the amplifier. That's not usually done. There actually is good logic in this. I dunno if it will detect "slight" clipping or only gross clipping. (We should rarely care.)

I think the idea is when the transistor saturates it's on the positive swing with finite voltage from the rail.   The 10k + 1k on the transistor matches the 10k + 1k2 on the detector.    The idea only works if the negative swing is more than the positive swing (otherwise the transistor clips negatively without the LED lighting.)

The circuits got a lot of quirks.

[PRR]

R.G.> ... post the voltages on all pins of the transistors in question. That's not just me being uncooperative.

[EAsch]

Thanks again everyone for the help, haha and sorry to be Dick, R.G.

Rob, yes, that seems odd to me. Seems like it would indeed make much more sense to put a 10K pot in series with a 10K multiturn trimpot at the very least, to both make biasing less tricky, and enable more precise biasing on pots with the same number of turns.

As far as the clipping detectors, that's interesting. So the detector actually requires a negative DC bias to the audio signal in order to register as clipping? That seems very strange. I must admit that the clipping detector circuitry, in particular, is well beyond my comprehension. I've never used gates before and have only the most rudimentary and theoretical understanding of what they do, with almost no understanding of their actual electrical functioning or practical application.

It's unfortunate that the input load is so high. I guess I will find out after toying around, I need to pick up a few more audio cables tomorrow, but do you think there will be a significant issue using the reverb with other effects? Obviously being reverb, it will most likely come last in my signal chain before the amp.

PPR, interesting thoughts. Again, my understanding of circuitry is pretty basic, so I feel only vaguely literate in examing your reworked circuit. Is the biasing avoided merely by having appropriate value resistors in the circuit, or is it for some reason less crucial with the NPN 2N3904 than the PNP 2N3906 in the schematic? Your diagram does appear much more parsimonious than Mr. Cook's which my scientific education has taught me to value if nothing else.

As far as the clipping detector discussion, as I mentioned, that is pretty squarely over my head. But I appreciate the input and interest haha!

[Rob Strand]

Rob, yes, that seems odd to me. Seems like it would indeed make much more sense to put a 10K pot in series with a 10K multiturn trimpot at the very least, to both make biasing less tricky, and enable more precise biasing on pots with the same number of turns.

I like PRR's idea of not having any adjustment.  It should be possible to  bias a transistor sufficiently close to a desired voltage so no adjustment is necessary.   Once the worst-case clipping limits are established just tune the clip detector to that level.

I must admit that the clipping detector circuitry, in particular, is well beyond my comprehension. I've never used gates before and have only the most rudimentary and theoretical understanding of what they do, with almost no understanding of their actual electrical functioning or practical application.

The clip detector detects positive peaks.  When the input on the opamp +input exceeds the DC threshold on the -input the output will swing from low to high.   When the signal come out of clipping the opamp output goes goes low again.   The circuit with the logic gates is a monostable,
http://www.edutek.ltd.uk/CBricks_Pages/Monostable_LOGIC.html
It pulses the LED when the input goes *low*.   So the way the circuit is you get a pulse on the LED when the signal *come out* of positive clipping.    Normally you would expect the LED to pulse when the signal is rising and just enter clipping.  To me it doesn't make sense to pulse the LED out of clipping.  If I was reviewing this circuit I would think it was a bug.   The NOR gate version with the LED connected to ground would make more sense.  (If I've screwed-up in my logic please feel free to correct me (anyone!).)

There's even more weirdness to me.  The clipping detect is on the first stage but the output stage has has less voltage swing capability so it will probably clip before the input stage.

The first stage has quite a substantial gain.  IMHO it is not necessary.  The gain is set by the 27 ohm resistor on the emitter of the first transistor.  The 27 ohm resistor is what is causing the low input impedance.   If we lower the front-end gain by increasing the 27 ohm to a somewhat higher value, say 2.2k to 3.3k or so, then it will substantially increase the input impedance and decrease the gain.    The biasing would need to be tweaked.  The next thing to do would be to increase the 100k resistor on the base to say 470k.   (Now, by reducing the front-end gain we may need to decrease the 47k resistor that goes to/near pin 2 of the opamp to compensate. Maybe best leave it.)   The clip detection threshold will need to be tweaks to match.

