Parametric EQ with State Variable Filter, why which topology?

Hi All!

As I'm growing more intrest to build a parametric EQ and compared different designs, I couldn't find the answers to some questions. Mainly, why is which topology used? And how does a switchable bell/shelf work?

Most (all?) designs I've found do something like this:
https://www.dropbox.com/s/ssmbwdla6wfz8hk/Schermafdruk%202018-08-09%2020.45.12.png?dl=0

Except the populair SSL 4000 - 242EQ does a different trick:
https://www.dropbox.com/s/44luiw2qh6rot1k/Schermafdruk%202018-08-09%2020.58.22.png?dl=0

Could anyone explain what the difference between those two topologies is?

Next to that;
Most filters use the same cutoff frequencies for the high-pass and the low-pass, IE: the capacitor values are the same. Though, when I searched for a switchable bell/shelf EQ I mostly find SVF's with different cutoff frequencies. For example this one:
https://groupdiy.com/index.php?topic=62157.msg812575#msg812575
Could anyone explain why the capacitors have a different value and why the schematic would result in a switchable bell/shelf EQ?

Thanks

PS: How can I insert an image?

..probably just so they can make do with an inverted version of the filtered signal, thus saving an opamp for polarity inversion?

Jakob E.

NickApparently said:

Hi All!

As I'm growing more intrest to build a parametric EQ and compared different designs, I couldn't find the answers to some questions. Mainly, why is which topology used? And how does a switchable bell/shelf work?

Most (all?) designs I've found do something like this:
https://www.dropbox.com/s/ssmbwdla6wfz8hk/Schermafdruk%202018-08-09%2020.45.12.png?dl=0

Except the populair SSL 4000 - 242EQ does a different trick:
https://www.dropbox.com/s/44luiw2qh6rot1k/Schermafdruk%202018-08-09%2020.58.22.png?dl=0

Could anyone explain what the difference between those two topologies is?

Click to expand...

there are far more than just those two, but the major benefit of #2 is that you can realize more boost/cut with nower noise gain (less noise is good in EQs).

In #1 you have a BP version that you then must boost or cut X times, in #2 by subtracting the BP signal from the negative feedback you can in theory get infinite boost, while the amp is nominally low gain wide band. or infinite cut subtracting the BP signal from input.

Next to that;
Most filters use the same cutoff frequencies for the high-pass and the low-pass, IE: the capacitor values are the same. Though, when I searched for a switchable bell/shelf EQ I mostly find SVF's with different cutoff frequencies. For example this one:
https://groupdiy.com/index.php?topic=62157.msg812575#msg812575

Could anyone explain why the capacitors have a different value and why the schematic would result in a switchable bell/shelf EQ?

Click to expand...

using different values for SVF poles shifts the nomical Q/bandwidth.

for SVF shelf you need to grab a HPF or LPF from the SVF and either add that to the BP, or use it instead for the shelf. This is not simple, I did it inside a console strip back in the 1980s. I don't recall all the details other than that trickery was involved.  8)

JR

Thanks

PS: How can I insert an image?

Click to expand...

Thanks for your reply!

JohnRoberts said:

there are far more than just those two, but the major benefit of #2 is that you can realize more boost/cut with nower noise gain (less noise is good in EQs).

In #1 you have a BP version that you then must boost or cut X times, in #2 by subtracting the BP signal from the negative feedback you can in theory get infinite boost, while the amp is nominally low gain wide band. or infinite cut subtracting the BP signal from input.

Click to expand...

Oh cool! Gotta play around with that one. Do you have any suggestions the the circuitry for the BP in #2? Will a normal SVF work? (maybe switch polarity) Or would another circuit fit better?

JohnRoberts said:

using different values for SVF poles shifts the nomical Q/bandwidth.

for SVF shelf you need to grab a HPF or LPF from the SVF and either add that to the BP, or use it instead for the shelf. This is not simple, I did it inside a console strip back in the 1980s. I don't recall all the details other than that trickery was involved.  8)

JR

Click to expand...

I get that different values for cut-off frequencies shift the bandwidth, but why do I mostly see the same cut-off frequencies for bell functions but different cut-off frequencies when the design features a bell/shelf switch?

I got a shelf working by adding the HPF to the BP. Very cool that all the functions of the SVF stay intact. You can get really cool shelf-shapes when messing with the Q pot. Here's the schematic. Sadly it's quite heavy on the op-amps.

https://www.dropbox.com/s/dnueqfaz24wl6jn/Schermafdruk%202018-08-11%2010.14.30.png?dl=0

NickApparently said:

PS: How can I insert an image?

Click to expand...

At the bottom of the message mask, there is a tab "Attachments and other options".

NickApparently said:

why do I mostly see the same cut-off frequencies for bell functions but different cut-off frequencies when the design features a bell/shelf switch?

Click to expand...

That's because the definition of boost/cut is different for a bell-shaped EQ than a shelf EQ. For bell, it's exactly the characteristic frequency of the BP filter. For a shelf, boost/cut is not measured at +/-3dB; it is measured at +/- 15-20dB, which results in a frequency that's distant by about 2-3 octaves above or below the characteristic frequency.

