Basic op amp inverting buffer questions

[kenjib]

Hi everyone. I am trying to use an inverting op amp (TL071) buffer at the start of my guitar pedal chain. It has to invert because there is a mixer/summing op amp at the end that inverts and ideally I want to preserve polarity from start to finish. I am using the attached schematic and unfortunately there is a very noticeable roll-off of high frequencies. Vcc is 9v and -Vcc is -9v via a separate ICL7660S circuit.

I tried a bunch of different things via breadboarding. What I found is that when I raise the value of the two 10k filters (R2 and R3), the roll-off gradually reduces as the value goes up. The signal seems to get close to unity at around 1M ohms, however as I raise the filter values in this way the noise floor also increases considerably with it and is pretty insufferable at 1M.

1. So what is going on here and what am I doing wrong?
2. Should I get rid of the 10uf C2 cap at the end? I think maybe that was to remove the DC current for a 4.5v biased circuit and it is no longer necessary now that I am using -9v and 9v instead of 0v and 9v. Maybe it isn't needed anymore???

Thanks a ton for any assistance!

-Kenji

[Rob Strand]

I am trying to use an inverting op amp (TL071) buffer at the start of my guitar pedal chain. It has to invert because there is a mixer/summing op amp at the end that inverts and ideally I want to preserve polarity from start to finish. I am using the attached schematic and unfortunately there is a very noticeable roll-off of high frequencies.

It's not a good solution.  The 10k input impedance will load the pickups.   A better solution is to increase the the resistances to 100k or 220k. Some commercial pedals do this. There is still a small loading effect.  The problem now is the higher value resistors significantly increase the noise.

Better is to maintain a non-inverting front-end then try to move the extra signal inversion to another point in the circuit which has a low output impedance.  That way you can keep the 10k feedback resistors and keep the noise low .  In some case you might need to do that and change the resistor values to 22k to 47k in order to find a trade-off between loading and noise.  If there's no low-impedance point you might need rethink the circuit, or even live with the signal inversion!

[kenjib]

Ah okay. I knew it had something to do with impedance but I am still learning that part. Thanks! Sometimes the best solution is to rethink the whole approach from the start. I am sure there are other places I can move it. Is it possible to simply put an inverting buffer right after a unity input buffer in series? If so I guess I would still need to grapple with the question of whether or not preserving polarity is worth that anyway...

Also, if I convert it to a voltage follower unity buffer, will a 10uF cap at the end still serve any purpose when I am using +9/-9 vdc for the op amp and ground for bias?

-Kenji

[PRR]

> there is a mixer/summing op amp at the end that inverts

Why? Change that, not your precious first stage. Try a passive mixer plus a buffer. (Active mixers are for studios.) Or build an inverter after your active mixer.

I'd also question a "need" for absolute phase. What is the phase of your pickup? IS there a phase?

[duck_arse]

your circuit show at R1 2.2m . this means milli, when you really want M for meg. this won't matter except in a sim, really. however, your R4 shows "100" to ground, which is a very heavy load for a pulldown resistor. have you built with 100R, or is it mistype?

[kenjib]

> there is a mixer/summing op amp at the end that inverts

Why? Change that, not your precious first stage. Try a passive mixer plus a buffer. (Active mixers are for studios.) Or build an inverter after your active mixer.

I'd also question a "need" for absolute phase. What is the phase of your pickup? IS there a phase?

The pedal is a two channel loop switcher. Input buffer -> 4 sends -> output buffer. Each loop has a channel select and phase switch and the channels are mixed together at the output buffer. The output phase inverts because I am using a virtual earth mixer to keep the two volume pots from interacting with each other (or so I read).

Those are good points. I do have a studio mixer mentality which I'm probably bringing to this too much. The reason why I wanted phase to be the same for input and output is that I'm dealing with phase issues on each of the four sends and having to correct for other pedals not always preserving phase. So it seems strange for this pedal to not preserve phase when I am having to correct it for other pedals. It's possible I'm just needlessly hung up on it though, so for the moment I'm not going to worry about "correcting" phase.

your circuit show at R1 2.2m . this means milli, when you really want M for meg. this won't matter except in a sim, really. however, your R4 shows "100" to ground, which is a very heavy load for a pulldown resistor. have you built with 100R, or is it mistype?

