having the cake and eating it too - passive notch filter design

[iainpunk]

so i want to cut a notch 6dB of 4.1kHz, as narrow as possible. (the exact frequency isn't that important)
when i use the Twin Tee calculator, whatever i try, the bandwidth is too wide at such a low attenuation depth. are there any other ideas to get a nice and narrow bandwidth at such a shallow cutting depth?

cheers and thanks in advance,
Iain

edit: preferably passive, but active will be an option if parts count is kept low.

[PRR]

Active filter.

[iainpunk]

Active filter.

do you have something specific in mind?
thanks in advance.

cheers

[ElectricDruid]

What about a passive wien bridge network? Again, it probably goes too deep for you, but it's less complicated that the Twin T. Couldn't you degrade the performance by mixing some straight signal back in afterwards, part-bypassing the notch.

[iainpunk]

damping a deeper cutting filter, does that keep the shape while being less deep?
would something like this work?



my back of the napkin calculations say its -6.2dB at the center point

thanks and cheers,
Iain

[antonis]

so i want to cut 6dB of 4.1kHz, as narrow as possible.

n^th order Eliptic filter..??

[iainpunk]

Eliptic won't do what i want, the ripple is to big and they are generally in a LP or HP form, i have yet to come by a Eliptic BS (notch) filter.

does anyone know a good learning source for LTspice? just downloaded it after having decided that i need a better circuit analyser than the one in my head, hahaha

cheers
iain

[Rob Strand]

Here's a scaling scheme which increases the Q of a Twin-T notch filter.
I worked it out a few years back.





To get higher Q's than that you need an active filter or an LCR filter.

FYI, the splitting is done in order to develop the method.  Obviously you can combine the parallel R and C in the legs and use a single part.


For a finite notch you can use a Bridge-T but the Q is lower.  It's possible to use the scaling scheme on a bridge-T as well.   Off-hand, and don't quote me on it, the maximum Q for a scaled Bridge-T is 1/(2*sqrt(2))  = 0.35.

You have to be very careful about what you mean by Q with notches.  With an infinite notch it's fairly clear.  With a finite notch you can use the same Q as the infinite notch *or*  you can use the band-pass like Q.    A finite notch can be seen as the inverse of the peaking equalizer.  The peaking equalizer has a Q but it's not the same as the Q you get for the infinite notch.

[iainpunk]

Here's a scaling scheme which increases the Q of a Twin-T notch filter.
I worked it out a few years back.





To get higher Q's than that you need an active filter or an LCR filter.

FYI, the splitting is done in order to develop the method.  Obviously you can combine the parallel R and C in the legs and use a single part.


For a finite notch you can use a Bridge-T but the Q is lower.  It's possible to use the scaling scheme on a bridge-T as well.   Off-hand, and don't quote me on it, the maximum Q for a scaled Bridge-T is 1/(2*sqrt(2))  = 0.35.

You have to be very careful about what you mean by Q with notches.  With an infinite notch it's fairly clear.  With a finite notch you can use the same Q as the infinite notch *or*  you can use the band-pass like Q.    A finite notch can be seen as the inverse of the peaking equalizer.  The peaking equalizer has a Q but it's not the same as the Q you get for the infinite notch.

definitely not going for an infinite notch, as i need no more than 6dB of cut before oscillations and resonance start to peek around the corner a bit to much.
its all part of a fancy feedback system, meant to linearise an ''hifi'' amp and speaker combo, but we wanted a bit more 4kHz brilliance, and a normal twin T gave a bit to wide of a boost. Too much of the harsher 1kHz range was pushed along with it with the lower Q filters.

thanks, ill go with an active notch filter in this case. im looking in to Gyrator circuits, this thing is already way over-engineerd anyways.

chees, Iain

[ElectricDruid]

does anyone know a good learning source for LTspice? just downloaded it after having decided that i need a better circuit analyser than the one in my head, hahaha

There's a great website called "DIYstompboxes.com" where loads of helpful people use LTspice all the time and are happy to answer questions on it?

[Rob Strand]

definitely not going for an infinite notch, as i need no more than 6dB of cut before oscillations and resonance start to peek around the corner a bit to much.
its all part of a fancy feedback system, meant to linearise an ''hifi'' amp and speaker combo, but we wanted a bit more 4kHz brilliance, and a normal twin T gave a bit to wide of a boost. Too much of the harsher 1kHz range was pushed along with it with the lower Q filters.

thanks, ill go with an active notch filter in this case. im looking in to Gyrator circuits, this thing is already way over-engineerd anyways.

