Parametric EQ questions

[dave h.]

i know i talked a bit about Par. EQ's already but its been a while.

after reading GEO's article on "easy parametric EQs" and such, i wondered how i could get the ultimately adjustable EQ. obviously, the following things are required for each band:

adjustable center frequency
boost/cut
bandwidth/Q

and this is pretty common knowledge. but heres some food for thought: imagine your normal graphic EQ box. ive got a big old DOD from the late 70's. turn it sideways so that the sliders are horizontal. imagine that each slider didnt boost/cut, but panned between left and right channels of a stereo rig. i read a few articles about some old "stereo imagery" gear that worked sort of like that. since the freqencies are divided between two speakers (depending on the setting, it could be 30left/70right, 40L/60R, etc.), we get two differently "shaped" signals from each side but our ears "piece them together" to form the whole. interesting concept. so say we have a 4 band parametric EQ. each band gets a pot for:

Frequency
boost/cut
Q
left/right

but back to the frequency response curve.
for the midrange frequencies, it's usually logical to use a notch/peak filter like the ones weve described. however, on the top and bottom ends, it's sometimes advantageous to have a shelf type filter. this way, you can boost or cut all the frequencies above the frequency you select for treble and boost or cut all the frequencies below the center freq. you select for the bass.

so on the outer 2 bands, we add a switch to select either a peak/notch or a shelf filter.

Lets skip back to the GEO article. go ahead and picture the frequency response chart of a peak/notch filter in your mind, or go look at the article if necessary. http://www.geofex.com/Article_Folders/EQs/paramet.htm

the thing i would REALLY like to adjust to make things interesting is the symmetry of the filter. so far, all of the stuff ive listed is something that if i didnt know how to do it, i could look it up and find out how relatively quickly. this i don't know about.

take the idea of asymmetical clipping. you have two different diodes for pos. and neg. so the sound wave gets clipped at different values for the peaks and troughs. this is a similar idea, only more vertical than horizontal. imagine now the image of a peak/notch filter. you adjust the Q to make it look "sharper" or "duller." well, what if you want a steep transition between the low mids and high mids? a single notch or peak doesnt help much.  we want something that boost a wide bandwidth on the bass side and a narrow bandwidth on the treble side. like a HI-Q and LO-Q instead of just an overall Q. instead of /\ you could get /| or |\ to use some ASCII imagery. any ideas on how/whether this could be done?

so now were left with the following layout:

bass and treble bands:
freq.
shelf/peak|notch
Q (slope in shelf mode)
boost/cut
left/right

midrange bands:
freq.
HI-Q
LO-Q
boost/cut
left/right


so can the Q be split into two adjustments? your thoughts?

[onboard]

A shelving notch?

[Mark Hammer]

Not to rain on your parade or anything, but people get confused between what is NEEDED for equalization, and what is MARKETED for equalization.

If you know your precise needs, often you can accomplish what you want with a very simple system.  For instance, the presence of variable Q in parametric EQs is largely because the company manufacturing it has absolutely no idea if your intent is to zero in on specific "wolftones" or resonances in a concert hall and tame them, or provide broadband adjustments. Same thing goes for 31 band EQ.  The manufacturer has absolutely no idea what bands are of relevance to you, so they give you control over lots of them.

Is it "wrong" to want more controls and finer ones?  Not at all.  But it's a bloody nuisance, though, to try and replicate something that took you 387 knob twists to nail.

When I learned how to operate a polygraph for recording EEG, cardiac or muscular activity in the mid-70's, one of the things I found neat about the preamps on polygraphs was the manner in which specific bands could be isolated by manipulating a cascaded highpass and lowpass filter.  Rather than using a selective-gain-type Q control, the passband could be narrowed or widened simply by setting the corner frequencies of the two filters closer or farther apart.  The same thing can be done for musical signals, and in fact the pitch-to-voltage convertor on the old Korg MS-20 mono synth used just such an arrangement for zeroing in on the note range that the P-2-V was supposed to tackle.  The nice thing is that nobizarre gain arrangements are needed.  Manyhigh-end bandpass filters for lab measurements work this way too.

