EMRFD Message Archive 2089
Message Date From Subject 2089 2008-09-19 22:25:55 kerrypwr Crystal Filter Failures
expended) to design and build a crystal filter.
Specs are; 5MHz, Gaussian-to-6, 3000Hz BW.
All my attempts have failed. The results of each attempt are fairly
consistent; the upper section of the passband is "inverted". The top
of the passband is concave instead of convex.
I have re-tested my crystals and I have used capacitors that are very
close (perhaps 1 or 2%) to the design values. This has involved
measuring with my AADE meter and, in some cases, combining two
capacitors to get a value.
I have created a folder named Crystal Filter in the Files section; it
contains the XLAD design, the crystal parameters (I used the Resonance
method ones for design) and the results of testing the latest disaster.
Would anyone with more brains than me (that should be everyone :) )
please have a look at this and tell me what I'm doing wrong?
I think I must be using XLAD wrongly but I can't see what the problem is.
I'm exhausted from fighting this thing and quite despondent; if anyone
has a set of golf clubs they'd like to swap for an electronics
workshop, please let me know.
Kerry.2091 2008-09-20 02:16:54 rotfunkblau Re: Crystal Filter Failures
I just tried to design your filter with the AADE filter design
program, it also failed. I can design butterworth and other shapes,
but not Gaussian, even with higher number of xtals it is not possible.
Maybe your xtals are simply too good for Gaussian. Does anyone has a
better idea?
Best Regards
Guenter
dl5sdc
2092 2008-09-20 02:16:59 rotfunkblau Re: Crystal Filter Failures
I just tried to design your filter with the AADE filter design
program, it also failed. I can design butterworth and other shapes,
but not Gaussian, even with higher number of xtals it is not possible.
Maybe your xtals are simply too good for Gaussian. Does anyone has a
better idea?
Best Regards
Guenter
dl5sdc
2093 2008-09-20 02:39:08 kerrypwr Re: Crystal Filter Failures
XLAD designs it OK although 9k termination is required.
GPLA gives a reasonable passband shape but I can't reproduce that in
the real world.
I have just tried the design in the AADE program; it has provision for
Gaussian-to-6. The Analyse function gives the same "upside-down"
effect that I see in a real filter.
Can anyone tell me what is happening here?2097 2008-09-20 13:23:08 Wes Hayward Re: Crystal Filter Failures
Wow, sorry for all of the pain that you have had to endure. Really,
XLAD was not intended to be an instrument of torture.
I took a look at the filter you were designing and essentially
concluded that this is not a practical design. First, 3 kHz
bandwidth is so wide as to be taxing what one can do with a simple
ladder filter at 5 MHz. This goes back to the work published over a
decade ago by Bill Carver, and before that by Jacob Makhinson. And
those guidelines were for a Butterworth or Chebyshev filter where the
3 dB bandwidth is not that much different than the bandwidth a few dB
further down. The Gaussian to 6 dB filter is worse. It will have
a 6 dB BW that is half again that at 3 dB. So your 3 kHz wide
filter will be 4.5 kHz wide at the 6 dB points where it makes the
transition from a Gaussian shape to a more Chebyshev like skirt
response. That is going to be a really difficult filter to
realize and I can believe that you encountered problems. Sorry for
the pain! This same filter might be much easier to realize with
crystals at 10 MHz or higher.
There is some "good news" this here. You were doing the right
things with GPLA. That is, you were driving the program correctly.
I went through the design with the latest version of GPLA and XLAD.
I now have routines in there that will tell you the minimum R needed
to terminate a filter and I ended up with a termination of 6100
Ohms. This is a pretty high impedance and is going to be difficult
in practice. Even though the filter simulated OK in GPLA, it was
extremely asymmetric in shape.
I have not seen the inverted passband that you describe, but can
believe that you got there.
There is a simple way to examine a crystal filter and see if it might
be possible or not with crystals that you want to use. Start by
designing the filter using an arbitrary, very low crystal parallel
C. I used 0.1 pF. The resulting filter I got looked good in GPLA
with a generally symmetric shape, but with the mentioned 4.5 kHz 6 dB
BW. But the shunt coupling capacitors were often very close to the
crystal actual parallel C of 4.25 pF. This tells me that this is
not going to be a practical filter, at least with XLAD.
