I have a few CW rigs that have lousy filtering. Rockmites and other low part count direct conversion rigs naturally don’t have much in the way of selectivity, but this is also (unfortunately) true of my unmodified Yaesu FT-817nd. I should probably invest in the an IF filter for the 817, but for other rigs, it makes sense to have an external audio filter that I can use across projects. A natural choice would be the New England QRP Club’s NESCAF filter.
The NESCAF filter is designed around a family of versatile switched capacitor filter chips. The grand-daddy of the family is the MF10, but more modern versions exist (e.g., the pin-compatible LMF100 or the LTC1060) and have better performance. Each of these chips has two 2nd order filter stages, which can be configured as low pass, high pass, all pass , notch, or band pass depending on how it is wired up and what external components are used. One stage can serve as input to the other to create 4th order filters.
In the case of the NESCAF, the chip is configured as a Butterworth filter — maybe not the sharpest band edges, but minimal ripple in the pass band. Referring to the MF10 datasheet, it looks like the NESCAF implements this through two sequential mode 1 filters with a Q of 10. The filter bandwidth is controlled by two ganged 50k ohm potentiometers, with a section devoted to each filter stage. The center frequency is driven by dividing a time base by 100. A 555 timer outputing 70 kHz will result in a 700Hz center frequency. The 555 rate is controlled by a panel-mounted potentiometer. The center point of that potentiometer is set an onboard trimmer potentiometer. Output from the filter is amplified by an LM386. Some builders make this an onboard trimmer set to yield a gain of 1 relative to the input signal, but I made it panel mount to allow adjustment of volume.
Power input is nominal 12V and the LM386 uses this level directly, with only a couple of capacitors for smoothing. The rest of the circuit operates at 9V regulated by an 78L09. The MF10 chip is used in single-supply mode via a resistive voltage divider.
This design is sold as a kit through the New England QRP club, but schematics are openly available. To try it out, I decided to build my version on vector board, which was great in terms of character building, but next time I would certainly save a boat load of time and get their PC board.
The dual-ganged bandwidth pot takes up a lot of room in a case, so ideally the project would be housed in a deep enclosure. I had a bud box on hand, so I used that. The project schematics call for a 4.7k resistor on the power indicator LED, but that value results in such a dim LED that it would only be useful in a dark room. I knocked the value down to 1k in one build and 1.5k in another, and both worked fine. The marginal power cost is insignificant to me — I’d rather have a reasonably bright LED to let me know that the unit is powered.
I had no difficulty machining the cast aluminum case with titanium coated bits, but a stepped drill bit made it a pleasure to bore out the holes for pots and switches, leaving a burr-free hole with minimal effort. The rectangular hole for the power pole connector is still something of a pain. I drill a few holes, enlarge them, and then refine it with some tedious file work. It amazes me that there is no elegant way to panel a single power pole pair. There are snap-fit panel mounts for 2×2 and larger arrays, but I would think the most common requirement would be a simple pair. The only solution I have found so far is to order very overpriced metal retention brackets that clamp on both sides of the socket. With a single screw in each bracket, they have a tendency to tilt, which gives the project an unprofessional look.
For once, I remembered to drill first and paint later. In this case, two layers of white spray paint. I made the lettering by printing text on water gilde decal paper using my laser printer. I filled the sheet with words, numbers, abbreviations, etc., that I thought I might need in future projects, since the paper is somewhat costly. After applying the decals and letting the cure for half a day, I went over everything with two coats of clear matte acrylic spray and then assembled the project.
As a functional test, I played back a recorded session of PSK31 signals and output to a computer running MixW. I made a video of this session and also took the photo at left. In that photo, events at the bottom of the waterfall are the oldest. So, proceeding upward, I had the filter off with audio bypassing it entirely. When the filter was turned on, signal was lost until I turned the AF gain up to the point where the level was similar to the source volume. The filter was already set to narrow bandwidth and the curvy line resulted from me dialing down the center frequency. Towards the top of the waterfall, I had opened up the bandwidth and then dialed up the AF gain to greater than unity, oversaturating the waterfall.