NickApparently said:

Thanks for your reply!

Oh cool! Gotta play around with that one. Do you have any suggestions the the circuitry for the BP in #2? Will a normal SVF work? (maybe switch polarity) Or would another circuit fit better?

Click to expand...

I have designed enough parametric EQ over the years that I don't remember them all..  first was a parametric kit in Popular Electronics in 1979.

SVF is the best/only way I've ever made fully parametric EQ, while simpler sweep EQ can be made using less parts.

The subtractive approach does deliver superior S/N performance. Another consideration when messing with SVF is to take care to not saturate any of the multiple passbands of the SVF even if not using them for signal, because clipping there will reflect into your audio path...  I have seen a lot of different variations on the basic SVF each with different pros and cons.

I get that different values for cut-off frequencies shift the bandwidth, but why do I mostly see the same cut-off frequencies for bell functions but different cut-off frequencies when the design features a bell/shelf switch?

I got a shelf working by adding the HPF to the BP. Very cool that all the functions of the SVF stay intact. You can get really cool shelf-shapes when messing with the Q pot. Here's the schematic. Sadly it's quite heavy on the op-amps.

https://www.dropbox.com/s/dnueqfaz24wl6jn/Schermafdruk%202018-08-11%2010.14.30.png?dl=0

Click to expand...

nah.. you need to add an extra polarity inversion to get BP and HP/LP back in sync. Yes, the Q gives a lot of variation in shelf mode, but I never figured a good way to describe/label/apply that... I only did one console strip with parametric shelf and that was back in the 70s/80s... I never revisited it.  It worked but.... IIRC I may have limited the effect of Q in shelf mode because it changed the EQ too much, but my recollection is not crystal clear these decades later.

One of my primary rules about circuit design is make it hard for the customer to get bad sounds out of any path (effects can break those rules, but even then some moderation is advised.)

JR

One of my primary rules about circuit design is make it hard for the customer to get bad sounds out of any path

Click to expand...

.

+1. JR


If you haven’t already, You might want to also look up an AES paper from George Massenburg. 

His ITI equalizers and later Sontec EQ,s. Ruled the studios back in the 70’s

abbey road d enfer said:

That's because the definition of boost/cut is different for a bell-shaped EQ than a shelf EQ. For bell, it's exactly the characteristic frequency of the BP filter. For a shelf, boost/cut is not measured at +/-3dB; it is measured at +/- 15-20dB, which results in a frequency that's distant by about 2-3 octaves above or below the characteristic frequency.

Click to expand...

Thanks! That inspired me to try something else. "Why don't I switch one of the cut-off frequencies out of the audible spectrum for a shelf function?" And so I tried. I built something like the attached file. This way, with the switch in the High-Shelf position, the cut off frequency moves to about 71kHz, which results in a great High-Shelf function. The Q adjust does work, but isn't as usefull as with the summing of BP+HP (In general it might be a bit weird, but I like the weird shapes). It just seems to move the shelving frequency instead of changing its shape. In this case I would also eliminate the Q pot.

JohnRoberts said:

One of my primary rules about circuit design is make it hard for the customer to get bad sounds out of any path (effects can break those rules, but even then some moderation is advised.)

Click to expand...

Nice! Great way of looking at audio electronic design

Next I'll try the SSL style (subtractive as I've heard?). I like the better S/N ratio and the possibility to infinite boost/cut Ofcourse, whilst trying to avoid clipping.

NickApparently said:

Next I'll try the SSL style (subtractive as I've heard?). I like the better S/N ratio and the possibility to infinite boost/cut Ofcourse, whilst trying to avoid clipping.

Click to expand...

In practice even if infinite cut or boost was possible (boost will be limited by open loop gain, cut by component accuracy), good design generally limits this to say +/-20dB. 

One desirable  attribute of fully adjustable EQ is symmetrical boost/cut curves, so one could in theory undo existing EQ... while that is rarely done (if ever).

JR 

JohnRoberts said:

In practice even if infinite cut or boost was possible (boost will be limited by open loop gain, cut by component accuracy), good design generally limits this to say +/-20dB. 

One desirable  attribute of fully adjustable EQ is symmetrical boost/cut curves, so one could in theory undo existing EQ... while that is rarely done (if ever).

JR

Click to expand...

I've noticed the component accuracy, or better yet, the lack of. Still, the general limit of +/-20dB seems logical. Personally I'm aiming for about 15dB. Most designs I've seen were restricted to 12dB.

Another question: When eliminating the Q pot, will a SVF still be most desirable as a Band Pass Filter? Or would a more simple RC+CR filter (where the resistors will be pots) be good enough/cheaper? Like this one:
https://www.electronics-tutorials.ws/filter/filter_7.html

NickApparently said:

Another question: When eliminating the Q pot, will a SVF still be most desirable as a Band Pass Filter? Or would a more simple RC+CR filter (where the resistors will be pots) be good enough/cheaper?

Click to expand...

This type of filter is too broad for parametric EQ, which most appreciated feature is being capable of quite narrow action (often called "surgical).

abbey road d enfer said:

This type of filter is too broad for parametric EQ, which most appreciated feature is being capable of quite narrow action (often called "surgical).

