Inverted output from panning control to mix wet/dry signals

[Steve Conner]

Looks good to me, except you have a mix of single supply and split supply techniques. You need to make your mind up which you are using. For instance, if you are using +/-9V, you don't need R10 and R12 because the inputs should be biased to 0V, not 4.5.

If you are using a single 9V rail, then you need to connect the wiper of VR1 and the non-inverting inputs of U2 to 4.5V, and put a DC block capacitor on the output.

[pdf64]

Take another look, I don't think the pan control will do anything?
R3 & R4 need to be in series with the main signal flow.

As Steve mentioned, get rid of the 4.5V bias - remove R10 and R12.

Pete

[ReadyTeddy]

As Pete mentioned, as drawn, you won't have complete panning. Also, the ckt looks overly complex. Do you really need the buffers? By adjusting the impedances in the panning section, you may be able to get rid of them. What's driving this stage? Also, why the 2 stages after the panning section? If you are concerned about phase inversion, the panning section can be connected to the noninverting input of the output op amp. So, it may be possible to do the whole ckt with one op-amp stage.

[eschertron]

I agree the 10K spec for the linear pot must be a typo if the other resistors in the voltage divider are 15k. Even with a 100k pot, there won't be absolute signal rejection from the other channel.

Speaking of which, the design looks more like a standalone 2ch mixing board than a simple piggyback affair. The cheap mixers from Radio Shack (back when...) were basically this in a silkscreened enclosure, just 8 channels instead of 2.

Thinking of alternatives, doesn't the above-mentioned stereo chorus unit have a wet/dry control on it? Most do. You didn't state, but perhaps the goal is to make the panning control (technically: segue) a foot treadle for live manipulation. If that's the case, there are passive foot controls available on the market. I picked one up used for about $25 that contains two ganged 100k pots and a selector switch to allow cross-panning and other signal manipulation. If it's placed between a stompbox and an amp, I don't think it needs to be an active component. Maybe a clearer picture of the design goal and signal chain will help us see what you intend to do.

edit (additional stream-of-consciousness): Why even alter the amplitude of the dry signal? If you 'duck' the dry signal while adding wet (processed) signal, there will be an audible dip in the volume overall. Simply adding processed signal to an existing 100% dry signal might be easier to implement and more aurally satisfying.

[R.G.]

Had to take a break due to other priorities, but finally got back to this project. Here is my (hopefully) final schematic incorporating buffered inputs and a final non-inverted output: [ATTACH=CONFIG]25539[/ATTACH]
Would appreciate any glaring errors being pointed out.

I believe the panning section is drawn incorrectly. Try relating it back to the original article.

pdf's comments are correct.

Also, the ckt looks overly complex. Do you really need the buffers? By adjusting the impedances in the panning section, you may be able to get rid of them. What's driving this stage? Also, why the 2 stages after the panning section? If you are concerned about phase inversion, the panning section can be connected to the noninverting input of the output op amp. So, it may be possible to do the whole ckt with one op-amp stage.

As Einstein was reputed to say, everything should be as simple as possible, but no simpler.

The original article this was taken from from National Semi did indeed have a single opamp stage, but noted the low input impedance and phase inversion. It does not work well if the panning section is connected to the noninverting input, as the two sides of the panner no longer feed a virtual ground, and hence interact. The inverter after is to reinvert the phase. It can be left out if phase inversion is OK, as was noted earlier in the thread. The first opamp after the panning section is to reintroduce the signal level lost to the panning section for a gain of about unity.

The buffers are to make it be more "polite" and predictable. With no buffers, you need a driving impedance under about 2K or so to be able to neglect loading effects of the panning network. Buffers make it work with impedance up to about the value of the biasing resistors on the front of the buffer. With buffers, you can ignore the question "what's driving this stage" in most cases.

It may indeed be possible to do the entire circuit with one opamp - if you're OK with the loading and phase inversion and always know what will be driving it. Otherwise, $0.25 opamp sections are pretty cheap insurance against feature creep. Actually, I'd be tempted to put a switch-hitter selectable phase inverter on one of the buffered inputs so if one of the inputs is nominally out of phase with the other, I could correct the bass loss that would give. For blending the dry and chorused outputs of a pedal, that may be needed. Chorus pedals usually are a wobbled time delay mixed with some dry signal already. The wet signal by itself is a vibrato. The dry mixed gives it the chorus effect. It's not always clear what phase the original pedal designers gave you as "dry".

