Shelf/bell topologies in SVF parametric EQs

[atavacron]

I’m researching State Variable parametrics and looking at how different designs approach shelving. Post examples!

Main questions:

1. Other than the PQ1549, what designs use the SVF to produce a shelving response from the cut/boost amp?

2. Has anyone directly compared the sound of a Sallen & Key shelf with an SVF-derived shelf? Q of the SVF would have to be .5 for an apples-to-apples evaluation.

Thoughts on two very different approaches:

Net EQ -

Shelves are single pole SK, which entirely replace the SVF when engaged.

Anybody know why Porter chose this route? I believe noise floor was a consideration, but also: His bell Qs range from 1 to 3, and he relies on the input resistor & BP feedback resistor for extra attenuation for better headroom within the filter, so I’m guessing that setting the SVF to an even lower Q and then deriving a shelf from it on the back end would have been more complex than the simple SK substitution.

Calrec PQ1549 -

Low shelf achieved by shorting the LP integrator’s feedback cap, which appears to A) negate the feed from the BP integrator and B) silence the feed back to the HP mixer, thus turning the BP integrator into a LP integrator for only the original signal (whew). That single pole low-passed audio is delivered to the cut/boost mixers via the very same BP output. High shelf seems to be achieved by tapping the HP output, correcting phase, and sending that high-passed audio into the cut/boost mixers instead of the BP output. [EDIT: Like above, the LP integrator is negated by a direct tie from output to NI input.] Both shelving switches also lock the filter on its lowest Q setting (which I haven’t calc’d) but would love to know).

That’s not very complicated. You don’t need an extra pot section controlling a separate filter, nor an extra op amp. It’s also less complicated than summing outputs, as is commonly mentioned (i.e. LP+BP or HP+BP). I have not yet seen a schematic with SVF outputs summed, but maybe that’s because when people mention it, they’re talking about a series application (???), and we generally don’t use SVFs in series. Does anyone have an example of this handy?

 

[TwentyTrees]

I thought the Neumann W492 also uses state variable filters for all four bands, including the high and low shelves:

https://groupdiy.com/threads/neumann-w492-equalizer-diy-project-pcb-docs-added.28332/
Audioxs' W492 schematic attached (I find it clearer than the original).

Edit: seems I might be wrong on the SVF point here!

 

[JohnRoberts]

I have posted about this before (years ago). Back in the 80s I put a parametric EQ in a console input channel with switchable shelving for the bottom frequency band. As I recall I summed the LP together with the BP (requiring a polarity inversion). IIRC the Q was a little dodgy in shelving mode.

JR

 

[gyraf]

1 - Focusrite isa? Porter? Amek2500?

 

[atavacron]

Neumann W492 also uses state variable filters for all four bands, including the high and low shelves:

 

[atavacron]

I have posted about this before (years ago). Back in the 80s I put a parametric EQ in a console input channel with switchable shelving for the bottom frequency band. As I recall I summed the LP together with the BP (requiring a polarity inversion). IIRC the Q was a little dodgy in shelving mode.

JR

I’ve read all your references to the Loft EQ…pity that no one’s dug it up! I have a scratchpad drawing of the BP and LP outputs feeding an equal value diff amp. Or BP and HP, same technique. The same switch that enables the diff amp sum also bypasses Q control and locks Q to its lowest value. Shall I post it? Would something like that achieve the same ends? And what is the purpose of summing the outputs rather than just using the individual HP or LP output to feed into the cut/boost network (with corrected polarity and disabled Q)?

 

[atavacron]

1 - Focusrite isa? Porter? Amek2500?

I’m working with the Porter SVF form, generally, because it leaves the inverting input alone save for the two basic equal value resistors from HP and LP o/p. He does all his input signal, Q, and noise gain biz on the non-inverting input. Seems like the best starting point AFAICT.

I’ll check out the ISA and 2500, thank you!

 

[JohnRoberts]

I’ve read all your references to the Loft EQ…pity that no one’s dug it up! I have a scratchpad drawing of the BP and LP outputs feeding an equal value diff amp. Or BP and HP, same technique. The same switch that enables the diff amp sum also bypasses Q control and locks Q to its lowest value. Shall I post it? Would something like that achieve the same ends? And what is the purpose of summing the outputs rather than just using the individual HP or LP output to feed into the cut/boost network (with corrected polarity and disabled Q)?

That full parametric LF shelf was not very satisfying IMO. If anything too much LF boost is not that desirable in a tracking environment. I never repeated that experiment for a later large console design.

JR

 

[atavacron]

That full parametric LF shelf was not very satisfying IMO. If anything too much LF boost is not that desirable in a tracking environment. I never repeated that experiment for a later large console design.

JR

ohhhhh i understand. it was for a resonant low shelf option. yes, i can see how it would have limited application. what were you doing for high shelf in that circuit — did you attempt the same feat?

 

[abbey road d enfer]

1. Other than the PQ1549, what designs use the SVF to produce a shelving response from the cut/boost amp?

dbx 905, Neve 83022

2. Has anyone directly compared the sound of a Sallen & Key shelf with an SVF-derived shelf? Q of the SVF would have to be .5 for an apples-to-apples evaluation.

I have compared high-pass filters. The big point for me was the ability to control the resonance of the SVF version. I must say that, predictably, for a similar response, both versions sounded similar.

