Using an attic space as a test chamber, three coils were tested in eight experiments over a period of 30 days during which a wide range of environmental conditions were experienced. The attic was well ventilated to the outside air so that the coil's temperature and humidity was reasonably well described by the conditions measured at a university weather station thoughtfully sited just across the road.

Measurements of the frequency and Q factor of the lowest three coil modes were taken automatically every five or ten minutes by pinging the coil with a pulse of 2.5 kV, capturing the resulting coil base current using a scope (averaging over 64 pings), and logging the trace data to a computer. The signals were then analysed by tcma to extract the Q and F values, which were filed along with readings of the temperature and humidity obtained from the weather station.

In these notes we use the abbreviations Qn and Fn to refer to the Q factor and frequency respectively of the n-quarterwave resonant mode. Frequencies and Q factors are plotted as percentage deviations from their mean values.

Experiment 0

An initial run intended to shakedown the test setup lasted just under 4 days and worked without problems. This run immediately gave some very large Q variations:

This is a fairly large sonotube coil placed within a few feet of the sloping roof of the attic. Almost immediately there is a huge drop in Q1, which falls from 163 to 28 in the space of about an hour during the night, followed by a recovery taking a good six hours. The Q factors of the two higher modes where also affected to a much lesser extent. This particular hiatus was associated with rain during the night and within 15 minutes of the rain starting the Q had halved. The associated frequency F1 also fell in unison with the Q dropout - by about 0.4%, suggesting that the extra loss was coupled to the resonator by an additional capacitive reactance.

During the early hours of the 9th, the temperature fell to the dew point and condensation formed on the roof. This brought a response from Q1 by way of a 30% drop in a nice ramp over 3 hours, again with much less impact on Q3 and Q5. During the remainder of the experiments we continued to observe events similar to this which confirmed the coincidence with either rain or condensation on the roof. A confirmation came on the afternoon of the 11th when the roof was hosed with water, resulting in the sharp Q1 dropout with similar characteristics to the others.

Experiment 1

The next experiment involved a smaller coil located more centrally in the attic. This coil did not demonstrate the large variations in Q1 that we saw in the first experiment. It did however deliver a very odd step change in both F1 and Q3 at about 10am on the 12th. This was quite large - several percent for Q3, and the coil did not recover to its previous value. Why q3 should be affected and not the other modes remains a mystery.

The gap in the data during the 13th was due to a computer problem. Further small step changes and glitches of around 2% occured in Q3 during the rest of this experiment.

Experiment 2

The coil in expt1 was replaced for expt2 by an almost identical coil in the same location in the attic.

With this coil, we see a number of distinct Q3 dips of around 2-4%, each lasting for some 30 mins. These are curious because there is no sign of any change in the Q factor of the other two modes and it's hard to imagine what factor might affect one mode without affecting modes above or below in frequency. The large jump in Q1 at midday on the 18th occured when another coil was removed from the test area. Removal of this potential source of mutual coupling did not prevent another Q3 blip later that evening.

Towards the end of expt2 we observed a very interesting phenomena involving Q1. At 19:00 on the 18th we see Q1 fall due to roof moisture as the temperature falls to the dew point. Approaching midnight, everything freezes and Q1 rises to approximately its original level, but this time with a very rapid and erratic variation. The suspicion is that this is due to frost on the roof, and perhaps ice forming between the roof tiles. The weather was pretty dry throughout expts1 and 2 but this changed with expt3.

Experiment 3

The third experiment was interrupted by failure of the pinger, and it's not possible to say for certain whether the Q variations seen during that experiment were real or simply due to deterioration of the pinger. Therefore we'll omit these results.

Experiment 4

Expt4 continues from expt2 with the same coil, but with some nearby stuff moved out of the way. We also begin to take a temperature reading from a sensor near the coil, in addition to the one from the weather station. From now on, all graphs involving temperature will use the local attic temperature unless stated otherwise.

The results from this run are quite curious, because we see a great deal more variation of q3 and q5 compared with q1. As with expt2, we see q3 taking some dips of several percent lasting for a few tens of minutes, without the other q factors being affected. Q1 hardly altered at all, until the night of the 26th/27th, when the temperature fell to the dew point and there was rain during the morning. All three q factors were pulled down by this, with the highest frequency modes being affected most of all - by up to 20%. A few minor problems occured during this run, which left gaps in the data. We suspect that spikes from the pinger were crashing the scope/computer data transfer by coupling through the temperature sensor leads. Reorganising the grounding arrangements around the scope seemed to cure things.

Experiment 6

Expt 5 was cancelled and we moved on to expt6 which is the same coil again, in the same location, but now fitted with a John Freau 4"x13" toroid. Things were pretty stable for the first two days until 5pm on the 1st, when rain and snow coincided with a sudden fall of q1 by around 15%. No immediate effect on q3 and q5, but these fell gradually by a similar overall amount, but took around 12 hours to do so.

