Feedback and oscillation - how to eliminate it?

[DWBH]

So I'm working on this Ibanez TM-5 rehousing project.

[TM5 Schem http://files.muziq.be/schematics/ibanez_tm5.gif (drive pot isn't shown but it's in parallel with R20)

After I decided to rehouse it I already had done some mods (only to the 1st opamp, the 2nd was out of the circuit), and with the drive pot above 2/3 o'clock it'd start to oscillate like crazy. Well, I thought it was something about the switching/opamps, because I could also hear the noise when it was off.
So, I eliminated the buffers and true bypassed it. Also, I did some other mods (final schem -> http://andrebispo.googlepages.com/hybriddirt-full2.JPG/hybriddirt-full2-full;init:.JPG), and it continues to feedback and oscillate like crazy when I turn up the drive (and also the filter) pots.
It sounds great (when not oscillating), but I'd like to know whether there's a way of diminishing it. Perhaps too much gain (oh no, there's never too much gain!!!)
(chip is a 4558)

[Cliff Schecht]

At first glance, I notice that your input op amps feedback has extremely small resistors. This would be the first thing to change, otherwise you start drawing a LOT of current through the op amp and subsequently increase noise and potential to feedback. 10 and 120 ohm resistors need to be changed, plus those capacitor values are quite large for guitars range of frequencies. Completely get rid of C10 and R18 as these serve no purpose (R17 and C9 only need to be here), change R17 to 2.2k and change the R20 resistor to 470k. You also need to understand how to calculate cutoff frequencies.. F = 1/(2*pi*R*C) will give you the cutoffs, but it also helps to understand whether you are using a low-pass or high-pass filter. The R17/C9 forms a high-pass filter, btw .

Edit: Also, too much input gain (you're looking at over 45 dB of gain currently) will cause your circuit to oscillate. Try spreading out the gain over a few stages.

[DWBH]

Thanks Cliff!

I have some doubts, though:
a) why do R18/c10 serve no purpose? Is the frequency upon which they actuate beyond our needs here?
b) by changing R7 won't I alter the frequency response of the pedal? I really need to understand this cutoff frequency subject. Are there any graphs?
c) how does one calculate input gain?

[R.G.]

Andre, no offense intended, but you're in deep water here.

Oscillation is simple to understand, but very complex to deal with when it happens accidentally because so many things can cause oscillation. For something to oscillate, it needs only
- a feedback path for feeding some portion of the output back to the input
- a proper phase relationship, so that the part of the feedback going back to the input aids (is in phase with) the input
- enough gain so the losses in the feedback path are at least made up for, making the gain from input, through the gain, and back to the input through the feedback path is greater than one. This means that an input will always return more signal than created it in the first place through the feedback path.

What makes dealing with feedback hard is:
- high gain; the more gain available, the lossier the feedback path can be and still cause oscillation
- high impedance; the higher the input impedance of the amplifier, the higher impedance the feedback path can be and still get over unity gain at some frequency
- no physical separation; if the wires carrying the output and the input are near one another, this forms a capacitor feeding the output back to the input
Feedback can also be through power supply wires and ground wires.

a) why do R18/c10 serve no purpose? Is the frequency upon which they actuate beyond our needs here?

The rolloff frequency of 10uF and 10 ohms is F = 1/(2*pi*10R*10E-6) = 1592 Hz.

b) by changing R7 won't I alter the frequency response of the pedal? I really need to understand this cutoff frequency subject. Are there any graphs?

Welcome to why using calculators for RC rolloffs hinders learning. The rolloff frequency of any single R-C network is F = 1/(2*pi*R*C).

But I don't see an R7 in your diagram.

c) how does one calculate input gain?

For inverting inputs, F = Zf/Zi, where Zf is the total feedback impedance and Zi is the total input impedance.  And for oscillating circuits, yes, the distinction between Rf/Ri and Zf/Zi matters.

It is highly likely that  your problem is either board layout issues or wire routing, though.

[DWBH]

Well, I have both jacks on the same side of the enclosure, and the wires to the switch do run close together, even though I'm using shielded wire. And the pedal is not yet finished, so the guts are still flying around inside the enclosure.
Will get rid of R18/C10 and I'll increase R20. The gain pot is 20k, which I think is too low. I'll probably increase it to a 100k pot, as well as the filter/tone knob (which is 50k stock).

[DWBH]

Just got rid of R18/C10 and most of the feedback/oscillation/whatever is gone. Cool, although I'll have to check later if there's any difference in tone.

ADDED LATER:
Well, from my calculations, it WILL have an impact in sound. R18/C10 and R17/C9 form the same frequency rolloffs as the Rat. So, by eliminating one of those pairs (in this case R18/C10), it will affect the the gain of the frequencies. If I want to keep these values (1592Hz and Hz accordingly), should I mod the pedal to Rat values (47R/2,2uF and 560R/4,7uf instead of the 10R/10uF and 120R/22uF)
Are the lower resistor values and higher capacitor values (compared to the Rat values) I used in this RC network causing the oscillation?