stupid question on AC coupling caps...

[armstrom]

I'm working on a little dry/wet mixer circuit with some gain on the front end. I have a stupid question about using polarized caps for the AC coupling between two OP AMP stages (single supply).

The one I'm concerned about is C9 (in the big red circle ) Will I have a problem with the VR (4.5V in this case) being tied to the negative side? Do I need to go for a lower value cap so I can avoid polarized electrolytics?

This is just a snip-it of the circuit but does anyone have any other comments? Glaring errors? It's operation should be pretty easy to figure out. The first stage serves as the input buffer and adjustable clean boost (up to ~20dB). The net with C10 on it is the "dry" signal. The second op-amp buffers the effect input which is then tied to the bottom of VR2 which acts as a dry/wet mixer. I chose a blend pot approach since some of the effects this DSP offers work best with a "fully wet" signal while others work well with a mixture of dry and wet.

[R.G.]

The one I'm concerned about is C9 (in the big red circle ) Will I have a problem with the VR (4.5V in this case) being tied to the negative side? Do I need to go for a lower value cap so I can avoid polarized electrolytics?

Polarized caps should never be reversed, even by signal.

First of all, C9 doesn't need to be that big to start with. The input impedance of U2b is huge, being a JFETgate. The effective impedance is that 470K resistor, R8.

The low frequency rolloff of C9 and R8 as it sits is F = 1/(2*pi*r*c) = 1/(2*3.14*10E-6*4.7E5) = 0.0338Hz. Nobody hears that low. If you wanted full hifi range, you could pick C9 = 1/(2*pi*f0*R8) = 1/(2*pi*20Hz*470K) = 1.6nF, or 0.0016uF. You're about a factor of ten thousand bigger on C9 than you need to be.

Just as a curiosity, have you computed the frequency rolloffs of the other RC networks in your magic secret sauce? 

[armstrom]

Thanks for the reply, I'll use a lower value. The reason I'm using 10u is because it is the value suggested in the datasheet for the DSP board I'm using.
As for the other RC networks, I basically pulled the gain stage directly from the "Roll your own distortion" article posted on GGG.

I chose C8 to give more low frequency response (15Hz roll-off I believe) while C7 should roll-off everything above 33Hz. While I know both of these are outside normal audible range, I didn't want to restrict things too much as this circuit could easily be used by a guitar or bass. I suppose I could lower the 33kHz filter, but why bother?

Same thing for the networks on the input. R4 and C6 rolls-of below 15.9Hz, C5 and R5 roll-off above about 16kHz which I would guess is what you're pointing toward. The input network limits the frequency response to about 16kHz while I'm filtering out above 33kHz in the feedback loop. I'll adjust that.

Another quick question then... Can I simply omit C9 and allow the bias voltage to pass-through? to the next stage? C10 will keep it from reaching the dry side of the blend pot. Or will the gain of the first stage screw with  the bias voltage? I ask because I've seen that technique used on circuits like the marshall cab sim.

Just out of curiosity, why do you call it "magic secret sauce"??  I'm not trying to keep anything secret. I simply posted a schematic snip-it to avoid cluttering things up with the power supply circuits and other I/O pins that are not directly related to the question I was asking.

[armstrom]

Ok, so now you've made me go back and rethink the RC networks in the feedback loop of the first gain stage. I'm struggling trying to find a good value for C8. I realized I had just run the numbers at full gain (VR1 all the way up). As I reduce the gain the low cut frequency creeps up too high.

Now I'm not quite sure I've identified all the networks in the feedback loop that I need to keep track of.  I'm pretty new to this, could you provide some pointers? Would I be better off increasing the size or R7 and VR1? maybe 5K and 50K? Just trying to learn here

[R.G.]

The (-) input of an opamp is a virtual ground, driven to almost exactly the same voltage as the + input. That being the case, the rolloff of C8 and R7 are not affected by VR1. The rolloff is F = 1/(2*pi*R7*C8) = 159Hz.

However, the rolloff ov VR1 and C7 changes a lot, from 34kHz on up - not audible, thankfully.