Thanks for catching those errors. It looks like I need to do more studying of how input and output buffers work and redesign all of that. Right now I am not sure when/why to use pull-down resistors vs series load resistors at the end of the buffer (both input and output) so it's time to study up!

Maybe after I try to clean that up I can post the full circuit to give some context.

[kenjib]

Okay here is the whole circuit. I changed the input buffer to non-inverting, fixed the lowercase letters, and used larger pull-downs.

I haven't assembled the four sends yet, as I have to order some 4pdt switches (though maybe there is a  more clever way to do that using dpdt switches?). The sends aren't tested yet as a result. So far I have bread-boarded the input, output, and power sections. The in/out buffers sound very transparent now and I don't perceive any signal loss.

A few questions now:

1. Do I need the 100 Ohm resistors R8, R3, R16, R25, R30 at the sends and the output to set some output impedance to the external pedals and amp? It seems to sound the same whether or not R8 is there before the amp.
2. Do I need the pull-down resistors R9 and R37 after the decoupling capacitors? Again it sounds the same either way but maybe they will help with switching pops on the outputs once I add bypass (or something like that)?

Thank you guys for being patient with me while I learn.

[ElectricDruid]

The input buffer doesn't have any DC bias. I'd move R9 to the pin 3 +ve input.

The splitters on the other hand have two doses of DC bias, once through the big 2M2 resistor R4, and again therough the 22K/100K R19/R20 pair. I'd say the 2M2 is redundant. What you should probably have is an input cap though. Currently, you're not blocking any DC offset that comes in. It'll be fine as long as the pedals you plug in all have good output caps and no offsets, but since you don't know *what* might get plugged in, it'd be better to be safe.

The 100R on the send I would regard as protection for the input buffer output, rather than anything to do with input output impedance. If someone shorts that send socket to ground, that input buffer is trying to drive a shorted output and it won't like that at all. Having a 100R there limits the current that can flow and might stop the op-amp getting fried. I would probably go a bit bigger for this reason, maybe 560R. You won't hear the difference between 100R and 560R because the input impedance of whatever you plug in will be several hundred K at least, so it's a fractional difference.

What's the intended purpose of the channel switches? They seem to duplicate the function of the bypass switches to me.

HTH

[antonis]

I'm a bit confused wuith your wiring diagram, but..

1. Those 100R resistors are usually there to prevent op-amps oscillation due to possible load significant capacitance..
2. R9 & R37 make no harm..

P.S.1
C1 DOES need a path to GND (non-inverting input DC bias stability), so use a 1M or so resistor between GND and C1/pin3 joint..

P.S.2
Can't see an obvious reason for R4, R10, R11 & R12 existence..
(other than setting R20, R21, R26 & R31 respective equivalent resistance values to 95.65k..)

edit: Tom is faster...!!!

[merlinb]

Hi everyone. I am trying to use an inverting op amp (TL071) buffer at the start of my guitar pedal chain.

You could use a TL072. Then you would have two opamps, one for buffering, followed by one for inversion. No impedance problems that way.

[kenjib]

I'll try and explain what the circuit is for and what the bypass and channel select switches do. Basically the signal splits into two channels, A and B -- kind of like a wet/dry setup. Then for each of the four sends/returns, you get to choose whether they insert into channel A or channel B. In the end channels A and B are mixed back together to a single output.

The channel select moves the insert for each send onto either channel A or channel B -- the other channel is passed through unaltered. The bypass, on the other hand, removes the send entirely and just sends an unaltered signal directly through both channel A and B.