For narrow (high Q) and only -6dB cut you will need to go active for sure.   There's active "Q-enhanced" versions of the twin-T and bridge-T.  For the twin-T you get a finite notch by passing both Twin-T ground connections through a single resistance to ground.   The Q-enhanced feedback adds this resistor for free.     Gyrators will also work since they are essentially LCR filters.

It's not actually over-engineered there's no other way I know of to do it!

FWIW, some of the more recent speaker sims for guitar use narrow finite notches.   Each of the three types of notches have been used since there's very little between the choices.

[phasetrans]

does anyone know a good learning source for LTspice? just downloaded it after having decided that i need a better circuit analyser than the one in my head, hahaha

cheers
iain

Hey Iain, what you want is ltwiki.org and to join the mailing list at groups.io (ltspice@groups.io)

There's basically an unlimited amount of depth to the tool.

You will get well acquainted with cntrl-right click

[Rob Strand]

You will get well acquainted with cntrl-right click

Ah yes, there's a lot of annoying hidden behaviors in Ltspice.   Another one is how to copy from one schematic to another.  There's many more.

I've found the help pages by Gabino Alonso on the Analog Devices very useful for some of this hidden stuff.    Without that you will be stuffing around for hours trying to work out how to do it.

[ElectricDruid]

Ah yes, there's a lot of annoying hidden behaviors in Ltspice.   Another one is how to copy from one schematic to another.  There's many more.

A straight google search is good for most LTspice problems too, since it's been around for ages and it's as awkward as hell, which means about a million people have already asked the question!

[PRR]

....does anyone know a good learning source for LTspice? just downloaded it after having decided that i need a better circuit analyser than the one in my head, hahaha....

Don't doubt the one in your head. SPICE is an idiot.

Installing and using LTspice....

[iainpunk]

....does anyone know a good learning source for LTspice? just downloaded it after having decided that i need a better circuit analyser than the one in my head, hahaha....

Don't doubt the one in your head. SPICE is an idiot.

Installing and using LTspice....

just learned that if i get the job i applied for, ill get a MultiSim license so learning Spice would be a bit redundant.

Also Paul, you're right as usual, the circuit sim in my head was right, you can have a higher Q on a shallower filter if you mix the input back into the filtered signal, this doesn't change the ''sharpness'' of the filter.
i can totally get away with a simple Bridge-Tee filter where the bridge resistor is a resistive divider!

thanks for the help and cheers!!!
Iain

[PRR]

> get a MultiSim license so learning Spice would be a bit redundant.

"SPICE" is a back-end FORTRAN/c black box we never talk to directly.

pSpice, nSpice, MultiSIM, LTspice, Nutmeg, etc etc, are SPICE packaged with user interfaces. They all do about the same thing. One difference is how much Shift/Alt/Ctrl finger-twisting is needed for which immemorable command. Another is the depth of the stock library and how well the company addresses issues.

[Rob Strand]

A straight google search is good for most LTspice problems too, since it's been around for ages and it's as awkward as hell, which means about a million people have already asked the question!

Very true.

Despite its quirks I still think LTspice is a great tool.    It also runs quite well under Linux using Wine.  From a developer's perspective making a free product I can understand how the quirks get in.   It's not like they had a team of people working full time on it for years like the commercial Spice softwares.

[iainpunk]

the main thing is that i was taught MultiSim in school and am quite versed in the software, the head engineer graduated 3 years before me from the same school, so he introduced MultiSim in to the company.

cheers

[Rob Strand]

the main thing is that i was taught MultiSim in school and am quite versed in the software, the head engineer graduated 3 years before me from the same school, so he introduced MultiSim in to the company.

If you are familiar with something there's little motivation to change unless you need specific features.

I used the original MicroSim (now Cadence) for years.   When you dig deeper into it it's extremely powerful and has many features not readily available in other packages.   However, 90% of people would not have a use for these finer details.    I use parameters all the time (see my example in this thread) and I was very happy that LTspice supported it.   LTspice also supports "soft" models where you can easily change the model for a circuit element like a transistor without deleting it and reselecting a new transistor from the library.   For example you just change the model text from BC547A to BC547B instead of deleting the component then finding the BC547B part from the list of a zillion transistors out of the the library each time.

The basic stuff is fairly similar across those simulators.