Click to expand...

Okay, but when the Q-pot is eliminated, wouldn't the average EQ-band have a Q of 0,7071? And the type of filter I suggest gives a nice Q of 0,7071, right??

I believe I already shared, for sweep EQ using fixed Q there are simpler (read cheaper) topologies.

Perhaps search sweep EQ, there should be numerous schematics to study.

JR 

NickApparently said:

Okay, but when the Q-pot is eliminated, wouldn't the average EQ-band have a Q of 0,7071?

Click to expand...

If by "average", you mean an equal resistors SVF, you're right, but nobody in his right mind would do that. Equal resistors is a particular case of SVF, that is not really useful; that's why some resistors are tweaked or added.

[/quote] And the type of filter I suggest gives a nice Q of 0,7071, right??
[/quote] Exactly, and thus almost useless (as a midrange swept EQ).

JohnRoberts said:

I believe I already shared, for sweep EQ using fixed Q there are simpler (read cheaper) topologies.

Perhaps search sweep EQ, there should be numerous schematics to study.

JR

Click to expand...

Thanks! I'll look into that

abbey road d enfer said:

If by "average", you mean an equal resistors SVF, you're right, but nobody in his right mind would do that. Equal resistors is a particular case of SVF, that is not really useful; that's why some resistors are tweaked or added.

Click to expand...

By 'average' I ment 'the most common eq's', in like small table mixers with Low-Mid-High eq's.
I might be in way over my head, and I'm just trying to understand, but why would nobody in his right mind build an equal resistor SVF? Obviously variable resistors make the parametric EQ usable, but the resistors for the integrator op-amp should be as equal as possible right? Or am I missing something?

abbey road d enfer said:

And the type of filter I suggest gives a nice Q of 0,7071, right??
Exactly, and thus almost useless (as a midrange swept EQ).

Click to expand...

How come a Q of 0,7071 is almost useless as midrange swept EQ? Is it to broad for Parametric EQ, as you earlier mentioned?

NickApparently said:

Thanks! I'll look into that
By 'average' I ment 'the most common eq's', in like small table mixers with Low-Mid-High eq's.
I might be in way over my head, and I'm just trying to understand, but why would nobody in his right mind build an equal resistor SVF? Obviously variable resistors make the parametric EQ usable, but the resistors for the integrator op-amp should be as equal as possible right? Or am I missing something?

Click to expand...

Indeed the RC products of the integrators should be equal (it's not necessary though, but makes manufacturing easier). When I say "equal value", that relates to the other resistors, the ones that defind the gain and feedback coefficients. When all values are equal, the "Q" of the filter is too small (I don't want to use the term Q in relation with parametric EQ, but I'm referring to the BP filter's ), which results in too large a bandwidth.

How come a Q of 0,7071 is almost useless as midrange swept EQ? Is it to broad for Parametric EQ, as you earlier mentioned?

Click to expand...

Yes.
[/quote]

Surgical eq for notching out interference, reducing explosive "s", etc. => small filter bandwith (high Q) is needed
General eq for modifiying tonal balance => broad filter bandwith (small Q) needed

Parametric can do both, but is usually too advanced/expensive to include in cheap consumer products.
For general eq my go to circuit (for simplicity and good sound) is the eq-section of the Soundcraft 200B
Hi/Low/2 swept mid bands...
https://groupdiy.com/index.php?topic=44786.msg561727#msg561727

L´Andratté said:

Surgical eq for notching out interference, reducing explosive "s", etc. => small filter bandwith (high Q) is needed
General eq for modifiying tonal balance => broad filter bandwith (small Q) needed

Parametric can do both, but is usually too advanced/expensive to include in cheap consumer products.
For general eq my go to circuit (for simplicity and good sound) is the eq-section of the Soundcraft 200B
Hi/Low/2 swept mid bands...
https://groupdiy.com/index.php?topic=44786.msg561727#msg561727

Click to expand...

The 200B EQ can hardly qualify as surgical.Its filter has a Q of about 0.3, but the EQ itself has a relative bandwidth of 1.5 octave, which suggests a Q of 0.6. *
BW can be narrowed, up to a point, by adding a resistor from the inverting input to ground. That was used on one of the later live mixers, and considered too wide, even in its narrow setting, for being of real practical use, as a surgical EQ.

* As I wrote earlier, I don't recommend using "Q" as a parameter of EQ's. I know I'm rowing against the current because that's the current usage, but I stand that "Q" is valid only for response that falls asymptotically to -infinity, which is not the case of an audio EQ, where the response tends to unity.
I'd rather express this as "(relative) BW", i.e. BW expressed in octaves, set at max boost/cut, measured at half max efficiency. That makes more sense operationally and perceptionally. There have been tentatives by the AES to standardize this, but never found completion.

abbey road d enfer said:

The 200B EQ can hardly qualify as surgical.

Click to expand...

As surgical as a butcher´s knife...
Sorry, I was unclear, it was my example for a "swept" as opposed to "fully parametric"
eq, that works very well for me for tonal balance adjustments, regarding center frequencies, sweep and cut/boost ranges and build simplicity!