For another article with more measurement and analysis of these filters, see the 31 May, 2019 article - Revisiting the 20 meter "helical resonator" band-pass/notch filters- linked here.
As with many groups, the Utah Amateur Radio Club (UARC) operates both a CW and SSB station on 20 meters during Field Day. Being that 20 meters is typically one of the "prime" bands where most of the contacts are made during daylight hours and often into the night, it would make sense to optimize operation on that band as much as possible.
Despite separating the antennas as far as is practical with the available geography - about 300 feet (approx. 100 meters) - and arranging the Yagis north-south of each other so that they are pointed generally broadside each other (east/west) for U.S. coverage, we occasionally encounter a bit of interference between the SSB and CW stations on the same band. This isn't too surprising because, at times, the transmit frequency of one station is just 100-150 kHz away from the receive station of the other and with stations so close together - in terms of frequency and proximity - they are bound to "interfere" with each other, at least occasionally!
Unfortunately, "normal" filtering schemes that one might initially consider don't really apply here. The methods and means of filtering that one might see applied at HF stations include:
- Tuning stubs using pieces of coaxial cable - typically in 1/4 and 1/2 wavelength segments. These are typically used to "notch out" signals from, say, 20 meter operations that are bothering a 15 meter station. This can happen because many radios filter the band(s) below the rather poorly on receive and/or a 15 meter transmitter will often generate low-level noise on 15 meters and all bands below it which can often be problematically radiated by multi-band antennas such as Yagis. The "noise problem" is mentioned in more detail, below.
- Commercially-available bandpass filters (e.g. "Dunestar" ). This filter would be used to mitigate the same sorts of problems as the notch/stub coaxes mentioned above, but in smaller, more convenient packages.
- Tunable L/C traps using transmitter variables and standard inductors. Using the sort of components that one might use for high-power antenna tuners, it is possible to construct L/C filters through which one can transmit that could be used to tune notches to eliminated "problem" signals. Unfortunately, these are somewhat complicated to use and rather large, so they are rarely seen.
What's more is that most transmitters produce a low-level, broad-band noise spectra when keyed up that can blanket the entire band - not to mention, sometimes, low-level spurs related to who-knows-what that can sometimes land on the other station's receive frequency. Even if these spurs are 100 dB down, they may still be strong enough to be an annoyance to the other station on the same band.
A similar noise problem can also exist in a receiver, seemingly created out of nothing even if the transmitter is completely "clean" - a response to various noise sources in the receiver chain in the presence of strong signals.
The situation can seem hopeless, then, if you can't manage to get a huge amount of geographic isolation and/or higher-end rigs that purport to have (or actually have) ultra-clean transmitters and receivers with super-high dynamic range and ultra-clean oscillators.
Or is it?
A few years ago I decided to see if it was possible to make a "20 meter Helical Resonator" so I dug out the formulas, crunched some numbers and observed that such a filter would be possible with practical (and relatively cheap!) materials - just.
A friend of mine gave me a chunk of #6 solid copper wire, but copper tubing of a similar outside diameter would have worked just as well, would have been lighter weight and probably have been easier to work with: Aluminum wire would have worked reasonably well, too, although a larger size (say 4 AWG) would be preferred to overcome the losses compared to those of copper - but it would be a bit more of a challenge to make low-loss, reliable electrical connections.
I wound the wire on a smaller form - some plastic pipe that was 3-1/2" (approx. 9 cm) diameter - and then forced it over an empty "Adams" (tm) peanut butter jar which was just about the right height and diameter to fit inside the 1 gallon can and being made of glass, was already a low-loss material that would have no trouble at all standing up to the high RF voltages that would be present. Initially taping the wire to the glass, I used RTV ("Silicone") adhesive in several places to secure the turns, avoiding its excess use as I wanted to keep the losses as low as possible.
Comment: The dimensions of the glass form - the peanut butter jar - are approximately 4-1/8"(10.5 cm) outside diameter and 7-1/2" (16.5 cm) tall. The coil was wound on approximately 4-1/2" (11.5 cm) of this height.
