Questions regarding Gyrator based GEQ

[Vivek]

I tried simulating some transistor and IC based Graphic Equalisers

The SPICE analysis shows what seem to be well-known features of this type of circuit :

A) The Q is fairly low, and the Q changes based on the amount of boost or cut. Is this found acceptable in real life ? If not, what can be done about it ?

B) Most of the action seems to be at the ends of the pot travel

Is there any special taper pots for this application ?  The Schematics I found on the net say Linear pots.

Do the pots actually sound linear acoustically ?

C) Both my simulations have a Manta Ray type tail at the high end. I have not completed my Analysis, but I feel this might have something to do with the highest filter band not having any gyrator. Any other ideas ?


MXR-10 Band GEQ


Boss GE-7 GEQ

[antonis]

Is there any special taper pots for this application ?

S taper..

https://www.diystompboxes.com/smfforum/index.php?topic=33186.0

[kraal]

W-Taper ? https://www.taydaelectronics.com/potentiometer-variable-resistors/rotary-potentiometer/w-taper.html

[Vivek]

If SPICE shows with linear pot: low action at centre of pot
and then lots of action at extremes of pot

Would I need
Pot that has slow changes at extremes but rapidly changes in the middle ?

[Ripthorn]

Would I need
Pot that has slow changes at extremes but rapidly changes in the middle ?

Yes.

Regarding the change of Q with gain and whether it is acceptable or not, it is what everyone is used to, even if it is theoretically suboptimal. I don't think anyone has gone through the effort to make one with a steady Q and compare to a traditional unit, so I'm not sure that we can say until you try it 

[ElectricDruid]

Would I need
Pot that has slow changes at extremes but rapidly changes in the middle ?

Yes.

+1 Agree. The pot response needs to compensate the gyrator response.

Regarding the change of Q with gain and whether it is acceptable or not, it is what everyone is used to, even if it is theoretically suboptimal. I don't think anyone has gone through the effort to make one with a steady Q and compare to a traditional unit, so I'm not sure that we can say until you try it 

It's true that that's what everyone is used to, but it's not true that there's nothing with steady Q. There's this from Rane:

https://www.ranecommercial.com/legacy/note101.html

They built it, so of course they argue that it's better than the competition. It's different, and that's often helpful in some situations and awkward in others. Usually, each solution has strengths and weaknesses.

[Rob Strand]

It's all very normal.

See this as a reference,

http://www.muzique.com/lab/gyrator.htm



And this common circuit,



Consider an EQ only set-up for boost - where you have a feedback resistor and a gyrator (to ground).

Using AMZ's numbering, you can have a gyrator with the gyrator series resistor R2.   The maximum boost is 1 + Rf / R2 where Rf is the feedback resistor.   R2 determines the inductance and Q of the gyrator.

Now add a resistor Rs in series C1.    The additiion of Rs does not change the gyrator inductance.   However it does reduce the maximum boost to 1 + Rf/(R2 + Rs) and it does reduce the Q.   Q reduces in proportion to 1/(R2+Rs) and the maximum boost reduces approximately by the same factor.   The Q and maximum boost are nearly proportional which why these equalizers are called proportional Q equalizers.

Next notice what happens when we change Rs.  When Rs is about the same magnitude as R2 the gain starts to be noticeably change.  For example  Rf = 10k, R2 = 2.2k would give 14.9dB boost.  Then if we add Rs = 2.2k  the boost drops to 10.3dB.  And when Rs+R2 is in the order of about 2Rf to 4Rf the amount of boost is quite small, 3.5dB to 2dB.    What this is saying is the usable range of Rs is going to be in the order of R2  to about 2Rf (to 4Rf).

The other thing is when Rs is close to R2 the rate the max boost changes with Rs is quite fast but as you get to Rs=Rf the rate of change slows down.

The complete equalizer acts pretty much like the clockwise half of the pot is in series with gyrator.    When boosting it's almost like the counter clockwise part of the pot is cut.    With this in mind you can see if you have a large valued gain pot  only a tiny part of the rotation is going to be in the order of R2, as R2 is small.  If the gain pot is somewhat larger than Rf then even the flattening out point where Rs become close to 2Rf could occur at small rotation of pot eg.  Rf = 10k and Rpot =100k  the pot behaviour is flattening off 20k from the end and the middle is useless.

The only way to compensate for the uneven control is the funky tapered pots.

However, it *is* possible to get OK adjustment with a linear pot.   The trick is not to use a large value for the boost/cut pot.  Make the value something like 2Rf to 4Rf and the control will work  OK-ish over the control range.  If you have 20dB boost then you might see some bunching-up of gain near the extremes however with 12dB or 15dB it's OK.

Another point is when you have the boost/cut pot connected in a full equalizer the pot does end-up affecting the Q in a multiband equalizer.   Having large a large boost/cut pot value reduces that effect.  However,  this is simple making the circuit agree with the *simple* theory.   In reality you need to design the equalizer gyrator with higher Q than you are targeting.    If you choose a lower pot value you will need to make a larger deviations from the theory than with the larger pot value. (The aim is to get the result not to match the simplified theory!)

If you want a graduated dB scale then what you need to do is use the funky taper pots *and* select the pot value so the amount of boost and cut matches the scale.
   

[Vivek]

Are schematics of constant Q GEQ released to the public ?

Any patents on this which we can study ?

[ElectricDruid]

Are schematics of constant Q GEQ released to the public ?

Any patents on this which we can study ?

There's further implementation details (a full discussion of several methods, in fact) in one of the papers linked from that first page:

https://www.ranecommercial.com/legacy/pdf/constanq.pdf

The basic principle is shown in Figure 8. Any bandpass filter will do the job, so choose one you like.

