Dual Amplifier Band Pass

[atavacron]

Anybody done anything with the DABP? Page 211 of the Electronic Filter Design Handbook. Originally Sedra & Espinoza, 1975. Simplified at this link by Hank Zumbahlen of Analog:

https://www.analog.com/media/en/training-seminars/tutorials/MT-209.pdf

Seems handy as a State Variable alternative in parametric EQs. Independent Q and frequency (?). One less op amp than the SVF. Non-inverting.

Fixed gain of two, so you would (generally) want to make a gain-of-half voltage divider out of the Q set resistor at the front end, and follow that with a variable resistance. Super high Qs clearly generate Johnson noise, but at regular parametric EQ Qs it doesn’t seem like an issue.

 

[JohnRoberts]

Looks like one too many opamps for a simple filter.

For full parametric control SVF is great. In consoles we see sweep frequency EQs using only one opamp per frequency band. Variable Q is overkill for every channel input. Back in the 90s when I was still designing consoles/mixers I pioneered a concept of including a few "super channels" loaded with full features. The mix engineer had to patch problem inputs into one of the super channels if the extra functionality was needed. This was far cheaper than putting full features on every channel, or using outboard gear.

JR

 

[user 37518]

From the transfer function it looks like the gain is not independent from Q, so adjusting Q would also vary the gain of the filter. This would require some modification to compensate for gain changes if you want to have a fully parametric EQ. For a semi-parametric (fixed Q, adjustable freq.) it should be ok.

 

[user 37518]

Looks like a good circuit. At first I thought that gain wasn't independent from Q, I missed the '2' in the numerator. But yes, at first glance it looks like it would be a good option to have variable Q and freq. However, notice that you would probably have to add at least another op-amp to make it able to cut as well as boost, in terms of ICs having three op-amps is practically the same as having four, since dual and single op-amps come in the same packaging. So you might as well go the SVF route.

 

[atavacron]

Looks like a good circuit. At first I thought that gain wasn't independent from Q, I missed the '2' in the numerator.

It’s interesting. Seems a wee bit more flexible than a S&K or MFB band pass. A little more noise than the former, a little less than the latter. I think gain is fixed at two, so I’d split the input resistor into an attenuator.

I stared at this for a while longer and realized that while the various books that describe it tout the “independent” variability of Q, they really mean at a given frequency, which pretty much blows the parametric application. I did the most basic of numbers and facepalmed.

You could wrap it in a Q multiplier, but then you’re up to the same parts count as a SVF. The Q multiplier was new for me, I didn’t realize that that’s what the first amp of the SVF actually is, and that you could do basically the same thing with a MFB band pass filter (or any other inverting unity gain filter) after that amp. Like…you don’t have to have two integrators, necessarily.

 

[atavacron]

However, notice that you would probably have to add at least another op-amp to make it able to cut as well as boost,

i was thinking of the feed and return being from and to a cascaded inverting stage, like the W492 or similar. just setting it for unity.

 

[atavacron]

"super channels"

Point!

I was looking at the 9098i schematic and wondering how often the notches on the mid bands actually got used. Interesting that the high and low bands are SVF-derived shelves, though. Sort of a maximalist-minimalist approach.

 

[atavacron]

Jim Thompson, 2002. Gain-at-resonance mod on the back end. [EDIT: I couldn't get Thompson's mod to work]

 

[atavacron]

I modified the DABP for parametric use. Here's the simulation.

I added two amps, a range-limited linear pot for positive feedback, and a triple gang rev log pot (25K/10K/10K) to maintain constant maximum Q so the feedback pot can do its job (.40 CCW, 1.2 CTR, 2.7 CW). More feedback widens the bandwidth but leaves output the same gain as input. Maximum Q is supposed to be that first pot section divided by either of the next pot sections, but in sim it comes out a little higher. Obviously a stepped 3-deck frequency control of 12- to 24 steps would be better tracking and less obscure than the pot. In low quantities, a 3-deck Grayhill 71 is not that much more costly than a custom 3-gang pot from State Electronics.

It is also possible to use a 4-gang 10K rev log pot with two sections in series (20K/10K/10K), though accuracy will degrade and Q range will be a bit limited. With a bit of a noise tradeoff, a 3-gang 50K/20K/20K would net the same Q range as drawn, with a 20:1 frequency range. This might be a good solution for general purpose tracking. A three-band 500 series EQ might look like 40Hz-800Hz / 400Hz-8kHz / 800Hz-16kHz. Low shelf is possible using this filter as a source, and for high shelf one could add a diff amp on the back end that subtracts the LP output from the original signal.

This filter's main advantage over the State Variable is that it has a lower noise floor (I think?), particularly when set up for a low Q range. Very low Q SVF sections require a fair amount of attenuation-then-amplification. It can be set up for very high Q very easily, with Johnson noise of the input resistor being the only additional penalty. The main disadvantage (aside from the spendy control method) is that you can't come in to the filter at full level; it requires 6dB headroom. Pretty similar to an SVF in that way, where you have to lose some signal on the way in if you want to avoid overloading the mix amp and LP integrator at sub-1.0 Qs.

I could not figure out a way to use slugged linear pots in this architecture — maybe someone can point out a method? Alternatively, if there is a way to make a resistor seem larger than it is, in series with a non-inverting input, this would be an ideal application - specifically to be able to get a healthy Q range out of a more-readily-available three-gang 10k pot. I looked at negative resistance but that’s negating a load — different problem. I also contemplated bootstrapping to make the input amp see a larger impedance (say, x5), but i’m a noob at calculating such things. Any ideas?