Though I've not heard of anyone using them for musical instrument purposes, you may still be able to get some of the many pseudo-stereo synthesizer chips made by Philips and others during the 70's and 80's for taking mono recordings and making them "sound like" stereo, using what were essentially custom-tailoring non-sweeping phase shifters.  A lot of the recordings made with them ("mono electronically reprocessed for stereo listening") sounded worse than death, but they may yet be interesting for other single-instrument applications.

[Chico]

I am onboard with Mark on this one.  When designing EQ, I personally find it best to keep in mind the context of the application, and the value that each additional feature/option provides.   In view of that, I tend to keep my eq pretty simple, with notable exceptions.

That said, I think that there is nothing wrong with getting creative with building an EQ.  In that regard, you may also want to consider:

gyrator based EQs per RG's article are not constant Q.  As you can see in his charts, the designed for Q is not realized until the gain is maxed (boost or cut).  So, does this fit in with your design objective?

If not, you may want to consider constant Q as an alternative, or asymmetrical Q(different Q for boost and cut).  See the tech bulletins at rane.com to get a good idea on how to build each of these.  Also, there is some info on the web concerning design of parametrics using state variable filters.

While you are at it, also consider that interesting results can be achieved with a few simple EQs in parallel.  Craig Anderton has a design for a parallel EQ that provides a remarkable range of adjustments with a simple set of unassuming controls.

Also, lots of fun is to be had with different EQ topologies.  For example, as noted in RGs article, notch filters can be bootstrapped for depth control.  A cool EQ that I have found REALLY useful is where I bootstrap a twin T notch - seeRGs article, but also provide controls to attenuate/boost the  treble and adjust the notch frequency - STM  has a post on this subject a while back, but I also add fixed frequency compensation to the bass and treble sides of the notch in a buffer amplifier post twin T.  Completely different sound from a bandpass based filter.  Better?  Worse? depends upon the context.

I also have found another EQ topology extremely useful (I am really suprised that I have not seen it more).  It is a modified version of the Tilt control for hi fi stereo systems.  Checkout Headwise.com (and other various circuits) for details here.  The circuit for hi fi tilt is easily modified to suit guitar.  A single knob tone control that does not muck with your signature sound.  Yea!   Basically, you design a pivot frequecy, and the single control varies the frequency response (somewhat) linearly in a "tilt" fashion about the pivot frequecy (within a reasonable range of frequencies).   This makes a great tool where you do not want to alter your overall tone but need to make "room" adjustments etc.   A bass player friend absolutely loved this control in a version that I modified for bass.

Also, there are several variations of the passive tone stack - see Duncan Amps for a great tool in this regard.

Also, do not forget ladder filters, e.g., resonant low pass filters etc. - The synth guys have tons of great resources here.

I think that others out there can add numerous other EQ topologies.

My point (If I ever have a point to anything) is that the web is littered with cool cool cool eq designs with enough alternatives to keep you scratching your head for a long time.  I do not think that an analog EQ designed to date is "ultimate", but each EQ has its own advantages and shortcomings

In other words, don't forget to consider that there is more than one way to skin a cat.

Best of luck in your design.

Tom

[Mike Burgundy]

I agree with Marks point, but have to point out that the balance thing isn't too hard I think.
Have a look at a regular parametric:
http://burgundy.users.balpol.tudelft.nl/bin/schem/parametricbass1a.gif
now, duplicate the mixer opamp circuit (second opamp) and connect it's inputs parallel to the first opamp's. Now you have two outputs. Either you can use the output pots to set a "balance" of sorts, but that doesn't work per band, and doesn't change the amount of eq effect, just overall loudnes. So we install a dual-ganged balance pot (say 100k) across the inputs. Center lug to mass, outer lugs to both negative inputs, and both positive inputs respectively.
I think that should work.

[dave h.]

i think the first miscommunication was that this wouldnt be so much of a pedal project, as it would be a studio type tabletop thing. ive actually seen a lot of the mentioned links (but not the "tilt" design, which is a very interesting idea....)

i guess the big question (for me) remains, is it possible to have a bandwidth control for each (treble and bass) side of a notch/peak filter?

[Rob Strand]

is it possible to have a bandwidth control for each (treble and bass)

Not really, well not at least for shelving bass/treble equalizers.   The idea of bandwidth doesn't make sense for "fixed order slope" equalization.  The only option is high slope shelving equalizers (as done my Mr Bohn at Rane).

