Can this non-inverting stage be converted to inverting?

[decc]

Here's a simple gain stage in a non-inverting configuration. (Values would probably change in actual use, but you get the idea.)



Z_in = R1 = 10k
High Pass f_in = R1 & C1 = 72Hz
Gain = 1 + R2 / R3 = 11x
Low pass to limit gain bandwidth = R2 & C2 = 3300Hz
High pass to limit gain bandwidth = R3 & C3 = 16Hz

Let's make it inverting:



Z_in = R1 = 10k
High Pass f_in = R1 & C1 = 72Hz
Gain = R2 / R1 = 10x
Low pass to limit gain bandwidth = R2 & C2 = 3300Hz
High pass to limit gain bandwidth =

Where does this last piece fit? It would probably not be audible, at least with these values, but I would like to understand the equivalent way to limit the gain on the low-end with an inverting stage.


Thanks

[anchovie]

It's not so much a case of fitting in the last piece on the second diagram, it's taking out the extra piece on the first. In the inverting configuration, what you've called C1 and R1 would be the equivalent of C3 and R3 in your non-inverting setup. You'd add a passive filter before C1 if you wanted the two roll-offs.

C1 and R1 on the non-inverting stage are not part of the active part of the circuit - R1 affects the input impedance of the non-inverting input and C1 would be chosen to make a passive high-pass filter according to R1's value. So with that circuit you'd most likely want R1/C1 to have the roll-off at 16Hz and R3/C3 to provide the 72Hz so that it's touching a frequency range that hasn't already been squished.

[decc]

It's not so much a case of fitting in the last piece on the second diagram, it's taking out the extra piece on the first. In the inverting configuration, what you've called C1 and R1 would be the equivalent of C3 and R3 in your non-inverting setup. You'd add a passive filter before C1 if you wanted the two roll-offs.

C1 and R1 on the non-inverting stage are not part of the active part of the circuit - R1 affects the input impedance of the non-inverting input and C1 would be chosen to make a passive high-pass filter according to R1's value. So with that circuit you'd most likely want R1/C1 to have the roll-off at 16Hz and R3/C3 to provide the 72Hz so that it's touching a frequency range that hasn't already been squished.

How would the extra passive filter be arranged exactly? I tried simulating a few things and could not get similar results between the two circuits.

Would this not be equivalent anyway as  R1/C1 and R3/C3 serve two different purposes in the non-inverting input? The first limits the low end in to the circuit, while the second limits the gain applied to the low end. As I understand it the two are not equivalent, at least in practice, as there can be low-frequency noise injected via the power rails and thereby skipping the input filtering.

[anchovie]

A passive low-pass filter would be just like your C1/R1 setup on the non-inverting stage, except with the resistor going to ground rather than Vref. Like I said, though, this setup is more to do with setting input impedance without necessarily affecting the tone.

I'm just going by all the tutorials that I've ever read, and they've always said that for bandpass filtering on an inverting gain stage you'd have an RC lowpass on the input and the highpass in the feedback loop. Here's an example: http://www.physics.unlv.edu/~bill/PHYS483/op_amp_filt.pdf

Power supply filtering should be carried out in its own domain, not as part of a gain stage that should be just dealing with an audio signal.

[decc]

A passive low-pass filter would be just like your C1/R1 setup on the non-inverting stage, except with the resistor going to ground rather than Vref. Like I said, though, this setup is more to do with setting input impedance without necessarily affecting the tone.

OK, a little experimenting and I came up with this:



12k seemed to get the plot closer to the non-inverting one. It's a dB or two off from about 10Hz to 100Hz but that's it. I tried connecting it to Vref instead of ground too, which simulated OK. Any reason it should go to ground instead?

I'm just going by all the tutorials that I've ever read, and they've always said that for bandpass filtering on an inverting gain stage you'd have an RC lowpass on the input and the highpass in the feedback loop. Here's an example: http://www.physics.unlv.edu/~bill/PHYS483/op_amp_filt.pdf

Power supply filtering should be carried out in its own domain, not as part of a gain stage that should be just dealing with an audio signal.