[Rob Strand]

If you take Bohn's figure 8 you will notice you have the freedom to choose the pot value.   A high value pot will cause bunching of the control near the extremes (boost vs position) even though it keeps the constant Q-ness (Q vs boost).  So there is a right value pot to suit the pot taper.

Also note he uses tapped pots.   This trick can be used on the gyrators circuits as well.   For the constant Q equalizer the tapped pots help with the constant Q behaviour.   When the pots are not tapped the response is no longer a 1 + k*band-pass response it's something (roughly) like a (1+ k*bandpass)* [1/(1 + kleak(1-k)*bandpass)].  The part is the brackets [] is leakage though the pot back to the cut circuit which causes non constant Q.   You can reduce the leakage by making the boost/cut pot larger relative to the feedback resistors but then you end-up with more control bunching.   You can scale the pots value so the bunching matches the funky pot taper gyrator however its more common to give up some constant Q-ness and just use the linear pots.    Keep in mind large pots values will cause control bunching.
 

[Vivek]

Any guesses on why there is a High boost shown in my Frequency response graphs ? What is causing it ?

[Rob Strand]

Any guesses on why there is a High boost shown in my Frequency response graphs ? What is causing it ?

You already got it,

C) Both my simulations have a Manta Ray type tail at the high end. I have not completed my Analysis, but I feel this might have something to do with the highest filter band not having any gyrator. Any other ideas ?

Some guitar equalizers have shelving type EQ on the highest band instead of peaking EQ.   Look at some Hi-Fi or Studio EQ schematics and you will see a gyrator on the highest band.

One thing you will see on those gyrator equalizers is some slight asymmetry with the boost characteristic.  At high frequencies the boost tends to flatten-off towards a slight positive boost.   That effect it due to the RL, the resistor to ground on the gyrator.   The simple gyrator circuit doesn't model an ideal inductor it models an inductor with series and parallel resistance.  The parallel resistance is what cause the slight rise at HF.  If you replace the gyrator with an inductor (+series resistor) the HF effect boost will disappear.

EDIT:
You can see the RL effect here,
https://www.diystompboxes.com/smfforum/index.php?topic=123011.msg1161976#msg1161976

[MG]

C) Both my simulations have a Manta Ray type tail at the high end.

There's the name of your new pedal!

It's normal to use linear pots for circuits that have both boost and cut (above and below a 0db baseline).  Otherwise, when set at the centerpoint, response would not be a straight line.

Having said that, I personally never worry about textbook response when dealing with guitars.  The history of guitars and amps is filled with stories of great-sounding design errors.

You'll also find that high Q peaks don't sound great unless you're looking for a special effect.  High Q notches are fine.  Peaks not so much.  You probably found that already.

Also the optimal center frequencies for mid boost and mid cut are arguably not the same.  Fender amps cut mids around 300 Hz, which might seem low, but that's a key part of their sound.  Stock Fenders do not actually boost mids. The mid control just decreases insertion loss and makes the response a tiny bit flatter, compared to their normal scoop.  But if you want true mid boost, try 800 hz or so.  Boost at 300 hz will sound tubby.

If you really did want to have a continuously variable boost with high-Q peak, you could run two parallel signal paths in phase; one flat, the  other through the balls-out peak filter.  Pan between them.

For mid cut:  Fender controls basically comprise a bridged T.  With the mid control at 0 (or the 6800 ohm mid resistor jumpered out), there is a deep phase notch at 300 hz.  Not as noticeable or obnoxious as a sharp peak, of course.  You can see that if you can find a Fender/Marshall/Vox tone control simulator.

I wrote various versions of tone control simulators in C++ back in the glory days of the alt.guitar.amps newsgroup (Hi, RG!) but I'd probably have to recompile them to run them now.  I do have screen captures somewhere if you can't find anything else.  Main point being the mid cancellation at 300 hz being visible.

Anyway, the gyrator idea is cool, but linearity is not always linearity when it comes to guitar. :-)  Try two separate filters:  A relatively low-Q boost circuit centered around 800 hz, which will work with a log pot.  And then try your gyrator for the cut circuit at 300 hz.  The transition from lower Q to sharper Q will sound good on the mid cut control.  But not on the boost side.  Choose the pot taper that sounds good on the mid cut side, according to the circuit you come up with.  Regular bridged T mid cut works well with log taper pots.

Nice thing is that both of those controls can be used at the same time.  Put them in series.

There's everything that you didn't ask about :-)  but hopefully the question of pot taper will be more clear, at least.

[PRR]

... if you can find a Fender/Marshall/Vox tone control simulator. I wrote various versions of tone control simulators in C++ back in the glory days of the alt.guitar.amps newsgroup (Hi, RG!) but I'd probably have to recompile them to run them now. 

https://www.duncanamps.com/tsc/  --- for Windows

https://www.guitarscience.net/tsc/info.htm   --- browser-based (javascript)

[MG]

... if you can find a Fender/Marshall/Vox tone control simulator. I wrote various versions of tone control simulators in C++ back in the glory days of the alt.guitar.amps newsgroup (Hi, RG!) but I'd probably have to recompile them to run them now. 

https://www.duncanamps.com/tsc/  --- for Windows

https://www.guitarscience.net/tsc/info.htm   --- browser-based (javascript)

Good to see that Duncan's version is still there.  I hadn't seen the JavaScript interface before.

I toyed with the idea of writing web-facing versions of a couple of my simulators, but it got too time-consuming.  My program updates the screen and graphs instantly as each control or component value is changed, so the interface wasn't simple.  I don't care for JavaScript anyway.

For now though, the OP will be able to look at Duncan's output graphs to see examples of Fender's frequency response.  Perhaps that will be helpful when designing the mid cut circuit.