State Variable Filter Design questions

[JohnRoberts]

Agreed no very heavy lifting here, but I could imagine a design using a few different parts.

It is pretty common these days to use canned solutions for input/out like THAT chip sets. Porter invested some effort as I recall in making an output driver that didn't suck. He didn't have the options we do today, to use simple solutions.

I have the THAT in/out chipsets in a few units here and they are indeed very fine sounding.  In conceiving this new build as part of my dream chain I'm pretty certain I'd like it to be unbalanced. I have a couple of things on my side

- My chain is (and will be) short, 4 or 5 units in close proximity.
- Al source and load impedances are predictable (and easy to work with).
- the broad goal is to change the sound as little as possible when run flat.



Cheers,
Ruairi

I suspect there may be some confusion about cost/benefit in balanced interfaces.

Or perhaps I should ask, what do you have in mind for an unbalanced input?

The distinction "balanced"  only means two signal lines with a differential receiver and similar source impedance to ground on both signal lines. Inexpensive "impedance" balanced sends using only one opamp are still "balanced" . 

The distinction "un-balanced" only means that the two signal lines are not the same impedance to ground. There can be unbalanced two wire and 3 wire interfaces.

While most true active balanced inputs require two or three opamps, mainly to deliver the like termination impedance to both lines, you can make a single opamp differential input that performs very close to a true balanced input in all respects except same impedance to ground and therefore degraded noise rejection of CM noise picked up in long wire runs. 

Lets not confuse "single ended" interfaces using a single audio line and shield, with perhaps unbalanced interfaces using differential receivers and 2 audio wires inside a shield.

I believe very strongly that single ended interfaces have no place in professional audio. Likewise full balanced interfaces are perhaps more than needed for typical short runs. IMO always use 3 circuit wiring (2 hots and ground), and differential receivers, even if just one opamp, and therefore technically unbalanced. 

JR

 

[ruairioflaherty]

John,

I've been mixing all day so just getting to this now.  Again you've made a light bulb go off for me and you've presented some very simple facts in a way that I haven't quite seen before.

To sum up I was considering going 2 wire single ended for several reasons

- My set up has very short runs in a controlled (not patching in and out) environment
- I was aiming for low noise (a single opamp being the goal)
- I was hoping to plan my eventual switching console to be unbalanced to simplify the design

Bruno Putzey has gone on record has also gone on record to say that he believes differential inputs should be standard and that impedance balanced outs are more than enough, with fully balanced outs being in most cases overkill.

It's obvious now that of course even a single opamp differential input with 3 wire interface is preferable to a single ended input.

I've always though of single opamp differential inputs as being somehow inferior but of course full balanced inputs need more opamps and complexity and end up being noisier.  I have been studying Douglas Self's low noise balanced in and discussing it off list with another member here.

Thank you,
Ruairi


 

 

[JohnRoberts]

OK, put on your seat belt and helmet because this could get esoteric.

[esoteric

If you analyze the signal current flows, between chassis for sundry interfaces there are several trends.

-For a differential output (both lines active and opposite polarity) to balanced input (same impedance to ground), there is zero net current flow. The current coming out of the + input will equal the current coming into the - so no electrons have to find there way home at the end of the day.

-For impedance balanced output (with only one active output), feeding a true balanced input. There is no opposing current coming from the cold leg. The signal current nets out to zero over time, but in the short term there is audio signal current flowing out of the driver chassis, that must find a path back home.

-For Single legged output (impedance balanced or not), to single opamp differential input, it is possible to select values for the differential, with active signal line applied to the plus differential input, such that the current flowing out the active signal line, is equal to the current flowing back into the signal return. So not for all differentials, but this one case, all the current leaving the send chassis, come right home.  Note: I am describing the differential with 4 equal value resistors. One from XLR pin 2 to opamp +, one from opamp + to receiver local signal ground, one from XLR pin 3 to opamp - input, one from opamp - to opamp output.  