Having initially wound the wire over a smaller form caused it to fit fairly tightly over the larger, glass form, so it "took" well to staying in place while I carefully separated the turns equally and secured their positions. Note: After the RTV had cured, the tape that had been used to temporarily hold the turns in place was removed.
Winding more turns than I expected that I would need, I set the coil aside for a day or two to allow the RTV to cure and constructed a capacitive coupling probe from a copper plate that was bent to a radius that (more or less) matched that of the coil - See Figures 2 and 6. This plate was actually made from the shield of a scrap of "Heliax" (tm) cable that was cut length-wise and then pounded flat. Between the relatively thick copper of the shield material and the "work hardening" from manipulating this copper, it was pretty stiff - especially after bending it to the radius. Copper flashing could have also been used, the edges rolled away from the coil form to increase rigidity.
For a connection to coaxial cable I cut a piece of 3/16" (4.8mm) thick aluminum plate that was about 2 times the length of an SO-239 connector and about 1.5 times its width, formed it using a hammer and piece of heavy pipe to the same radius as the paint can, mounted the SO-239 connector to it and then pop-riveted the assembly over a hole that I'd made in the side of the paint can that aligned with the top of the coil. See Figure 3.
Inside, I connected the center conductor of the SO-239 connector to the copper coupling plate using a length of #12 AWG solid copper wire, forming a capacitive coupling probe which was adjustable in position relative to the coil by virtue of bending both the coupling plate and the wire connecting it to the SO-239 connector. Picking the same spot on the opposite side of the coil, I duplicated the connector/coupling probe arrangement and put one there as well.
Now, the tricky part: Setting the frequency tuning range.
Placing and centering the coil in the paint can, the bottom end was grounded via a hole that I'd drilled in the side of the can for that purpose. I also placed the coupling probe about 3/8" (9-10mm) from the top of the coil as a starting point for determining resonance and coupling, knowing that this was likely to change. Using an antenna analyzer I swept the frequency up and down and saw a very prominent deflection that I believed to be the self resonance of the coil at around 12 MHz in my case - and placing my hand near the top of the coil and causing the frequency to shift downwards due to the added capacitance confirmed this. Taking a fraction of a turn off at a time, I soon moved the frequency up to around 14.5-15 MHz.
In the center of the paint can lid I drilled a hole large enough to accommodate a 1/4"-20 nut and de-burred it. On the top side of the lid I soldered a brass 1/4"-20 nut using plumbing solder and a hot soldering iron rather than a torch as the latter would have likely ruined the can's plating and made soldering much more difficult.
For the paint can lid I cut a round piece of wooden paneling about the diameter of the lid with a hole saw and made a hole in its center large enough to accommodate a 1/4"-20 nut. and used RTV to attach it to the top side - this, to stiffen it and prevent it from wobbling during adjustment - see Figure 5.
Onto some threaded rod (brass rod is easiest to work with) I soldered a disk of glass-epoxy circuit board material - also cut with a hole saw - to form a capacitive plate that would go up and down to allow tuning, making this disk as close to a right angle to the rod as possible (e.g. parallel to the lid) so that when rotated, the disk maintained an even distance to the top of the coil as it went up and down. If the disk isn't perfectly "flat", the frequency will be seen to "wobble" up and down while adjusting the tuning. This same sort of thing can happen if the disk isn't perfectly round and the hole isn't centered, but cutting it with a standard hole saw reduces those problems.
The metal from which this disk is cut isn't really important as it could be also be aluminum or even steel: It just needs to be stiff, flat and round. If it cannot be soldered to the threaded rod - which may be the case if steel/stainless threaded rod was used - then the disk may be electrically "connected" using 1/4"-20 bolts and some lock washers as the protrusion of the bolt through the center should have minimal effect except for the fact that the bolt on the "top" side will prevent the disk from being adjusted as close to the lid of the paint can and slightly reduce the tuning range. Note: In Figure 4 I used double-sided circuit board material. If single-sided circuit board material is used, place the copper side down, toward the coil.
