I have a nice off-center-fed dipole (OCFD) antenna that is my mainstay for multi-band HF work, but I would like to understand it better. With a wire antenna tied to trees it is not a simple matter to quantitatively determine the antenna’s radiation pattern experimentally. One way to get a handle on this is to compare signal levels against an isotropic antenna. Vertical antennas are nice this way, since you just need a wire straight up in the air and you naturally get an azimuthally symmetric pattern. This was my motivation to build a vertical dipole.
When you look around at how people put up vertical antennas, most of the talk is about radials – how many, how long, raised versus buried – so it goes. My gut feeling is that all this effort is misplaced. The problem with vertically polarized radiation is that when the EM wave impinges on ground it induces “radial” currents. If the antenna is close to the ground we can provide radial wires so the RF currents can return, a counterpoise for the currents in the 1/4 wave vertical section. I have tall trees. This means I can just pull the vertical antenna up into the trees so that the EM wave isn’t centered on the ground plane. I’d rather not launch a wave into the ground anyway — I want the wave to propagate across an ocean or two! Traditional verticals with radials will have the same problem with the EM wave propagating near the ground surface once the wave has moved past then ends of the radials, so why spend all of this effort on connecting the antenna into the ground for a few meters when you can’t do it for the next 1000 km anyway?! As soon as I pull the antenna up into the trees by at least 1/4 wavelength, I can now have a vertical dipole if I drop a wire from the feed point back toward ground. But rather than using another wire, realize that I already have a wire coming back down from the feed point, namely the outside of the co-ax feed itself. The important thing to do is to define the end of that radiating segment for RF, so there we place a choke. That’s it– our feed-line coax with a choke about 1/4 wavelength from the feed point and another 1/4 wavelength wire continuing on up — pretty easy!
With 4NEC2 I model because I can. This simple antenna system is a great sandbox for learning the tricks of the NEC codes and some of the basic principles of antenna design, and to find all the pitfalls of this nice idea. My NEC description for this vertical antenna model is located here, to use with 4NEC2 available here. I’ve included some “extras” like transmission lines, chokes, and wire coating in addition to the antenna wires.
When I write the model definition I like to use variables rather than just specify fixed numbers. I try to express the dimensions as functions of things I might like to change, for example the height of the top of the antenna. I’m going to pull this thing up with a rope and I can stop anywhere. If I want to model the effect of height I only have to change one variable. Hence everything is described in terms of a few variables which are meaningful when it comes time to build — height, overall length, and the length of the top driven element, angles, etc. Besides the antenna I also modeled the feed transmission line to the transmitter. NEC does this using an “ideal” transmission line with a specified impedance and length, the “inside” of the coax where only what happens at the ends of the line matter. The outside of the coax is part of the antenna system itself, and includes the RF choke that defines the lower driven element of the antenna. I then run the rest of the feed line down to ground where NEC assumes it is grounded and which could be where the radio is located.
With vertical antennas, the ground model will have an enormous effect on the level of losses, far field radiation pattern, and gain. For comparison purposes, I use the 4NEC2’s “Average” “Real Ground” model. You will discover that it is difficult to get more than about 50% radiation efficiency with this model – but that is life with a vertical antenna!
Now for some question we might hope 4NEC2 can answer.
- How much gain will this antenna give me?
- How high up should I pull this antenna?
The radiation pattern for this antenna at various heights is compared with the conventional 1/4 wave vertical antenna with buried radials in the figure below.
Longer radials help a little for the conventional ground-based antenna. Rasing the antenna higher improves the elevated vertical dipole. The gain of elevated antennas is especially better for low radiation angles less than 15 degrees. If you have a tall tree, there seems to be little reason to spend too much time burying radials in the lawn!
- What about that choke?
The choke both defines the length of the lower section of radiator, which is the coax cable shield, and also keeps the RF from traveling down the feed line back to the radio. The more inductance the better for these tasks, but we have to build the inductor, either air-core or with ferrite, to hang on the feed line and there will be capacitance. I spent an afternoon with a few ferrite cores, some coax cable, signal generator and oscilloscope and can report that even good ferrite is not very suitable for this application. Any choke I made with ferrite was lossy and had a self resonant frequency below my desired operating frequency of 14 MHz. Much better is just a simple solenoid made from the coax feed line itself.
Unlike a simple choke on a feed to keep the RF out of the shack, the choke for this application has a much more difficult job to do. A choke near the feed point for a normal dipole is trying to choke off the RF compared to the drive impedance at that point – which may be only a 50 to 150 ohms. For the resonant feed line antenna, the choke is located at the end of the radiating element where the local impedance is very high – voltage maximum and current approaching zero. This is probably the main downfall to this antenna design because it depends so strongly on getting this choke right. If we knew the reactance of the choke we could model it with NEC. I chose to model the inductor as a parallel LCR circuit where the tuning is not perfect for the band, but the inherent Q is high so parallel R can be >20 kΩ. The total reactance of a tuned trap can be many kΩ if carefully constructed and tuned.
