Software Used in EMC Testing


We occasionally get e-mail on specific topics, and sometime people just wanting to vent their spleen on a given topic, the favourite being the availability of vehicle manufacturer (VM) own specifications.  However recently the site got a small chain of discussion going about the use of software running on a piece of Equipment Under Test (EUT) during EMC testing.  Below is the edited version of the essential parts of the threads (we've protected names and companies as this is a general point of issue, not a specific dig at any one company or individual).  We'd welcome further input to the debate.


Respondent 1


Under the heading of "the OEM's can't agree on the colour of s***".


A colleague and I were already scheduled last week to go to VM1 for this meeting about the new test specs. The boss decided that since we're already going down there, you both can go to VM2 the day before for a design review meeting and to discuss the EMC test plan on a controller we make. During this meeting a couple of issues came up. One was whether to use production intent software during EMC testing. The VM2 guy said no, that wasn't necessary as long as the module was awake, the I/O lines were all exercised and functioned normally. The other was to test while the motor was under load. (We make the control module, not the motor). He said no, as long as the motor moves as far as it is supposed to or say puts out the correct number of hall effect pulses, its okay. 


The next day at VM1 we were told the exact opposite. Yes, you must use production intent software. That is always a pain because the software is constantly evolving (ask any Windows user) and may go through 2 or 3 revisions between design and production validations. Yes, test with the motor under load. There is an explanation for this ... motors produce more noise with a load than without. EMC = Even More Confused.


Anyway at the end of the meeting, I ran into a fellow lab rat who I had been chatting with during the breaks. When I mentioned that VM2 told us the opposite the day before, he laughed and said "Well maybe that would explain my little experience a couple of weeks ago. I was heading for Detroit in my VM1 rental car. I plugged in and turned on my FM iPod to listen to some music. A while later it started to rain so I turned on the windshield wipers. The wipers only went up half way, paused then went back down. Thought there was something wrong until I turned off the iPod and then the wipers worked normally. Perhaps VM1 doesn't test hard enough."


Some days all you can do is shake your head and wonder if you picked the right profession. Especially since the auto industry seems to be in a downturn now. 

Respondent 2


I used to design Sat-Nav, spoken word only.  In a test chamber I would have the software permanently speaking to get the most exercise of the I/O, but in application these things hardly ever speak.  I proved that there was over 15dB difference between speaking and non-speaking once when I retested for another application that used the identical hardware but had no speech, except for through-put of phone voice, this 15dB was centred on 100MHz in the radiated emission spectra measured, not in the audio band (due to use of D-Class switching audio amp).

Respondent 3


I also did something similar (by accident) at a previous company.  I was doing some test work and I had a board that was running a given programme to exercise the EUT, there was a problem with a single line, but the programme was essentially OK.  This line was fixed and the same code recompiled, but with a different compiler option (compile order), the difference was pretty remarkable with some peaks shifted by +10dB, others by -10dB and all at different frequencies over the 400MHz to 600MHz area of interest.  They were also just odd spots, nothing broadband?  Fortunately both sets were some way off the limit lines but it made me realise just what a difference software does make to EMC, but who takes any notice?


Respondant 4: AutoEMC


There is a lot there I have come across before and the software issue in particular is one I have a bit of a conundrum with myself.  I actually think that a separate test suite of software exercising the functions is best, but the letter of the law (re. most OEM spec and EU Auto Directive) suggests it should be the production software.  There is also an expectation that any software change should include a re-test for EMC (like that ever happens). 2007                                                           TOP OF PAGE                                                                 HOME