SVF parametric EQ. It works – fine tuning needed!

[lion]

I've build the ESP SVF parametric EQ and it's working. My next step is to fine tune it a bit for my needs, and I'm doing some experiments on this. Here's the circuit.


I have the filter caps C2-C3 as 15n, which gives me a total range that covers my "target" range – plus some more. With a 10k liniar pot a test proves that I don't need the first 2k span at the beginning of the pot rotation, and also not the last 1k span at the end of the pots rotation – in other words my target control range lies in a 7k span within the total 10k pot range.

The spread and "feel" of the target range (with relation to pot taper) doesn't feel too bad. So, in theory, if such a thing as a 7k dual potentiometer existed I could tag a 2k resistor at one end and a 1k at the other, maintaining a total of 10k, and I would have my target range spot on, and spread out over the total pot rotation.

One thing is theory – another is practice. The next dual pot value lower than 10k is 4.7k (obviously). Any alyternative solutions?

Looking at the total resistance of VR3A/B + R12/13 there's 12.2k in each string going to the C4/5.

I can divide that into subsections like this:

2.2k resistor >> 2k (surplus) pot rotation >> 7k (target range) pot rotation >> 1k (surplus) pot rotation

I don't understand how the voltage divider freq control functions in the circuit – is it the exact resistance values that matters or is it about the relation of values (like in "normal" voltage divider)? If it's the later I could scale all 4 subsection with the same 1.42 factor, giving me:

3.1k resistor >> 2.8k (surplus) pot rotation >> 10k (target range) pot rotation >> 1.4k (surplus) pot rotation

and solve this further into 1 resistor – pot – 1 resistor:

3.1+2.8k = 5.9k resistor >> 10k pot resistance >> 1.4k resistor = the target control range will now fit the 10k pot!?

Maybe I'm totally off on this  

Any comments, advise or ideas would be highly appreciated. TIA.

[PBE6]

If the feel for your target range is already good enough, you probably won't see much improvement by changing component values.

If you really want to exclude the top and bottom range from your pot altogether, try increasing the cap values from 15nF to something like 18nF or 22nF and using the smaller 4.7k dual pot if you have one handy. (Increasing the cap size lowers the top end limit and shrinking the pot size increases the bottom end limit.)

Alternatively, keep the 10k dual pot but increase R12 and R13. According to the simulation this will also move the bottom end upwards.

[PRR]

It's no big sin to have a little excess range on a knob.

Unless you have the Q hacked WAY up, this type filter is so broad that micro-trimming is pointless.

> how the voltage divider freq control functions in the circuit

Short out the bottom stopper for analysis.

If the VR3A pot is full-up, the 10K R9 acts like 10K.

If the VR3A pot is tenth-up, the 10K R9 acts like 100K.

If the VR3A pot is half-up, the 10K R9 acts like 20K. (Actually the half-up pot has 2.5K to it, so "20K" is really 22.5K. But get the ends right and let the middle fall where it may.)

The total divider resistance should not be much over 10K (to minimize the half-up error) but can't be much under 2K (the opamp will strain).  

So if you wanted a +/-10% range around half-up, you could be thinking 5K fixed, 1K pot (+/-0.5K), 5K fixed. Or since error is worst around half-up and you have plenty of load-impedance to spare, maybe 1K fix 200r pot 1K fixed.

There is an optimum but working it out would take much longer than cut/try.

If 5K panel-pot is handy, that would be good, but 10K can work.

Put a 5K or 10K variable resistance at both ends of the panel pot top and bottom. 2 stages, these should be gang-pots, but single pots work if you keep both of a pair at the same rotation.

Turn knob up, set top trims. Turn knob down, set bottom trims. This will interact, so repeat. By the 2nd or 3rd up/down you should be really close and anticipating the interaction, 4th try should be spot-on (remember, let it be a little wide).

Remove trim pots carefully and measure. If a pair are not showing very-near resistances, set them to the split-difference and re-check the knob range. Then find the nearest standard fix-resistor values to replace the trimmers.

[lion]

Thanks for the replies.

Paul – I'm not sure I fully understand the "full- tenth- half-up" and the half-up error explanation, but I'm working on it, but I get your idea with the trimmers at both ends of the panel pot and I'll give that a try.

< It's no big sin to have a little excess range on a knob. Unless you have the Q hacked WAY up, this type filter is so broad that micro-trimming is pointless >

The excess pot range would be acceptable I agree, but I take it as a learning process – and would like to see to what degress I can optimize the function.

