Comments? email to
The most current version of the filter is generation 2; the original filter article, and another revision (rev 6) of the original filter board are also described here later for reference only. Generation 2 and rev 6 are basically the same filter, though the gen 2 PC board does provide additional configuration options, and the filter is easier to construct.
is how the filter would be constructed for a single-pallet (1kw) amplifier
And here is one constructed for a 2-pallet (1.5kw+) amplifier.
Here is a link to the generation 2 .pdf file with assembly instructions, bill of materials, and a complete description of this newest filter. The table below shows the performance measurements.
Even though my primary interest is VHF/microwave, I do operate HF on occasion, and wanted to replace my aging Drake L7 KW amplifier with a solid-state one.
If you enjoy building your own as I do, one of the other components you will most certainly need is a good low pass filter to keep harmonics within FCC specs. Here's an excerpt from the current FCC documents pertaining to the amateur radio service:
Such a filter can be a real challenge...it has to be able to handle a lot of power, cover a very wide range of frequencies, be small enough to fit into a cabinet, and still be able to suppress harmonics enough to keep you within allowable limits. SSPA's are usually of the push-pull design, and by nature have good 2nd harmonic suppression, up to 40db, so we don't need a lot of attenuation of the 2nd (6 to 10 db is usually enough). However, it is not unusual to see the 3rd harmonic at only -10 or -11 DBc. This means if you are not using a filter, your 1kw SSPA on 40 meters is also producing 100w on 15 meters. We need at least another 33db suppression on that one. The analyzer plot below shows the unfiltered output of my SSPA running 1150w out on 40 meters (into a dummy load, of course):
The odd harmonics present an ugly scene, don't they? After passing through the filter, the spectrum looks like this:
This is one of the tougher bands (40 meters) to filter if we are to include 10 MHz in this segment, which is the case in this design, but the suppression is good enough for compliance.
Most filters of this type are reflective, meaning the harmonics are bounced right back into the device generating them. There are designers who suggest that reflecting 100w of 3rd harmonic energy back to the device might cause it to be damaged, and that certainly is a valid concern for some of the older devices; but I haven't found this to be necessary with modern LDMOS as long as good efficiency is maintained. A diplexing approach, using a reflective filter/diplexer network with an absorptive "dump" port (usually a high power 50 ohm resistive termination) is one way to deal with this; the harmonic energy reflected by the LPF section is channeled into the dump port, where it is absorbed.
I decided to keep it simpler than that; the newer devices available to us are very rugged, and can handle a lot...they are often rated at 65 to 1 SWR survival, and easily handle moderate reflected energy as long as reasonable efficiency is maintained (60% or so). Considering this, I opted for a more economical and less complex approach. The end result was a series of 5-pole Chebyshev filter segments, with the correct filter segment switched in to match the band in use.
Some of the bands need their own personal filter; 160m, 80m and 6m are good examples. Harmonic relationships prevent their use on other bands. However, one can combine 7 and 10MHz, 14,18 and 21MHz, and 24 through 30MHz in single filter segments; other combinations are possible. The 5MHz band can also be covered with different component selection in the 80m filter, but since only low power is authorized on that band, I decided there was no need to sacrifice the extra suppression on 80m just to include it. If the FCC were to authorize high power on 5MHz at a later date, this band could be included by changing the value of just two capacitors in the 80m filter.
To do all of this requires 6 different filter segments, and some method of switching them in and out for the band currently in use. Fortunately, even up to 1.5kw, we can use general-purpose relays and FR4 PC board material (.094 thick) in these frequency ranges and still keep losses very low (the typical insertion loss on any band is only a tenth of a db).
The switching arrangement is handled by 5 pairs of 12v G2 series SPDT relays with 16 amp contacts (for more info on these relays, see this article). These relays have excellent insulation resistance, are inexpensive, and can handle more RF power than we can legally produce. The arrangement is such that with no relay pairs energized, the 160m filter is active. To activate any other filter segment, the control line for that relay pair is switched to ground; this also disconnects the other filter segments. This switching arrangement can be slaved to the band data output from a driving radio for automatic band selection (requires the proper interface cable, something I'll be working on later).
The filter board measures 6 by 8 inches, and should be mounted on stand-offs at least 1/4 inch above conductive surfaces. Once constructed, you'll need to tune it, and the easiest way is with a network analyzer; if one is unavailable to you, it can be tuned with your barefoot radio and an SWR meter; just tune for lowest SWR with a good dummy load on the output of the filter. Adjustment of each segment is done by squeezing together (or spreading apart) the turns on each toroid core in that segment. If you are building this project from a kit I supplied, the original assembly instructions are here. An addendum for the version 5 configuration is here.
At 1 KW+, the filter coils will run hot if used in continuous duty modes like FM or data, so some airflow across the board is advisable in these cases. You don't need a hurricane, just a gentle breeze. Even this isn't really necessary for SSB or CW, but doesn't hurt to plan for it.
Schematic diagrams, measurements and a bill of materials are listed below; in some of these measurements, I noticed the roll-off of the bandpass degrades in a couple of places on the lower frequency filters. This was traced to some bleed-through across relay contacts for the filters on the higher bands, but fortunately these fall in areas where the relevant harmonics are already weak, and cause no problems.
The general description of the filter itself, and the requirements for harmonic suppression regulatory compliance are detailed in the original article following this description; I'll just add the new data for this (rev 6 ) version here: