Call Search
     

New to Ham Radio?
My Profile

Community
Articles
Forums
News
Reviews
Friends Remembered
Strays
Survey Question

Operating
Contesting
DX Cluster Spots
Propagation

Resources
Calendar
Classifieds
Ham Exams
Ham Links
List Archives
News Articles
Product Reviews
QSL Managers

Site Info
eHam Help (FAQ)
Support the site
The eHam Team
Advertising Info
Vision Statement
About eHam.net

   Home   Help Search  
Pages: [1] 2 Next   Go Down
  Print  
Author Topic: Homebrew Crystal Filters  (Read 9852 times)
AJ4SN
Member

Posts: 20




Ignore
« on: March 26, 2013, 02:20:51 PM »

I recently completed work on a homebrew 80-10 meter SSB transmitter/receiver (it shares a common VFO, but has separate front ends, IFs, and filters). I built 9 MHz crystal filters for both the transmitter and the receiver IF, but I have been dissatisfied with the performance of the filters.

If I have a narrow filter in the receiver, the audio is poor. If I widen the bandpass, the selectivity suffers. I see a similar problem on the transmitter. I am using a Flex Radio 1500 to view the transmitter output, and when the transmit audio is satisfactory, I see part of the lower sideband being transmitted. If I try to tighten up the passband, the audio suffers.

I know that it is possible to construct effective crystal filters, but my results are not impressive. I don't have sophisticated test equipment--only a scope and my Flex Radio. I am beginning to believe that I should scrap the homebrew filters and buy a commercial 8-pole filter with a nice shape factor. I have used homebrew crystal filters in simple QRP CW rigs before, and they performed satisfactorily. I just don't know if the average builder such as myself can hope to construct effective filters for a SSB transceiver. Has anyone had experience with homebrew crystal filters similar to mine?
Logged
G3RZP
Member

Posts: 4391




Ignore
« Reply #1 on: March 26, 2013, 02:32:50 PM »

There have been a number of articles published on home brew crystal filters. If you want to build a filter at 10MHz, I have a load of 10MHz crystals that were rescued from the dump when a company lab closed. For 50 crystals, you will need to pay postage from the US (I'll mail when I go to Dayton), or you can collect (or get a friend to collect) from the IARU booth (in the ARRL area) at Dayton. Or just make a suitable donation to the ARRL Frequency Defence Fund.

73

Peter G3RZP
Logged
WB6BYU
Member

Posts: 13038




Ignore
« Reply #2 on: March 26, 2013, 03:10:56 PM »

Quote from: WA0VPR
...
If I have a narrow filter in the receiver, the audio is poor. If I widen the bandpass, the selectivity suffers. I see a similar problem on the transmitter. I am using a Flex Radio 1500 to view the transmitter output, and when the transmit audio is satisfactory, I see part of the lower sideband being transmitted. If I try to tighten up the passband, the audio suffers.



There is no solution to this, because there are two conflicting requirements.
If you want "full" wide bandwidth audio, you can't pass it through a narrow
filter.  If you have good low frequency response then there isn't enough
filtering to remove the carrier and part of the opposite sideband.   That's due
to the limitations of crystal filters built with real world parts.  You can't get
a lot of audio through at 400Hz while suppressing the carrier (only 400Hz
away) and the opposite sideband (800Hz away) with a typical filter roll-off.
(Asymmetric filters may be better in this regard than symmetric ones.)

For good audio "punch" you want to emphasize the speech frequency range
that contributes the most to communications effectiveness - that usually
means filtering out the low frequency components.  The point is to be
effective, not to sound "natural".  That allows you to put more power into
those portions of the audio spectrum that contribute to intelligibility at
the far end.  The typical 2.1kHz bandwidth for SSB intentionally sacrifices
some audio range for better selectivity.

For SSB to sound good the BFO frequency has to be carefully positioned relative
to the filter passband (in both the transmitter and the receiver), or you clip too
much on the high or low end of the audio.  The narrower the bandwidth, the more
important this is.


Meanwhile there is a lot of information on homebrew crystal filters available from
folks such as G3UUR, W7ZOI, and others.  Here is a useful summary page that
might point you in some useful directions:

http://www.qrp.pops.net/crystal-tester-2010.asp

With some fairly simple measurements of crystal parameters and some math you
can build pretty good filters, especially making use of some of the software that
is also available.
Logged
W9GB
Member

Posts: 2600




Ignore
« Reply #3 on: March 26, 2013, 03:13:42 PM »

Larry Benko, W0QE has this article on Crystal Bandpass Filters
http://www.w0qe.com/Projects/crystal_bandpass_filters.html

Cohn-type crystal ladder filter
http://epic.mcmaster.ca/~elmer101/cohn.html

INRAD sells two 9.0 MHz crystal filter kits (#350, #351) for DIY filter builders
These low cost filter kits include four specially cut and color coded crystals along with five surface mount capacitors. Center frequency is 9000 kHz and impedance in and out is 200 ohms.

