Link This |
Email this |
Blog This |
Comments (0)
Reader comments on EMC issues
June 13, 2007
Jiri Polivka of
Spacek Labs., Santa Barbara, CA, wrote to comment on an article on the June 2007 issue of Test & Measurement World. Do you agree? Will EMC problems get worse?
=========================================
With much interest I have read your article in TMWorld, "Numerical Models Predict EMC Performance," June 2007. While your contribution is very interesting, and the reference paper by Mr. Hubing must also be interesting, I feel I would rather agree with your concluding sentence you took from Mr. Archambault [“There’s no replacement for the insight of an experienced EMC engineer.”]. I am a radio and microwave engineer with a learned insight in how the electromagnetic fields behave. My hobby is "noise fields," a phenomenon I can use but there is no theory so far.
Concerning numerical modeling, I have not seen any model I would trust. At microwaves, the authors of such models use mainly electrostatic models, and even the best ones are quite limited; the authors also do not show what equations and boundary conditions were used, and the worst is that the users of those models are not interested in their limitations.
The engineers and physicists became so fascinated by those simulations that they now fail to understand how nature works. Engineering education is limited to learning how to use various models, but the hands-on experience is mainly missing. Radio amateurs are becoming rare animals, and many professionals are either not interested, or, due to learning much about simulations, have no time for experiments.
Now coming to computer-related EMC, I can see that in the fast development of faster and faster processors, modern PCs claim that their processors "run at 1...2...3...5 GHz" while the real data speed is slower. Nevertheless, circuit boards have been mostly developed by guys who simply need to compress many multipin and complex ICs on the smallest boards available. Only upon putting all the stuff together, everybody is surprised that there is a lot if interference around and on the boards. Fast pulses are generated so that there are harmonics: The switching transistors can switch within several nanoseconds, which means that spectra overflow to more than several gigahertz.
With that, any conductor and bundles of conductors become poor antennas and pickup sticks, and strong interfering signals appear in unexpected places. Instead of stopping the spurious signals at their sources, they are left to propagate in PC cases, and they tend to overflow over cables, often quite far. The next problem is that due to a very dense IC organisation, low-voltage logic was introduced; this makes the fast chips more vulnerable to stray signals.
Gradually we are coming back to the >50-years-old TV receiver concept: while the line generator with the flyback HV rectifier filled the volume with 15-kHz noise, the front end of the signal receiver had to be enclosed in a metal case (tuner) to stay sensitive for VHF and UHF signals.
From my point of view, it is time to teach the data-processing engineers some knowledge of RF technology. They should know that rectangular pulses contain harmonics, that signal-generating and switching circuits be better enclosed in a good shield to prevent unwanted radiation. Sensitive circuits can be also protected in metal cans, and DC power should be filtered by RC or LC filters.
If most of the noise is contained in a PC enclosure, one bad problem remains in connectors. Any PC has many multipin connectors, and they are almost all quite "open" and do allow spurs to travel far away. It may be late but still worth trying to design better, shielded, and where appropriate, filtered connectors.
From my point of view of the "old-style" analog designer, switching power supplies add to all the noise. I wonder why their switching frequency cannot be locked to one common frequency, like the clock in the computer?
Many years ago, I tested a computer-controlled telephone exchange, a small box for 50 or 100 subscriber lines. I used a test receiver and a spectrum analyzer: the surprise was that the microcontroller had a 12-MHz clock oscillator, but the thing heavily radiated at harmonic as well as unrelated frequencies covering 20 kHz up to >50 MHz. Spurs were typically >30 dB over allowable levels.
I rejected the device; only then the owner admitted he purchased a number of such things as a cheap surplus, with an intention to sell many to Eastern-European customers.
We are going to face many more EMC problems, with the faster PCs, with "data over power lines," with growing market of gadgets working at 2.45 and 5.8 GHz. People are VERY concerned about cellular phones with 0.5W RF power, but do not care having a microwave oven generating almost 1 kW, and nobody cares if his/her oven radiates or not...and it often does!
Posted by Martin Rowe on June 13, 2007 | Comments (0)
