EMRFD Message Archive 1856

Message Date From Subject
1856 2008-08-02 18:11:10 wimmie262000 DC-compensated notch filter
Hi all,

I am not sure if the following has been done before and has a certain
name. So I thought I would share it.
Recently I was fooling around with SPICE, modeling a direction
coupler. It occurred to me that the matching/mismatch effect of a
directional coupler might be put to use to enhance a notch filter.

What I entered was the circuit here:

The directional coupler provides a signal on the reflected port when
the load on the (normal) output is of different impedance than the
load on the coupled port. So what if we deliberately provide a
mismatch for most frequencies, except for a narrow band? What the
circuit shows is a series resonant load on the output. To investigate
the circuit I used a crystal as a series resonant element.

The problem with a usual single L/C notch filter is that it has a
certain series resistance. The series resistance limits the notch
depth. E.g. a 5 ohm crystal has about a -15dB limit.

The directional coupler is set up to match that series resistance.
When the coupled port has the same impedance, maximum notch depth is
reached. Essentially unlimited, but that would be only with ideal

I have build the circuit, apart from the Rm in the schematic. Rm is
meant to optimize the return loss and provide a proper 50 ohm
impedance. The lower Rs, the higher Rm can be, and the lower the
insertion loss of the whole thing.

The load I tested was a 13.5MHz crystal with about 5 ohm series
resistance. The two transformers of the DC were MiniCircuits TC1-6-75
1:1 transformers I found cheap on eBay. The coupled port resistance
was trimmed with a 100 Ohm multi-turn trimmer parallel to a 10 Ohm

The bandwidth numbers:
* -3 dB / 949 Hz
* -10 dB / 305 Hz
* -20 dB / 79 Hz
* -30 dB / 27 Hz
* -40 dB / 7 Hz
Maximum depth was at -44 dB. And -46 dB when I also added a capacitor
on the coupled port of about the crystal Cp.

Since no Rm was used, the return loss was measured around -16.5 dB (67
Ohm). The wide band insertion loss was around 1.75dB. This varied
about 0.3dB from 1 to 180 MHz. Would a matching Rm be added, the RL
would increase, but also the insertion loss.

All in all I like the result. The maximum notch depth varied a bit
with touching. So I would expect it to go deeper when input and output
are screened. Currently it is an open Manhattan style thing.

I guess with a crystal the circuit could be useful to examine
oscillator phase noise on a spectrum analyzer since the notch is so
narrow. Or to filter out the fundamental frequency to examine
harmonics of a (DDS) generator.

With a normal series L/C one could improve notch Q by using a higher L
to C ratio by using a (higher u) core. The possible increase of Rs due
to core loss can be matched away with the DC. Also the usual low Rs
(lower than a crystal) helps to minimize insertion loss and impedance
mismatch of the notch filter.

Any more thoughts on the usefulness of this circuit?

Joop - pe1cqp
1988 2008-08-25 17:40:43 wimmie262000 QEX Jul/Aug (was DC-compensated notch filter)
1989 2008-08-25 22:17:55 victorkoren Re: QEX Jul/Aug (was DC-compensated notch filter)
I have a subscripti