State Variable Filter LPF noise

[MrStab]

thanks for the info. i'll try that combination, but unfortunately it'll have to be after new year. i might have a spare 5 minutes to try grounding the input when i get up later, though. FWIW, the noise isn't there when the connection between U3 & U1A is cut. as far as i can tell, the SVF itself keeps functioning as usual in either state.

here's something that may or may not matter: if a cap is used between one of the shelving outputs and the inverter stage, and it's too small, then it oscillates, too. that's going from memory so i can't remember if it's quite the same type of oscillation, or whether this was from U1B, U2B or both. i thought it was just the small size making the inverter unstable at the time, more an issue with the actual capacitance than the effects of the frequency.

the low band, using U2B into the inverter, is a-ok (i'm just naming the op-amps instead of filter names so i don't confuse myself with semantics because of all the inverse effects). lol

[MrStab]

...also, the Frequency pot is microphonic, again just like what i'd expect when the Q is high in peaking mode.

[Rob Strand]

i'll try that combination, but unfortunately it'll have to be after new year.

From what you saying, especially your second paragraph, I'm thinking the problem isn't high freq type oscillation, which those mods address, but some sort of in-band oscillation which points to a fundamental problem.
i might have a spare 5 minutes to try grounding the input when i [/quote]
That is still worth trying.

FWIW, the noise isn't there when the connection between U3 & U1A is cut. as far as i can tell, the SVF itself keeps functioning as usual in either state.

OK, that at least makes sense.  (It might rule out some layout issues later on.)

here's something that may or may not matter: if a cap is used between one of the shelving outputs and the inverter stage, and it's too small, then it oscillates, too. that's going from memory so i can't remember if it's quite the same type of oscillation, or whether this was from U1B, U2B or both. i thought it was just the small size making the inverter unstable at the time, more an issue with the actual capacitance than the effects of the frequency.

I'll try to analyse this case.  If you have an specific examples of good and bad cap values it might be useful.  If you has a smallish cap + a second order high-pass filter your get a third order high-pass filter which has more phase shift.   Now my gut feeling is it should not be able to cause oscillation because the loop gain should be less than one - this is what I will check.

Thanks for the update/info.   It's a tricky problem.

[Rob Strand]

i'll try that combination, but unfortunately it'll have to be after new year.

Maybe give it a go.   I really do feel some parts like this are required.  At least the cap across R2 but the others may help.

My gut feeling is something isn't what we think it is.  A wrong part value or something.  Maybe around U1A or U3.  If the overall gain around the loop is higher than what we think it will show the symptoms we are seeing.   *but* if that  was the issue then I would expect cut to be worse than boost - which is not what you observe (grrrr.).

[MrStab]

ok, i've figured out part of the problem. i should've been more meticulous, i must've tested without the CD4053 with a higher-gain resistor in place, and not dug deeper because the other channel worked fine. seems the Q was jammed within some margin by the 4053. no peaky sound anymore.

i thought wiring the Q pot lugs to Common instead of the other way around would avoid switch pops for some reason, in case you're picturing it the other way around. maybe the other way would be better.

still a lot of white noise, but i'm about to quickly try a cap across R7 to see what the status is now, with a lower Q. don't have any time for anything else, unfortunately.

[MrStab]

no more oscillation with a 510p across R7! it's hard to compare noise levels with the peaking mode without the switch in place, but the signal now outweighs the hiss a lot more, and the BC pot is more responsive. still some bleed-through with Frequency turned high and BC in the middle, but maybe it'll play nicer with the other measures now. i'll retry some stuff when i next get the chance. shame the CMOS will have to be removed or reconfigured.

update: here's what i think happened so far: there's another, fixed R6. while i was probably able to lower the CMOS' high Off resistance by paralleling across it with the other R6 (i'll draw a diagram of this convoluted setup later), that one still kept it at 10k. so i'll try lowering that before abandoning the 4053.

[Rob Strand]

Cool. It's pretty much done.