Re the Q
The ESP/SVF article states the bandwidth as variable from 2.7 – 0.17 octaves (Q of 0.46 - 8.2).
However from a prelim freq test the Q doesn't seems to get that high in my circuit:


The plot shows the response with max boost and the MAX Q setting. As far as I can see the bandwidth is aprox 1 octave  (the LF and HF roll off seen are not the SVF but due to other circuitry I forgot to bypass).

The plan was to set up the bandwidth switchable between aprox. 0.7 and 1.4 octaves (Q 2-1)
Apparently my build doesn't get the Q higher than 1.4 (= 1 oct). which is far from the stated 8.2 Q. Could be an error in the build on my part -  but apart from the Q pot and the 1k res to ground (which checks OK) I'm not sure how to troubleshoot Q function – or modify to get the desired sharper Q of 2?

[lion]

I'm an idiot -  

R11 should be 1k (as I correctly wrote above) and not the 10k I 've wrongly soldered in (and checked and double checked as 10k).
I'd bet it solves the Q issue. Will make new freq test/plot.

[tubegeek]

I'm an idiot

If you're an idiot then I'm Rush Limbaugh. You're doing fine.And an order-of-magnitude (wrong third stripe) error is just about the most common. Check out the Craig Anderton Super Tone Control for a very full-featured SVF implementation.

[lion]

Thanks tubegeek

The Q issue is solved now with the correct R11/1k res in place.

I have actually been looking at CA's STC - but I decided on another route. The plan is to cascade two SVF's for a low mid and hi mid boost and combine these with individual variable lo- and hi-cut Salen-key filters to give me a custom purpose EQ box. Here's the lo-/hi-cut circuit:

I'm not totally decided though how to combine the circuits for best performance?

Re the input Z of the SVF (schem in 1st post). I recon it's set by RI (22k) - that's pretty low. So I was thinking to break the lo-/hi-cut circuit between the output of the buffer stage 1C1A and R6 - and insert the cascaded SVF parametric bands here. Wouldn't the output Z of the SVF circuit be low enough to drive the Salen-Key LP-/HP-filters?

[tubegeek]

I think that the NFB reaching the input end of the 22K (R1) input resistor acts to bootstrap it and raise its effective value. Since the voltage across the resistor is being reduced (the opposite polarity of the returning feedback loop will subtract from the input signal) the resistor behaves as though it is a higher value than its actual resistance.

Somebody double-check me on that - I only recently learned about "bootstrapping" and I'm not sure I completely understand it.

What is the application you are designing for? In the article presenting the Super Tone Control Anderton points out that the circuit is versatile enough to be able to get some terrible sounds out of it - the extra complexity you are designing in may lead you to the same problem, that it might be tricky to use well, and as a guitar effect it might be more complex than is ultimately useful.

I've built the STC and I can vouch for this statement - there are definitely some useless settings on it.

[lion]

I remember reading something about bootstrapping an inverting stage – I'll have to do so research. Thanks for the tip.

<What is the application you are designing for?>
It's a guitar EQ – to sit on top of my amp for a custom/final tone shaping. I get your point about complexity – and I agree.

Here's the idea behind the build/experiment. For a certain type of music I want a certain signature tone. I can get close to that "target tone" using just guitar > reverb/echo > amp, but for the last percents some additional EQ is needed. Tried a 10 band graph EQ and a single band parametric in the chain – but the conventional filter bands doesn't quite do what I need – nor did the single band para EQ (although the later got me some of the way).

Out of curiousity set up test. I recorded the amp - to get the sound I hear standing in front of the amp the best I could (that's not as easy as it sounds). Then I post processed the track in my DAW to get my target tone (trying to find the simplest possible EQ corrections needed).

After some experimenting I found that (without going into the exact details) a 3 part EQ gave me my target tone:
- a steep(er) LF roll off
- a low mid boost
- a high(er) mid boost (both mid boosts at specific freqs and Q obviously)
Measuring the total resultant freq response turned out to form a twin peak "Camelus Bactrianus" curve.

To take the experiment one step further I though it could be interesting to take the "Camelus" processing from my DAW experiments and "put it in a box". so to speak. I have no idea if the DAW EQ settings will work the same "live" as it did in post processing. In case some changes are needed in real life I'll keep all the variables in the circuits I've decided to use accessable for a start – low cut, lo-mid Fc/Q/gain, hi-mid Fc/Q/gain (and in addition also a variable hi-cut filter which I think could be usefull - and was part of the Salen-Key LP/HP filter circuit I found). If it turns out as I hope some/most of the variables can be preset internally – or controlable in a limited range – leaving as few controls as possible for simplicity.