Figuring out the self capacitance of the choke coil took a little research. Attempts at direct measurement of capacitance are confounded by parasitic capacitance of scope probes and by resonances with the physical length of coil wires and connections themselves, so that the measured resonances do not reflect the desired stray capacitance of the coil but rather the length of wires. Internet research yielded the following formula, which seems to be plenty good for what we want. Note that the capacitance is almost entirely just a function of the coil diameter.
Cs = 0.46 * D
where Cs is in pF and D is in centimeters, good for coils with aspect ratio 0.5< l/d < 2.0. There were a surprising number of just plain incorrect quite recent theory papers on this subject, which lead me astray before finding this simple empirical formula discovered by Medhurst in 1947.
When all was said and done, I went for a about 19 feet of coax wound on a 5″ form, 14 turns. Calculated inductance is 22.5 μH with about 6.2 pF of capacitance and theoretical self-resonance at about 13.5 MHz.
Because the choke is so critical, I decided to model several different versions and mis-tunings to see the effects. If too far from self-resonance, the choke will leak RF current to the feed line and change the effective length of the radiator, which in turn, can degrade the SWR.
To drive this antenna several questions come up.
- Should I use 50 ohm or 75 ohm coax?
- What will be the SWR? How can I tune that?
- If I use 75 ohm coax, how long should the transmission line be?
The driving impedance of a dipole in free space happens to be about 73Ω. My transmitter would like to see 50Ω. If I just use 50Ω cable to the antenna, the NEC model finds the expected SWR of around 1.5, which could be acceptable. (With a perfect choke NEC could optimize to SWR 1.4 – with a more realistic choke I could only get to SWR ~1.7). I have a large free spool of 75Ω CATV cable so I would like to show how you can win twice using the higher impedance cable. The basic principle is that the feed line should be an odd 1/4 wavelength multiple of the operating frequency. Reflections from the ends of the cable will tend to cancel each other out with this special length cable. When the cable’s impedance is the geometric mean between the input and output terminations, the cancellation of the reflected waves will be perfect! In other words, the ratio 75Ω cable / 50Ω transmitter should be the same as ZANTENNA / 75Ω cable, or about 112 Ω impedance at the feed point. We can find such a higher impedance feed point if we don’t feed in the center of the dipole, but move off to one end or the other. Notice in the model figures that the feed point is substantially below the mid-point. The numbers I came up with were 24 ft. of #14 THHN wire for the upper section and about 9 ft. of RG-62 coax between the feed point and the choke.
4NEC2 can help get the correct dimensions with the “Optimizer” that is built in. Any variables that you define, and to which you assign a value, can be used as a variable to optimize. In this case we want the optimizer to tune the wire lengths to give us low SWR. The model is set up with the 3/4 wave long 75Ω transmission line. I start off with a guess for L and L1 and let the optimizer hunt for the best values. For this antenna, I also optimize for the antenna gain at 12 degrees elevation. There are two solutions for SWR ~ 1 that correspond to the two equally off-center points of the dipole. The optimizer can converge to either solution depending upon starting conditions. The solution with the long upper wire gave slightly better low-elevation gain. The optimizer is a nice tool, but not magic! Keep in mind the physics and know what answer you are expecting before you run the optimizer. Choose only one variable at a time to begin with. Remember, too , that you do not want a design that is sensitive to small parameter variations. Experiment, but make sure you can understand the optimizer’s results based upon physical principles and you will not be led astray by numerical magic.
I just pulled this up in the tree today. Initial tests look pretty good. The SWR without additional tuning is around 1.2; signal levels average a few db less than the OCFD, which is what I expected. The uniform pattern can help, however, and I already found some DX into Kazakhstan, which is in the northerly null of the OCFD.
It’s early Spring and times are changing quickly. Although not quite officially Spring yet, the weeds are greening up in the garden. That means that we get fresh salad again! I just planted some flats of lettuce starts, but no need to wait for them if you can grow the right weeds. My favorite for this time of year is Mache, Rapunzel, Lamb’s Lettuce or Corn Salad. Whichever of its many names you wish to call it, this green is happy to grow as a winter weed in the garden. I never plant it on purpose anymore, but I do make sure that some of the volunteer plants will grow to maturity and go to seed again to provide for the next season’s weeds. You will always have weeds — better those you want to eat!
Between rain showers today, I plucked a few of the nicest rosettes for a salad, weeding up the chickweed and dandelions to provide a little more space for the growing Rapunzel.
The key to eating this delicious green is to pick the whole plant; don’t bother separating the leaves. There is no bitter in this green, just a touch of perfume. It is great with a light lemon salad dressing.