In conclusion this one case of differential input shares the same signal current balance as the full dual output balanced, balanced input.  The only downside is this interface is not truly balanced beyond the gain bandwidth of the opamp, so HF CMRR will be inferior to true balanced, so a good idea to add some passive RF filters.

If not obvious, there is merit to reducing signal current flowing between chassis, This can compromise crosstalk in a multitrack environment, or stereo separation in a simpler interface.

Note: Well designed interfaces using any of the above topologies should ignore ground potential differences between the chassis, so this is just a little extra something to keep signal current from increasing ground potentials.

OK, you can take off your helmet now...   /esoteric]

JR

edit] I need to mention that transformer coupled outputs (or inputs) also share the characteristic that zero net signal current flows between the two chassis. I forgot to mention transformers since I don't think of them for high quality interfaces, while some people still do.  /edit]

 

[Gold]

Nice post John. I've never seen this explained this way. Interesting way to look at.

 

[ruairioflaherty]

Excellent post John, as Paul said I've never seen it described that way before.  It's unlikely but I'll ask anyhow, is there any book/resource that works through what you've described?

Thanks,
Ruairi

 

[JohnRoberts]

Excellent post John, as Paul said I've never seen it described that way before.  It's unlikely but I'll ask anyhow, is there any book/resource that works through what you've described?

Thanks,
Ruairi

No... not that I know of, but to be honest, i don't read many "how to" books about circuit design. I had to figure this stuff out decades ago. In my earlier days I devoured other peoples schematics voraciously, and did more than little what-if bench work. These days not so much.

To be fair, I did qualify this as esoteric... managing ground currents between chassis is already well controlled by proper use of differentials and pin 1 bonding.  This is just some extra subtle information about interfaces.


JR

 

[ruairioflaherty]

Esoteric for sure but very interesting.  Self on Audio arrived yesterday and that has a pretty good section on interfacing so that should keep me out of trouble for a few lunch breaks.  Next thing I need to do is build up this Porter circuit and start from there.

Cheers,
Ruairi

 

[Gold]

The gold nugget in there for me was the explanation of why a four resistor differential input is not balanced. I'm still trying to get a handle on common mode impedance vs. impedance to ground and how that impacts equipment interface. I can't read the circuits well enough to figure this out myself.

 

[JohnRoberts]

There has been a lot of good work done characterizing and describing the mechanics of balanced interfaces,

http://www.thatcorp.com/datashts/AES6261_New_Balanced_Input_IC.pdf   here is a good link from the THAT website, no doubt pimping their IC.

Whenever active electronics are involved they loose their their mojo at some very high frequency, one reason that transformers have superior RF rejection.

Just to remind, this whole discussion about using a simple differential is in the context of short runs in a controlled environment. For difficult environments real balanced is still superior.

JR

edit/ curiously I participated in a recent thread  (on another forum) asking why consoles weren't fully balanced on the inside.  ;D  /edit

 

[Fenris]

Hi all,

I know that a state variable topology can output HPF, BPF and LPF at the same time so why do designers not use this to give a switched shelving option on eqs? (I couldn't find any examples of it anyhow). 

(if I remember well)
because the filter has different features by using its different outs. Using the filter in BDF mode its gain does not depend by the Q pot, but the same filter in HPF or LPF mode has the gain depending by the Q pot.
The problem is that to get a good Q in HPF or LPF mode (starting from the filter set for BDF mode) to have the shelving feature, the gain is too low and the signal needs to be amplified some times, so many designer prefer use an other unity gain filter to avoid other noise.

I'm trying out some mods on a Wheatstone console with 3 band state variable EQ. I want to turn the high band into a shelf. By "good Q" do you mean narrower? A first-order slope is fine, would I get a decent amount of gain this way?

It doesn't have to be switchable, I can convert it permanently to a shelf. In that case, would I change the feedback resistors to get more gain?