600 Hz, 9000 kHz 4-pole crystal filter kit
http://www.inrad.net/product.php?productid=239&cat=149&page=1

2400 Hz, 9000 kHz 4-pole crystal filter kit
http://www.inrad.net/product.php?productid=240&cat=149&page=1
« Last Edit: March 26, 2013, 03:31:28 PM by W9GB » Logged
VK2TIL
Member

Posts: 310




Ignore
« Reply #4 on: March 26, 2013, 03:21:28 PM »

It's possible to design very good crystal filters but it takes some effort; the crystal parameters are important.

Many simple projects use a few crystals of unknown specification; sometimes it works, sometimes it doesn't, sometimes the builder is unsure.

Crystal parameters can be measured with simple equipment; a signal source and a detector.  You have those in your radio and 'scope.  Suitable attenuators will embed the crystal-under-test in a 50-ohm or 12.5-ohm environment (see the K8ZOA paper referenced below).

The G3UUR method doesn't even require that, just a simple oscillator, but it does require a frequency counter.

These methods are somewhat tedious if you are sorting large batches of crystals; better equipment speeds the process but is not essential.  It's usually necessary to measure a lot (20 - 50) of crystals to sort them into reasonably-matched sets for filters but, if you want good filters, that's what you have to do.

Jack Smith K8ZOA has written the best paper on crystal measurement that I know; go to http://www.cliftonlaboratories.com/Documents.htm and find the Crystal Motional Parameters paper.

A fixture similar to Jack's is excellent for the series-resonance method; the G3UUR oscillator is simple to build if you want to try that.

Filters can be designed once you have the parameters; the free program from the AADE site is a good one but there are several others available for downloading.

It's perhaps counter-intuitive but good narrow filters (CW bandwidths) aren't difficult to make; the wider SSB filters are quite difficult.  The essential reason is the small frequency difference between the series resonance and the parallel resonance of a crystal; you can only make a filter with somewhat less width than that difference.

The difference increases with crystal frequency which is one reason for using 9 or 10MHz crystals instead of lower-frequency ones.

I've looked in my computer files and found a few articles that are available on the 'net;

www.arrl.org/files/file/QEX.../QEX_Nov-Dec_09_Feature.pdf

http://www.giangrandi.ch/electronics/crystalfilters/xtalfilters.shtml

http://www.w0qe.com/Projects/crystal_bandpass_filters.html

Patient searching using various search terms will find a good deal more.

I've found working with crystals to be fun; I've worked my way through all the measurement methods described by K8ZOA, from simple series-resonance to the final graduation to VNAs.




Logged
AJ4SN
Member

Posts: 20




Ignore
« Reply #5 on: March 26, 2013, 06:07:03 PM »

Thanks for all the great responses to my post. I'm inspired to go back to the drawing board and see if I can improve my results. Really appreciate your willingness to help.

Stan
Logged
KB8E
Member

Posts: 24




Ignore
« Reply #6 on: March 27, 2013, 09:51:35 AM »

Stan,

I have a commercial, 8-pole, 9 MHz, SSB crystal filter that I'll part with if you want it. Been sitting in my junk box for years, but still looks good; I just swept it a month or so ago. I may be able to send plots if I can remember how to capture my network analyzer's screen display.

Sam
Logged
AJ4SN
Member

Posts: 20




Ignore
« Reply #7 on: March 27, 2013, 06:32:09 PM »

Sam,

I'm very interested. What do you want for it?

Stan
Logged
KB8E
Member

Posts: 24




Ignore
« Reply #8 on: March 28, 2013, 03:01:57 PM »

Stan,

I'll donate it to the cause, it isn't doing me any good sitting here. It is a KVG XF-9B. The terminations should be 500 ohms // 30 pF. I'll provide a plot as well.

Sam
Logged
AJ4SN
Member

Posts: 20




Ignore
« Reply #9 on: March 28, 2013, 03:37:06 PM »

Thank you. Can I at least send you something for shipping? Make a donation to the defence fund in your name?
Logged
KB1GMX
Member

Posts: 714




Ignore
« Reply #10 on: March 28, 2013, 03:50:27 PM »

Ok, the usual problem is the shape of the filter.  If you doing the common 4 crystal filter therein
lies the problem.  Those filters if say designed for 2.4KHz to 6db are typically wider than 8khz a
mere 50db down.  yes we are talking shape factors greater than 2 and approaching 3 (ratio of 6db
band width to 60db bandwidth).