But don’t stop with the corn salad. Look around for some of the other classic early spring volunteers like arugula (rocket), and lettuce seedlings. Arugula also responds well to scattering of seed pods from plants gone to seed the summer before. You can go to the effort to actually save the seed and replant in the normal manner, but with these winter weeds it is better to just help nature along and improve the odds of more random plants by making sure some seed finds a place to grow over the winter. Parsley and cilantro can also respond well to this type of treatment. When the plants finally go to seed, you can save a few seeds in an envelope in case nature doesn’t do what you want, but the best thing to do is to just toss the seeds into the garden to start the new weeds. When the corn and beans need weeding, you will find a few friends to nibble or nurture depending on your whim and in the spring your salad will be there for the picking.
Our comment letter to EPA follows below.
Environmental Protection Agency Docket Center
1200 Pennsylvania Ave. NW.
Washington, DC 20460-0001
Re: Pollinator Ecological Risk Assessments: Imidacloprid Registration Review
To whom it may concern,
We would like to offer some comments on the regulatory endpoints that are proposed in the first Preliminary Pollinator Assessment to Support the Registration Review of Imidacloprid (January 4th, 2016).
Below is a figure from our paper (Rondeau et al. 2014), with the toxicity endpoints and the colony-level feeding exposure endpoint shown along with our summary of time dependent toxicity data for imidacloprid.
Indeed, the EPA has chosen the 10-day LOAEC and 48-h LD50 to be safely on the low side of most of the reported results in the literature. However, we feel that the colony level exposure of 25µg/L as determined from the Tier 2 study results should be viewed with much caution. The recommendation of <25µLa.i./L is shown as the red vertical line on the plot. Nectar consumed at 25µLa.i./L with typical daily consumption of 20µL/day yields a daily dose of 0.5ng a.i. per individual. A single day’s exposure, 0.5ng/bee, exceeds by a factor of 2 EPA’s accepted LOAEC of 0.24ng/bee. It is very hard to imagine that the cumulative dose to individual bees at this level of contamination would not give rise to neuro-toxic symptoms.
Curiously, the 25µLa.i./L exposure rate corresponds on our graph to the LT50 time of about 20 days, roughly the age when nurse workers enter the field foraging force. With the colony getting much of its nourishment from the contaminated nectar, there may be little need for a robust work force aged much more than that. Clearly the colony is providing services that transcend the health of individual bees. In fact, it can be expected that there may be colony-wide paradoxical effects. The colony response to a shorter lifespan of individual bees could very well be acceleration of brood production. If colony health was measured by the amount of capped brood, such a seemingly positive effect would be interpreted incorrectly.
In our study we found that the results from many diverse toxicity studies for imidacloprid on insects could be understood in terms of a simple power law scaling,
LT50 ∝ D tP
Where D is the dose rate, t is time and P is the power law exponent. We discovered that much of the honeybee toxicity experiments can be unified with such a scaling law where the time exponent is approximately 2.
Although there is an abundance of toxicity data for imidacloprid and honeybees, there is still missing a good study that can determine the time when such t2 scaling is likely to come to an end. Absent such a study, the prudent policy is to assume that the scaling we present continues indefinitely until individual bees die. If you use bees as proxies for solitary pollinators and other beneficial insects, then you are left without the benefit of robust colony services to maintain health in the presence of pesticide residue. Instead the best estimate we would have would be individual toxicity scaling, with a maximum time based upon natural insect lifetimes or reproductive cycles. This will vary dramatically from species to species, and could approach good fractions of a year for insects on an annual cycle. Unfortunately, the regulatory process is not as simple as assigning a threshold guideline concentration for which to stay below, since the organism life span becomes part of the equation.
The EPA’s continued adherence to the concept of threshold toxicity levels should be questioned with pesticides like imidacloprid that are designed to bind strongly to synaptic receptors and which act to directly stimulate the post synaptic junction. Relatively few molecules of the pesticide can have long lasting effects, as we demonstrated with a model in our paper. This is in contrast to acetylcholine esterase inhibitors, where the action of a just few molecules does almost nothing, since there would still be many acetylcholine esterase sites to clear the junction of neurotransmitter. Hence, for this latter class of chemicals (organophosphates), there is a natural threshold built into the toxic effect, namely the point where concentration of pesticide is sufficient to inhibit a large fraction of the acetylcholine esterase sites.
EPA’s history of successful regulation of the organophosphate insecticide can be seen as a vindication of the threshold theory for the acetylcholine esterase inhibitors. The neonicotinoids present the EPA with a major change in mode of action. Hence a fundamental change in the regulatory framework that backs away from the threshold concept and looks more deeply at the accumulation and persistence of toxic effects is required (Sánchez-Bayo & Tennekes 2015).
