passive EQ idea

[jhaible]

Here's something that's going around in my head for some time.

Think of a typical passive EQ with a L C series connection to create a bell courve.

Usually, you have a rotary switch to select the center frequency, the switch connecting different L and C values to the circuit.

We all know the center frequency is calculated as fc = 1 / (2 Pi SQRT(LC)).

For a certain center frequency, we have multiple choices. We can pick an L value, and then calculate the C value accordingly.

But there's also the Bandwidth, or it's inverse, the Q value. So depending on the impedance of the rest of the circuit, and the Bandwidth you want, you must choose a specific L value *and* C value. In other words, with the product L * C being constant (for a certain center frequency), a higher L (and lower C) gives you a higher Q factor, and smaller Bandwidth.

Some passive EQ circuits have a variable R in series with the L and C. This is a means to vary the Bandwidth while keeping L and C constant. Unfortunatey, this also means that you normally get less cut / boost range at higher bandwidth than at smaller bandwidth.

This is nothing new so far - I just tried to sumarize the situation without being too technical.

Now I think of this: We could have much better control over Q (and Bandwidth) in a passive EQ, if we use *separate* switches to select the L and the C values. One could start with the typical values of a commercial EQ of rather high Q factor, but instead of adding a variable resistor in series with the L and C, we could simply switch Ls and Cs with two separate rotary switches.

The problem is just, how to label this switch.

I can imagine 3 ways to do this:

(1) Engineer's method (bad) :? :
Just write down the C and L values. One switch to choose your Nanofarads, one switch to choose your Millihenries.

(2) Overkill method (overkill) :shock: :
Use a bank of relays, and some control logic, then you can choose fc with one switch, and Q with a second switch.

(3) Pseudo-Frequency method :idea: :
Label both, the C selector knob and the L selector knob in frequencies.
If you want the "nominal" Bandwidth at 1kHz center frequency, set both switches to 1kHz. Want to increase Q factor, set turn up to C "frequency", and turn down the L "frequency" by the same number of positions.
Example:
1kHz, nominal Q: C-Switch = "1kHz", L-Switch = "1kHz"
1kHz, higher Q: C-Switch = "2kHz", L-Switch = "500Hz"
1kHz, even higer Q: C-Switch = "4kHz", L-Switch = "250Hz"
1kHz, lower Q: C-Switch = "500Hz", L-Switch = "2kHz"

Looks complicated? It isn't. Just start with 1kHz on both knobs, and then move one up and the other down by the same number of steps.

Does this make sense? Has it been done before?

JH.

 

[gyraf]

Example (3) is what Klein&Hummel does in their UE100/1000 eq's, I think. They do that without inductors - by setting highpass and lowpass frequencies individually for the bell filters...

Also, I think that this way of altering Q will have exactly the same influence on cut/boost maximum levels, as it will be the L/C pair's Xlc (collective reactance) that is altered, and it will be acting to do what is done by the Q-pot..?

Jakob E.

 

[jhaible]

[quote author="gyraf"]Example (3) is what Klein&Hummel does in their UE100/1000 eq's, I think. They do that without inductors - by setting highpass and lowpass frequencies individually for the bell filters...
[/quote]

I don't know these EQs. But just to clarify, I wasn't speaking about separate HP and LP filters to make a BPF function - it was just about labelling the selected Ls and Cs in a certain way. It's still a single 2pole BPF, with complex poles. If you go for a HPF+LPF configuration, I'd expect it to _either_ have only real poles (i.e. limited Q), or more (at least 4) poles (i.e. resonant HPF and LPF).

Also, I think that this way of altering Q will have exactly the same influence on cut/boost maximum levels, as it will be the L/C pair's Xlc (collective reactance) that is altered, and it will be acting to do what is done by the Q-pot..?
Jakob E.

The maximum cut/boost capability is determined by the series resistance in the resonant case.

Just think of a simplified circuit. Resistor divider R1 (series) and R2 (shunt).
And a series LC that can be dialed it to go parallel to R1 for maximum boost. At the resonant frequency, the Z of the LC becomes real. For an ideal L and C, Z would be zero. That means maximum boost is unity gain (while flat response means R2/(R1+R2) ). With a real life L, you have some losses, i.e. a small R in the resonant case. So the maximum boost means not-quite, but almost unity gain. Regardless of the LC combination that is chosen. So you can choose bandwidth by altering L and C inversely (keeping L*C constant), without decreasing the maximum boost range. (Except for the little detail that a larger L might also have a slightly larger R).
OTOH, if you add an external resistor Rq (pot) in series with L and C, in the resonant case this external Rq is parallel with R1. For low Q, Rq is rather large, so you have less maximum boost.

