Bainter notch filter, anyone?

They did, eons ago:

"INTUITIVE IC OP AMPS" National's Semiconductor Technology Series Thomas M. Frederiksen

I'm mentioned on page 14 ;-)

I received an E-mail from Walt almost exactly a month ago claiming that....

"IB-Cancellation-WithTwoOpAmps.pdf" on the S.E.D/Schematics page of my website was the same as a circuit he published in Ideas for Design in

It isn't, it looks more like Howland's "pump" method.

So Walt is aware of my site.

...Jim Thompson

Thanks Jim, I'll take a look.

Hmm... I like how National Semi

---Joel

Try also...

...Jim Thompson

Or simply a schematic?

I haven't been able to find anything by surfing.

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

Folks,

OK, I went to that horribly anarchronistic institution known as a library :-) and dug up Bainter's paper. The idea is novel (at least to me)... the generic bi-quad, of course, has a numerator that's in the form of as^2+bs+c, and for a notch filter b is ideally zero. Bainter's topology ends up creating a filter where 'b' is inversely proportional to the gain of the amplifiers used, so as long as you can maintain decent gain at the notch frequency, you get a really nice, deep notch. He points out that with most other topologies, 'b' only has terms related to component values, and therefore the depth of the notch obtained depends on how well components can be matched and will remain matched over time.

This is very much the kind of a filter I was after... although I am still comparing it with a simple RLC circuit with Jim's gyrator as the L.

I scanned in the paper so that others can take a look; point your browers to:

---Joel Kolstad

Thanks, Joel!

I'll run some Algebra on it and see what it has to offer.

...Jim Thompson

Jim,

Those .sym files that you include with your various PSpice model... what version of PSpice are you using? I'm using the flavor of PSpice that comes with Cadence ORCAD 9.2 and I can't for the life of me figure out how to import them. The .sub files import into the model editor just fine.

---Joel Kolstad

..SYM = raw symbol file. It must be imported into a .SLB symbol LIBRARY.

(This applies to the old original-flavor MicroSim Schematics. But I think there's a way to import into Capture also. Look in the help files for instructions on importing MicroSim Schematics/Symbols.)

I use PSpice v10.3i

...Jim Thompson

And you can certainly build a less-cantankerous notch filter than the Bainter configuration, as shown in "GyratorFilterNotch.pdf" on the S.E.D/Schematics page of my website.

(If you want to control the depth of the notch, add an R between node "OUT" and C2)

...Jim Thompson

Hi Jim,

OK, Capture comes with a tool to convert MicroSim symbol libraries (.slb) into Capture part libraries (.olb), but it (1) wants a Micro configuration file (msim.ini) -- that I don't have and (2) unless just renaming a .sym file to .slb can 'fake' a symbol library, I wouldn't have the .slb library anyway.

Since it's just an op-amp, I re-drew the symbol, which is trivial enough... but as far as I can tell if you just have Capture installed you can't get from .sym to .olb.

Thanks for the help,

---Joel

Can I definie 'less-cantankerous' as 'response doesn't move around as much when you perform a Monte-Carlo simulation?' Or did you have something a little more specific in mind? :-)

Fewer (as in none) interactions between Fo and Q. Far less complex expression for Fo and Q.

...Jim Thompson

I like those reasons...

Hey, in your notch filter file on your web site... on the schematic it says that Q=2*pi*F0*R3*C2, but since the way you've drawn it is as a series R-L-C, I'd claim Q is really 1/(2*pi*F0*R3*C2), since you're after 'reactance of the cap at resonance'/'resistance'... right?

If you buy this I'll claim my complimentary autographed 1488 at the next BBQ. :-)

---Joel

Sure, but from Q=omega0*L/R and omega0=1/sqrt(LC), you can solve Q as

Great, let us know the date of the next one!

---Joel

Isn't "Q" always defined as XL/R(series) ??