Gary Rondeau, Applied Scientific Instrumentation, Oregon, USA
Francisco Sánchez-Bayo, The University of Sydney, NSW, Australia
Henk A. Tennekes, Experimental Toxicology Services (ETS) Nederland BV, Zutphen, The Netherlands
Axel Decourtye, ITSAP-Institut de l’Abeille, Avignon, France
Ricardo Ramírez-Romero, Universidad de Guadalajara, Jalisco, Mexico
Nicolas Desneux, French National Institute for Agricultural Research (INRA), Sophia-Antipolis, France
Rondeau, G. et al. Delayed and time-cumulative toxicity of imidacloprid in bees, ants and termite. Sci. Rep. 4, 5566, doi:10.1038/srep05566 (2014).
Sánchez-Bayo, F. & Tennekes, H. A. in Toxicity and Hazard of Agrochemicals (ed Marcelo L. Larramendy) Ch. 1, 1-37 (InTech Open Science, 2015).
Once I had the RTL-SDR radio working as a pan-adapter for my IC-751A, I quickly became interested in seeing what I could do with all of those signals that were apparent on the SDR radio’s waterfall. My first goal was to see if I could receive the complete JT65 and JT9 signal band, with both modes usually in a band region up to 4 kHz wide on the digital sections of the main ham bands. The WSJT-X program is happy to decode signals with this audio bandwidth, but the native IC-751A only has about 2.8kHz audio bandwidth on the upper side band mode. So this could be a simple enhancement of the existing receiver if it really worked. I set up the RTL-SDR and various software components as described previously. But first I wanted to know just how good is the SDR receiver really?
There is a big advantage to have the RTL-SDR on the IF of a good multi-band receiver. By the time the RF signals get to the 1st IF mixer, they have already been filtered by the low-pass band filters of the parent receiver, which effectively eliminates interference from local high power FM stations that can degrade the performance of the RTL-SDR. Hence, the RTL-SDR should be able to perform just about as well as it ever will in this situation.
Questions immediately arise when trying to set the proper receiver gain. There are many places to turn up or down gain levels. Here is a list starting at the antenna:
- The antenna tuner can dramatically increase the RF signal level when the antenna is resonant.
- The IC-751A has an RF preamp before the 1st mixer that can be deployed.
- The RTL-SDR control allows “Tuner AGC” or fixed Tuner Gain from 0 to +50 dB.
- The RTL-SDR also allows “RTL AGC.”
- The HDSDR software has AGC and Volume controls that affect the demodulated output levels.
- The sound card channels have level adjusters.
- WSJT-X has an input audio gain setting slider.
What I discovered was that the RTL-SDR receiver needed as much RF gain as I could give it to get comparable results to the IC-751A. A tuned antenna helped the RTL-SDR much more than it helped the IC-751A for clean reception. With the antenna well tuned, the IC-751A RF preamp was less important than if the antenna was not resonant. The RTL-SDR Tuner Gain needed to be all the way up, +50 dB, for best performance. “Tuner AGC” seemed to accomplish the same thing as just setting the gain at 50 dB. Once past the Tuner Gain setting, all of the later gain adjustments for the demodulated audio had little effect unless you turned things down completely. I followed the WSJT-X recommendation to set audio levels such that the quite band noise level showed about 30 dB on the WSJT-X level indicator.
The figure above shows what I manage to receive using the two receivers on the same signals. From the antenna to the 1st IF stage of the IC-751A, the signals travel the same path. After that, the RTL-SDR dongle’s tuner and 8 bit ADC generate digital signals that HDSDR converts to an audio stream, which is then sent on to one instance of WSJT-X. The other WSJT-X window is decoding signals processed by the IC-751A in the conventional way. Hence, just as I did before, we have a simple method to compare the relative receiver performance using the reported JT signal to noise reports generated by the WSJT-X software as the metric. One thing you can immediately notice is that the HDSDR bandwidth is large enough to cover both the JT65 and JT9 signal regions of the band whereas the 751A’s filters limit the bandwidth to about 2.8kHz. The other thing you might notice if you look carefully, is that some of the splatter that you see on the IC-751A in the presence of a very strong signal is not present on the SDR radio. Notice, at time 19:46, the strong signal at ~1200Hz, and the splatter 2200 – 2500 that is absent on the upper image. Goes to show that arithmetic doesn’t generate inter-modulation images, where as real mixers and amplifiers can.
So on to the noise performance data summary from thousands of JT65 and JT9 signal reports, generated by simultaneously decoding the signals on the two radios. The table below summarizes the experimental results. The first column is how much better the IC-751A is compared to the the dongle using HDSDR.
With the non-resonant antenna, especially on 40m, the RF signals are weak and need amplification to achieve the same performance as the IC-751A. However, you can see once the RTL-SDR dongle is given enough gain by tuning the antenna or using the preamp, then there is really very little difference in the quality of the decoded signals that WSJT-X provides between the software and hardware radios.