The need driving this design is simplifying my current pedal board. I have a signal splitter then I run a pretty simple tone chain with just overdrive, delay, and slight reverb to keep a "normal" guitar sound going. Off of the second side of the splitter I put lots of crazy ambient stuff including delays on the edge of self-oscillation, pitch shifters, heavy reverbs, PLL, modulation, dirty robot, etc. So the second chain ends up with a giant wash of ambient sound and noise in which the guitar gets totally lost. That's why I mix the more normal first signal back in at the end using a little mixer pedal. I end up being able to generate different washes of ambient textures and then play along with myself using the more "clean" channel. I can also put overdrive/distortion on the first "normal" channel but leave it out of the second "ambient" channel so that the background textures stay more subdued with the distorted guitar out in front. Here's the idea:

https://soundcloud.com/kenjib-364738420/ambient-guitar-sample?si=5fd2a9c1ab6a4b078f17833fe2123c57&utm_source=clipboard&utm_medium=text&utm_campaign=social_sharing

Does that make sense? So I am making this pedal to streamline this process and manage it more efficiently, add buffering because all of these pedals (many true bypass) are causing a lot of tone suck, control phase issues, etc.

Anyways, I'm realizing now that after each of the returns I need another full input buffer, since I'm coming back from the return via who knows what is plugged in and need to treat it just like the input phase of the pedal. I think that means that I need a two stage buffer within each send if I want to be able to control phase for each send, right? So I will have to revise each send pretty heavily.

Unfortunately, I have learned from experimenting with this that I definitely do need phase switching individually for each send (rather than, say, just for the two channel outputs before mixing), since any given send/return can be switched between channels. That means that *all* of the sends need to preserve phase on their own, or else the final two signals could keep going in and out of phase with each other as each send switches between channels and/or is bypassed.

[kenjib]

The input buffer doesn't have any DC bias. I'd move R9 to the pin 3 +ve input.

Done! Does C2 need a path to ground too though? Maybe I need it in both places?

The splitters on the other hand have two doses of DC bias, once through the big 2M2 resistor R4, and again therough the 22K/100K R19/R20 pair. I'd say the 2M2 is redundant. What you should probably have is an input cap though. Currently, you're not blocking any DC offset that comes in. It'll be fine as long as the pedals you plug in all have good output caps and no offsets, but since you don't know *what* might get plugged in, it'd be better to be safe.

You are very right there. I need a high impedance there due to the inputs coming back from the returns and tried to shoehorn it in, incorrectly, by just adding another 2.2M resistor. Unfortunately, it looks like I'll need to remove the 2.2M and then add another non-inverting buffer there before that stage for each channel since this is the exactly the same problem I had in my first post with the input stage, except that it is also happening at each return.

The 100R on the send I would regard as protection for the input buffer output, rather than anything to do with input output impedance. If someone shorts that send socket to ground, that input buffer is trying to drive a shorted output and it won't like that at all. Having a 100R there limits the current that can flow and might stop the op-amp getting fried. I would probably go a bit bigger for this reason, maybe 560R. You won't hear the difference between 100R and 560R because the input impedance of whatever you plug in will be several hundred K at least, so it's a fractional difference.

What's the intended purpose of the channel switches? They seem to duplicate the function of the bypass switches to me.

HTH

That is a super helpful explanation. Thanks. I'll bump them up to 560. I explained the channel switches in the last post.

I'm a bit confused wuith your wiring diagram, but..

1. Those 100R resistors are usually there to prevent op-amps oscillation due to possible load significant capacitance..
2. R9 & R37 make no harm..

P.S.1
C1 DOES need a path to GND (non-inverting input DC bias stability), so use a 1M or so resistor between GND and C1/pin3 joint..

P.S.2
Can't see an obvious reason for R4, R10, R11 & R12 existence..
(other than setting R20, R21, R26 & R31 respective equivalent resistance values to 95.65k..)

edit: Tom is faster...!!!

Thank you! I will add the resistor and, yep, I need to make a lot of changes to the send return buffers. You caught my clumsy mistake. 

You could use a TL072. Then you would have two opamps, one for buffering, followed by one for inversion. No impedance problems that way.