Those q3 dips that we saw plenty of in the previous experiments are conspicuously absent from this run. During the evening of the 3rd, the roof was dampened lightly around the test coil. Ignoring one set of data readings because the experimenter was near the coil at the time, we saw q1 drop by 20% and took around 20 minutes to recover. Q3 dropped by only 1.5% and recovered in 5 or 10 minutes, and q5 was not noticeably affected.

Experiment 7

The test coil was moved outside for the next two, fairly short experiments. The aim was to get a rough idea of the kind of Q variations seen on an outdoor setup.

Expt7 produced some interesting short term variations of all three q factors. These are noisy variations of around 1% amplitude. In the case of q5, this variation is about the same size as the experimental noise, but in common with the other two modes, the variations are correlated quite a lot from one sample to the next, so they are likely to be genuine variations. In the case of q1 and q3, the variations are at times an order of magnitude higher than the noise floor. These noisy variations of q1 and q3 abruptly cut off in the early hours of the 6th, and gradually come back again during the day. The sudden temperature rise on the local sensor coincided with the sun hitting the coil. A most interesting behaviour is that throughout the experiment, q5 trends in the opposite direction to q1 and q3, which remain more or less in step with each other. It's worth looking at a graph of frequency against temperatures for this experiment,

We see that although q5 trends oppositely to the others, all three frequencies trend together, although by significantly different amounts, with the higher modes changing by the highest percentage.

Experiment 8

Expt8 continues on from expt7 after a gap of an hour, during which the lawn was watered and q1 increased by 2.5% and both q3 and q5 increased by 5%. Over the next 8 hours q1 fell by around 7%, and q3 by half that. q5 continued to do its own thing.

A most notable point about q1 and q3 in these last two experiments is that, contrary to the indoor experiments, these two q factors are varying in step with temperature. We expect Q factor to fall as the winding resistance rises with temperature, so the contrary behaviour of q1 and q3 in these outdoor experiments is interesting.

We've had around 25 days worth of data collection, involving three coils and two locations, with various events of rain, snow, icing, with the full envelope of temperature and humidity being visited. That's enough for us to get a feel for the typical q and f variations. We've observed a number of effects:

Indoors:

• 1) Gradual variation of q and f with temperature;
  Most of the time we see roughly the expected variation of Q with T. However the variation is often a factor of 2 or more higher or lower than the value 0.4%/degC expected on the basis of copper losses alone.
• 2) Big, sudden, q1 dropouts due to moisture;
  These occur within minutes of the rain starting and are much more pronounced with the coil near the attic roof than with the coil near the center. q1 can take several hours to recover. Condensation may have an even greater effect than rain because it can affect sheltered surfaces too.
• 3) Slower q3 and q5 moisture response;
  The higher mode q factors often seem to take quite a lot longer to respond to rain or condensation. They are often still falling as q1 is recovering.
• 4) Mysterious q3-only dips;
  Lasting around 30 mins and affecting only q3, these occured on two runs of the same coil. Another similar coil (expt1) showed sudden step changes to q3. Not present on the toroided coil.
• 5) Eratic q variations during icing;
  Occuring in expt2 when the roof was frosting. Rapid variations of q1 up to 5%.

and outdoors:

• 6) Fairly small, rapid q variations when outdoors;
  Not present all the time, and seem to affect all the modes about equally.
• 7) Unexpected variations of q1 and q3 with T;
  In which q1 and q3 rise and fall in step with temperature rather than against temperature.

Some of these results are not at all surprising. For example, except for q1 and q3 on the lawn, the general trend of q with temperature is of about the expected size and direction. The response of q1 to moisture on the nearby roof is also not surprising, but the amount by which the q factor can be reduced can clearly be very high in some situations, and might be enough to cause problems when trying to work a coil in a shed or garage on a rainy night.

Perhaps most interesting are the short term variations, on a timescale of a few minutes, seen during icing of the roof and when the coil was set up over the lawn. Clearly the coil is responding to changes in its environment taking place on the same short timescale. Random variation of the combined resistance of the ground return paths may be involved, with multiple paths continually forming and decaying in the ground. We are also at a loss to explain the short term q3 dips seen in several of the attic runs. These are quite noticeable and last for 20-30 mins, with a roughly gaussian profile. One possibility is that another coil, somewhere in the building, happened to have a resonant mode that was drifting in and out of tune with our mode 3.

With this fairly casual set of experiments, it has only been possible to briefly sample the territory ahead, but this is quite sufficient to indicate that there are interesting things to be brought out. These tests have allowed us to iron out some of the problems with the equipment and we can now expect to do 5 day runs with only an hour or two of downtime, if any. In order to progress further we will have to plan some more carefully controlled tests in order to isolate the various factors involved, and these will hopefully proceed in the near future.

For further details on the above experiments, consult

• expt0. Attic.
• expt1. Attic.
• expt2. Attic.
• expt4. Attic.
• expt6. Attic, toroided.
• expt7. Outdoors.
• expt8. Outdoors.