The rod was then threaded through the bottom of the lid of the paint can and a lock (e.g. "split") washer and a flat washer along with another nut used as a "jam" nut were spun onto the rod to give it a bit of tension and the threaded rod was then lubricated with a drop of oil to make its operation smooth. As can be seen in Figure 5, a piece of scrap wire from the coil was soldered to this nut and secured with small straps to maintain the tension on this jam nut.
I set the disk as high as it would go (against the lid), installed the lid and then re-checked the frequency and as expected found that it was again low (around 13 MHz) due to the capacitance of the metal lid itself as well as the disk and then did more trimming of the coil, a process that required repeatedly taking the lid off, trimming the coil, and putting it back on again to check.
Eventually, I got to the point where I could tune through the 20 meter amateur band and was ready to do some initial testing.
Using an HF transmitter set to just a few watts and a VSWR bridge and combination power meter on one side and a 50 ohm dummy load and another power meter on the other side, I tuned through resonance and noted that I had both a high amount of insertion loss and a high VSWR at resonance.
I moved both of the the coupling probes equally closer to the coil itself and taking a bit more of its winding off - as the proximity of the probes actually lowered the resonant frequency. After a few more iterations of trimming the coil I found a point where I was able to get around 80% of power to pass through the filter at resonance (about 1dB loss) and a reasonable VSWR - less than 1.5:1 - and achieve a bandwidth of only a few 10's of kHz: This would be my CW station bandpass filter!
- Had I cut too much wire off and raised the frequency too high I would have simply soldered - using a very hot iron - a short piece of pre-bent, to match the curve of the glass - section of #12 or #14 wire, to extend the top of the coil: At the very top, the thickness of the wire in the coil at the top ("high voltage") end is less important in determining its loss characteristics and a few inches/centimeters of smaller-diameter wire here will have no ill effects.
- Avoid the temptation to increase the coupling too much to reduce losses much below 1dB or so. If you do this, the resonator will be over-coupled and its filtering effects in the SSB portion of the band will be reduced. If the transceiver has a built-in antenna tuner to help sort out the mismatch that may be caused by the insertion of one of these filters - particularly as you move away from the center frequency - by all means, use it!
- It is in the "narrowness" of the filter that the importance of using large outside-diameter conductor for the coil becomes important. The first attempt at a bandpass filter used 12 AWG wire and dramatically inferior results were experienced, with the filter offering only 3 dB or so of attenuation 100 kHz away on 20 meters. The #6 wire used in these coils was probably a bit overkill and #8 AWG would have probably been fine, as would copper tubing of similar outside-diameter (e.g. 3/16" or approx. 4mm outside diameter.) Remember: RF flows on the outside of the conductor, so there's no real need to use a solid conductor for the coil - but make sure that you use only clean copper or silver-plated material for it. Beyond a certain point, however, the "loaded Q" - that is, our coupling (via probes) into the coil to put our transmitted energy into it - and then take it back out again - becomes dominant and "improving" the coil itself even more by using still-larger wire, for example, reaches a point of diminishing returns.
A "Notch" filter:
I then replicated the above - but this time, I built a filter with just ONE coupling probe: Instead of a bandpass filter, this would be a "suck-out" (e.g. "Notch") filter.
For adjusting the coupling of the notch filter, the easiest way is this:
- Connect, using a UHF "Tee" connector", to a radio and a 50 ohm dummy load with a VSWR meter connected between the radio and the Tee connector.
- Set the notch at 14.050 MHz (or the desired 20 meter CW frequency) using an antenna analyzer.
- Tune the radio to 14.150 and connect it inline with the notch filter.
- Transmit into the dummy load with the notch inline and note the VSWR.
- If the VSWR is lower than about 1.3-1.5:1, move the coupling probe closer to the coil, noting that you'll have to re-tune the coil - possibly removing some wire. The idea here is to increase the coupling - something that will raise the VSWR a bit - the target being around 1.3: to 1.5:1.