Thanks for the link. That explained the situation a lot better than some of the more involved write-ups I've read. And while I agree power-supply filtering is it's own thing it never hurts to cover all your bases when dealing with home-made stuff. (I also think reducing the bandwidth that the gain applies to helps with noise and stability somehow, and this must be accomplished in the feedback loop itself.)

[grapefruit]

You can omit R4 and C4 in that last diagram. C1, R1 are limiting the gain at the bottom end. If you want the same gain as the non inverting circuit you posted you'd have to make R2 110k.

Stew.

[decc]

You can omit R4 and C4 in that last diagram. C1, R1 are limiting the gain at the bottom end. If you want the same gain as the non inverting circuit you posted you'd have to make R2 110k.

Stew.

As anchovie said, C4 & R4 are a second filter to increase the roll-off to match the steeper one in the non-inverting config.

[anchovie]

They don't have the same effect impedance-wise as the input filter on the non-inverting stage, though. I still don't think it does a lot from an audio point of view. If you're worried about 50/60Hz hum (depending on location) getting into the audio path from the power line then you need a better designed PSU.

[grapefruit]

But the -3dB point of RI C1 is 72 Hz, where R4 C4 is 16 Hz.
Power supply noise or ripple shouldn't be an issue. R4 C4 is pretty much redundant.

Stew.

[decc]

They don't have the same effect impedance-wise as the input filter on the non-inverting stage, though. I still don't think it does a lot from an audio point of view. If you're worried about 50/60Hz hum (depending on location) getting into the audio path from the power line then you need a better designed PSU.

You're right, one would probably not hear the difference in this case. I was just trying to figure out if there is an exact equivalent between the non-inverting and inverting circuits. Looks like one can get close, but they aren't the same thing.

[PaulC]

If you're trying to make the two as close as you can to each other you first need to factor in the gain difference between the two designs.  You showed that the non-inv was (R2/R3)+1 = 11, so you need to give the inv opamp a gain of 11.  Make R2 (100k) = 110k.  Then you need to change the 470pf by the same amount to 427pf.  Not critical in the real world (hard value to find anyway), but if you're trying to match up plots you need to do this.

If you want to match the extra pole of filtering that you get with the 10k/1uf you can do it the way you showed, but the two poles will interact in the inv setup.  The poles are isolated in the noniverting.  One way to get you close is to change the 12k back to 10k, but to change the value of c1(.22uf) to make up for it being in series with C4 (1uf).  C1 should be .28uf

In the real world you might have a buffer in front of the inv op because of it's low imp, and you could just put the 1uf/10k pole at the input of the buffer.  This would work more like the non invertering setup due to the isolation of the poles. 

The other thing to remember though is the slope of the non inverting gain setting pole is somewhat gain dependent.  At low gain settings it's more of a shelf EQ responce because the slop can't extend below unity, and at higher gain settings it get's closer to a 1st order responce.  You have to compensate for that responce if you're trying to nail the curves.

Hope this helps

PaulC

[decc]

If you're trying to make the two as close as you can to each other you first need to factor in the gain difference between the two designs.  You showed that the non-inv was (R2/R3)+1 = 11, so you need to give the inv opamp a gain of 11.  Make R2 (100k) = 110k.  Then you need to change the 470pf by the same amount to 427pf.  Not critical in the real world (hard value to find anyway), but if you're trying to match up plots you need to do this.

I was actually aware of this originally but left that detail out thinking it would simplify my question. Sorry about sending people down the wrong path.

If you want to match the extra pole of filtering that you get with the 10k/1uf you can do it the way you showed, but the two poles will interact in the inv setup.  The poles are isolated in the noniverting.  One way to get you close is to change the 12k back to 10k, but to change the value of c1(.22uf) to make up for it being in series with C4 (1uf).  C1 should be .28uf

...

The other thing to remember though is the slope of the non inverting gain setting pole is somewhat gain dependent.  At low gain settings it's more of a shelf EQ responce because the slop can't extend below unity, and at higher gain settings it get's closer to a 1st order responce.  You have to compensate for that responce if you're trying to nail the curves.

This is the key I was looking for. There just isn't a place for a second, non-interacting filter on the inverting config nor is there a way to make this gain dependent.

In a real-world circuit this doesn't really matter, but I'm glad I understand how things work better now.

Thanks to everyone