ARS Dives Into Crystal Filters

 

 

In our recent article "ARS Explores the AIM 4170 Impedance Analyzer," we took a look at some the unusual features of the AIM4170. We felt that one of the best aspects of this box is its ability to characterize crystals. In one swoop, the AIM delivers a set of numbers that can power several powerful filter design programs. Since crystal filters are probably the single biggest hurdle to homebrewing radios, AIM paves the way to inventing and constructing all sorts of interesting radio projects.

 

This article (part two of a two part series) concentrates on designing and building the filter. You will be startled to discover how straightfoward and satisfying it is.

 

I am aware of four programs that can design crystal filters:

 

1. a primitive program in Hamcalc that facilitates the design of Butterwork and Chebyshev filters (although I haven't tried it, this program seems almost irrelvant in comparison to its competition).
2. a  DOS program, written by Wes Hayward, W7ZOI, which is included in the diskette that accompanies Introduction to Radio Frequency Design.
3. a Windows program, also by Wes Hayward, which is included in the CD packaged with Experimental Methods in RF Design.
4. the AADE Filter Design and Analysis program, which can be downloaded from the Almost All Digital Electronics website.

 

The amazing thing is that all four of these programs are free. Most of my experience has been with Wes' DOS program, but I've fiddled around enough with programs 3 and 4 to be impressed.

 

As usual, the most accessible technical background for crystal filter design is Experimental Methods in RF Design. (See section 3.4)

 

I'm not going to drag you through the details of operating these programs. They are all explained in the help files. But I will mention one little thing about Cohn filters. For a long time now, Cohn filters have been the overwhelming favorite for homebrew radios. That's because they are simple and have remarkably low insertion losses. Programs 2 and 4 automate the design of Cohn filters, but number 3 currently does not. However, the solution is simple. Just look for the k and q data in Table 5 of section 3.4 of Experimental Methods in RF Design.

 

Of course, the first thing is to choose a filter frequency. That decision will be driven by your choice of an IF frequency (one of these days we'll write an article on that too) and by the standard frequencies of your favorite type of crystal. (Once again, I will plug the Epson CA-301 crystal. It is so much better than the other stuff you will find in the Digi-Key and Mauser catalogs.) Then you need to buy a mess of them. And finally you will measure the frequency of all of them and characterize a reasonable number of them.

 

Here is a spreadsheet I prepared for two Epson crystals, with nominal frequencies of 9.83 and 10.6244 MHz. The data here was gathered by the G3UUR method I described in Part 1 (and which is also described in section 3.4 of Experimental Methods in RF Design). If you use the AIM4170, your data will be much more simple and direct. But the main point is that it is necessary to average the results of a reasonable number of characterizations, so that you can feed the crystal design programs with average data.

 

 

 

Then you will sort your crystals by measured frequency, so that you can combine compatible crystals in each filter. I put my crystals in a little plastic box (which, in a former life, housed dry files). You will notice the handmade chart that documents the frequency range held by each little compartment.

 

 

 

Next, you will design the filter, using one of the programs listed above. Here's a 6 pole SSB Cohn filter I designed, using the program that comes with Introduction to RF Design. You will instantly see that Cohn filters are really easy to construct, and are ideal candidates for the venerable "ugly" or "dead bug" method of construction.

 

 

Then you will either trust to luck and install the filter in your project, or you will use some method for examining its performance. Once again, Experimental Methods in RF Design is full of lots of good ideas for measuring your creations. I used four methods for assessing this filter: the reflection coefficient response with the AIM4170, the S21 reponse with the TAPR/TecTec VNA, the response on an HP spectrum analyzer using a noise generator as a signal generator, and the response on a Tektronix spectrum analyzer using a tracking generator as a signal source.

 

Another word about Cohn filters. The good news: they are simple and absorb very little of your precious signal. The bad news: the "passband" (the shape of the filter at the top) is pretty darn raggedy, especially in the case of a six pole filter like this one. It's hard to know how much practical difference this makes, but it isn't pretty.

 

Here is what the reflection coefficient looks like on the AIM4170, which is a strong hint that messy, complicated things are happening in the passband:

 

 

The story is even more dramatic in Smith chart format.

 

 

Now, let's take a look at the passband with the TAPR/TenTec VNA, at 5 dB per division. This ought to be interesting to a lot of you, because it's one of the few affordable network analyzers on the planet. The uppermost "horns" on both sides of the passband are a famous attribute of Cohn filters. (I don't know if the lower horns are real or artifacts.) In my experience, crystal measurements with this VNA are challenging. You have to make slooow scans and gather lots of data. But the results, both in terms of insertion loss and the overall shape of the passband, seem to be reasonably consistent with the other measurements presented in this article.

 

 

Next, we'll look at what a farily modern HP spectrum analyzer produces when it's driven by a broadband, lab quality noise generator. Not too bad, but at this small bandwidth, the "noisy" quality of noise degrades the detail, even with a lot of averaging. The bright horizontal line is "reference" (the signal produced by the noise generator when connected directly to the spectrum analyzer).

 

 

 

Finally, we have the results of a "classic" Tektronix spectrum analyzer with a tracking generator. Perhaps the most useful graphics of all. The first image is 10 dB per division (with the reference being set at -40 dBm), the second is a "close up" at 2 dB per division (with the reference being set at -44 dBm), and the third is a "far view" (with the reference being set at 0 dBm), which allows us to see the "stop bands" of this filter. Looking at the stopbands in this manner would have been an impossible task for the TAPR/TenTec VNA (because it lacks the required dynamic range) and a difficult task for the HP, because the noise generator doesn't produce enough power in the required bandwidth.

 

 

 

 

Interesting stuff, isn't it! The great news is that there is now an easy way to characterize crystals and a number of free design programs ready to leap into service. You are by no means limited to Cohn filters. The AADE and Hayward programs give you lots of intriguing options. Go for it!

 

Russ Carpenter, AA7QU

 

 

 

 

 

 

 Create your own free wiki!HelpPublic, Silver wiki Privacy Policy RSS