- Go back to step 2 as necessary.
From experimentation, it has been determined that at about 100 kHz away from the notch, when the notch filter itself causes a VSWR of about 1.5:1 to occur in an (otherwise "flat" 1:1 match) it will notch out the energy at the notch's center frequency.
As can be see in Figures 2 and 6, I lined one paint can with copper foil. The other filter, I left alone, with its original gray protective paint inside the pain can: There was no difference in apparent "Q" or performance between the copper and steel!
This result was not unexpected as the current is rather widely distributed along the inside of the "cavity" and the ohmic losses there are of less importance than those of the coil itself. Any future duplications of these filters would not use copper foil lined cans.
How the two filters are used at a Field Day site:
With both a bandpass filter and a notch filter available at a Field Day site, the two are used in combination thusly:
- The Bandpass filter is placed on the CW station, in series with the transceiver. With its usable passband of +/- 15 kHz or less, it covers a reasonable chunk of the 20 meter CW passband, but it is easily retuned by a operator who simply turns the knob on top while watching the reflected power and will attenuate energy from the SSB station at 14.150, the amount of attenuation increasing as one moved up the SSB portion of the band. This passband response will reduce not only the signal level from the SSB station to minimize the possibility of receiver overload, but will also attenuate any broadband noise that the transmitter might produce that could degrade the 20 meter SSB operation. Most CW operation occurs within a 10-15 kHz of a central "spot" - particularly if one "runs" a frequency - so frequent retuning is usually not required.
- The "notch" filter is connected "across" the SSB station's transceiver coax with a UHF "Tee" connector, but is tuned for the center of the CW passband where such operations will take place. At SSB frequencies, it has practically no effect whatsoever, but at the CW frequencies will be attenuated, minimizing both overload by the CW station and also notching out broadband noise that might be produced by the SSB transmitter itself that might degrade reception at the CW frequencies. As long as the SSB operator doesn't transmit anywhere below 14.125 (which should never happen!) the VSWR will be practically unaffected.
Does it work?
Yes, actually. While we haven't needed it every year (we don't know why we can get away without it some years even though we have been using the same rigs and antennas for years now...) when placed into service, it does completely remove - or very much knock down - inter-station interference between 20 meter operations while easily handling the 100 watt output of the radios that we are using.
Notes on these filters:
- Paint cans are quite flimsy. As you can see from the pictures, several measures had to be taken to "beef up" the paint cans with pieces of paneling to overcome their inherent mechanical instability. Without these steps, tuning could be radically affected by just touching the top of the filter, the weight of the coax hanging on the a connector, or just bumping it while handling it! Even after doing all of these things they are still quite fragile and likely to be de-tuned (or broken!) if dropped!
- They are a pain to tune the first time - and a bit of a pain to set up at each Field Day! After the initial tune-up, you need to have on-hand an antenna analyzer to know where the notch is and a reflectometer (e.g. VSWR meter) is imperative for the CW station to know when the bandpass filter is centered on the operating frequency.
- There's no practical reason why similar filters couldn't be constructed for the 10 and 15 meter bands - or other frequencies in between. For lower bands (e.g. 40, 80) you would need much larger cans: Think "bucket" sized!
For follow-up article with more measurement and analysis of these filters, see the 31 May, 2019 article - "Revisiting the 20 meter "helical resonator" band-pass/notch filters" - linked here.
Sweeps of these filters and analysis:
The Notch filter
After several years of having been used, I finally got around to "sweeping" them using a spectrum analyzer. Not having a tracking generator on-hand, I used a broadband noise source, setting the resolution bandwidth to 3 kHz to provide the needed frequency resolution and the video bandwidth to 10 Hz to "smooth" the trace. The result of sweeping the notch filter may be seen in Figure 8.
Admittedly, the depth of this notch is as great as I'd hoped and while I was at it, I did a bit of experimentation - moving the capacitive coupling probe closer and farther away from the coil. As it turned out, moving the probe closer to the coil widened the notch and reduced its depth even more while moving the probe away narrowed the notch and also reduced its depth. This result implies that the notch depth is, at least in part, limited by the unloaded "Q" of the coil within: You may recall that this effect was briefly mentioned above when I originally wound the coil with 12 AWG wire.
How do I improve the performance? There are several options:
- Use a larger coil. As noted above, 1 gallon paint cans are a bit small for 20 meters, so a "full sized" coil would probably have a better unloaded "Q", but something larger than a paint can would be required.
- Use larger conductor for the coil. The 6 AWG wire is 0.162" (4.1mm) in diameter and being wire, it is solid - and heavy. Because of the "skin effect", RF energy flows only on the outside of the conductor, so tubing may be used, instead with the added advantage of being lighter and a bit easier to handle. Obvious alternatives are "1/4" inch tubing (actual O.D. of 0.375", 9.5mm) and "3/8" inch tubing (actual O.D. of 0.5", 12.2mm) which would increase the surface area by a factor of and 2.3 3, respectively, significantly reducing the RF resistance - but it would probably be difficult to fit a "larger" coil within a 1 gallon paint can.
- Reduce the loss of the coil former. As noted, the coil is supported by a glass jar with the windings held in place with blobs of RTV. While glass is a pretty good choice for this task, a "skeletal" coil former made of polycarbonate or other low-loss plastic would offer slightly lower loss, be much lighter in weight and allow the coil to be trimmed/lengthened somewhat more easily. Even if it has only a small effect as compared to that of using a larger conductor, it would be worth doing owing to the savings in weight and being designed to be specifically suited for the task at hand.
- Use more than one pass/notch element. Cascading notch and/or band-pass resonators will improve off-frequency rejection and allow better control of the shape of the pass/reject band.
One thing that somewhat improved the notch (by 3-4 dB) was using a stub of less than 1/4 wave (electrical) length piece of coaxial cable between the signal source and the notch cavity: The lower impedance of the notch is transformed to a higher impedance at the "radio" end by this stub section, slightly deepening the notch.
While the notch depth is "only" around 10dB this still represents a significant decrease in the energy from another transmitter. If one takes the oft-quoted advice that states that 1dB increase in signal represents 3dB in undesired mixing products at face value, things like intermodulation distortion in the receiver can be easily reduced by 20 dB or more.
Because it is identical, the bandpass filter's response is also a bit worse than I was hoping - and for the same reasons as the notch filter. In doing experimentation, moving the probes closer expectedly increased the bandwidth, but moving them farther away did not narrow the bandwidth much and significantly increased the insertion loss - yet another indication that the unloaded "Q" of the coil is a bit poorer than desired.
As can be seen from Figure 9, the attenuation at a frequency 220 kHz removed (about 14.250 MHz) is only about 6 dB, or 1 S-unit while the insertion loss at the center frequency was just under 1 dB. To increase the "Q", the same sort of things mentioned above could be done to improve the coil and this would further increase the off-frequency attenuation while maintaining the same amount of loss - but at the expense of commensurately reduced pass-bandwidth. As noted above, in the case of the CW station, it is very easy to retune the filter as one moves around in frequency (e.g. watching the reflected power while transmitting and adjusting the tuning) or one can, to a limited extend, rely on the antenna tuner to take some of the resulting mismatch into account without causing much additional loss.
Figure 9 reveals a few other things as well: It functions well to remove energy from stations operating on other bands. For example, the filter will attenuate 30 and 40 meter signals by more than 60dB while knocking down 17 and 15 meter signals by more than 40dB. Another thing to note is that the passband sweep is somewhat asymmetrical: The attenuation at a given frequency is somewhat greater below the center frequency than above. To a degree, this is due to fact that the percentage difference in frequencies above the center frequency as represented on the plot is lower than those below the plot, but it is also typical of capacitive coupling. If inductive coupling were used, this response would be somewhat more symmetrical - but experimentally adjusting the amount of coupling in such a filter is decidedly more difficult to do!