 

[JohnRoberts]

For the record do not assume that all SVF EQ are the same.  While the general topology of two integrator sections in series will be found, how that bandpass result gets added/subtracted to the dry path can be accomplished multiple ways.

Further IIRC the Whetstone stuff is probably a Gary Snow (?) design, derivative from his old Audio arts work.  He did some interesting variants on the Q adjustment in SVF (tweaked the BP gain up then padded output down to get narrow Q) that may make a shelving conversion tricky. 

While simple in concept, shelving conversion means adding the BP to HP (or LP) or perhaps ignoring the BP entirely and using just the LP/HP output.

varying the Q will have a different affect on filter shape for HP/LP than BP but you'll see that. 

It sounds like a good learning experience ( for you), already been there and done that so long ago I forgot about it.  8)

JR

 

[Fenris]

Further IIRC the Whetstone stuff is probably a Gary Snow (?) design, derivative from his old Audio arts work.  He did some interesting variants on the Q adjustment in SVF (tweaked the BP gain up then padded output down to get narrow Q) that may make a shelving conversion tricky. 

I think you're right. The circuit is a SVF but it's not the textbook version, and the Q is narrowed by increasing the value of 2 resistors.

it's an oddly designed console (TV-600), they made the EQ post-insert and bell only. The mono channels use SVF's on all 3 bands, but the stereo channels use a Baxandall variation on the hi and lo bands that can easily be converted to shelf.

 

[Twenty Log]

Be savvy to section crosstalk betwixt multiple section (dual/quad) op-amps if the sections/channels on the same chip may be different and the outcome be sensitive...

 

[JohnRoberts]

Be savvy to section crosstalk betwixt multiple section (dual/quad) op-amps if the sections/channels on the same chip may be different and the outcome be sensitive...

Crosstalk between dual/quad opamps is generally very good. I don't think I've ever experienced it as a problem, and I've used lots of duals and quads.

About the only thing I can imagine is a soft unregulated power supply rail, since the PS is common to all opamps inside the same package.

Good design practices should keep interaction via this common vector under control.

JR

 

[Fenris]

I'm still struggling to understand how SVF's work and I could use some help on this. Scanner is on the fritz, but here's a schematic from an SP-8 with identical EQ. The 3 opamps in the hi eq section are U4a, U3b, and U3a from left to right. I tried disconnecting the output of U3b (bandpass) from to the input of R19, and connecting the output of U4a (highpass). Instead of an EQ, I got a HF oscillator.

I just want some kind of useable high shelf, it doesn't have to be switchable.

 

[JohnRoberts]

In general the three opamps in common SVF are 1=HPF out, 2=BPF, 3=LPF.

That schemo looks like some Rs may be missing.

Too early in the morning if not....

JR

 

[Fenris]

Here's the TV-600 schematic, showing the entire EQ section. Do I simply disconnect the BPF output (U6b) from the boost/cut pot (CR13) and connect the HPF output (U1a), or is it more complicated than that?

 

[JohnRoberts]

The polarity of HP is inverted wrt BP so boost/cut will be reversed.

JR

 

[joaquins]

SVF HP or LP filters are 2nd order, BP is 1st. In the opening of the topic there is a schem that has 1st order filter for shelving... just to have in mind, not a problem...

JS

 

[Fenris]

I got it to work! I had to do two things: 1. Replace C30 (the cap on the first integrator) with a very small value, 0.5 nf or smaller. 2. Bypass the first section of CR12 (the frequency control) so it's always 0 ohms.

I tried lowering the reistance of R104 (on the input of the first integrator), but if you lower the value too much, the circuit misbehaves and you get DC on the output. You can't remove or bypass the cap either.

Basically, you're fixing the LPF frequency above 20k. I'm not sure what would happen with an even smaller capacitor, it might cause misbehavior. A more elegant solution would be nice.

The result is a good sounding, sweepable, 2cd-order shelf. It has the same frequency range, and it can easily be made switchable. I still have to do some pink noise tests and try it with a Q pot. When I do I'll post the results.