Hi John,
? ?I appreciate your comments.? I agree that the first option isn't going to fly.? I saw the changes in the noise floor but at this point in my 'crystal kindergarten' I don't grok the significance of the floor.? When I'd first started (this was my 2nd crystal filter ever), I saw the glove...er...ripple and think I learned it was heavily influenced by input/output impedance.? I'd still like to find a way to measure the impedance of filters but haven't figured out how yet.
? ? In the meantime, I've learned? that I didn't gather enough information from those crystals (the holder capacitance, for a start), and I didn't put them in any sort of frequency order.? I have another set of crystals that I have properly measured and modeled in DISHAL.? DISHAL suggests the finished filter will have a 69¦¸ Z with a 2.5 KHz bandwidth.? We'll see.? (Is it correct to model this as an 8 crystal filter, even though the first and last are paralleled?? I hope so as that's what I'm doing... )
? ?I'm interested in how? you arrived at the 400¦¸ estimate - that's intriguing.? I'm also surprised that I keep hearing 2K¦¸ for crystal filters, yet DISHAL suggests low numbers.? Lots to learn.? ?Sleep first. ;)
? ? Best,
? ? ?Andy? (a maintainer's worst nightmare:? an operator with tools)
toggle quoted message
Show quoted text
On Tue, Dec 19, 2017 at 08:49 pm, John Backo wrote:
That's nice, Andy. But look at what happens to your noise floor
in the third instance.
The first instance shows a peaks and lows over a 5+ dB range, perhaps a bit
too much for a good filter.
The second shows a good ripple and a pretty good noise floor. I probably would
go with that one. Note that the G3UUR 8-xtal filter has different characteristics than
those of a plain 4 or 5 xtal ladder filter. You probably should assume (from your
measurements) that the output impedance is about 500 ohms and design a transformer for
that 10:1 match. It's worth a try anyway. Remember that the impedance ratio is the square root of
the number of turns, so your turns ratio would be around SqRt(10): SqRt(1). Try it and see what happens.
john
AD5YE
