Help verifying lo-hi cut circuit dual supply

[lion]

I planning to include this lo/hi cut filter circuit (credits on the schem) in another project, and for that I've tried to adapt it to work from a +/- supply.
Not totally sure I've got it right though.

If anyone will take a look I'd much appreciated it. Thanks.

[PBE6]

This circuit looks a bit odd to me in some places, but it may be correct.

The input buffer built around IC1A seems fine.

The filter built around IC2A appears the be a 2nd order Sallen-Key low pass filter with a cutoff frequency variable between 15.9 kHz and 1.4 kHz. The part that looks odd is the voltage divider in the feedback loop formed by R8 and R9. Typically a low pass filter would have a unity gain buffer here. With the current configuration the gain at very low frequencies will be about 1 + 56k/100k = 1.56 = +3.8 dB. Is that to make up for the signal loss from the filter itself? If so it's probably fine, but if not a unity gain buffer will leave the very low frequencies essentially unchanged.

The filter built around IC2B appears to be a 2nd order Sallen-Key high pass filter with a cutoff frequency variable between 2.3 kHz and 280 Hz. The same question applies regarding the use of the voltage divider.

The output seems fine as well, although the pot value is not specified. Any value above 20k will prevent undue bass loss in guitar signals, while 50k will work do the same for bass signals. Of course, you're already cutting frequencies below 280 Hz so you can probably get away with an even smaller pot if necessary.

With regard to power, I think an alternate method may save some battery life:



This method should provide better power efficiency and power filtering than the one on your diagram. It will also let you get some use out of IC1B which is just sitting idle right now. I got the idea from PeterPan in a thread about oscillation problems:

http://www.diystompboxes.com/smfforum/index.php?action=printpage;topic=109445.0

(Someone please double-check this! I've never built a dual supply design before and I could be overlooking something really obvious that would make this approach incorrect...)

[Keppy]

With regard to power, I think an alternate method may save some battery life:
This method should provide better power efficiency and power filtering than the one on your diagram.(Someone please double-check this! I've never built a dual supply design before and I could be overlooking something really obvious that would make this approach incorrect...)

I think that's incorrect. The point of providing Vref from an opamp output is that it makes an extremely stable virtual ground. Circuits with bipolar power don't need a virtual ground or Vref, though, because actual ground sits halfway between the power rails.

I planning to include this lo/hi cut filter circuit (credits on the schem) in another project, and for that I've tried to adapt it to work from a +/- supply.
Not totally sure I've got it right though.

If anyone will take a look I'd much appreciated it. Thanks.

Your power supply and ground connections look good to me. Anything else should be the same as the original single supply design. I think you're good to go.

[lion]

Thanks for the replies, much appreciated. I'm glad I got the dual supply wiring correct as I was almost finish doing a vero when I got worried I had missed something.

....The part that looks odd is the voltage divider in the feedback loop formed by R8 and R9. Typically a low pass filter would have a unity gain buffer here. With the current configuration the gain at very low frequencies will be about 1 + 56k/100k = 1.56 = +3.8 dB. Is that to make up for the signal loss from the filter itself?

This is how it was original designed by R.G. (for another circuit). For my purpose I think I could do without the gain - IF it's better for noise etc?
On the other hand I was looking at a couple of Salen-Key filter calculator sites not long ago, and I seem to remember that the resistor in the FB has an effect on both Q and damping, but maybe I misunderstood (and I can't remember where I read it)?

I'll finish the vero and build it in the next couple of days.

[PBE6]

The voltage divider shouldn't affect noise much at all. The main effect will be from the boost to the passed frequencies (which in this case will be a 2 x 3.8 dB = +7.4 dB to the mid frequencies since both filters contribute) but I doubt you'll hear any noise over the boosted signal, especially as you'll be rolling off the high end hiss. The boost will probably even help combat hiss reducing the need for gain stages further down the line.

Increasing the internal gain of the filter will cause the circuit to resonate near the cutoff frequency, but the resistors values given won't cause any issues.

Looks like you're good to go on all fronts!

[PRR]

> looks odd is the voltage divider in the feedback loop formed by R8 and R9. Typically a low pass filter would have a unity gain buffer here.

The gain is correct for this filter.

Using *equal* component values, unity gain would give a very low-Q gentle corner. Musically little different from a single-pole filter.

Put in a lot of gain and you get a big hump. Small gain brings-up the corner without going bump. IIRC the gain for Q=1 is about 1.6. The indicated 1.56 is the same for practical purpose.

There is *another* way to go: un-equal values. By staggering the values significantly you can again hit Q=1 with unity-gain. However this makes varying the values difficult -- find a pot with 5K on one track and 20K on the other.

Q=1 is a nice solution with a mathematical description and a guy's name on it. It is also "usually" the near-best response for musical band-limiting. Q=0.7 gets another name and has lower phase-shift but is mighty gentle around the corner. Q=1.4 usually gives an obvious boom/ring at the corner but is sometimes useful to distract your ear from the stuff that is missing. Start with Q near 1, or gain=1.6 in the equal-value filter.

> an effect on both Q and damping,

Spot-on. FWIW, Q and Damping are flip-sides of the same coin. Low damping is high Q. High damping is low Q. Sometimes the math is easier if you invert Q to get "d", damping, to get factors all on the same side of an equation.


For guitar-level work, I *might* prefer scaling the 56K:100K down to 5K6:10K. The 100K is on the edge of where resistor hiss may be heard above guitar hiss. Probably a minor detail, but the common opamps can easily drive the 15K6 load, so why not?

I do not see any power or hiss advantage in the rail-splitter, and it often leads to unexpected consequences when integrated into a system.

[lion]

Thanks for the replies, much appreciated. I'll look at the 10/5.6k resistor issue and the Q possibilities later, and will probably also need to fine tune the filter cap values as well for the response I need.

As for now. I have built the circuit on vero and it works – but two issues maybe.

Firstly the offboard wiring and the output wire seems to be very noise sensitive. I have some noise/buzzing, but maybe it'll go away when I get it boxed and tidy up the wiring. I have a 100K output level pot.

Secondly, something odd. Couldn't find my 2x battery holder with wires, so for a test I did the -/0/+ connections with 3 alligator clip leads. Verified the supply voltages on first the IC sockets, then with the TL072's in, as +9V on pin 8 and -9V on pin 4. However, during my short test the 0/ground connection  between batteries and the vero fell off –  but much to my surprise the circuit kept working!? Measured on the IC pins – still +/-9V! How can that be?

I have 2x 47uF decoupling caps on the vero on the supply feed and 4x 100n ceramics directly on the IC's pins 4/8.

Obviously I'll have to look closer at it today, but something seems wrong at first impression.

[antonis]

during my short test the 0/ground connection  between batteries and the vero fell off –  but much to my surprise the circuit kept working!? Measured on the IC pins – still +/-9V! How can that be?

It depends on what you mean by "working"..

IC1A, High Cut A&B, Low Cut A & OUT do work (although the Output Impedance is affected.)
Low Cut B, R8/R9 & R12/R13 shouldn't work..

Unless you mean momentarily working (untill capacitors discharge..)

I have 2x 47uF decoupling caps on the vero on the supply feed and 4x 100n ceramics directly on the IC's pins 4/8.

Good enough for stability...
(allthough I'll add 2x 100R resistors in series with Vcc and Vee for de-Q the potential resonace and "fusing" from accidental shorting output to ground..)

[slacker]

It probably works with the ground wire removed becuse the 47u caps are forming a voltage divider across the battery. There's 18 volts across them so if you measure from the mid point to either end you get + or - 9 volts. The midpoint looks like ground for AC so everything works. There's no guarantee it will work if you power it up with no ground wire attached though.

[lion]

Yes - I get it now! I felt quite tired last night when I had finished soldering the circuit, but was eager to verify the function - not the best time to solve "problems" 

Haven't had the time to look at it again today.

Next step in the build of custom purpose EQ box is to compliment the lo/hi cut with 2 parametric mid boost filter sections. I intend to use the ESP/westhost SVF circuit (currently being discussed in another thread): http://sound.westhost.com/articles/st-var-f5.gif

I was thinking simply to put the two SVF filters in series with the Hi/lo cut somehow. Placing them after the buffer IC1A (before R6) seems to me to be possible - but could also be first sections before the lo/hi cut, or even at the end (before the level pot). Any thoughts on the pro cons or the best way to do it for best performance?

[lion]

Spent a bit more time with the circuit today. Added 2x 100R resistors before the 47uF caps, cleaned up the wiring a bit and got it partly boxed up (it's a test/proto build). Filter functions verified - and the noise is gone. Increased the filter caps in the low cut section to 47n to lower the control range. Will leave the scaling of the 56/100k resistors, pot tapper issue etc for the final fine tuning.

Next step is to add a 2 state variable filters as a 2 band mid boost feature to the lo/hicut module. How would the experts go about this - any advise for best performance or what to consider would be much appreciated.