JH.

 

[Mendelt]

I had a similar idea for a passive eq but instead of making the user choose a separate inductor and capacitor i want to have two bandwidths per eq section. So that's two sets of inductors and capacitors for a wide and narrow bandwidth.

I tried this in the simulator and the results look really different from just adding a resistor to alter the Q. When you add a resistor you just push down the peak of the filter without really affecting the width of the rest of the bell curve. When you use different inductors and capacitors the peak stays the same (inductance is still 0 at the resonant freq. ) but the bell gets wider. To me this seems more usefull allthough i havn't tried it yet.

The problem with my implementation is that i'll need lots more inductors and capacitors. Your idea seems a bit more cost-effective :grin:

 

[gyraf]

Hmm.. I think I'll need to breadboard this to see if I understand it correctly.. :?

 

[jhaible]

[quote author="gyraf"]Hmm.. I think I'll need to breadboard this to see if I understand it correctly.. :?[/quote]

You can quickly try it with the tonestack calculator
http://www.duncanamps.com/tsc/

The "Bench" tab of this program has a simple passive EQ with LC midrange.

JH.

 

[chrissugar]

Hey Juergen

I understand your idea.
It is something like in photography. For a specific film sensivity and light you have many pairs of diafragm and time values that will let the same quantity of light to reach the film. In this case the equivalent for the Q (bandwidth) in the filter example is the depth of sharpness. Different pairs will produce different depth of sharpness.
Very good idea. I like it. :thumb:

chrissugar

 

[peterc]

Here's a pic of the UE 100.

Peter

 

[jhaible]

[quote author="chrissugar"]Hey Juergen

I understand your idea.
It is something like in photography. For a specific film sensivity and light you have many pairs of diafragm and time values that will let the same quantity of light to reach the film. In this case the equivalent for the Q (bandwidth) in the filter example is the depth of sharpness. Different pairs will produce different depth of sharpness.
Very good idea. I like it. :thumb:

chrissugar[/quote]

Thanks - and I like your photograph analogy!
The dual switch method is good old manual camera, while the relays control version would be a new-fangled automatic ...


Thinking about this a little more, I had another idea:

One could even set up a "mismatch" deliberately. In my example above, I always had the two offsets - L selector "frequency" and C selector "frequency"- (logarithmically) cancel each other: Set one to twice the value, the other to half the value, get same center frequency and different Q. But you can also use the same configuration to select in-between frequencies, like SQRT (500Hz * 1kHz) = 707 Hz by setting one to 1k and the other to 500. At the price, now, of slightly varying Q factor. (Some of the commercial EQs even share one L value for two adjacent frequencies, also having a different Q for both bands.)

So with one set of Ls and Cs, with just the additional expense of a second rotary switch (and getting used to the concept), we could either have true control over Q independent of Cut/Boost, or the possibility to dial intermediate frequencies with some compromise about Q, or a mix of both.

Of course you don't have to stick to the 0.5 factor between Ls and Cs. In fact I suggest something smaller, like 2/3 or SQRT(2). You just have to keep it equal between all switch positions.

JH.

 

[Scodiddly]

The creative audio types seem to thrive on equipment where you can abuse the concept a bit, rather than having strict engineer options.

Here's a goofy idea: Have a binary switching system. Have capacitors ranging in value 1 - 2 - 4 and also inductors the same way. Each one has a simple on/off switch. You want C=5, turn on 1 and 4.

 

[chrissugar]

Scodiddly

I like your idea, with condensers that have the power of two you can cover a wide range of values. With three relays you can have 8 values.
If you do the same thing with the inductances you can cover 8x8=64 variations of freq and Q setings with only 6 relays.
I think it is very simple to implement something like this.

chrissugar

 

[Learner]

hey, sounds like the circuit relies on selectable L&C individually keeping R constant to difference the reactance in order to influence the Q and select the resonance freq, however I think the challenge in this circuit is to be able to find the right value that actually exist on the market to provide the freq you are looking for. Especially if the freq are to match of those in musical notation, it might be worth while to look into winding your own inductors if you are into the pa$$ive EQ thingy..... you might be able to get an even finer control over Q if the R is variable too, which allows you to expand the tweakability on each L&C setting. However, I donno how audible that extra tweakibility would be......I think IC filters are much more friendlier to the bank acc not to mention less calculation and weight, passive EQs are a real luxury to have.

In regards to labling the EQ. I don't think it would be a problem how you lable it if you know the principle of what the circuit does since you know the relationship of L and C to determine Fr and influence the Q but for someone who does not know the principle for the circuit, may be they'll have to rely on their ears to decide what sounds "good" to them and just tweak'em knobs? :shock: :green: