An Intriguing Look At Crystal Filters. A Special Report from The ARS Lab
By Russ Carpenter, AA7QU
For The ARS Lab
There is good news in the realm of crystal filters. Filters are fun to design, computer crystals are cheap, and with just a little stretching you can build filters that are better than any commercial rig on the market.

This article will set the stage. We'll take you on a tour of the filters in familiar transceivers, to help you distinguish among the good, in-between, and ugly. In a later piece, The ARS Sojourner will teach you how to build your very own world-class filter.

It's not really practical to remove a filter from a transceiver and test it as a single, isolated component. Therefore, our testing will concentrate on the composite AF and IF response of the receiver. The IF response is, of course, largely determined by the quality of the IF filter. The AF response is affected by all kinds of things, which may or may not include intentional efforts to shape the receiver's AF response.

We'll start with the goofs. Sad to say, there are some IF/AF responses in well known radios that are distinctly suboptimal.

The worst is the TenTec Argonaut II. It is fashionable to treat this radio as a QRP classic, but in reality its "Jones Filter" is the weakest performer we have ever tested.
The graph shows the IF/AF response with four different positions of the filter bandwidth control—12 o'clock, 11 o'clock, 9 o'clock, and 7 o'clock. What's wrong with picture?
  • Good responses have a symmetrical, "brick wall" appearance. These are about as asymmetrical as responses can get.
  • There is far too much response in the low frequency, "bass" region.
  • As the filter is narrowed, the center of the passband changes frequency.
  • There is too much insertion loss in the narrow settings.

Next is a really interesting case. It is the IF/AF response of the Kenwood TS 870, which uses DSP throughout the radio. (We cheated a little, because this is an IF DSP filter, not a crystal filter.) The curves are for nominal bandwidths of 1000 Hz, 600 Hz, 400 Hz, and 100 Hz.
The responses look perfect. Symmetrical, steep sided, centered on each other, and with minimal insertion loss. In reality, we found the two narrowest settings to be plagued by ringing. In our opinion, they were not usable. We checked the Internet and found similar comments by other owners of this radio.

The next radio is one we respect and have used incessantly. Nevertheless, we have always felt that the adjustable filter of this box—the Wilderness Sierra—was the Achilles Heal. The graphic below will show you why. The responses are with the bandwidth setting at maximum, 1/4 closed, 1/2 closed, and 3/4 closed.
As you can see, the shapes in all four cases range from awful to pretty bad. In addition, the passbands are not lined up as well as they should be. Insertion losses seem excessive.

Our next case is typical of inexpensive QRP rigs—the SW-20 from Small Wonders Lab.
Even though the "walls" are not very steep, the response is nevertheless useful and reasonable for a simple radio. We will point out, however, that computer crystals only cost about 80 cents each. With a few cheap parts, you can do much better.

Our final example comes from the ever interesting K-2 from Elecraft. This radio fixes almost everything that was wrong with the Sierra's adjustable filter. The responses are at nominal settings of 1500 Hz, 700 Hz, 400 Hz, and 100 Hz.
In truth, the 1500 Hz response isn't so great. But the others are very good. Steep sided, reasonable shapes at the top, good centering of the passbands, and good insertion losses. Our experience with the K-2 also indicates that ringing with these filters is about as tolerable as it gets. Most people love using the K-2's filters, and now you know why.

Stay tuned for our follow up article on crystal filters. Soon, you will be a virtuoso with these fascinating devices!

****

Here are some important links:

The lab's goals and equipment.

The research materials we use in preparing these reports.

The ARS comment page

russ@natworld.com