I don't know if I want to give up my "stash" of MC1488's or MC1489's, but if you bring one to the BBQ I'll autograph it ;-)

...Jim Thompson

I read in sci.electronics.design that Jim Thompson wrote (in ) about 'Bainter notch filter, anyone?', on Tue, 29 Mar 2005:

No, not always. It's equally valid to use XC/Rs or Rp/XL or Rp/XC, since XL = XC at (one definition of) resonance.

This doesn't seem to be unqualifiedly true. According to the PDF on your web site, for your circuit, Fo=1/(2*Pi*SQRT(r4*r5*c1*c2)) and Q=2*Pi*Fo*r3*c2.

This seems to me to indicate that if Fo is perturbed, Q will change, but Q can be perturbed without changing Fo. So your claim (Fewer (as in none) interactions between Fo and Q, you said) is only true in one direction.

Your circuit's Fo is given by: Fo=1/(2*Pi*SQRT(r4*r5*c1*c2))

The Bainter circuit's by: Fo=K1/(2*Pi*SQRT(r1*r4*c1*c2))

Pretty much a toss-up on this one.

Substituting the expression for Fo of your circuit in the expression for Q so that we have the full bore expression like the one Bainter gives, your circuit's Q is given by: Q=(1/SQRT(r4*r5*c1*c2))*r3*c2

The Bainter circuit's Q=SQRT((K2*c2)/(r3*r4*c1))*(r4*r5)/(r4+r5) but if we notice that (r4*r5)/(r4+r5) is just the parallel combination of r4 and r5, we can use a common symbolism for parallel components and make the Bainter expression *look* simpler: Q=SQRT((K2*c2)/(r3*r4*c1))*r4||r5

Another toss-up I would say. And even without this symbolism, I wouldn't say that your circuit's Q expression (having substituted the Fo expression in the Q expression, putting it on an even footing with Bainter) is "Far less complex". Of course, that's just my opinion.

With the Bainter circuit, Fo can be changed by varying r1 without affecting Q, but unless K1 also varies, this changes the character of the transfer function from a pure notch to a low-pass or high-pass with notch, as Bainter notes in his description.

r4 appears in the Q expression but not in the Fo expression. Varying r4 changes Q without changing Fo, but, again, unless K2 also varies, this changes the character of the transfer function from a pure notch to a low-pass or high-pass with notch.

But I have another question for you. If I understand your circuit correctly, the part of the circuit to the right of c2 is a gyrator simulated inductor. So r3, c2 and the simulated inductor (whose value is L=c1*r4*r5, I think. Correct me if I'm wrong) form a series RLC circuit. Assuming an idealized series RLC circuit with the components arranged in the order R---C---L, left to right (right end of L grounded), the transfer function with input applied to the left of R and output taken between R and C as your circuit does, can be derived from a simple voltage divider point of view. So Vout/Vin is given by (1/sC + sL)/(R + 1/sC + sL) and after some algebra, I get Vout/Vin = (s^2 + 1/LC)/(s^2 + s(R/L) + 1/LC). From this, the pole Q can be found to be (1/R)*SQRT(L/C), and Fo=1/(2*Pi*SQRT(LC)), from which an alternative expression for Q is (2*Pi*Fo*L)/R. The equivalent for your circuit would be Q = (2*Pi*Fo*c1*r4*r5)/r3. It appears to me that your Q expression is incorrect. Have I made a mistake?

for Q is

incorrect. Have I

Looks like I flipped over a term... probably wrote it down from sight instead of calculation.

My "independent" comment was in regard to component tolerances... setting Q doesn't influence Fo, and *fine* tuning Fo does not influence Q... obviously if you change Fo substantially you affect Q (by its definition).

I'll fix the PDF and add some further explanation.

Thanks!

...Jim Thompson

expression for Q is

incorrect. Have I

Aha! I see what I did. On the S.E.D/Schematics page of my website, "GyratorFilterBP.pdf" shows a bandpass connection. A simple reconnection changes it to "GyratorFilterNotch.pdf", but I didn't change the "Q" equation.

...Jim Thompson