Dear Cortland People can't simply say: "ordinary semiconductors won't demodulate RF levels produced by an unintentional radiator" – even the smallest amount of RF can be demodulated – there are no hysteresis or threshold effects in a PN semiconductor junction or FET that is biased into its conduction region (at least not until you get below signal levels equivalent to less than a single electron).
What I am sure most engineers would really mean to say is: "ordinary semiconductors exposed to RF levels from an information technology product which is fully compliant with all relevant EMC emissions standards and is at 10 metres distance will generally not demodulate a sufficient level of interference to make an appreciable difference to most electronic systems." Now we have a statement which has some scientific rigor and some engineering validity to it. (Although I do worry that in Europe our harmonised EMC standards only test emissions up to 1GHz, so what does that say about the possible emitted fields strengths from a PC with a 1.2GHz clock frequency?) Let's see if we can put some meat into this discussion with a real-life example... I once tested a blood sample incubator for RF field immunity. The incubator was used during screening programs (for cancer and other diseases) and kept about 100 test tubes at 37.1C (normal blood temperature), while the reagents in the test tubes changed colour. After 24 hours of incubation medical staff would inspect the test tubes and write letters to people telling them they were sick, or that they were clear of the disease. I don't know what temperature tolerance the reagents had to give an accurate medical diagnosis, so assume ±0.1C. On the front panel of the incubator was a display of its temperature, which was of course 37.1C. We found that field strengths as low as 1V/m would cause the incubation temperature to range over full scale, from heaters fully off (in which case the temperature would decline to ambient) to maximum (in which case the water used to incubate the test tubes would boil). We could use the RF test frequency to control the temperature between plus and minus full scale over the frequency range 80 to 1000MHz at 1V/m (and did not test beyond 1GHz). Most worryingly, the front panel display would only show temporary variations from its 37.1C when the RF field was turned off or on, and would continue to show 37.1C even when the water in the incubator was stone cold or actually boiling. Most demodulation effects in bipolar and FET devices approximate to a square law - for example a 1dB fall in the field strength (keeping everything else constant) would typically result in a 2dB fall in the demodulated 'interference' error signal, as John Woodgate has recently pointed out. If we assume that the 1V/m field strength was causing a 60C temperature error, how low would we need to make the RF field to get down to the 0.1C accuracy of the front panel display? Assuming square-law characteristics for the device doing the demodulation I calculate a field strength of around 40mV/m or 92dBmicrovolts/metre. You will notice that I have been generous to the incubator and assumed that the 1V/m field just about caused its temperature error to increase by 60C to boil the water, whereas it could have been overdriving the internal circuits by a considerable margin and still suffered a 60C error at 0.1V/m. We didn't test this possibility as our focus was (as in most of these cases) on quickly modifying the product so it passed the immunity test - which we did. 92dBmicrovolts/metre is not a very high RF field level for a PC without any EMC precautions at a distance of 10 metres. How many people reading this would be now be quite happy to place even a fully-compliant PC (compliant at 10 metres distance, that is) right next to the unmodified incubator? If it helps, imagine that it is your young daughter whose blood sample is in the incubator to discover which drugs she needs to survive. Shall we have a vote on how close we would be prepared to place the PC? Might be interesting. Let's not even think about the problems of proximity to cellphones and other intentional radiators. I didn't mention that the incubator was a small model used for mobile screening, for installation in a truck adapted for medical screening purposes which travels to various communities and parks there for a few days while it tests the local people for disease - hardly a very well controlled electromagnetic environment. What does the above imply for similar incubators in countries that do not have mandatory EMC immunity standards? Or for older incubators in the EU that have never had to meet the EMC directive? (Please don't reply with the old chestnut that "we haven't heard of any problems so far, so everything must be OK" - people who should have known better were using that phrase before September 11th. It is just not an acceptable argument where safety issues are involved, as any expert in safety law will tell you. Try: "I've been driving past that school at 40mph for ten years and haven't hit a kid yet, so it must be safe mustn't it?" as a test of the concept.) Regards, Keith Armstrong In a message dated 04/01/02 17:54:23 GMT Standard Time, [email protected] writes: > Subj:Re: EMC-related safety issues > Date:04/01/02 17:54:23 GMT Standard Time > From: [email protected] (Cortland Richmond) > To: [email protected] > CC: [email protected] > > I don't believe this is what people are saying here. What they are saying > is, ordinary semiconductors won't demodulate RF levels produced by an > unintentional radiator. Cortland > (What I write here is mine alone. > My employer does not > Concur, agree or else endorse > These words, their tone, or thought.) > > [email protected] wrote: > >> Does anyone else think that ordinary semiconductors doesn't respond to RF? >> I have tested a product which was little more than an LM324 quad op-amp >> for RF immunity using IEC 61000-4-3. This op-amp has a slew rate of >> 1V/micro-second on a good day with the wind in its favour. It was housed >> in an unshielded plastic enclosure. Demodulated noise that exceeded the >> (not very tough) product specification were seen all the way up to 500MHz >> at a number of spot frequencies that appeared to be due to the natural >> resonances of the input and output cables. >>
