Blogospheric Refraction

If you ever decide to put any kind of sensitive receiver on a bicycle, you’ll want to avoid this ugly duckling of the bicycle lighting world:

Nitelights Illuminator 900

These lights are great from the illumination point of view, and they’re not badly priced either.  However, from an EMC viewpoint, they stink.  I was given one as a present some time ago.  The first night I got it, I mounted it on the helmet, charged up its battery, then went to work the next day.  That evening, tried using the radio on the bicycle as I rode home.  The interference rendered the radio totally useless.

At first I couldn’t figure out why the signals sounded so bad on receive.  I was stuggling to hear repeaters that were normally quite strong.  The only thing that was new was the headlight.  I got home, switched on the set in my room and tuned to 2m sideband, then put the headlight on flash.  The tell-tale static from the radio gave away the headlight as being the culprit.  Worst of all, the emissions weren’t conducted, they were radiated.

Since then I’ve spent the last few months trying to figure out ways to make this headlight less noisy.  The following page serves as a notepad, and I’ll keep adding to this as time goes on.

Initially when I used it I kept the battery in my shirt pocket.  This proved to be a fatal mistake, since the roughly ¼? power cable proved to be a very effective radiator of this interference.  I found winding the power lead up into a tight coil and moving the battery to the top of the helmet right behind the headlight helped things a bit.  Adding a 2.2mF capacitor in parallel to the headlight further reduced emissions to make it barely tollerable.

Then the battery pack died.  After a few weeks of non-usage, the cells collapsed.  So for a while the problem solved itself, I could no longer use the headlight.  The headlight’s battery pack runs at a nominal 7.2V (typical 8V).  Since I run a 12V supply on the bicycle, I set about making a step-down power supply that could drop my 12V supply down to 8V approximate to run the headlight.

Initially I tried using a LM7808 linear reg.  This worked, with one major drawback: the linear reg got bloody hot.  Turns out this headlight draws about 1A of current when at full brightness.  That means in order to drop 4V, the reg was dissipating 4W of power.  Ooops!  Poor thing.

I designed a crude switchmode power supply to do the work.  Using a voltage divider to provide a voltage reference, I used a LM311 comparator to detect when we were under voltage.  The output signal from this would pull down on the gate of a IRF9540N MOSFET which acted as the main switch.  Hysteresis was used to fine-tune the switching behaviour.  Capacitors at both sides would smooth the waveform.  470µF was used on the input side, 330µF on the output (we still have that 2.2mF capacitor not far away).  I also made liberal use of 100nF decoupling capacitors to try and control the rate of switching.

On the breadboard with a dummy load, my circuit performed pretty well.  A fairly smooth output with a bit of ripple at high load.  I mounted it in a box and tested it with the headlight, and presto, the headlight was back in service.  I mounted both on the bicycle directly, so as to minimise cable length and therefore radiation.

Since now the radio shared a power rail with the light, I knew there was potential for conducted emissions to cause problems as well as radiated.  The next evening I tried it out… 2m was lousy with the headlight turned on.  The good news is that it wasn’t much worse than before, but it still rendered the radio useless at times, particularly if the signal was weak from the repeater.

Recently, I added some 470µH inductors in series with the headlight and my switchmode power supply.  I also tried common-mode chokes to no avail.  A 1mF capacitor has been added in parallel to the 470µF capacitor in my power supply to further try and reduce the noise.

Low-pass filtering on headlight: 2.2mF capacitor in parallel, 470µH inductor in series

Part of the problem was that I was shooting blind to try and find the interference.  Today, whilst tuning up the HF antenna, I decided to experiment and see what bands this headlight had an influence on.  To my (unpleasant) surprise, interference was severe right down to 80m.  I used the spectrum analyser built into the FT-897D to hunt for the culprit, and found it lurking at around 400kHz.  When the power supply initially starts up, it’ll be up around there, then it sinks down to 392kHz as the case warms up:

Noise from the headlamp detected around 400kHz

This appears to be the fundamental frequency for the switchmode power supply built into the headlamp.  It would appear to have a fairly sharp square-wave type pulse, as it contains very strong odd-order harmonics.  The same interference can be observed at around 1260kHz (3rd harmonic).

The fact that this frequency is so low, probably suggests it is being intermodulated with a parasitic oscillation at some higher frequency.  This I have not yet found, the CRO showed some other hash over the signal, but I will need to do some further investigation, probably with a more sophisticated spectrum analyser than the primitive one found in my transceiver.

I’d be interested to hear from others who have had issues with these headlamps.  In particular, interference to AM/FM radio reception or transceiver operation would be quite useful if the need to persue this with the ACMA ever comes up.  The manufacturer of the headlights has so far been unresponsive to my queries, so in the meantime I can only recommend that people avoid using these headlights if they intend to use any kind of radio receiver whilst riding at night.

In the meantime, I’ll be chucking further notes here as I find more on this issue.