"Pete KE9OA" wrote in message
...

The 1st IF (45 MHz) filter is a 2-pole crystal filter with 15kHz bw
at -3dB, 100kHz at -24 dB. I acquired a 30kHz/-3dB, 120kHz/-40dB
matched pair that I was thinking of substituting in order to widen
bandwidth for DRM. Note that the centre of the 2nd IF tunes across a
5kHz segment of the 1st IF passband so that a 10kHz or wider bw at the
2nd IF rolls off on one side or the other because of the shoulders of
the 15kHz 1st IF. Is this a bad idea for intermod? The stopband
attenuation is going to be poorer out to maybe 50-60 kHz bw but should
be better beyond that, apart from stray coupling due to squeezing in a
pair of filters where one would ordinarily be.

You should be ok...........the main thing that would be affected would be
the close-in IP3.



I am looking for a wider, flatter response around the uniform spectral
density of the common 10kHz DRM channel for better group delay but I'm
not sure that DRM is so very susceptible because there is pronounced and
dynamic group delay inherent in ionospheric propagation. A 1dB ripple is
considered ideal so I imagine that a flat amplitude-frequency and linear
phase-frequency response in the radio have a small beneficial effect on
the rate of successful decoding. There is also a 20kHz wide DRM mode.

The 4-pole cascaded filter is spec'd to have minimum attenuation of 70dB
at +/- 910kHz and 800ohms/1pF terminating impedance. The filter it
replaces is unspecified at +/-910kHz but I would guess that it is 35-40
dB; its term impedance is 560/6 ohms/pF. So the 910kHz image suppression
should be improved by up to 35 dB less strays. Filter loss will increase
by 1.5 dB per spec. If the circuit currently matches the stock filter, is
the ripple and increased loss liable to be severe?

73, Tom

Tom, those specs are fine...........the thing that catches my attention is
the relatively low impedance of the crystal filter. Typically, these
filters have an IN/OUT Z of between 1500 and 7000 ohms. I would surmise
that the current filter has around a 3000 ohm impedance. Take a look at
any series/shunt resistors around the filter area and it might give a
clue.
A series L, shunt C would do the trick if you need to do any impedance
matching.
I have created a dual-purpose spreadsheet that does these calculations; a
second page of this spreadsheet does PLL loop filter calculations that are
based on Tom Wheeler's equation set. This is for the double-ended phase
detector output circuit that uses the Phi V and Phi R outputs.

Pete

I'm interested in your spread sheet, Pete. Can you e-mail it to me?

The specs are as I described. The stock 2-pole filter is the TEW MF45R2
http://www.tew.co.jp/crystal/mcf/e_19.html and the 4-pole (dual cascaded
2-pole) with which I replaced it is the Fox 45F30B
http://www.foxonline.com/pdfs/filters.pdf . Much to my surprise and
disappointment, the improvement in image rejection looked to be at best 2
S-units (hard to tell with the coarse meter on the DX-394), not the 30-35 dB
I was hoping for. Having searched again for the TEW specs, I found that they
had been updated since I last searched and now specify the attenuation at
45MHz-910 kHz (44.09MHz) - it's a guaranteed minimum of an amazing 60 dB.
The 2-pole Fox equivalent is only 40 dB and the matched pair is guaranteed
to be 70 dB minimum. That's only 10 dB better than the stock filter or,
coincidentally, about 2 S-units on an accurate scale. There was also some
loss of sensitivity - not sure how much but no evidence of gross ripple
(just using the S-meter).

I did the substitution with no change in the surrounding LRC's. The filter
looks into a 100pF cap (Xc= 35ohms) in series with a 1k that is paralleled
with the input impedance of the 1st IF amplifier. So we know the result is
going to be under 1k. The source for the filter is a tunable transformer. If
the 1st IF amp input impedance was around 1k, then the stock filter would
see something close to its desired 560 ohms. Increasing the 1k resistor to
3k3 would raise the load impedance to near 800 ohms, the desired load for
the Fox pair.

73, Tom