In contrast, I used your crystal parameters and got a reasonable
looking (in GPLA) 1 kHz BW filter. One at 2 kHz BW was getting
pretty ugly.
I've loaded two files in the Files section. One is the design,
titled xlad_g-to-6_20sept8.jpg. The other is gpla_g-to-
6_sept8.jpg. You may note that the versions of the programs that
I'm using have some new features, there has been no change to the
basic guts of the program. So if you have XLAD and GPLA from the
EMRFD CD, you have the ability to do what was done here.
One final comment: If you have questions about operati2098 2008-09-20 19:50:45 Allison Parent Re: Crystal Filter Failures
I had similar experience before using the filter programs and
by actual experiment concluded that wide filters using lower
frequency crystals are hard to realize. At 8mhz they worked
but not at 5 or lower. At 9-20mhz it was much easier. So
right now if I want narrow SSB filters (less than 2.4khz) I
use 8-10mhz and if I want filters that are wider I go up in
frequency. Favorite frequencies are 8, 9, 9.6, 11, 12 and 16MHz.
Those are easy frequencies to get good series resonant crystals
for cheaply.
To the thread originator.. Try the same filter with 12mhz
crystals.
Allison
2099 2008-09-20 21:00:44 kerrypwr Re: Crystal Filter Failures
Please don't apologise; it's not XLAD's fault. XLAD is just a tool
and, as I suspected, the fault lies with the user, not the tool.
I was beginning to suspect that it was a bandwidth problem but I got
really frustrated yesterday after a lot of careful & patient work
produced another failure so I thought I'd seek advice on the Group.
"Elmers" are non-existent, at least in respect of EMRFD-type work,
where I am.
Your explanation was, as ever, crystal-clear. (Sorry for the pun!!!)
It covers more ground than other papers on filter design that I've
seen in that you have illustrated the importance of the slope of the
BW plot. I felt fairly certain that I was pushing the limit with the
5MHz design but I couldn't see just why; now I do.
My first (and usual) instinct was that I was doing something wrong;
perhaps somehow misusing XLAD.
You would have got the same XLAD plot as I did; not pretty but I
thought I might get away with it. If you want to see the "real-world"
result go to the Excel file Test Results in the Crystal Filter folder
and look at the plot.
Now that I understand a little more, my confidence is renewed; I will
hold-off on the golf clubs proposal for now!!
Thanks again.
Thanks also, Allison. I have been at the bench testing a few 10MHz
crystals I have; they are el-cheapos and Q is only in the 60 000
region. I will have to see if I can find some decent ones and 12 MHz
might be a good target.2100 2008-09-21 11:46:13 timshoppa Re: Crystal Filter Failures 2101 2008-09-21 14:34:26 kerrypwr Re: Crystal Filter Failures
What you said is what I have learned; I had to learn the lesson the
hard way, on the bench!!
The filter is for my version of the Measurement Receiver; it will be
more elaborate than Wes' design.
I built a very nice 200 Hz Gaussian-to-6 filter as in Wes' design and,
flushed with success, decided to include a wide filter.
The actual bandwidth isn't important for my purposes; the important
part is knowing what the bandwidth actually is and what its shape is.
The final design will have narrow (c.200Hz), wide (3-5kHz) and very
wide (50-100kHz for noise measurements) filters. Subject to testing,
the very wide bandwidth might be achieved with a simple tuned circuit.
I will now change the design to use a 10 or 12 MHz IF.2102 2008-09-22 05:45:04 timshoppa Re: Crystal Filter Failures 2104 2008-09-22 17:30:32 kerrypwr Re: Crystal Filter Failures
I'm awaiting delivery of some crystals; I hope they test better than
the el-cheapo 10MHz ones I had. They are surplus stock from a quality
manufacturer.
They are an odd frequency, about 11.9MHz; that (I hope) is why they're
cheap.
My receiver is like Wes' in that it uses a signal generator as the
VFO. Mental/calculator arithmetic will be required to allow for the
"odd" IF but, since filters built with even-integer crystals
(5.000000MHz etc) don't come-out exactly on frequency anyway, that
would always be necessary.
I could actually have widely-different centre frequencies for each
filter (like 5MHz for one, 10MHz for another etc) and just adjust for
different IFs but I'm using a diplexer after the mixer to get "clean"
performance.
