1-band parametric EQ

[kothoma]

Does anybody know of a pcb for the following circuit?



(Based on http://geofex.com/Article_Folders/EQs/paramet.htm)

Any Diptrace/Eagle wizzard interested in doing a layout for oshpark?
Preferably with a quad opamp like the TL074. The fourth opamp could be used to buffer Vb.
Or if it saves space then one could replace the first opamp with a JFET or bipolar transistor and go with a dual opamp.

[RobA]

I don't use Diptrace or Eagle, but I could probably put this one together in KiCAD in an afternoon. It's pretty simply and I've been playing with the very similar graphical EQ circuits lately.

You probably don't need to buffer the Vref on this one and the op-amp in the simulated inductor circuit could be replaced with a transistor. I'd need to test how much that would alter things, it has a bit of a change on how sharp the peak can be. I'd think though that you could get this to a single dual op-amp and one transistor easily.

[kothoma]

I don't use Diptrace or Eagle, but I could probably put this one together in KiCAD in an afternoon. It's pretty simply and I've been playing with the very similar graphical EQ circuits lately.

You probably don't need to buffer the Vref on this one and the op-amp in the simulated inductor circuit could be replaced with a transistor. I'd need to test how much that would alter things, it has a bit of a change on how sharp the peak can be. I'd think though that you could get this to a single dual op-amp and one transistor easily.

Thanks for your answer!

Hm, I thought if you use a transistor for the gyrator then you can't have the Q-pot?
Edit: OK. Maybe that impression was wrong. That pot is only damping the peak anyway, so you can apply it (or leave it off) in both situations, I guess.

[RobA]

I think it's possible to have the Q control with a transistor there. When I was playing with these circuits a bit ago, the transistor versions just had a bit less Q to start with compared to and op-amp version. I think it has to do with the op-amp buffering and ability to be a good unity gain follower. Another possible consideration is noise. At very high Q, a really good op-amp could have less noise and these circuits can be a bit noise prone.

Another problem with these as parametric EQ's is that the Q varies with frequency. So, how good this particular type of parametric is depends on what you want to use it for. There are other designs that take a couple more op-amps and some dual-pots, but they would be better as a generally useful one band EQ.

Where you looking for an Eagle or Diptrace layout in particular, or is getting the PCb up on OSH Park the important part?

[kothoma]

Typically I'd use a SVF based design (like http://experimentalistsanonymous.com/diy/Schematics/OOP%20Japanese%20Electronics%20Book/parametric-eq.gif) which requires one of these infamous dual reverse log pots.

But I'd also like a smaller design to put in any dirt box at the input (for example turn a Big Muff Pi into more of a Deluxe BMP) or output, with pots or fixed values.

Sorry, forgot to mention KiCAD wizzards... Any way to a PCB would be great.

Even better would be to have the option of an additional treble shelf (like in a tube screamer) and this would be much more flexible than what I often use now: a resonant lowpass where I often miss the possibility of decoupling the resonance from the treble roll off.

[RobA]

Typically I'd use a SVF based design (like http://experimentalistsanonymous.com/diy/Schematics/OOP%20Japanese%20Electronics%20Book/parametric-eq.gif) which requires one of these infamous dual reverse log pots.

But I'd also like a smaller design to put in any dirt box at the input (for example turn a Big Muff Pi into more of a Deluxe BMP) or output, with pots or fixed values.

Sorry, forgot to mention KiCAD wizzards... Any way to a PCB would be great.

Even better would be to have the option of an additional treble shelf (like in a tube screamer) and this would be much more flexible than what I often use now: a resonant lowpass where I often miss the possibility of decoupling the resonance from the treble roll off.

OK, I've got the picture of what you are looking for then.

A couple of questions: Board mounted pots? If so, size and configuration? Do you have a limit on the size of PCB to be useful?

On the high shelf, it's really easy to add into the circuit. You could put a pot in for the slope of the high shelf. It works but isn't ideal. I had an idea when playing with this that kinda makes the really simple high shelf a bit more useful. If you put the cap in a capacitive multiplier circuit, you can tune the frequency of this too. It does add another op-amp to the circuit. I breadboarded the idea and remember it working pretty well. I could test that idea again if you want to put it in the circuit.

[kothoma]

A couple of questions: Board mounted pots? If so, size and configuration? Do you have a limit on the size of PCB to be useful?