The other weird thing is it has four level controls.  For guitar, if the gain of the first stage is chosen correctly you could get rid of the input level control alltogther.  That would remove more loading.

Anyway as you can see there's many areas that need tweaking and it starts to get a bit out of hand.  You almost need to go over the whole thing and come-up with a new version!

[PRR]

> Is the biasing avoided merely by having appropriate value resistors in the circuit, or is it for some reason less crucial with the NPN 2N3904 than the...

Many millions of transistor stages self-bias without fiddling.

The first problem is the wide variability in transistors even of the same type. There are several techniques. One is to just set the bias so solidly that "any" transistor will work. This means setting the bias currents large which detracts from signal gain. A cleverer technique is to use negative feedback. The bias is derived from the output. The transistor amplifies its own bias error and drives itself back to a happy (for us) bias point.

Consider the plan I showed. If the collector goes high, current in the 2Meg goes high, and pulls the base high. This turns-on the transistor, increasing drop in the 10K, and the collector voltage drops. If it gets way low, the opposite: base goes low, turns-off the transistor, collector goes high.

If you want half-voltage (as we are told, and often a good plan), then the base resistor should be hFE larger than Rc+Re (or just Rc, since Re is relatively tiny). I recall these parts being hFE 100-300, so aimed for 10K*200, makes 2Meg. (2.2M or 1.8M standard values) Wide variation of hFE cause small variation of collector voltage, operating current, and maximum output.

(A disadvantage of this is low input impedance. The 2Meg is divided by stage gain to look-like around 20K. But as Rob showed the base impedance is already quite low, so this may not be dominant.)

The '04 and '06 parts have essentially the same hFE spread and are nearly mirror-images of each other. My main point is that with a positive supply an N-type is "more natural". If we needed PNP elsewhere and didn't want to buy another part number... no, he has an NPN also. Being cheaper by the dozen, why not use all the same?

There is a notion that PNP has lower hiss. This is not true outside a test lab, and really only in theory, not in practical devices. For decades the reference low-hiss parts were dual NPNs, mostly because NPN fabrication dominated the foundries and they knew how to control the process (and could run an un-thrifty layout without disrupting the large-market NPN process). Anyway with jelly-bean 2N390X parts the hiss level may depend on the Monday vs Wednesday variable (Monday morning, the machines are cold and the workers are grumpy), and the price does not justify caring.

Getting a little deep here: the hOE (emitter impedance), critical for gain, at 0.6mA is 44 Ohms. This makes the 27 or 10 Ohm added resistance very small. At high levels hOE varies with current, which makes distortion. By the time these "clip detectors" blink the THD will be over 5%, possibly near 20%. This may not sour a reverb but seems pointlessly high.

And... what is cheaper, a 20-turn pot or another jellybean transistor? A 2-transistor compound has more gain to squander in bias and signal-gain control.

> trouble differentiating a few colors in the color codes

Use your Ohmmeter!!

Mild color-blindness is common. Many types of light don't have all colors. Many resistor makers let the paints go dull and lifeless. (And 10 in a million, they just get the colors wrong.)

[PRR]

Anyway.... I would strongly suggest Spring Reverb Unit, Rod Elliott.

It is configured for dual +/- supplies, which may not fit your plans or parts. It is thoroughly well designed.

[Rob Strand]

Anyway.... I would strongly suggest Spring Reverb Unit, Rod Elliott.

Looks a lot like one I did back in the 90's.  I'm sure I used a non inverting stage for the mixer at the output  Used a quad opamp. The extra opamp was used as a buffer for the front-end.   Something like that.