Different IFs might also cause difficulty in the IF amp & detector; I
think that using the same IF (+/- a little for each filter) will cause
less pain.
I can live with a BW of more than 200Hz; anything up to, say, 1kHz
will do as my "narrow" filter. The important thing is good shape, a
clear peak and known characteristics (including noise BW); as long as
centre frequency & width are known, it doesn't matter what they
actually are.
This receiver will not be used for reception of radio signals; it will
be a test instrument.2106 2008-09-22 19:33:58 ha5rxz Re: Crystal Filter Failures
I am learning). If it is so difficult to design a filter for SSB
working at 5 MHz then how did Elecraft do it on the K2? This rig uses
a 4.9152 MHz I.F. and they even made the filter variable width!
HA5RXZ2107 2008-09-22 20:26:11 kerrypwr Re: Crystal Filter Failures
> working at 5 MHz then how did Elecraft do it on the K2? This rig usesA good question; it made me think!!
> a 4.9152 MHz I.F. and they even made the filter variable width!".
I looked-up the K2 Manual;
http://www.elecraft.com/manual/ELECRAFT%20K2%20Owner%27s%20Manual%20Rev%20G%20WEB.pdf
It contains schematics and explanatory notes; to their great credit,
it seems that Elecraft keep no secrets.
The filter shape is controlled by capacitors; Varicap diodes in fact.
Very clever!!
The explanatory notes contain the following;
"A 5-pole variable-bandwidth crystal filter is used on CW (X7-X11).
This filter is optimized for use at low bandwidths (~200 to 500 Hz),
but can be set both narrower and wider as needed with only a small
additional loss. The shape factor and passband ripple content are
optimized at around 300 Hz. (On SSB, a separate fixed filter is
switched in; this filter is located on the SSB adapter)".
It seems that the variable BW is only for CW and that it is optimised
for c.300Hz BW. No magic but a very intelligent compromise.
Many filters that work very well in practice are not "perfect"; we
often read about a person building a filter and being very happy with
the on-air results.
In my case I desire a "perfect" filter for a measurement instrument.
My designs have been subjected to most rigorous testing with good test
equipment and the slightest deviation from "perfect" is the cause of
much anguish!!
For instance, Fig 3.44 of EMRFD shows the response of some Min-Loss
filters; these apparently work well in practice but, for my particular
purpose, I am aiming at much smoother passbands.2109 2008-09-23 02:15:16 Leon Re: Crystal Filter Failures
2112 2008-09-23 07:06:05 bobtbobbo Re: Crystal Filter Failures
some commercial filter cans to see what I could learn. One neat trick
was how one company arrived at precision coupling capacitors of
non-standard values. They selected an NPO ceramic disc cap of the next
lowest standard capacity value and ground a little off the top of the
disk, opposite the leads, until they got up to the required value. I
tried it, using a Dremel tool with a very fine grind stone and it
worked very well. It takes practice to remove very little at a time so
you don't take too much off and end up with a higher value than
needed! Also, let the cap cool down after each grinding before
measuring. I use my trusty AADE L/C meter.2114 2008-09-23 07:14:06 k5nwa Re: Crystal Filter Failures
>When I started building crystal filters many years ago I opened upShouldn't that cap be re-sealed to prevent moisture in the future
>some commercial filter cans to see what I could learn. One neat trick
>was how one company arrived at precision coupling capacitors of
>non-standard values. They selected an NPO ceramic disc cap of the next
>lowest standard capacity value and ground a little off the top of the
>disk, opposite the leads, until they got up to the required value. I
>tried it, using a Dremel tool with a very fine grind stone and it
>worked very well. It takes practice to remove very little at a time so
>you don't take too much off and end up with a higher value than
>needed! Also, let the cap cool down after each grinding before
>measuring. I use my trusty AADE L/C meter.
affecting the value? If so what is recommended besides cements like Q-Dope.
Cecil
K5NWA
www.softrockradio.org www.qrpradio.com
"Blessed are the cracked, for they shall let in the light."2119 2008-09-23 18:40:35 bobtbobbo Re: Crystal Filter Failures 2120 2008-09-23 18:40:35 bobtbobbo Re: Crystal Filter Failures