At first I found this quite surprising, bearing in mind that the RTL dongle has merely an 8-bit ADC. Realize that the range of signal levels the WSJT-X can decode, and I accept for comparison, is about 20 dB for JT65 and >40 dB for JT9 signals. Now bear in mind that with 8 bits, the RTL’s ADC will be limited to a maximum of 48 dB dynamic range. So, I guess this “just fits” more or less. Very seldom to we bump into the top end of the dynamic range of the JT signals. Rather we are almost always struggling to pull them out of the noise. So as long as there is sufficient gain at the front end, the RTL tuner can do its job. It begs the question about any advantage an SDR with a higher resolution ADC might have. Maybe someday I’ll get one and I can find out!
There is little doubt that the RTL-SDR tuner dongle radio receivers are the hottest new thing for the amateur radio experimenter. Re-purposing the $10 tuner dongle to be an extraordinarily wide-band software defined radio is the subject of countless internet articles and videos. Here I am adding my experiences to the crowd. What I will describe are the steps required to use one of these devices as a full-band pan-adapter for a conventional receiver. Once that is working, then we can look more closely at the actual RTL-SDR receiver performance, compared with the native performance of the parent receiver. I’ll discuss the performance of RTL-SDR receiver in another segment. For now, let us get this up and running.
First a diagram of what we are about to do:
My RTL-SDR is the common RTL2832U chip with R820T tuner. The device can tune from about 30 to 1700 MHz. To be useful for the HF ham bands, some up-conversion is required. Many HF receivers employ a first IF stage, mixing upward to a 45 to 75 MHz intermediate frequency. This is perfect for the RTL-SDR tuner. Tapping into the IF after the first mixer is easy on the IC-751A since there is an unused connector on the RF board that is designed for scope monitoring of signals at this location. On my old IC-745 I had to solder in a resistor at the right spot to get the signal. Usually a little study of the receiver schematic will make clear where you need to tap into the 1st IF signal, and whether the radio design will make this easy or not. It is important to tap in right after the mixer and before the 1st IF band-pass filter, since otherwise the pass band that you can tune with the RTL-SDR as a pan-adapter will be severely curtailed by the IF filter. Ideally, use a short piece of 75 Ω coax to match to the RTL-SDR tuner input impedance (but 50 Ω is okay too).
Now we can bring up the RTL-SDR in the HDSDR software. If you are doing this for the first time, some good instructions are located here. Of the various SDR radio programs that are floating around the internet, HDSDR seems to be the best for interfacing the RTL-SDR with another radio and then being able to co-ordinate the tuning of the two systems. When you first bring the dongle up in HDSDR you should be able to see HF band signals when you tune HDSDR to the IF frequency. This makes it confusing to remember where you are tuned, so HDSDR has the ability to shift the tuning display by the IF frequency offset (RF front-end frequency options & Calibration tab). I suggest tuning the receiver to a well-known strong AM station (WWV at 5.0 or 10.0 MHz is a good choice) and then playing with the IF-offset until you have HDSDR tuned correctly. Including a global offset of 10kHz is a good idea so that the IF zero beat is not in the middle to the HDSDR’s audio output band. Additionally, it is important to make sure that the side bands are not mirrored, and the corrections for tuning USB, LSB, CW, etc. match that of the parent receiver. Tuning JT65 signals can help clarify this to make sure they are not mirrored (and un-decodable!). For whatever reason, I needed to check the “Swap I and Q Channel for RX Input” to achieve the correct result. I also set a USB offset of -2940 Hz so that HDSDR and the 751A would tune together in USB mode.
HDSDR will be the master of all, so we set up the HDSDR Omni-Rig interface to talk to the IC-751A via the serial RS-232 port. At this point, it should be possible to tune with either HDSDR or with the receiver directly and have them follow each other. As a pan-adapter, you are ready to go. You can open the band in HDSDR and see signals across the band. A simple click and you will tune to that signal on the 751A. For phone operation, you are done – have fun.
With everything working, the pan-adapter begins to provide useful information. The example snapshot above shows a smattering of signals on the USB phone section of the 20 meter band where the HDSDR is tuned. It appears that the IF pass band of the 751A drops the signal level on the RF waterfall a little, so you see less background in the center of the waterfall where the parent receiver is tuned. With the pan-adapter synchronizations working, clicking on a signal will immediately return the receiver so you can copy the signal. In fact, you can hear the audio on either the parent receiver, or on the computer speaker from the HDSDR demodulated audio output.