Net EQ -

Shelves are single pole SK, which entirely replace the SVF when engaged.

Remember that the Net EQ, for correct operation, must use the center-tapped pots. Failure to do that results in unpredictable interaction between bands. Since shelves have usually a wider BW than bells, this interaction may be more noticeable.

I have not yet seen a schematic with SVF outputs summed, but maybe that’s because when people mention it, they’re talking about a series application (???), and we generally don’t use SVFs in series.

I don't understand what you're suggesting. What do you mean by "SVFs in series"? If you mean cascaded, there are many examples of EQ's based on cascaded SVF elements.

 

[JohnRoberts]

ohhhhh i understand. it was for a resonant low shelf option. yes, i can see how it would have limited application. what were you doing for high shelf in that circuit — did you attempt the same feat?

I literally don't remember what I did for the HF section, probably the same thing.

Not to veer off topic, I found it useful to incorporate dynamic or adaptive HPF into low boost sections of EQ. It is possible to incorporate a HPF pole in series with the input that shifts the low frequency cutoff higher when more LF boost is commanded.

Over the years I have learned to avoid very loud, very LF content. Modern digital is more tolerant of LF but I am out of the trenches .

JR

 

[atavacron]

Not to veer off topic, I found it useful to incorporate dynamic or adaptive HPF into low boost sections of EQ. It is possible to incorporate a HPF pole in series with the input that shifts the low frequency cutoff higher when more LF boost is commanded.

That’s snazzy. Great CYA to avoid overload conditions. I would hope, though, that most users know to leave the low band on “bell” in event of big bass boosts.

 

[atavacron]

dbx 905, Neve 83022

Ok, will check these out.

I have compared high-pass filters. The big point for me was the ability to control the resonance of the SVF version. I must say that, predictably, for a similar response, both versions sounded similar.

Bingo. Very helpful. I’m going to try setting minimum Q to .5.

Remember that the Net EQ, for correct operation, must use the center-tapped pots. Failure to do that results in unpredictable interaction between bands. Since shelves have usually a wider BW than bells, this interaction may be more noticeable.

Totally. I’m working in an entirely switched environment, and it’s looking like I can accomplish the same isolation as Porter did by feeding different fixed resistor values from each filter output directly into the cut and boost inverting mix inputs, respectively. Make sense?

I don't understand what you're suggesting. What do you mean by "SVFs in series"? If you mean cascaded, there are many examples of EQ's based on cascaded SVF elements.

Yes, cascaded, thank you for clarifying. I’m focused on mixing rather than cascading.

 

[atavacron]

Here’s part of where I’m going with this. People who don’t know analog switches can just think of them as toggles, changeover switches, or rotaries….which is why they’re drawn this way.

I haven’t included the frequency switching because it’s a separate topic, really. But I do realize that there may need to be a cap change with the shelf engaged.

 

[abbey road d enfer]

My only suggestion is that you translate that diagram into a simulation schematic and check what happens.
Now you must be aware that simulation may not reveal unstable topologies or DC runaway.

 

[JohnRoberts]

are you concerned about switching noise? Might be easier to add a momentary mute. One trick i've used in the past was instead of muting a circuit full off, mute the output to the input so the mute is actually a momentary bypass.

JR

 

[atavacron]

Here’s a version (with frequency switching drawn) that does exactly what the Net EQ does, with a HP in place of the bell when shelf engaged. Four amps here instead of Porter’s five. Controls are a dual concentric coded hex switch, a tact with latch for bell/shelf, and a SP3T toggle for Q….per band…five bands…mono or stereo…

Also I just realized I’ve been referring to buffered single pole filters as Sallen & Key throughout the thread. Whoops.

 

[atavacron]

are you concerned about switching noise? Might be easier to add a momentary mute. One trick i've used in the past was instead of muting a circuit full off, mute the output to the input so the mute is actually a momentary bypass.

JR

I’m not sure because I haven’t gotten there yet. The OPA2210 is an impressive BJT… mid-pA Ib, low-uV Vos. Would that ameliorate concerns? Any different than a FET amp? Or would the muxes themselves cause problems in their applications here? In the drawing of post #17, I’ve got safety resistors in parallel to all switches that I could think might need ‘em (so far). I’d like to avoid timed switching/muting tricks if possible but I am open to it. I know nothing of such things as yet.

 

[abbey road d enfer]

Neumann W492 also uses state variable filters for all four bands, including the high and low shelves:

https://groupdiy.com/threads/neumann-w492-equalizer-diy-project-pcb-docs-added.28332/

I don't think this qualifies as a SVF. I may be wrong, but for me SVF involves the use of integrators.

 

[abbey road d enfer]

I’m going to try setting minimum Q to .5.

IDK. I may have misunderstood your question...
Did you mean replacing a shelf EQ with a bell centered at the shelf's max boost/cut frequency?
As JR mentioned, shelf EQ at high boost present risks of overload at frequencies outside the hearing range.
IMO, a big shelf boost needs to be complemented with a good HPF/LPF.
I have experimented with my digital mixer EQ, which allows exactly that, and found that for settings that are not too different than a usual Baxandall, the sonic difference was not very significant. Probably because I tend to wipe out anything below 35Hz and CD format takes care of anything above 22k.
However, when dialling high LF shelf or low HF shelf frequency, midrange can be significantly different at the extremes of the audio range as can be seen from the attached graphs