* * *
Comments on the rigs that UARC uses for Field Day:
The rigs that UARC has used for the 20 meter stations for several years have been old Kenwood TS-450SATs - the ones with built-in antenna tuners. We have "standardized" on these since they seem to be relatively clean and able to withstand strong, nearby signals - even on the same bands - as compared to other rigs. We have "discovered" a few interesting things, however:
Most radios produce "Below Band" noise spectra when keyed up:
Almost all modern, solid state rigs from all manufacturers tend to put out a noise spectra that blankets all HF frequencies ON and BELOW the band on which they are being operated. What this means, is that if you are operating on 20 meters using a multi-band antenna such as a Windom or Yagi, when that radio keys up on 20 meters, those operating on 20 meters and lower and lower bands will experience an increase in the noise floor - even if there is no modulation!
Note also that this noise could also appear above the current band, depending on the low-pass filter. For example, it could be present on 17 meters when operating on 20 meters if those two bands use the same low-pass filter, but this depends on the radio's design.
In the case of the TS-450s with the built-in antenna tuners, we have found that simply enabling the antenna tuners - even when operating into a 50 ohm load and the tuner isn't really needed - will attenuate this noise by 15-40 dB on all other bands by causing the tuner to function as a low-Q bandpass filter for both receive and transmit and for this reason there is now a label on all of our TS-450s admonishing the operators to always keep the antenna tuner inline! After all, no-one really notices the extra 0.25-0.5 dB or so (measured!) loss that the tuner causes at or near 50 ohms!
As noted above, other ways to eliminate transmitter-related noise and QRM on other bands include:
- The use of 1/4 and 1/2 wave notch/stub coaxial filters. These are band-specific and notch ("suck out") the frequencies for which they are tuned will leaving other frequencies unaffected... more or less. One must make sure, however, that when changing bands, that the notch is removed and (if needed) replaced with the appropriate notch for another band to prevent a different interference problem or, above all else, you must make sure that you do not transmit on the band for which the notch you currently have installed is tuned!
- The use of commercial, high-power bandpass filters such as "Dunestar" (tm). These are fairly broad, bandpass filters that cover (more or less) an entire amateur band. These do a decent job of knocking down out-of-band energy on both receive and transmit, but they can be quite expensive to implement - particularly if you have a set of these at each station! As with the notch filters, they must be changed when you change bands!
A notable exception to this seems to be rigs with tube finals. This isn't too surprising as the Pi network output of these rigs is inherently narrowband - as is often the preselector/driver - so they are practically incapable of producing broadband noise that extends outside the currently-operating band! Because of the receiver's preselector front end that these same radios typically have, they also seem practically immune to QRM from operation on other amateur bands as well!
Certain radios are banned from our field day site!
The one radio that is explicitly banned from our Field Day site is the Icom IC-706 and its relatives of the same/similar vintage.
When this radio first came out, one was set up as the 20 meter SSB operating positions and within a few minutes of Field Day having started we had to take it off the air because it made operation of the 20 meter CW station - as well as the 40 meter SSB station - completely impossible. Additionally, its receiver was completely demolished by the other transmitters' on the air as well, no matter what band they seemed to be on.
I don't remember what we replaced it with - our old "backup rig", a Kenwood TS-820, I think - but after that, all was well.
The realities of Field Day:
(That is, we use the rigs that we have!)
Having said all of the above, I'm sure that someone reading this will say "You wouldn't have any of these problems if you got a bunch of (fill in the blank) radios!"
I do know of a few higher-end radios that do seem to co-habitate with each other without causing mutually-assured QRM, but unless you have a bunch of club members that happen to bring those same radios every time, or unless the club just happens to own such radios - either donated to them or because of a large enough budget - it must make do with what one has onhand.
That is the case with most of "us hams", isn't it!
This page stolen from ka7oei.blogspot.com