For digital modes, we need to run other programs to decode the digital signals. WSJT-X for JT65 and JT9 signals and FLDIGI for a multitude of digital mode transmissions are the programs I use most often. Besides running those programs concurrently with HDSDR, we need to plumb in the audio and control signals to these applications. For most straightforward operation, the digital mode program will be connected to the audio channels of the tranceiver just as you would normally for the stand-alone configuration, but the rig control must now come from HDSDR. To do this you need a Virtual Serial Port pair. There are a couple of possible programs that are available to do this for you. I’m using VSPM from Steve Nance, K5FR. You will have to write to him with your call sign to get a copy for you to use. Reading between the lines, it looks like Steve’s program, updated and expanded, is being sold as Eltima’s Virtual Serial Port Driver, which I’m sure also works, but is not free. A second free program, Eterlogic VSPE, also might work for you. VSPE has some nice additional features, like the ability to have port “splitters” as well as port pairs, but I also experienced some BSOD (Blue Screen Of Death) errors when using this program on my old Windows Vista laptop. Whatever you use, once you install a port pair you can now connect the HDSDR’s “CAT to HDSDR” port to the rig control communication port in WSJT-X or FLDIGI. HDSDR talks in “Kenwood TS-50S” language, so the digital rig interface should be set up accordingly.
So far, the RTL-SDR dongle is just showing you a wide view of the ham band and the parent transceiver is operating in the normal way. But there is more that we can do with the RTL-SDR dongle than just look at the big picture. If we pipe the output of HDSDR’s demodulated audio into a digital mode program, we can use the RTL-SDR dongle as the real receiver rather than just the panoramic view generator. To do this we need another little program called Virtual Audio Cable. This program is not free, but is inexpensive for what it does and how well it works. Once installed, the program can generate pairs of sound-device ports, virtual audio cables, that can be used to pipe audio streams between applications.
We can connect the “RX Output” from HDSDR into a virtual cable, then connect the other end of the virtual cable to, for example, the WSJT-X sound card input so that we could decode JT65 and JT9 signals directly from the RTL-SDR dongle. Once you discover the flexibility of Virtual Audio Cables and Virtual Serial Ports you will come up with many interesting configurations of software components that can be strung together in interesting ways.
I found that the RTL-SDR dongle can be used as a serious receiver of digital-mode signals. However it lacks any transmit capability so you must return to the parent transceiver for that function. Beware that the frequency calibration “going around the loop” between the RTL-SDR receiver and the conventional transmitter must be carefully maintained. I would fine tune with the USB offset number in the RF calibration tab. The wide audio bandwidth that the RTL-SDR can generate is not present in the conventional transmitter. Just because you can decode a JT9 signal up at 5 kHz on the audio of the RTL-SDR does not mean that you can point the transmitter there and have it work! You can use some of the same principles, using VSPs and VACs to connect other SDR radio programs together with a slew of audio processing and decoding programs. The possibilities are almost endless — you better go get one of these and start playing with it!
True understanding of a problem is confirmed when one can validate theoretical model predictions with measurements. It is too easy to believe pretty pictures that modeling programs produce, especially when making meaningful measurements is so difficult. In this article I try to take the bull by the horns and see how close I can come to declaring that I understand my antennas.
Last time we discovered that the noise performance of my two old radios was essentially identical to one another. This means that I can use the two receivers to listen to signals from my two antennas and be able to make comparisons between them. When digital mode JT signals are decoded, the software determines a signal-to-noise measurement for the received signal. We can use these measurements which we get for free with the WSJT-X software for every received signal to quantitatively make comparisons between the two antennas. In practice this means leaving the two receivers tuned to the same JT65 and JT9 band, have both receivers running an instance WSJT-X software, and recording all of the transmissions received by both radios.
In my case, I have a 40m loop that is about 30 feet high and an off-center-fed dipole (OCFD) at about 75 feet up in my tall Douglas fir trees. Both of these antennas are horizontally polarized and have distinct directional properties that are unique to each geometry. This means that I would expect signals from stations a various locations to be preferred by one antenna or another.
To begin with we will look at the predicted propagation pattern of the two antennas on the 20m band. Since I am now getting down to making comparisons for real antennas with a model, I spent some time adding in as many parasitic elements into my NEC model as I could, and included both of the antenna geometries in the same model – just substituting the driving point of the unused antenna with the matched transmission line impedance. Besides the two antennas I also included the aluminium rain gutters on the house in the model, and I did my best to correctly locate the antenna elements as they really are laid out. The NEC model and the two radiation patters for the two antennas are shown in the following figures.
The pattern for the 40m loop antenna is less complex at 20m, since there are fewer harmonic resonances on the loop.
Since the OCFD is considerably higher than the 40m loop, it generally has better low elevation gain; but the two antenna patterns are really quite difficult to directly compare to one another. Fortunately, 4NEC2 can give us tabulated data for both patterns. Since we will be looking at differences in signal levels, we can also look at differences in the predicted model radiation patterns. I don’t really have the tools to do this with the full 3D pattern, so instead I just looked at the difference in total far-field radiation at the 1o and 15 degree elevation ranges where long distance propagation is most often successful.
