Adapting a Commercial Narrow SSB Filter for Homebrew Use: ICOM FL-80

I am a long time follower of the Soldersmoke podcast with Bill Meara N2CQR, Pete Juliano N6QW, and now Dean Souleles KK4DAS.  Pete has been kind enough to correspond a number of times with me on idle curiosity and questions that have arisen from the various topics Bill, Pete, and Dean have discussed.  (Sometimes, these topics actually are of considerable use in my day job .. but I digress.)

Recently, Pete alerted me to the availability on eBay of surplus narrow SSB filters at useful IF frequencies, culled from people parting out broken radios.  One of the best/sharpest of these is a narrow 2.4 kHz (-6 dB point) SSB compatible filter, the Icom FL-80.  Under normal circumstances, this was paired with another narrow SSB filter further down in the receiver chain at 455 kHz, the FL-44A, to enable their famous Passband Tuning (PBT) control.  However, the FL-80 is useful on its own for home-brew applications.

It sports a very nice Q of (9.0115 MHz / 2.4 kHz) = 3,754.  My skills at IF filter construction have a long way to go to reach that level of performance, so on Pete's suggestion, I picked one up - at a bargain price - for RF playing around.

So three things came to mind immediately:

1. How to match the impedance in/out from 50 ohms behind and ahead of this filter?
2. How to come up with a convenient mechanical interface, both for characterizing the filter and for use in home-brew constructions?
3. How to verify that the passband is that narrow?

Let's tackle each in turn.

1. Matching in/out filter impedance

Since this filter came with no documentation, I first tried looking up schematics for rigs like the IC-781 which use this filter in the IF stages.  That ended up being not too informative.  

Upon further reflection, I realized that Inrad makes replacement IF filter stages for a number of radios, including a line of narrow SSB ones at the ~9 MHz IF frequency.  A quick email to Inrad revealed that their discontinued Icom 2.3 kHz SSB IF filter (#110) had an impedance of 1000 ohms, which is not that unusual, so I settled on that as the target filter impedance.

As Pete advised, the match from 50 ohms to 1000 ohms at 9 MHz can be accomplished with a mix 43 ferrite toroid, such as the FT50-43, using a solenoid winding (auto-transformer) with a 9:2 winding ratio.  This provides (81/4) = 20.25 times impedance match, which is almost perfect compared to 1000/50 = 20.

I wound two of these toroids in short order, leaving a loop at 2 turns for the 50 ohm side (the full 9 turns will go to the filter input or output).  My toroids all come from the "Toroid King", Diz at kitsandparts.com.  Handy tip: the toroid calculator also provides an estimate for how much magnet wire you'll need, so you can remain economical.

2. Interfacing the filter

Even though Pete does not completely trust it, I planned to use my NanoVNA-H4 with updated firmware to characterize the filter, which uses 401 points across the stimulus range.  I've compared the H4 to precision grade $$$ test equipment, and find that the Nano is an accurate instrument if you understand its limitations - one of which is significant inaccuracy in measurement if the impedance of the device under test is far away from 50 ohms (either direction, high or low).  Fortunately, this was not going to be the case since we are using toroids to transform impedance.

A few years ago, since my fingers don't always have maximum dexterity, I invested in a very convenient VNA testing jig for the series of bandpass and low-pass filters made by Hans G0UPL at QRP Labs.  The jig was created by Lex PH2LB, and looks like this:

There are convenient, quick jumpers on the board to use when you are calibrating the VNA for various start/stop frequencies (Open/Short/Load/Isolation/Thru) and the 50 ohm load is built in.  You can also see that the In and Out are set up to be perfectly plug-compatible with QRP Labs filters through 4 pin 0.1" spaced headers.  The jig was perfect for my needs, and the mechanical interface would also be useful later when applying the filter in a home-brew setup.

So I ordered a LP filter board/kit for less than $5 from QRP Labs, and it arrived in just over a week from Turkey!  (International FedEx shipping is remarkably quick including customs clearance).  This got me the following bare board:

The second step was to figure out how the board could be adapted for my use.  From the amazingly detailed kit instructions (a hallmark of all G0UPL's work), this is the schematic for the LP filter:

I was going to have no capacitors and only two toroids - L1 and L3 equivalent in the above diagram - with the FL-80 in the middle between them - the 7t here refers to the turns above the 2t, so a total of 9 turns on the filter side:

After a minute or two, I realized things could be hooked up this way (using the kit wiring diagram):

(The two connections for "2t" come about because I twisted a loop in the wires when winding the autotransformer; this way, I could just cut them to length and plug the two wires into the two marked plated holes.)

Out came the soldering iron, and this somewhat hacked-up board emerged:

You can see the two 9:2 wound toroids on either end and the wires going off to the filter in the middle, where C2 and C3's connections were.  (The In and Out of the filter connect between L1 and L3.)

Here's a view plugged into PH2LB's interface.  Yes it's a bit mechanically unstable; I'll try to fix that when putting it to actual use, and shorten up the connection wires where I can.

3. Filter Performance

Now that the board was ready, it was time to characterize it.  Here are the S21 (e.g. through response) results with the NanoVNA-H4, all centered at 9.011 MHz (OK, I was a little off in the center frequency.)

First at 1 MHz span:

Next zooming in to 100 kHz span:

Let's go in further to 50 kHz span:

Finally, 10 kHz span:

The filter looks very good and the markers in the last plot indicate that it is indeed about 2.1 kHz wide at 3 dB down; if I had remembered to check 6 dB down, I would have found it to be very close to 2.4 kHz.  There are some wiggles in the passband but I chalk those up to some of the deficiencies in my wiring, capacitive strays on the toroid, or other non-idealities.  Plenty good enough for experimentation!  

Filter loss is just over 8 dB, which I will take care of in actual implementation through a preamp in this IF filter stage.

Finally, the filter is probably cutting off well below 30 dB on the wings, but the NanoVNA's limitations in measurement are showing here.  It's more than good enough for me to conclude things are working properly.

Conclusion

This approach is a pretty good way to get a spectacularly narrow IF filter for your own RF creations.  It's also good because the Icom people did the hard part!  Yes, one can build your own IF filter, but this is also a nice approach to use and to compare with your own home-brew efforts.

Thanks as always to Pete N6QW for information and wisdom.