K2 modification of the variable BW Cohn CW crystal filter

[NU7B]

I ve been using me K2 (s/n 6770) in the digital modes lately and have wanted to make an improvement in the passband ripple, which is almost 10 dB p-p when the BW is set at 2 KHz, with the RX audio viewed spectrally. So I simulated the filter using the crystal parameters given here:  http://www.arrl.org/files/file/QEX_Next_Issue/Nov-Dec_2009/QEX_Nov-Dec_09_Feature.pdf , using the spice simulator Qucs at http://qucs.sourceforge.net . The simulation agrees with the measured passband when I set the termination impedances between 200-300 ohms, with the ripple being more pronounced as the impedance is lowered.

I found a reference on the web to a modification of the Cohn filter which preserves the equal capacitances and crystal parameters for all elements, documented by G3UUR and described (page 5) here:  http://www.warc.org.uk/filestore/projects/xtal_bpf_203/eDishalHelp.pdf . I haven't gone into the mathematics of this alteration, but the effect of placing a crystal in parallel with X7 and X11 is a de-Q-ing of these two elements and substantial reduction of ripple at the passband edges. This effect is apparently not unknown to Wayne, who uses it in the BFO resonator and considered it for increasing the pullability of the PLL reference. I don't have the 2010 ARRL Handbook, but G3UUR has an article in it describing his modification of the Cohn ladder response.

So - the mod I tried in simulation was to place a crystal in parallel with X7 and X11, and to replace the coupling varicaps D29 and D34 with 4700pF fixed caps. The varicaps D30-33 are untouched. My simulation results:


K2 design:
With Cvc = 44pF:

6dB BW = 2.07 KHz, 60 dB BW = 4.55 KHz, ripple = 9 dB p-p, shape factor = 2.2.

With Cvc = 200 pF:

6dB BW = 0.57 KHz, 60 dB BW = 1.52 KHz, ripple = 1.5 dB p-p, shape factor = 2.67.

Modified K2 design, G3UUR topology:
With Cvc = 56pF:

6dB BW = 2.03 KHz, 60 dB BW = 5.31 KHz, ripple = 3 dB p-p, shape factor = 2.61.

With Cvc = 205 pF:

6dB BW = 0.57 KHz, 60 dB BW = 1.96 KHz, ripple = 0.5 dB p-p, shape factor = 3.50.

So it looks like I get a substantial reduction in passband ripple and a flatter group delay with the modification, at the cost of a bit of increase in the shape factor. If I used the K2 strictly for CW or psk31, I might not want to make a change. But I use wider digital modes like Olivia where group delay distortion can make a difference in BER. So I think I'm going to try the mod. Don has suggested in the past to narrow down the passband when using the digital modes, but this makes enough of a difference above 1 KHz BW (and I see surrounding signals in the waterfall), that this mod is worth a try. I'm curious if anyone on the list has considered it.

Steve NU7B
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Don Wilhelm-4]

  Stephen,

If you are doing mostly digital modes and not too much SSB, you might
want to change the OP1 filter instead.  The SSBCAPKT contains capacitors
to make it either a 1.9k, 2.1k, 2.4k or 2.6k filter.

73,
Don W3FPR

On 7/17/2011 6:28 PM, Stephen Rector wrote:

> I ve been using me K2 (s/n 6770) in the digital modes lately and have wanted to make an improvement in the passband ripple, which is almost 10 dB p-p when the BW is set at 2 KHz, with the RX audio viewed spectrally. So I simulated the filter using the crystal parameters given here:  http://www.arrl.org/files/file/QEX_Next_Issue/Nov-Dec_2009/QEX_Nov-Dec_09_Feature.pdf , using the spice simulator Qucs at http://qucs.sourceforge.net . The simulation agrees with the measured passband when I set the termination impedances between 200-300 ohms, with the ripple being more pronounced as the impedance is lowered.
>
> I found a reference on the web to a modification of the Cohn filter which preserves the equal capacitances and crystal parameters for all elements, documented by G3UUR and described (page 5) here:  http://www.warc.org.uk/filestore/projects/xtal_bpf_203/eDishalHelp.pdf . I haven't gone into the mathematics of this alteration, but the effect of placing a crystal in parallel with X7 and X11 is a de-Q-ing of these two elements and substantial reduction of ripple at the passband edges. This effect is apparently not unknown to Wayne, who uses it in the BFO resonator and considered it for increasing the pullability of the PLL reference. I don't have the 2010 ARRL Handbook, but G3UUR has an article in it describing his modification of the Cohn ladder response.
>
> So - the mod I tried in simulation was to place a crystal in parallel with X7 and X11, and to replace the coupling varicaps D29 and D34 with 4700pF fixed caps. The varicaps D30-33 are untouched. My simulation results:
>
>
> K2 design:
> With Cvc = 44pF:
>
> 6dB BW = 2.07 KHz, 60 dB BW = 4.55 KHz, ripple = 9 dB p-p, shape factor = 2.2.
>
> With Cvc = 200 pF:
>
> 6dB BW = 0.57 KHz, 60 dB BW = 1.52 KHz, ripple = 1.5 dB p-p, shape factor = 2.67.
>
> Modified K2 design, G3UUR topology:
> With Cvc = 56pF:
>
> 6dB BW = 2.03 KHz, 60 dB BW = 5.31 KHz, ripple = 3 dB p-p, shape factor = 2.61.
>
> With Cvc = 205 pF:
>
> 6dB BW = 0.57 KHz, 60 dB BW = 1.96 KHz, ripple = 0.5 dB p-p, shape factor = 3.50.
>
> So it looks like I get a substantial reduction in passband ripple and a flatter group delay with the modification, at the cost of a bit of increase in the shape factor. If I used the K2 strictly for CW or psk31, I might not want to make a change. But I use wider digital modes like Olivia where group delay distortion can make a difference in BER. So I think I'm going to try the mod. Don has suggested in the past to narrow down the passband when using the digital modes, but this makes enough of a difference above 1 KHz BW (and I see surrounding signals in the waterfall), that this mod is worth a try. I'm curious if anyone on the list has considered it.
>
> Steve NU7B
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html