With the model results now in hand, let us look at the results of the JT65 & JT9 signal comparisons. After collecting a few days worth of JT signals and spending some time pouring over the Excel spread sheet I was able to generate the antenna difference radiation pattern. Rather than having a nice uniform distribution of signals, there are large concentrations of data points from US, Japanese and European hams and otherwise a fairly sparse azimuthal distribution.
Is there agreement? Hard to say, but some things are clear. Usually the OCFD does better than the Loop, often by 5 dB or more. We see various “lobes,” although they do not necessarily line up very well with the ones predicted by the NEC code. The lobe aimed at the EU at 30° seems in place, and stations due North are suppressed on the OCFD as we would expect along the wire direction. The polar plotting is best for comparison with the NEC model. Below we show the same data as function of compass bearing angle along with the a measure of the statistical errors present in the data, and some geographical reference points.
The graphs above shows the result after quite a bit of data manipulation in Excel. Remember that peaks in the above chart can come about because of either a strong transmission lobe for the OCFD antenna or because of a null on the Loop antenna. Many factors can give rise to a spread in the gain difference at a particular bearing angle. The actual propagation path for a particular transmission might be at any of a range of propagation elevation angles, all of which are it different for each antenna. The transmitters have unknown polarization and the antennas will respond differently according to the actual polarization vector. The radiation patterns I’ve presented are just the total gain and do not consider this added subtlety. Nevertheless, there are definitely directions that favor one antenna or the other.
The data file is included here as an example and template for others to use. Excel Antenna Comparison Calculation Template The template file has some useful Excel formulas embedded in the worksheets that will calculate Latitude and Longitude from the Grid location code, and will then calculate bearing and distance from your location.
Several steps are required to get the cleanest results.
- Only signals received simultaneously by both receivers are considered.
- JT65 signals when both receivers showed signal levels less than -5 dB s/n are included. Strong JT65 signals are rejected from the data considered.
- Grid data is added to signals where it is possible to unambiguously determine the grid.
- All signals from a particular grid square are averaged to get a single number of each grid square.
Once the averaged signal differences for each grid point are determined, this data is further processed .
- Latitude and longitude are determined from the grid location code.
- Bearing and distance are determined from the latitude and longitude of the transmitter and receiver locations.
- Data for distances less than 500 km is excluded.
- Signal differences are plotted on a polar graph.
The details I presented above for the 20m band show the level of data analysis possible with the kind of information that is relatively easy to collect in just a few hours of listening with a couple of radios. Even a much more cursory look at the data is valuable, however. Merely taking averages for all matching transmissions on a given band can give a single-number figure of merit for comparison of two antennas.
Here is such data for my two antennas for the four bands where the loop works well.
First thing to notice is that although the OCFD up at 75 feet does significantly better on the lower bands, 40, 20 and 15 meters, the Loop wins out on the 10 meter band. This was a bit of a surprise, but goes to show that height is not always your friend.
Also notice the rather large standard deviation. In some sense, this is just a measure of the depths of the lobes of the two antennas.
Direct comparison to the NEC models was marginally satisfactory. However, there were many small uncertainties for both of my antenna geometries which, when all combined, proved to make the direct comparison with the theory less than perfect. The models are only as good as the input data, but the radio signals tell no lies.
In the future, doing a similar experiment with a vertical antenna as a reference might be a good plan. The uniform radiation pattern from the vertical would provide a constant reference that would allow the pattern of the test antenna to be more accurately determined.
I entered second heaven when I got my second radio. My original 32-year-old Icom IC-745 got placed on the back shelf when the new 29-year-old Icom IC-751A arrived. As much as I enjoyed the old IC-745, I was after a little computer control, and the 745 predates the idea that people use computers. The 751A, just a few years later at the dawn of the PC age, at least had an optional serial computer interface.
Once I had two radios I could not help but to start to make comparisons. Both radios do an admirable job, but I began to notice some small differences in the way they would handle strong narrow-band digital signals. When I used the IC-745, I would turn the AGC off and manually set the RF gain so that I was just picking weak signals out of the static noise floor. This meant that strong signals would be over saturated, but the 745 didn’t mind that too much. When I tried to use the IC-751A the same way, the strong signals would generate signficant amounts of in-band spurious interference, mostly third-order intermodulation products (IMD3).