You can make the bass and treble peaking types then vary the bandwidth of those.

[Mark Hammer]

If you live in Norman, probably the smartest thing to do is to contact PAiA in Edmond (is that near you?) and score a PCB for one of their 4-band quasi-parametrics that Jules Rykebusch designed for them.  Four tunable bands of boost and cut.  Simple flexible design.

[dave h.]

If you live in Norman, probably the smartest thing to do is to contact PAiA in Edmond (is that near you?) and score a PCB for one of their 4-band quasi-parametrics that Jules Rykebusch designed for them.  Four tunable bands of boost and cut.  Simple flexible design.

holy crap... they live in edmond?

my parents live in edmond. im there like once or twice a week. crrraaazy.... thanks for bringing that to my attention. ill have to check that out...

[dave h.]

is it possible to have a bandwidth control for each (treble and bass)

Not really, well not at least for shelving bass/treble equalizers.   The idea of bandwidth doesn't make sense for "fixed order slope" equalization.  The only option is high slope shelving equalizers (as done my Mr Bohn at Rane).

You can make the bass and treble peaking types then vary the bandwidth of those.

i dont mean for the shelves at each end of the spectrum, but for the "right" and "left" sides of just one filter. so if you had the frequency set at 350hz, the level set at +10db, it would make it possible for the slope coming from below 350hz was less steep than the slope going away from 350hz. extreme example would be a sawtooth, where only one half of the notch/peak is applied and the other is left at zero, or below.

i suppose that would be an interesting option as well, setting a center frequency, say 350hz as above, and then a lo-side-level (left of the center freq.) and a hi-side-level (right of center) and a "slope" or adjustable Q for each side and a "soften" knob that adjusts the separation of the lo- and hi-side levels from the center frequency.

this would be a lot easier to explain by drawing it out.

and perhaps it isn't marketable, especially not as a "stompbox" but it could provide interesting and ultracontrollable EQ shaping.

but work with me here. all of this stems from using fruityloops a few times and seeing how adjusting the parameters in the EQ would change the actual curvature. seeing as im all for ultra-adjustability (im building my first distortion pedal which utilizes 7 knobs and 13 switches) i think it would be an interesting, albeit complex, project to make a tabletop EQ that would be as flexible as possible. so perhaps, as people have pointed out, your average guitarist is not going to have any use for this stuff. but, in theory, it would be the next step to acheiving total control over the shape of an EQ's curvature.

and lets translate this outside of the "boring" application of Equalization. imagine if we could detect the slope of a sine wave. with an adjustable threshold, you could make clipping start when the slope reaches a number really close to zero, ie. at the peaks and valleys of the wave, and then adjust it further and further from zero for harsher clipping. you could clip it assymetrically both horizontally and vertically. what would that sound like? who knows? so i guess, more than creating a mass-marketable product, im searching for new ways to think about the way we control and modify sound.

i dont know much about electronics (yet) so most of the ideas i have for projects are based less upon components, circuits, etc. and more on theory of what you could do to a signal that doesn't fit under the same vanilla categories that are already being done by 30 different companies. id like to make pedals as a living some day, or possibly amps, but id rather avoid making "tube bombers" and clones of things, or boring 3-knob choruses and such.

[Rob Strand]

i dont mean for the shelves at each end of the spectrum, but for the "right" and "left" sides of just one filter

You can do it.  It's a little tricky if you want to do it with an analog circuit and also vary the frequency (costly multiple gang pots or a number of OTAs would be required).

Here's one way to do it with a graphic EQ circuit:

- Band n is the centre and it has it's own boost/cut control.
- Add bands n-1 and n+1.  Use a dual gang pot for the boost/cut control for these bands but wire the direction in reverse for one of the bands.

Basically as you rotate the new control it makes the profile assymetrical.  A better way would be for the new control to have no effect when the centre control is at the centre position but I can't think of an easy way to do it off the top of my  head (well not without some other non-ideal behaviour).

If you think about a whole equalizer made this way you might as well just build a graphic EQ!!!!  The number of pots isn't that much more.