I don't see the low frequency rolloff changing with VR1.

[armstrom]

So, would you say the configuration I had before was sufficient? I don't want to change the frequency response of the signal (at least in the audible range).

I'm also toying with the idea of moving the gain adjustment to after R7 and convert VR1 to a fixed value. Is there any advantage to one approach vs the other? The only difference I can see is that it won't be possible for the stage to go down to unity gain (aside from using a pot with a switch in it to cut the connection to ground) but that's not a big deal.

Sorry for the barrage of questions, and thanks for taking the time to reply to them

[R.G.]

So, would you say the configuration I had before was sufficient? I don't want to change the frequency response of the signal (at least in the audible range).

Make C8 10uF. That gives a low frequency rolloff of 15.9Hz.
The 1nF C5 may be a problem with some source impedances.

I'm also toying with the idea of moving the gain adjustment to after R7 and convert VR1 to a fixed value. Is there any advantage to one approach vs the other? The only difference I can see is that it won't be possible for the stage to go down to unity gain (aside from using a pot with a switch in it to cut the connection to ground) but that's not a big deal.

Either way is OK as long as you have the headroom in U2A. If you move the adjustment, then U2 is always doing gain, and may clip on big signals that you'd be padding down with a later pot. But you can put a resistor under the fixed value pot to ground that does not let you turn the wiper all the way to ground.

[armstrom]

great, thanks for that!  I've adjusted the schematic. Here's a link to the full schematic. http://www.mr2-power.com/verb.jpg

J4 is a DSP board header, J5 is a header for a binary coded rotary switch to select between the 16 different effects in the DSP. For now I'm using the unregulated 9V power for the op amps to retain as much headroom as I can. The DSP is powered by the regulated 5V supply. Technically speaking my headroom maxes out at 1V RMS/2.8V P2P due to the hard limit of the DSP.

Would choose a different value for C5? or just omit it all together since there's some filtering in the feedback loop already?


Also, while we're on the topic of op amps... What are your thoughts on biasing the inverting input vs. the non-inverting input on single supply circuits? I've read some app notes from TI and others that seem to indicate this is a preferred method. However, most of the circuits you see floating around bias the non-inverting input (as I've done here). Here's a link to the document: http://instruct1.cit.cornell.edu/courses/bionb440/datasheets/SingleSupply.pdf Pages 5-6 illustrate the type of circuit I'm talking about.

[R.G.]

Would choose a different value for C5? or just omit it all together since there's some filtering in the feedback loop already?

I would consider what was driving it. If it's a guitar, C5 will wipe off a lot of highs. If it's an opamp, or other low impedance source, it can supply enough current to make C5 almost irrelevant.

Also, while we're on the topic of op amps... What are your thoughts on biasing the inverting input vs. the non-inverting input on single supply circuits? I've read some app notes from TI and others that seem to indicate this is a preferred method. However, most of the circuits you see floating around bias the non-inverting input (as I've done here). Here's a link to the document: http://instruct1.cit.cornell.edu/courses/bionb440/datasheets/SingleSupply.pdf Pages 5-6 illustrate the type of circuit I'm talking about.

In a perfect world I'd want to be the last one to tell TI how to bias an opamp. However, my experience leads me to believe I'm either not too sharp today and missing something, or else there was an editorial mistake in Figure 3, which is the only one I can find that I interpret that way. It is possible that the writer is more informed about something than I am. But whenever I've tried that second form, the thing always drifted all over the power supply. Even if the + input of the opamp in the second example is perfectly matched to the - input, the input impedance is unreasonably high, being just the impedance of a JFET gate in parallel with the resistance of a capacitor. These are both so high that things like the resistance of dust and crud on the surface of the PCB and air humidity can couple in signals to make the input wander around.

On the other hand, I know that tying the + input to a bias voltage with even a large resistor fixes the bias point firmly. So I think that there may be some problem with Figure 3. Perhaps there was an R3 to Vbias that they meant to be equal to R1||R2 that got left off. That's kind of implied by the notes in the figure, but not made explicit, and certainly not in the drawing.