What happens if you don’t have your levels set right? Most hams know that if you overdrive PSK signals during transmission, there can be IMD3 spurious emissions that can disrupt the band. This appears on PSK signals as multiple “railroad tracks” that spreads out over the band. Less well-known are the effects that come from improper receiver level adjustments. The tricky part is that we need to process a very large dynamic range of input signals, all right next to one another. The IMD3 problem shows up on receive as well as on transmit. Here the problem is the mixing of harmonics of two in-band signals, where the mixing products appear back in the pass band. The “third order” products are generated by the mixing of the fundamental frequency one signal with a second-harmonic of the second and vise versa. With frequencies A & B, the in-band mixing products will be 2A-B and 2B-A. If A-B = Δ, then you will find spurs at A+ Δ and B – Δ.
The picture above shows a clear example of receiver IMD3 distortion. The IC-751A had the AGC turned off so when the strong signal in the center of the waterfall began broadcasting, the front-end stages of the receiver were overdriven. The same signals in the IC-745, with the AGC on, don’t have the problem. Note the strong “mirror” effect of spurs across the strong signal. If you were operating this way, you may be tempted to latch onto one of the traces on the right side ( above the IC-751A label in the figure). The nature of PSK is that these spurs are decodable, but if you try broadcasting there, the sender will not be waiting for you.
Before we move on to the measurements, lets look at another set of spurs that have a completely different origin.
Note the two spurs at 1600 and 2680 Hz on the IC-745 waterfall above. There is another strong signal in the picture, but the spurs don’t appear related to the frequency difference between them. In this case we are looking at audio harmonics, the third and fifth, of the audio signal at about 550 Hz. The odd harmonics will appear in the audio if you start clipping it. In this case, the problem was that the sound card was being driven too hard on that strong signal. The solution was simply to reduce the input level to the sound card.
It took me a while to realize that the IC-751A had significantly more RF gain than the IC-745. If I set the manual RF gain on the 751A at about 60% of full, I had similar gain to the IC-745 at full. But before I figured all of this out, I thought that the IC-745 had better noise performance the IC-751A. But how to know for sure?
My test consists of splitting the antenna signal with a 50 Ω power tee so that each receiver gets the same signals from my antenna. Then I set up both receivers to have almost identical noise characteristics on the FLDIGI waterfall. I run two copies of both FLDIGI and WSJT-X during the comparison. Running WSJT-X, I set the input levels to have about 30 dB noise level between transmissions on both receivers. This gives the decoders plenty of dynamic range and again matches the overall gain of both receivers for the digitized signals. Now we just listen for a while and accumulate JT65 and JT9 signal reception reports from both receivers. If everything is identical, you should get identical signal-to-noise reports on both rigs. Comparing the two waterfalls can show where there are problems with one rig or the other.
The JT65 and JT9 signals on a busy Saturday afternoon shown in the waterfalls above illustrate the dilemma. Many of the transmissions are very strong, and weak signals may be intimately interspersed among the strong signals. I found that I could get significantly better noise performance on a crowded band by reducing the RF gain below the point that the AGC would normally operate, especially on the 751A. The waterfalls above were recorded with only 40% of max RF gain on the 751A and 75% of max gain on the 745. Under these conditions, the signal reports back from the WSJT-X program are virtually identical between the two radios. Both of my receivers can be made to show strong IMD3 spurs if the front end gain is allowed to get too high. The AGC helps in this regards, especially with the IC-751A because this receiver has quite a bit more gain head-room in the RF amplifier than the IC-745. Under extreme conditions manually turning the RF gain down even further seems to help. JT65 signals can become a speckled smear across the band when there is IMD3 distortion from multiple strong signals.
In the midst of this exercise, my neighbor three blocks aways started calling on the PSK band just below the JT band in frequency. This caused some trouble for the IC-745. My neighbor’s +60 dB over S9 signals, only 6 kHz away from where I was tuned, made a mess of the strong JT signals. I found it curious that the problem was greatest with the strong signals. Clearly this is one place where the IC-751A had superior performance.
Measuring accurately the comparative noise performance of the two radios required analysis of the reception logs created by the WSJT-X programs. The reporting on JT65 signals is compress at the high-end, the program never sending more than -1 dB signal level report. JT9 on the other hand, robustly covers signal-to-noise levels from +15 dB to -28 dB. This makes the JT9 signals much better beacons for our tests, however there tend to be more JT65 signals to be received. I looked at both modes, but for JT65 I don’t consider strong signal reports larger than -5 dB in the receiver comparisons. In short order it is possible to collect several hundred report pairs that can be used to get good statistics. You can use my RX_compare spread sheet as a template to do something similar if you wish.
It is not too surprising that my two radios have essentially equivalent noise performance. They were both considered “hot” receivers during their day. I would be really curious to repeat this little test, comparing a modern direct conversion SDR receiver with the classic superhetrodyne architecture my receivers have. That might show some real differences because of better IMD3 suppression in radios without so many mixers.
Now that I am convinced I can adjust the two receivers to accurately reflect the antenna noise performance without introducing their own bias, I can proceed to compare my two antennas. Stay tuned!