Aiming for a utility board that could fit many situations off-board wired pots would give the best flexibility.
I imagine that the Q often is wired to a fixed value instead of a pot.

As for size: the smaller the better  
Let's say if the regular dirt pedal is of 1590B size then adding the eq should not exceed a 1590BB box.
Could the eq fit in a box of size 1590A?

On the high shelf, it's really easy to add into the circuit. You could put a pot in for the slope of the high shelf. It works but isn't ideal. I had an idea when playing with this that kinda makes the really simple high shelf a bit more useful. If you put the cap in a capacitive multiplier circuit, you can tune the frequency of this too. It does add another op-amp to the circuit. I breadboarded the idea and remember it working pretty well. I could test that idea again if you want to put it in the circuit.

So this would essentially work like a variable slope lowpass filter on the left half of the cut/boost knob? That would be fantastic!

[RobA]

I've got a lull in projects for a couple of days, so I can do a bit of breadboarding and layout on this now. What's the feature set for the whole thing? 1590A is probably going to be a tradeoff between features and size. I'd say that just a three control 1-band PEQ could be done pretty easily. But, and here's where it gets a bit interesting, because of the topology of the controls and bands, it's pretty easy to set this up so that one PCB gives multiple options. You could have a single board that could support one or two PEQ bands, a PEQ band and a high shelf, etc. The slick part is that if you don't want that band, then you don't populate that section of the PCB and it won't impact the sections of the circuit you do want at all. The same is true of setting a fixed Q or having a Q control in place, fixed frequency, ...

I mentioned the possibility of making the high shelf a movable frequency. I haven't tried that with a transistor as the buffer for the capacitive multiplier. I'd need to breadboard that for sure. Using the transistors would be helpful for making the thing modular. There is another possibility with just putting three cap selections on a toggle. This could be just as useful for what you are looking for. I'm guessing from what you mentioned, that you are looking to do some pickup emulation with the EQ. If that's the case, then doing the high shelf with the frequency selection caps on a toggle and a variable slope could be useful. Although, some testing with the variable slope would be needed to see how much we can really get from it and how much it adds to what's already available from the boost/cut pot.

The version with one three control PEQ and one two knob, one toggle high shelf (with boost/cut pot, variable slope, and frequency selector toggle) might be doable on a PCB that could fit in a 1590A with selecting just the controls you want but still make it useful in a bigger build as a utility board. This version should only need one dual op-amp and a single transistor (and the passives).

How's this sound? Anything that definitely needs to go in or out of the idea?

[kothoma]

This sounds fantastic. And I really appreciate the time you (have already) put into this.

The main thing would be controlable mids. It's usually good to have mid boosts before distortion, and mid cuts after.

As for the treble cut/boost this would be a bonus. Both can be useful before and after distortion, depending on the desired result. (Personally it's usually treble cut on either side for me.) A few switchable caps indeed may be all what is needed. Hm, variable slope (varying the R in the RC path to Vb) may not be too important as this overlaps somewhat with the boost/cut function anyway, not sure about that. All this wouldn't take much space on the board, the bigger part beeing pads for pots and a switch.

So that's probably the way to keep this thing small and still usable enough. I'm a bit bothered about the possible sweep range. If I remember right all designs I have seen so far seem to stick to a 10:1 ratio for C1:C2. But Keen uses other ratios. And he states "We only get about a 4.5:1 range". Maybe I need to wade through all the formulas he gives...

Edit: On the other hand, 400...1800 Hz should cover enough mids.

(Edit2: This http://sound.westhost.com/p28_fig1.gif shows (besides a treble shelf) how to turn a peak into a (bass) shelf. I wonder if something similar simple exists for treble?)

[RobA]

This is one of the projects I've had sitting in the back of my head and in the notebook for some time now. It's good to get motivated to actually work on them. My eventual goal is to get it into a guitar body, so I've got to figure out what's really needed.

I think the ESP schematic with the low shelf switch requires a dual power rail for the op-amp. When you switch in the shelf toggle, it's got a non-capacitive path to ground. Though, it might work to rearrange the whole thing to center around a Vref.

You could do a similar thing with the treble/high-shelf and you wouldn't have the problem with the ground path since there will always be the top cap in place. But, toggling to high shelf mode will take the op-amp or transistor completely out of the signal path, so I think it would need a DPDT so that you could ground the input of the amplifier. Pretty much, you'd want to toggle to a completely new path past the caps. It would work though, and you'd make one level pot do double duty. I need to pull out my notes on using the capacitive multiplier idea to move the high shelf frequency, but it might be possible to work that into the system for toggling between peak and shelf.   

