Dual EQ filter with a single pot?

[ElectricDruid]

I'm looking for soft, broad boosts and cuts). I could reconsider the mid cut (make it narrower, even though, I must say, with a Q=0.6 it works very well, also because the bass boost-low mid cut and high mid boost-treble boost are voiced so as to actually be used and never left flat, and they compensate each other very well), but I'm quite happy with the bass boost. The high-mid boost and high shelf, piece of cake comparatively speaking, but one thing at a time.

So do you have specific frequency and Q values in mind or are you experimenting?

And how many elements does this thing need? Initially you were talking of a bass boost and mid cut, but now you've mentioned a low-mid cut and a high-mid treble boost too, so are those additional?

Sorry, I'm losing track of where we're trying to get to.

[fryingpan]

Yes, I'm sorry, I wasn't clear. Basically it's going to have two pots, one doing bass boost (up to 10dB, 80Hz, Q=0.32) and mid cut (up to 6.5dB, 250Hz, Q=0.6), and the other doing high mid boost (up to 10dB, 2kHz, Q=1.20) and a high shelf (up to 5dB, about 6.5kHz). The second EQ stage is easier, probably, but the first is giving me more trouble.

[ElectricDruid]

Yes, I'm sorry, I wasn't clear. Basically it's going to have two pots, one doing bass boost (up to 10dB, 80Hz, Q=0.32) and mid cut (up to 6.5dB, 250Hz, Q=0.6), and the other doing high mid boost (up to 10dB, 2kHz, Q=1.20) and a high shelf (up to 5dB, about 6.5kHz). The second EQ stage is easier, probably, but the first is giving me more trouble.

Ok, thanks, that's great.

In that case, here's my revised attempt with the values adjusted to get as close as I can for stage one, the bass boost/mid cut.

[fryingpan]

That's great. May I ask you a few questions?

- what's the purpose of R1 and R4, and especially of C5 (and why it's on the bass boost but not on the mid cut)?

- what topology is this? I'd like to be more independent in the future.

Thanks a lot!

[ElectricDruid]

- what's the purpose of R1 and R4

U3 is basically a non-inverting op-amp. Instead of two resistors to set the gain, it has R1 and a gyrator, which gives a frequency-dependent resistance (a reactance). Since the resistance of the gyrator drops at it's tuned frequency, the feedback goes down, and the gain goes up and the U3 stage makes a boost.

R4 and the second gyrator do the same thing, except that this time it's set up as a passive voltage divider. Hence when the resistance of the gyrator drops, the volume drops, and it gives a cut.

and especially of C5 (and why it's on the bass boost but not on the mid cut)?

It's just a small cap to limit the high-end response of the op-amp, a stability cap. I was taught that should never amplify anything that you don't want amplified, so you should limit the top and bottom of an amp's bandwidth to only the range you're interested in. In this case, I just stuck a tiny cap in as a reminder, but the rolloff doesn't happen till 32KHz.
That's why there isn't one similar on the mid cut - no op-amp, so no potential oscillation. You could add one across R9, and if I was building this for real, I probably would.

- what topology is this? I'd like to be more independent in the future.

The two gyrators are standard stuff that I designed by throwing values at the AMZ tool:

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

U3 is a standard non-inverting op-amp, as mentioned above.
U1 is an active panning/crossfade circuit, best described by RG here:

http://geofex.com/Article_Folders/panner.pdf

[fryingpan]

So, I've actually got round to testing the circuit on LTSpice, but things don't seem to work as they should.
I tried designing your suggestions and they didn't work, I was already working on my circuit (I reshuffled the values to make them workable with common components) and this is the result:



If I take the clean buffered signal and add it to the Wien bridge (U2 + U5) and measure the output of U2 + U5, this happens:



Basically, a 9.7dB boost at around 80Hz. The Q is actually wider than I wanted, around 0.23 instead of 0.32, but it's workable.
The treble boost is to counteract the treble cut due to the following gyrator stage, which measures like so (U5 -> gyrator -> U1, measurement at the output of U1):



Which is fine, around 280Hz, Q=0.62, 11.4dB cut.

If I connect U2 to U5, and then U2 + U5 to the gyrator in series, I should have bass boost + mid cut, right? And I do, but the result is not what I expect, namely this (peak at around 35Hz, cut at around 500Hz, treble about 2.5dB lower than it should be):



instead of this:



(peak at around 50-60Hz, cut at around 300Hz).

I can't understand why.

(edit: wrong picture, wrong info)

[ElectricDruid]

The schematic has a problem, which is that U2 and U5 are fighting. U2 is trying to make its output the same as its input, which comes from the wein bridge network. Meanwhile, U5 is trying to make its output the same as its input, with a bit of gain and rolloff. But they can't both be right!

At a minimum, you should connect them both via resistors. maybe try adding another 13K resistor from the output of U5 to the top of R9 to meet the other signal coming via R1 from U2?

[fryingpan]

Nope, that's not it. If I put a reasonable resistor (~13kohm) on each opamp's output nothing changes, or actually the bells are even more shifted to the left and to the right (respectively).

[ElectricDruid]

Ok. That was just the first thing that jumped out at me.

The next problem I can see is that most of those op-amps have "assumed" DC conditions - that is to say, there's no solid biasing.

U3 is tied to ground, which is ok in the sim, since it can assume a bipolar supply. That'll break if you build it with 9V power.