Yeah I think I need to do that for each of the returns, so if I want to maintain overall phase I can just do it in the beginning as well. Thanks!

Okay I'll do some revisions and post another version soon. Thanks everyone. I am learning a lot already.

-Kenji

[antonis]

I need a high impedance there due to the inputs coming back from the returns and tried to shoehorn it in, incorrectly, by just adding another 2.2M resistor.

Particular configuration might be a bit tricky, as fas as input impedance concerns, but it should be more easy to "seperate" inverting and non-inverting inputs impedances, to get an idea of what's happening..

e.g. by considering pin2 (inverting input) AC ground and pin3 (non-inverting input) "open" (infinite resistance), you have pin2 impedance R18//R20 and pin3 impedance just R20 (phase switch open)..
(any other signal shunting resistor, like R4, further dominates the above impedances..)

In brief: For inverting op-amp comfiguration, input impedance equals to inverting input resistor where for non-inverting one input impedance equals to bias resistor..

[kenjib]

Okay here is another revision with cascading buffers. Now I am paying more attention to the impedance coming out of the returns. I also put capacitors in parallel with the feedback resistors after reading the other active thread about cascading inverting stages. Electric Druid -- I used your calculator to come up with the values.

1. Do C9, C10, and C11 look correct?
2. Should those be there with a dual power source buffer or are those filters really just addressing a problem only single source buffers have?
3. Also, should I add caps parallel to the non-inverting buffer feedback loops for U1A and UnA (and if so, what size)?

[ElectricDruid]

1. Do C9, C10, and C11 look correct?

C9 and C10 give 33KHz, which is fine, but you won't find a 500pF cap easily. 560p is more likely, which gives you 28KHz, still fine. C11 is in parallel with 22K instead of 10K, so the cutoff drops by an octave, which is perhaps starting to be a bit low - 22K/560p gives 12.9KHz. Try 470p or 330p in that position.

2. Should those be there with a dual power source buffer or are those filters really just addressing a problem only single source buffers have?

In general, we should only amplify signals we actually want. In our case, that's audio from 20Hz to 20KHz. Amplifying anything outside that range (subsonic thumps, DC offsets, ultrasonics, radio frequencies, etc etc) is just adding noise. The filters help limit the frequency range that the amplifier operates at. It has nothing to do with the power supply.

3. Also, should I add caps parallel to the non-inverting buffer feedback loops for U1A and UnA (and if so, what size)?

If you had non-inverting gain, then yes, you should limit that to the range you care about in a similar way for the same reasons. But your buffer has no gain, so you don't need to worry about it. You *should* choose input and output cap values appropriately to limit the signal to the bit you care about. You've got 100n/1M for C1/R35, which is a highpass filter at 1.6Hz. You could reduce that cap value without hurting anything. 10n would be 16Hz, 4n7 gets you around 35Hz.

[antonis]

It should also be a good idea to short (via a jumper) pin 13 & pin 14 and ground pin 12 of U1D (unused amp)..

[kenjib]

Okay so I incorporated those changes and the current revision sounds really great so far. The power, input buffer, and output buffer all sound pretty much identical to running straight into the amp and the volume knobs work great for both channels. Thanks so much guys!

I will put in a Tayda order soon so I can build the switchable sends and report back in a week or two on how those go as well. I'm thinking of upping the number of sends from 4 to 6 and seeing if that many 3PDT switches are not too cramped in the weird oblong 1032L enclosure they have.

[PRR]

What does R42 do for you?

[kenjib]

What does R42 do for you?

That part of the circuit originally came converting this one to mono:

https://sound-au.com/project107.htm

I think it is roughly the same as this R.G. Keen one:

http://www.geofex.com/article_folders/polarity_reverser/polarity_reverser.htm

So I guess R42 is a bias resistor. Do you think it might be the wrong value?

[antonis]

Bias resistors are connected on non-inverting inputs..
Here, there isn't any need for those resistors due to direct DC coupling..
(UnA output sits at GND and "injects" it, via R43, into pin 5..)