Although the "Camelus" EQ idea looks good on paper, I'm aware it might turn out less usefull – if at all in practise. Time will tell.

Re the circuit
The lo/hicut circuits has gain, I wonder if it would best to keep both before the SVF paramtrics (for better S/N) and put a level control at the end. Any suggestions?

[tubegeek]

Suggestion:

Use all the gain available up front unless the filter/boost stages clip the stages afterwards. If that happens, EITHER voltage-divide down until the clipping stops, OR lower the gain on those stages, OR enjoy your spiffy new combo distortion/EQ.

It sounds to me like you are going about the goal of boxing up your favorite EQ curve in a pretty reasonable way (in case you care what I think.)

The low pass filter will probably be useful to get some speaker-cabinet emulation - guitar cabinets are not really full-range.

[PBE6]

In my build I put an Orange Squeezer - complete with extra gain and an output volume control - before the SVF. Worked great unless the first stage output volume was set too high, which made the whole thing sound like a snoring walrus.

If you want to put the gain upfront to reduce excess noise, I'd suggest keeping the gain moderate. I don't think a standard opamp will swing much more than +/- 3V on a single 9V supply, so 2x to 3x pre-gain should allow all the transients through with a little room to spare.

Alternatively, you could add some limiting diodes/LEDs to the first stage to make sure the input signal distorts a bit more gracefully if it does go out of bounds.

[lion]

I appreciate your opinions and advise – thanks.

Having the gain stages up front makes sense. Here's the complete circuit as I have it now – incl. the switching setup:



The supply is a dual +/-12V and the gain in the HPF/LPS is only 2.3, so clipping will not be a problem I think.

If you spot any problems or bad practice in the circuit please let me know.

[PBE6]

Just did a quick scan, looks great! Nice job.

[lion]

Thanks. I have it complety wired together now - and everything seems to work. The filter *caps in the Hi-mid SVF should be 4n7 (rather than 2n2 stated on the schem). I had a wrong resistor (again ) which caused a bit of a problem and shifted the control range strangely until I found the resistor error.

Everything's still on the worksbench - no real listening/playing test yet - just verified functions and have been running some freq response tests.

I definetely need to work on limiting the control ranges of the FREQ pots as per PRR's suggestion - to not cover more than the variations I need, and to limit complexity.
Alternatively I noticed in another state variable filter thread a mention of "Rod's attenuator plan" - would that be Rod Elliot of ESP?

[tubegeek]

Alternatively I noticed in another state variable filter thread a mention of "Rod's attenuator plan" - would that be Rod Elliot of ESP?

Almost certainly - I think there are a lot of people on diystompboxes who follow the ESP pages closely They have certainly helped me learn quite a lot of really useful information and I think that site may be the best source for analog op-amp audio design advice on the Web.

Two really important supplements to ESP I can personally vouch for are two dead-tree sources: the Active Filter Cookbook by Don Lancaster and Douglas Self's Small-Signal Audio Design. The latter has recently come out in a very much larger 2nd Edition which I can wholeheartedly recommend to anyone interested in a really intensive exploration of best practices for little teeny signals and how to mess with them properly. With ESP and SSAD you could tackle a studio mixing console project with almost-confidence. The new edition of SSAD includes expanded information on discrete transistor design among other big changes. I'd have to say it's worth buying a brand-new 2nd edition over a cheaper used 1st edition, and I am a terrible cheapskate by nature (although I rarely resent spending money on books!) so take that for what it's worth.

There are extensive explorations in SSAD of gain staging and design of level and balance controls for lowest noise etc. so I'd say that resource could be helpful for you. If I had to sum up a lot of Self's advice in one sentence, it'd be: keep your resistances as low as you can without overloading the op-amp's ability to deliver current, so as to reduce the noise of the resistors themselves.

+/- 12V supply a good decision - for historical reasons a lot of "stompbox" designs feel locked into 9V battery compatibility but when you are designing for yourself only, there's no need to observe that convention and certainly many reasons not to.

[lion]

Thanks for the litterature recommendations. I've heard mention of the "Coookbook" before, but the SSAD is new to me. Sounds interesting and usefull.

Re "Rod's attenuator plan"
Except Rod's taper law changing resistor trick I can't seem to find anything related to this on the ESP site. I think it was Paul/PRR that mentioned the attenuator plan at some point. Is it anything but a rotary switch with reistors/fixed resistance - instead of a pot/variable resistance?