[ewd]

After reading the comments and doing some thinking about the design, I made some revisions. I removed the 4.5V bias on the buffers keeping the +/- 9V buffer power supply. That was clearly an error on my part. The R3/R4 positioning was a typo, redrew to put them back in series with the signal flow. Also removed what I believe were redundant capacitors I had incorporated and replaced with a single output cap. My intent is to be able to variably mix the dry signal and wet signals from a chorus box with no overall gain loss in a resultant non-inverted output. I’m using RG Keen’s “Panner” design, with added buffers on each channel (at RG’s previous suggestion), to attempt to accomplish this. I know I could probably find a mfg’d box somewhere to do this, but I want to try to build one. It’s intended for use by a friend to play his harmonica through, outputting to his amp. While I may have missed something, what I have seen so far are chorus boxes with both the dry and wet outputs, with control on how much chorus effect is produced out of that wet output. I have not seen where the signals could be actually be blended together. Again, I may have missed something that is already available, but I wanted to build this myself. The latest schematic revision is posted here:

[ewd]

Thanks RG. I noted in my response in the main thread that I did typo the drawing and posted a new revision with that corrected, and a couple of other changes. Glad you and the others spotted that. That selectable phase inverter is a great suggestion, I may have to work on that after I see how this project works out in actual usage. It would definitely make the circuit more usage-friendly.

[ReadyTeddy]

I believe the panning section is drawn incorrectly. Try relating it back to the original article.

pdf's comments are correct.


As Einstein was reputed to say, everything should be as simple as possible, but no simpler.

The original article this was taken from from National Semi did indeed have a single opamp stage, but noted the low input impedance and phase inversion. It does not work well if the panning section is connected to the noninverting input, as the two sides of the panner no longer feed a virtual ground, and hence interact. The inverter after is to reinvert the phase. It can be left out if phase inversion is OK, as was noted earlier in the thread. The first opamp after the panning section is to reintroduce the signal level lost to the panning section for a gain of about unity.

The buffers are to make it be more "polite" and predictable. With no buffers, you need a driving impedance under about 2K or so to be able to neglect loading effects of the panning network. Buffers make it work with impedance up to about the value of the biasing resistors on the front of the buffer. With buffers, you can ignore the question "what's driving this stage" in most cases.

It may indeed be possible to do the entire circuit with one opamp - if you're OK with the loading and phase inversion and always know what will be driving it. Otherwise, $0.25 opamp sections are pretty cheap insurance against feature creep. Actually, I'd be tempted to put a switch-hitter selectable phase inverter on one of the buffered inputs so if one of the inputs is nominally out of phase with the other, I could correct the bass loss that would give. For blending the dry and chorused outputs of a pedal, that may be needed. Chorus pedals usually are a wobbled time delay mixed with some dry signal already. The wet signal by itself is a vibrato. The dry mixed gives it the chorus effect. It's not always clear what phase the original pedal designers gave you as "dry".

Well, does the panner work poorly when connected to the noninverting input of the op amp? To test, I did a simple LTSPICE simulation. See the pix. below. The first 3 show the panner connected to the noninverting input of the op amp with the control at the 1/2 point, the 1/4 point, and fully to one side. The next 3 show the same thing, except with the panner connected to the inverting input of the op amp. I don't see much difference. For clarity the two inputs are of different frequency, and I delayed the start of one so they could be clearly delineated.







I'll also stand by my comment about eliminating the buffers. The OP knows what circuitry will be driving the panner. If the voltage levels are high enough, the resistors can be scaled to higher values to increase input impedance if necessary and not create any noticeable noise in so doing. Therefore, it may well be the case that in the application in question no buffers may be needed. Certainly, if one had no knowledge of the driving circuitry, buffers would be appropriate.

[R.G.]

I don't see much difference.

OK, you don't.

However, as a generality, it is simpler to get a mixer to work as you expected it to work when you mix into into a virtual ground. Yes, I know that no generality is worth a d@mn, including this one. For reference, go check some info on mixing board design, where mixing into a virtual ground is almost universal. It's usually simpler to get a mixer running into a virtual ground to work the way you originally intended than it is to do it into a high impedance input. Mixing into a high impedance input often needs extra parts to get the mixing to be non-interfering. Not that it can't be done. However, my experience has been that this often requires more than the two extra resistors to put an inverter after it.

If a single panner was the only application that panning fragment was used for, one could go do a non-inverting one. On the other hand, that panning section can be used interchangeably many ways; as a one-in, two-out panner, a two-in, one-out panner (as it is here), or extended to multi-in, multi-out. See http://geofex.com/Article_Folders/panner.pdf for some of how that might work with other situations.

I'll also stand by my comment about eliminating the buffers.

OK, you stand there.

The OP knows what circuitry will be driving the panner. If the voltage levels are high enough, the resistors can be scaled to higher values to increase input impedance if necessary and not create any noticeable noise in so doing. Therefore, it may well be the case that in the application in question no buffers may be needed.

Yep, he does, mostly. Yep, resistors can be scaled. Yep, increases in noise would be calculable, at least. Yep, it may well be that no buffers may be needed. May.