If you aren't happy with the noise you can add pre-emphasis/de-emphasis.
Look at the boss GE7 and the Korg parametric:
http://www.diystompboxes.com/pedals/schems/korgpeq1.gif

[MrStab]

happy new year!

i've found the how of the problem, but i'm a bit stumped as to why. that extra resistor was a minor oversight i would've probably remembered at some point, and the main issue is with the CMOS itself. this is how i'm switching the Q pot with the 4053:



(just imagine there's another fixed resistor to Vref to replace R6 [R1 above] for shelving mode!)

i chose this way as i thought it'd minimise pop and prevent interference with the SVF, as opposed to making the op-amp side common. the perma-high-Q would suggest that U1B's non-inverting pin has a permanently-low path to Vref when Off. i can't find any material on why this might be happening - at best i can find leakage current on the datasheet, which is little help as it suggests the really-high resistance i'd expect, and as far as i can tell doesn't refer to the separation between outputs.

a CMOS switch is vastly preferred for space and reliability reasons. maybe i should wire em the other way around, and switching between the pot or a fixed resistance. unless there's some way i can make the current scheme workable.

i tried to incorporate some kind of emphasis-ish system at one point, but then i thought i should iron this out first. i've been looking at the CE-2/GE-7 schems. cheers though!

[MrStab]

....aaand it's now i realised i mixed up pin ordering slightly. my proof-checking skills are seriously awful, i went over it about 4 times. lol

[anotherjim]

If the +in of the amp is really disconnected when the switch is off, then that connection is floating at a really very very high impedance and will pick up any radiant or leaked crap going. Inputs should have a tie resistor to Vref that's about x10 the source resistance when switched on, so it doesn't load the source or mess with pot range.

[MrStab]

interesting you should mention that, Jim, as i was about to post about an issue i'm facing in that region (edit: not quite!):

there's a DC offset when the Boost/Cut pot is turned. when fully cut (U1A output side shorted), the DC at U1B+ is roughly 300mV lower than Vref, which approaches 4.5V as the pot is turned. At the centre there's a small click, and the whole other half of the pot (boost) is a firm 4.5V. this problem also seems to extend into the switching, creating a big pop when that pot is rolled back.

i've tried a few things, some of which might've been a bit unorthodox, but hey. coupling and biasing the inverter output (what would be C6 on the ESP diagram), and coupling between U1A's output and the BC pot. i've tried fixed resistors of 10-100k to Vref (on my OP diagram) before R1, R14 and a few other places. i can't list any more because i have no idea wtf my notes say. lol

i think this is the final hurdle, and would iron out 2 issues for a well-behaved pedal! i've run out of ideas, so any are welcome. i don't suppose R12 & R13 on the ESP diagram could be to blame? when the frequency pot is turned near the 16KHz limit, the pop goes away. i've run out of caps so can't really try right now.

cheers!

[MrStab]

the switch never cuts signal to U1A+ completely btw, afaik the 4053 is break-before-make but generally in the Boost position, the switching is quiet

edit: coupling between B/C wiper and R4 makes no improvement either. i don't get it! the inverting input of U1B wiggles slightly, but nowhere near as bad as the non-inverting. so i assumed something attached to the + was to blame.

[Rob Strand]

interesting you should mention that, Jim, as i was about to post about an issue i'm facing in that region:

The other resistors (shown in your schematic) should prevent any issues here.

[MrStab]

gah! i meant it's actually U1B as the op-amp with the offset. sorry about that. resistance to Vref on either input, and no resistance at all, makes no difference. i've edited the posts as needed, if it helps - i still reference U1A correctly where appropriate

[Rob Strand]

gah! i meant it's actually U1B as the op-amp with the offset. sorry about that. resistance to Vref on either input, and

From your original schematic both U1A and U1B have feedback resistors which provide a path for DC bias.  Which is fine.