U5 is tied to an AC source, which varies around ground, so is sort-of ok, in the sim. the other op-amp input is separated from DC by C5, so has no influence. Again, this is going to be insufficient as soon as you build it for real.

U2 has the positive input connected to ground via R6, so it's in the same position as U3 - fine on a bipolar supply.

U4 is the same - R3 goes to ground, fine on a bipolar supply, will break as soon as you move to 9V single supply, where you'd need R3 to go to 4.5V Vref level instead.

U1 is fed whatever bias voltage comes from the previous amps. That might work, if they were behaving.

But that's not your actual problem, right? You're actually concerned because the two stages are interacting and messing with the frequency responses, is that right? Given that you're doing a passive mix into a gyrator stage, I'm not entirely surprised...

[fryingpan]

Yeah, I've used ideal opamps with no actual power supply considerations, since I'm just using them as building blocks (I'll probably build it as a mostly discrete circuit, with opamps only where they make most sense - for example, the level boosters may be JFETs or BJTs, don't know yet). Anyway the frequency response should not be skewed because I've isolated each stage with buffers, right?

[iainpunk]

Anyway the frequency response should not be skewed because I've isolated each stage with buffers, right?

no, they interact because they overlap, not because the filters themselves actually interact.

this is a cool bridge-tee based filter that boosts 84Hz and cuts 279Hz up to 15-ish dB

have you given this circuit a try in the simulator? you can change filter values to change the frequency's they cut and boost, and also the Q and depth, this one is quite basic, but i can assist you in calculating those filter values if you want.

cheers

[fryingpan]

As for your first response, also the digital EQ, a screenshot of which I've provided, is a composition of two different filters, and they do in fact interact (the bass boost is at 80Hz but the actual peak is at 55-60Hz, for example) but this isn't mirrored in my design.

Regarding the second, I will give it a try, but do you have any documentation on the matter? I'd like to understand myself how to do certain things.

[iainpunk]

sorry, i dont have much documentation, its based off of bridge T filters, they are fairly easy to understand. but the absolute basics of this filter are that its a simple notch filter that takes out a 'band' of frequency's (Europeans call it a 'band stop' filter, Americans call it a 'notch filter')

anyting you put in the feedback loop of an opamp ''flips upside down''. if the circuit in the feedback takes out frequency's, the opamp wil compensate and boost those frequency's. if you were to put in a high cut, it will turn in to a high boost, and a low cut becomes a low boost, crossover distortion becomes soft clipping, and hard clipping turns in to a non-linear expander.

with this basis, lets consider this next circuit,
i have chosen a filter that scoops around 80Hz, so in the feedback loop of an opamp, it boosts that 80 Hz region.
the other filter is around 300Hz, and it scoops that region as a normal bridge t would.



the variable resistor basically diminishes the 'depth' of the filters. less depth is less scoop and less boost

i dont know if many people tried this specific circuit, i haven't found any reference to this setup, but it does the job and creates that wiggle in the frequency response.
the first time i used it was when i was experimenting with a preamp for my DIY amp build, and i wanted a mid control that didn't change the volume to much. the filter frequencies were 1kHz boost and 300Hz scoop, but i re-calculated the boost for the 80Hz you want. i use this Twin T calculator with C3 set to 10F and R3 set to 999M so they act as not being there, so this calculator is able to calculate bridge T filters.

cheers

[marcelomd]

Had a few moments to spare and simulated Iain filters.



Looks like it works, but the insertion loss is big so I put a simple gain stage after it. This gain stage is not amplifying treble and I can't see what I'm missing right now. Anyway, the basic idea works.

[iainpunk]

Had a few moments to spare and simulated Iain filters.



Looks like it works, but the insertion loss is big so I put a simple gain stage after it. This gain stage is not amplifying treble and I can't see what I'm missing right now. Anyway, the basic idea works.

insertion loss? i din't have any insertion loss when i breadboarded it... (idk about spice but shouldn't the resistors be denoted 1M instead of 1Meg?)
also, try not using the mixer method for the sweep, but the original ground path blocking method when changing the 'depth' of the filters. it prevents the boosting opamp from amplifying that range all the time, lowering the noise floor and preventing clipping

if you want a broader or narrower filter, changing the capacitors by doubling one while halfing the other keeps the frequency the same while changing the Q

cheers

[fryingpan]

100k over 100 in the feedback path? That's... quite the gain stage  (1000x, 60dB!).

Anyway, LTSpice denotes megaohms as Meg instead of M (in order to avoid confusion with milli-, I think).

[marcelomd]

Mr Pan is right, Meg is for Mega-. I believe the underlying Spice is case insensitive (so M,m goes to milli-).

I did not save the simulation, nor did I calculate anything. I ran it just to kill a few minutes =)

I don't know if "insertion loss" is the correct expression here. Fact is the signal level was way, way, down after the bass boost (and a bit lower after the mid notch). That ridiculous gain is there to make it 1:1 with the clean side.

[iainpunk]

Fact is the signal level was way, way, down after the bass boost

that shouldn't happen normally, nor does it IRL. my falstad simulations dont show this either.

meh, its mainly to get the point across anyway, with the freq response, and thanks for that!

cheers

[m4268588]

Do not ignore DC volts.
If necessary ".ac" automatically calls ".op".

[marcelomd]

Do not ignore DC volts.
If necessary ".ac" automatically calls ".op".

*FACEPALM*

I'll simulate again in the morning...