The thing with the ratio's and frequency sweep is that the caps essentially set the max Q (with the resonance pot at zero) and then the Q changes with frequency. So, you have to limit the frequency range to keep the Q in a reasonable range and figure how big you can let Q get in one range to make it usable across the whole sweep. I think there's some room to play here, depending on the usage scenario.

[kothoma]

This is one of the projects I've had sitting in the back of my head and in the notebook for some time now. It's good to get motivated to actually work on them. My eventual goal is to get it into a guitar body, so I've got to figure out what's really needed.

That's a nice goal. It could add new dimensions to a guitar. (The only question is: how many knobs can a guitar bear...)

I think the ESP schematic with the low shelf switch requires a dual power rail for the op-amp. When you switch in the shelf toggle, it's got a non-capacitive path to ground. Though, it might work to rearrange the whole thing to center around a Vref.

Not sure I get that. If you added that switch to the split-rail design, wouldn't that have the same effect, Vref being "relative ground"?

You could do a similar thing with the treble/high-shelf and you wouldn't have the problem with the ground path since there will always be the top cap in place. But, toggling to high shelf mode will take the op-amp or transistor completely out of the signal path, so I think it would need a DPDT so that you could ground the input of the amplifier. Pretty much, you'd want to toggle to a completely new path past the caps. It would work though, and you'd make one level pot do double duty. I need to pull out my notes on using the capacitive multiplier idea to move the high shelf frequency, but it might be possible to work that into the system for toggling between peak and shelf.

That would give a nice circuit. It could be customised to all sorts of specialized eqs, or one super flexible yet compact one. But I somehow have the feeling that we are talking about two (related but different) projects simultaneously...

The thing with the ratio's and frequency sweep is that the caps essentially set the max Q (with the resonance pot at zero) and then the Q changes with frequency. So, you have to limit the frequency range to keep the Q in a reasonable range and figure how big you can let Q get in one range to make it usable across the whole sweep. I think there's some room to play here, depending on the usage scenario.

OK, this confirms my gut feeling about the Q of the filter. I think the exact Q isn't critical. No extreme Q should be needed and a bit of change in Q would be acceptable, may even be musical (if it should turn out to change in the right direction...)


Perhaps we should not restrict the discussion to this gyrator based design? There are other bandpass filters that could be used (multiple feedback bandpass, inverting bandpass filter aka active Wien bridge bandpass). They all would need dual pots ideally but you can get away with making only one resistor variable and accept a bit more change of Q along the sweep.

[RobA]

[...]

I think the ESP schematic with the low shelf switch requires a dual power rail for the op-amp. When you switch in the shelf toggle, it's got a non-capacitive path to ground. Though, it might work to rearrange the whole thing to center around a Vref.

Not sure I get that. If you added that switch to the split-rail design, wouldn't that have the same effect, Vref being "relative ground"?

Yes, this should work. You just need to be a bit careful about the details and how you couple to previous stages (if they are there).
[...]

The thing with the ratio's and frequency sweep is that the caps essentially set the max Q (with the resonance pot at zero) and then the Q changes with frequency. So, you have to limit the frequency range to keep the Q in a reasonable range and figure how big you can let Q get in one range to make it usable across the whole sweep. I think there's some room to play here, depending on the usage scenario.

OK, this confirms my gut feeling about the Q of the filter. I think the exact Q isn't critical. No extreme Q should be needed and a bit of change in Q would be acceptable, may even be musical (if it should turn out to change in the right direction...)

Perhaps we should not restrict the discussion to this gyrator based design? There are other bandpass filters that could be used (multiple feedback bandpass, inverting bandpass filter aka active Wien bridge bandpass). They all would need dual pots ideally but you can get away with making only one resistor variable and accept a bit more change of Q along the sweep.

I think it would be possible to rig it so that the Q would be maximum at a particular point in the frequency sweep and then fall off from there in either direction. Maybe it could be done with the peak close to one edge or the other.

The weak spot of the gyrator version looks to me to be the Q control pot. It's not as powerful as other filter topologies. But, some of the advantages for this particular application are pretty nice -- like being able to drop one band or adding the shelving toggles, etc. without disturbing the overall layout.

If you have or can install LTSpice, I can do a simulation that'll let us look at the frequency responses for various settings and that might help guide the choices.