Not worth arguing over. I like the predictability of buffers in experimental situations, which I took this to be. It's perfectly fine if you want to design without buffers by taking into account any special cases of parts/equipment/etc. in hand. It can be done other ways. I'm pretty sure I could do it with vacuum tubes or parametric amplifiers, and quite possibly with an entirely passive mixer that was followed up by one big amplifier to regain the signal level. Many ways to do the job. One might even come up with a single-transistor circuit to do it.

Certainly, if one had no knowledge of the driving circuitry, buffers would be appropriate.

Yep. Or if one was so taken by the beautiful performance of one's new creation that one started looking for other places to use it. People often find new applications and situations to use a circuit in. And they will almost always start with what they did before instead of starting from zero. Untying the special cases is often frustrating. I've done a lot of it, and gotten frustrating at trying to remember what *I* did to myself in terms of "Oh, I won't need *that* in this situation.".

So - you prefer noninverting, no buffers. On the basis of predictability and generality, I prefer buffers plus inverting mixer and a second inverter - and done in two dual opamp packages, which I hadn't said yet for layout reasons.

OK-fine.

[ReadyTeddy]

OK, you don't.

However, as a generality, it is simpler to get a mixer to work as you expected it to work when you mix into into a virtual ground. Yes, I know that no generality is worth a d@mn, including this one. For reference, go check some info on mixing board design, where mixing into a virtual ground is almost universal. It's usually simpler to get a mixer running into a virtual ground to work the way you originally intended than it is to do it into a high impedance input. Mixing into a high impedance input often needs extra parts to get the mixing to be non-interfering. Not that it can't be done. However, my experience has been that this often requires more than the two extra resistors to put an inverter after it.

If a single panner was the only application that panning fragment was used for, one could go do a non-inverting one. On the other hand, that panning section can be used interchangeably many ways; as a one-in, two-out panner, a two-in, one-out panner (as it is here), or extended to multi-in, multi-out. See http://geofex.com/Article_Folders/panner.pdf for some of how that might work with other situations.

OK, you stand there.


Yep, he does, mostly. Yep, resistors can be scaled. Yep, increases in noise would be calculable, at least. Yep, it may well be that no buffers may be needed. May.

Not worth arguing over. I like the predictability of buffers in experimental situations, which I took this to be. It's perfectly fine if you want to design without buffers by taking into account any special cases of parts/equipment/etc. in hand. It can be done other ways. I'm pretty sure I could do it with vacuum tubes or parametric amplifiers, and quite possibly with an entirely passive mixer that was followed up by one big amplifier to regain the signal level. Many ways to do the job. One might even come up with a single-transistor circuit to do it.


Yep. Or if one was so taken by the beautiful performance of one's new creation that one started looking for other places to use it. People often find new applications and situations to use a circuit in. And they will almost always start with what they did before instead of starting from zero. Untying the special cases is often frustrating. I've done a lot of it, and gotten frustrating at trying to remember what *I* did to myself in terms of "Oh, I won't need *that* in this situation.".

So - you prefer noninverting, no buffers. On the basis of predictability and generality, I prefer buffers plus inverting mixer and a second inverter - and done in two dual opamp packages, which I hadn't said yet for layout reasons.

OK-fine.

I don't disagree with anything you said, I was only trying to show that a very simple implementation was possible given certain conditions. :->

[R.G.]

I don't disagree with anything you said, I was only trying to show that a very simple implementation was possible given certain conditions. :->

Yep. That makes sense.

There is always a tradeoff between minimalism, refined "polite" performance, and rococo extravagance. Elegance falls somewhere between the first two.

[eschertron]

I don't disagree with anything you said, I was only trying to show that a very simple implementation was possible given certain conditions. :->

...for the very same reason Einstein gave us his 'special' theory of relativity before he gave us the *general* theory of relativity. Well, almost the same reason.

[ewd]

Yep. Or if one was so taken by the beautiful performance of one's new creation that one started looking for other places to use it. People often find new applications and situations to use a circuit in. And they will almost always start with what they did before instead of starting from zero. Untying the special cases is often frustrating. I've done a lot of it, and gotten frustrating at trying to remember what *I* did to myself in terms of "Oh, I won't need *that* in this situation.".

That's exactly what came to mind after seeing your sugestions about the incorporation of the buffers and the second inversion to regain the original phase of the signal. -Again, thanks RG, and everybody else, for the suggestions and error corrections.

[eschertron]

That's exactly what came to mind after seeing your sugestions about the incorporation of the buffers and the second inversion to regain the original phase of the signal. -Again, thanks RG, and everybody else, for the suggestions and error corrections.

I hope you do take the design and creative processes to further steps. As mentioned in post #19, you can add more buffered inputs to the basic design (controlling them with gain/pan/whatnot) and sum them ALL into the virtual ground connection. You are not limited to two (but you knew that). The design is flexible and expandable. The beauty of op-amp circuits.