Solution use more crystals with maybe more sophisticated filter shape.  I usually start 6 crystals
and have gone as far as 10 to get the steep skirts.  I did do one filter using 5 crystals at 8mhz that
was very good (used very high Q crystals) but it still suffers from fairly weak ultimate rejection
and its in a radio I use for occasional fun.

If the filter has poor skirts or tepid ultimate rejection (most of the 4 crystal fitlers will fail)
then opposing sideband will get by the filter (also happens if the layout or shielding is poor).
Also all of the crystal cases need to be grounded, they are next to each other and can easily
allow unwanted coupling lowering the high rejection skirts to mush.

If you have carrier leaking adjust your balanced modulator (or improve the layout and
shielding of the carrier osc)!

The ARRL press has Experimental Methods in RF Design, EMRFD is very worth having if your
a builder and covers said issues well.


Allison
Logged
WD4HXG
Member

Posts: 182




Ignore
« Reply #11 on: March 28, 2013, 05:20:59 PM »

You stated you are using separate filters for receive and transmit. Are
you using a common mixer in the transmit-receive chain. What type of
mixer is in use? Active, doubly balanced passive? I would expect
the mixer at 9 Mhz to suppress the carrier quite a bit. The Mini-Circuits
SBL-1+ is good for about 45 dB + LO (Carrier) attenuation at 10 Mhz
when properly terminated with a broadband match. If you use the
recommended +7 dBm Lo then the LO bleedthru at the IF should be
no higher than -33 dBm. If using a passive such
as the SBL1+ make sure to use either a diplexer between the filter
and mixer or an attenuator. If the mixer is unhappy with the match at
any port it can cause a lot of mischief.

Also filters can be persnickety about the source and load impedances.
WHat is the input impedance of you filter on the receive side? Is the
filter properly loaded at the output? If a filter is poorly matched the
usually it is typically most evident in the passband. Often the ripple
is really high. But it can also cause decreased ultimate performance
particularly close in to the passband.

Hope this helps.

Chuck

Logged
AJ4SN
Member

Posts: 20




Ignore
« Reply #12 on: March 28, 2013, 06:16:51 PM »

The only thing common to the transmitter and the receiver is a 5.0 - 5.5 MHz VFO. I'm using a separate mixer in the receiver and transmitter. All the mixers and the balanced modulator are the mini circuits ADE-1. I do have a 7dBm injection level for the bal mod. I am currently using four 9 MHz crystals in the filters.  Originally, I only planned to construct the receiver, but then I decided that I needed a transmitter to go with it. Most of the transmitter circuitry--the SSB generator, the VFO, the pre-driver--was taken from Experimental Methods in RF design. The power amp is the homebrew 2010 challenge winner from the June 2010 QST. I puts out about 50 watts on 80 through 20 meters and somewhat less on 15 and 10.
Logged
VK2TIL
Member

Posts: 310




Ignore
« Reply #13 on: March 28, 2013, 07:01:09 PM »

I think that you should consider actually measuring the response of the filters you have built.

Your Flex Radio looks like a useful signal source with either 5 watts or 1mW available.

I did say earlier that your 'scope would make a good detector but, considering the power/voltage levels involved, I don't think that's correct; it would not have the required sensitivity.

An AD8307 meter as described in EMRFD and elsewhere is ideal; I used mine a great deal for this kind of measurement, entering the DUT output at various frequencies into Excel and producing response graphs of filters, amplifiers and the like before I graduated to other test instruments.

Just the basic AD8307 with a DVM on the output is trivial to construct for low frequencies but a kit such as KA7EXM sells would be worth building.

If you measure your filters you will have information that will help you decide what, if anything, is wrong with them and how you can improve them.

Logged
KM3K
Member

Posts: 279




Ignore
« Reply #14 on: March 28, 2013, 08:28:51 PM »

........; the wider SSB filters are quite difficult.  The essential reason is the small frequency difference between the series resonance and the parallel resonance of a crystal; you can only make a filter with somewhat less width than that difference.
Regretfully and with all due respect, the above statements are incorrect.
SSB-filters are quite easy to make and I can not recall that the "small frequency difference between series and parallel resonances" ever entered into any of my design-engineering decisions.
73 Jerry KM3K
Logged
Pages: [1] 2 Next   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.11 | SMF © 2006-2009, Simple Machines LLC Valid XHTML 1.0! Valid CSS!