[kothoma]

If you have or can install LTSpice, I can do a simulation that'll let us look at the frequency responses for various settings and that might help guide the choices.

OK, just downloaded and installed LTspiceIV and successfully opened one of the example files.

I was tempted to do so in the past but always chickened out. Schematics done in LTspice always looked different and scary.

Good opportunity to finally jump right in.

[Thomas_H]

I have only seen this now, so I may be a little late to the discussion 
I already have a 1590a design and some pcbs lying around as I recently did a custom build for a bass player using a single param eq as discussed here.

Its an OpAmp design based on RG Keens idea. As a standalone device it needed an additional adjustable gain stage.

I will post some details later.

[RobA]

If you have or can install LTSpice, I can do a simulation that'll let us look at the frequency responses for various settings and that might help guide the choices.

OK, just downloaded and installed LTspiceIV and successfully opened one of the example files.

I was tempted to do so in the past but always chickened out. Schematics done in LTspice always looked different and scary.

Good opportunity to finally jump right in.

The schematic capture is a bit weird and limited. The standard parts are also limited, but OK for most of the stuff needed for basic modeling for effects circuits. The bigger issues are knowing the tricks to getting Spice to behave -- and I certainly don't know enough of those. Still, it works alright for getting ideas. I wouldn't trust anything without breadboarding and listening first though.

I adapted another circuit I had modeled to match the one from the RGK graphic you linked above. It's at http://rock.it-frog.com/Downloads/Code/PEQ_2Band_OPA.asc

There are three sections in there. The first two are what the PEQ section would look like in either low shelf or PEQ mode. The third one is a high shelf made so that it can be frequency swept using a capacitive multiplier circuit.

Notes
Pots: I haven't seen pots in LTSpice (I could have missed them easily if they are there), so I model pots as resistor divider networks or just as a resistor for variable resistor pots. So, R2 and R22 form one pot. R1 is the resonance pot. Notice that R1 is set to 1Ω. That's because you can't set a resistor to 0Ω in LTSpice and have the simulation work. R26 is a combination of the frequency pot and the 51k resistor in the above schematic. (I used 47k) So, you can vary the frequency by moving this resistor from 47k to 1047k.

High shelf: C4 sets the base frequency in combination with the 8.2k resistors and R7. R5 is a variable pot and a 470Ω resistor combined. The total capacitance of the section is C4 * R5/R8 and that will move the two corner frequencies for the high shelf. R7 is also a 470Ω resistor in combination with a pot. The shelving filters are a bit weird because the slope is really set by moving the upper and lower corner frequencies and these are influenced by the values of the gain network resistors (R29 and R13) as well as the caps and resistors in the shelf filter network. They are a bit fiddly to work with. I went to the 8.2k resistors to get a range for these that's easier to work with.

PEQ to High shelf?: If you look at the high shelf and the PEQ sections, they are really pretty close to each other. Basically, you need to flip the position of R5 and C4 and then add a cap at the top above R7. The problems here are that I think that'll take a 3PDT and the pot that is part of R5 is very different from the high shelf to the PEQ (10k to 1M or so). So, if you want to do a combo high shelf and PEQ, I think it would need to go to a simple shelf with some fixed capacitor values on toggles.

If you happen to use Python, here are a couple of functions I use in an interactive session to look at cap values to play with. the caps function returns the two caps needed to match a desired Q and frequency. The second function returns the range of frequency and associated Q from the extreme settings of a 1M frequency pot. 

Code Select Expand


from math import *

def caps(q, f):
c1 = 1.0/(q * 2.0 * pi * f * 470.0)
c2 = q/(2.0 * pi * f * 47.0e3)
return c1, c2

def f_q(c1, c2):
q2 = sqrt((c2 * 47.0e3)/(c1 * 470.0))
q1 = sqrt((c2 * 1047.0e3)/(c1 * 470.0))
f2 = 1.0/(2.0 * pi * sqrt(470.0 * 47.0e3 * c1 * c2))
f1 = 1.0/(2.0 * pi * sqrt(470.0 * 1047.0e3 * c1 * c2))
return f1, q1, f2, q2



The Q value for the PEQ is proportional to the square root of the frequency pot setting (plus the base 47k resistor), so it increases (a lot) as the frequency goes down. That's kind of the opposite of what I'd want it to do and puts a restriction on how high you can set the Q on the high end. The resonance control can help offset that though, so including it will give you a broader range of usable Q.