Twin T Notch Filter Anomaly

Presumably the filter presents a strange frequency-selective impedance to the output of whatever you connect it to - perhaps the output impedances of your oscillator and follower are not as low as you thought? You could try feeding the filter via an attenuator to see if that reduces interaction.

Incidentally, 'bridged T' is a different configuration from 'twin T' although a notch filter can be made using either.

Chris

It sounds like your notch isn't where it is supposed to be. If you are interested, I can send some data for creating a notch filter using an inductor, two caps and a pot. It will create notches in the 60db range. If the inductor is made adjustable, you can dial it in exactly where you need it. The inductor will typically be in the mH range at 2k . Almost anything you have in the junk box can be made to fit. bg

Yes, that's a good idea, I'll try a 20dB attenuator and see if there is any observable difference.

Yes, brain fart on my part ;-)

Cheers

Ian

That was my first thought but I checked the response and the notch is at

2075Hz, 4KHz is 11dB down, 6KHz is 7dB down and so one. Pretty much as expected.

If you are

60dB is plenty for what I need. My HP wavemeter can reliably read components over 70dB down so if I can reduce the fundamental even by 50dB I should be able to read harmonics down to -120dB.

That is VERY interesting, especially as it is adjustable. Can you email me details to ianbellATukfsnDOTorg please?

Cheers

Ian

Why not post it to alt.binaries.schematics.electronics for us all to see?

Chris

need

anything

Check back here and I'll post an address for my webspace as soon as I get it uploaded. My isp doesn't do binaries. bg

You will need the text document and schematic

The text file

The schematic and sweep

Optional Spreadsheet see the doc

Hope it downloads!

There's a bit more on this type of filter at

and an analysis of the twin capacitor one at

I have used the single-capacitor form successfully for a listen-through filter at about 145 MHz. This used a single, centre-tapped air-wound inductor, a variable capacitor across its ends, two taps on the coil for input and output, and a tiny cermet preset variable resistor between the centre-tap and earth/ground. This allows the loss resistance of the coil to be 'cancelled out', yielding the effect of an enormous Q over a small frequency range.

Chris

The twin tee is best and works better if driven from a very low impedance and has a very high impedance load. So use voltage followers on the input and output. A nasty trick is to take a percentage of the output from the output's voltage follower and use that to drive the common (reference) point of the tee. That sharpens the notch considerably.

At the moment it feeds my HP 3581A wavemeter which has an input impedance of 1Meg so the load would seem to be OK. Source impedance is another matter.

Yes, but this will need extremely low distortion in order not to invalidate the results but I guess that is not too hard these days at unity gain ;-)

Yes, I think I saw that in J. Linsley-Hood's book

Cheers

ian

Nothing personal, but that is a pretty sloppy notch for my typical uses. For me a 60 dB notch at 2 kHz; 3 kHz and 1.5 kHz are not more than 1 dB down.

standard

readings

need

anything

It doesn't quit fit into the brickwall category does it? bg

NP. It is completely passive and that is exactly the sort of perfomance a passive twin T gives. You can, however, considerably sharpen the notch using feedback and obtain the sort of figures you quoted.

Cheers

Ian

No, but then you canna change the laws of physics but if you add an op amp and some feedback you can improve a twin T no end.

Cheers

Ian

... but please do take a look at that LC bridged-T circuit I posted about earlier - it can offer a surprisingly narrow notch because of the effective cancellation of the loss resistance of the inductor at and around the notch frequency.

Chris