All opamps have some form of DC offset which varies from unit to unit.   What can happen is if you have DC connections around the opamps (ie. no AC coupling caps)  then the DC gain that results in the circuit will multiply the opamp's DC offset and you will see this larger value at the output.   For U1B the DC gain is set by R6, R7, and R8 and ends-up at about 7.8.  If your opamp has a large offset, say 10mV, then you would see 78mV at the output.  I don't expect it to be that big; also I have ignored the effect of the other opamps.

This doesn't necessarily represent a "problem", it just is.  If the offset cause pot scratching and pops only then it become a problem.   This is when it might be wise to add caps in series with the resistors which set the DC gain.

If the offsets are larger than expected then that usually means a circuit problem or a damaged opamp.

[MrStab]

thanks for the info Rob, appreciated as always! i didn't even think to check the datasheet for inherent offset! thanks to your explanation (and past help) i'm starting to realise exactly why these things are going on. pot scratching is there (albeit slight), and while it might be cool to have an "aural centre-detent", i'd rather kill it. also it's likely tied into the parametric switch pop.

the TLC274 (SVF) seems to have 2-3mV offset at worst, and the TL072 (input stage & inverter) are slightly worse, at up to 10. neither on their own would seem to account for a whole 300mV, though.

U1A receives a pre-biased signal from the input stage, with C8 separating the amp from DC. U2A and U2B are basic integrators and i'm reading a more-solid 4.5V out of em. hmm... narrowing things down, at least.

interestingly, Spice suggests that a cap from BC wiper to R4 would kill the offset, but it doesn't seem to in reality. Spice makes the offset appear to be a matter of microvolts, though. i'm not taking it as gospel for any of this.

i'm still at a loss tbh, but i'll try to digest the info and see if i can come up with anything else.

cheers!

[MrStab]

at this point i think it might be worth trying caps at the end of the Frequency pot - the ESP article does mention offset, albeit in BP mode, and the problem does have a relationship with the Freq pot. i'll score some caps tomorrow and give it a go.

[Rob Strand]

at this point i think it might be worth trying caps at the end of the Frequency pot

It might help.  There is something a little tricky going here with amplifying offsets: 
If U2A has an offset then, via the feedback loop via R5, it will force U1B'a output up in order to match that offset.  The integrator cannot have an offset at the input otherwise it will keep accumulating of difference.  Without the feedback via R5 the integrator output would continue to rise or fall depending on the sign of the of the offset.
If you have the frequency pot down at the 33k end, the voltage at the top of the 500k frequency pot will be about (500+33)/33 = 16 times the offset of U2A.   With the frequency pot caps in then the multiplying factor is only 1.

[MrStab]

U2B's output doesn't seem to wiggle but i'll double-check when it's a sensible time of day. just to clear up one thing: do you mean R5 should prevent (or at least limit) the issue you describe with the integrator's offset? if caps on the Frequency pots don't help things, maybe there could be some benefit to placing a cap between R5 & U1B+?

i think looking at the relationship between U1B-U2 makes a lot of sense now, as i've pretty much tried everything with U1A.

cheers!

[Rob Strand]

just to clear up one thing: do you mean R5 should prevent (or at least limit) the issue you describe with the integrator's offset?

No.  It's about *how* the circuit finds the final equilibrium state that you see.   Without R5 the circuit will go nuts (in that the integrators will continue ramping up to or down until they hit positive or negative saturation) as there is no means for the circuit to find set of DC conditions where an equilibrium is reached such that all the DC offsets are taken into account.   

It's like if you stand in the middle of your bed all the spring forces balance out and you reach a certain equilibrium height, if someone put an extra mass (an offset) on there, there is a new equilibrium height,  but if you stand in the middle of a pool there no upward spring force and you keep sinking.  (sorry think of a rock not a person as obviously you will float)

if caps on the Frequency pots don't help things, maybe there could be some benefit to placing a cap between R5 & U1B+?

The caps on the pots allows the DC to get through without multiplying it - that's a good as you can do.  If you brak the DC loop then it will go nuts.