[dave h.]

i dont mean for the shelves at each end of the spectrum, but for the "right" and "left" sides of just one filter

You can do it.  It's a little tricky if you want to do it with an analog circuit and also vary the frequency (costly multiple gang pots or a number of OTAs would be required).

Here's one way to do it with a graphic EQ circuit:

- Band n is the centre and it has it's own boost/cut control.
- Add bands n-1 and n+1.  Use a dual gang pot for the boost/cut control for these bands but wire the direction in reverse for one of the bands.

Basically as you rotate the new control it makes the profile assymetrical.  A better way would be for the new control to have no effect when the centre control is at the centre position but I can't think of an easy way to do it off the top of my  head (well not without some other non-ideal behaviour).

If you think about a whole equalizer made this way you might as well just build a graphic EQ!!!!  The number of pots isn't that much more.

graphic EQ's can be good for some things, but in my opinion, whether i can work out this assymetrical profile filter or not, a parametric would be better. i would choose a 3/4 band parametric over a 15 band graphic just because the par. is so much more tweekable. the ability to pin one exact frequency and adjust the amount of emphasis (Q+level) rather than just a "ball-park" freq. and make it louder is a very strong tool when used after any sort of heavily distorted signal. also, i believe it was the amptone.com site (? maybe im wrong) that had a large thesis about pre-preamp EQ (as in an EQ pedal right before the input on your amp) and how it affects the texture of the distortion. if you use a graphic EQ or even just turn your tone knob way down to emphasize the bass, you still get treble from the distorted signal, but its a whole different kind of distortion.

but people like the guy running amptone and I are rather EQ obsessed, in that both he and i think the first effect any guitarist should get concerning their tone is an EQ or three.

one of my next projects is going to be making a parametric EQ with trimpots and sticking it under a pickguard.

but back to the original question:
how would one approach getting an assymetrical filter? it seems like you would somehow have to divide the path to the "Q" knob into two, one treble and one bass, but that seems like a tough task to tackle when the center frequency is movable. so how could one split this signal accordingly? ganged pots do seem to help some. is there some digital device that will reroute a signal one of two ways if it is above a certain frequency?

i think this is an interesting theory question, whether or not it is applicable to anything useful.

[puretube]

pop-up...

[Rob Strand]

how would one approach getting an assymetrical filter? it seems like you would somehow have to divide the path to the "Q" knob into two, one treble and one bass, but that seems like a tough task to tackle when the center frequency is movable. so how could one split this signal accordingly? ganged pots do seem to help some. is there some digital device that will reroute a signal one of two ways if it is above a certain frequency?

i think this is an interesting theory question, whether or not it is applicable to anything useful.

Well I sort of answered it but I'll be more explicit.  You can't get that type of assymmetry with a "standard" filter.  You have to form/create such an EQ profile out of more than one filter, and the resulting filter will be of a higher order equalizer.   In my graphic EQ example the bands on the side of the main one are the "extra" filters required to provide the assymetry.

In principle you can do it with a parametric but there are practical problems.  The main one being that if you have to adjust the frequency you have to adjust the frequency of *all* the filters in the circuit.  That means the frequency pot requires many gangs (after all a single filter for a  parametric requires two gangs)  and you cannot buy those readily.  To get around that problem you would have to use an OTA based filter, switched capacitor filter, or PWM switched conductance filter.  These filters let you control the cut-off of a number of filter with just a DC voltage.  The circuit becomes very complex and the costly, and each technique isn't without it's side effects.

So in the world of analogue there are reasons why most equalizers follow simple  regular structures.

Of course if you go digital then life is a little easier if you want to deviate from the norm....

[seanm]

I also have found another EQ topology extremely useful (I am really suprised that I have not seen it more).  It is a modified version of the Tilt control for hi fi stereo systems.  Checkout Headwise.com (and other various circuits) for details here.  The circuit for hi fi tilt is easily modified to suit guitar.  A single knob tone control that does not muck with your signature sound.  Yea!   Basically, you design a pivot frequecy, and the single control varies the frequency response (somewhat) linearly in a "tilt" fashion about the pivot frequecy (within a reasonable range of frequencies).   This makes a great tool where you do not want to alter your overall tone but need to make "room" adjustments etc.   A bass player friend absolutely loved this control in a version that I modified for bass.

How did you modify the tilt for bass? Just move the center frequency?