Ed’s use of the current probe / H-field antenna examples brings up something I sort of implied earlier, but meant to discuss in more detail, but forgot.
Both of Ed’s examples function like transformers at the low end of their frequency range: the correction factor changes linearly with frequency. This is inherent in how the devices work. If I were to get data from the cal lab that showed departures from linear operation below, say 100 Hz, whereas from 100 Hz to 10 kHz I was seeing linear changes with frequency, I would not blindly input the non-linear calibration. I would likely check to see if the dynamic range of the measurement set-up had been exceeded, and I was getting a plot of the measurement system noise floor. That’s a crude example, but it gets directly to the more subtle problem of the OP and his subsequent post about including all local maxima and minima and also looking for inflection points, etc. That’s all well and good assuming that what you get from the cal lab is guaranteed the antenna performance itself. But if the swept data point step sizes are a tiny fraction of what it would take to get the real antenna factor to change a dB, and yet you are seeing dB changes, then what you are seeing is artifacts of the measurement system errors, not the antenna factor itself. We ought not be correcting our data for errors or uncertainty in the calibration process, and a knowledge of how the antenna works goes a long way towards “smoothing out” the raw data. Case in point, a standard gain horn. The antenna factor can be calculated >from the gain, and the gain is calculated from the physical dimensions. The measurement and tolerances on the physical dimensions are much better (orders of magnitude better) than the measurement errors on the antenna range, and it makes no sense at all to prefer measured rf data over measured dimensional data. The only reason for measuring the antenna gain is if you think there is a problem, which you ought to be able to check by visual inspection. Now a biconical is more complex than a SGH, but the same principal applies that there is some frequency interval over which the antenna factor will be flat, or at least smooth. Apparent ripple in the AF at frequency intervals shorter than what is theoretically possible is “noise” from the test setup. I strongly suggest that more emphasis be placed on understanding our test antennas, and less on manipulating data with unknown sources of error. Ken Javor Phone: (256) 650-5261 ________________________________ From: "Price, Edward" <[email protected]> List-Post: [email protected] List-Post: [email protected] List-Post: [email protected] Date: Mon, 8 Jun 2009 13:09:06 -0700 To: <[email protected]> Conversation: [PSES] Measurement Accuracy and antenna factors Subject: RE: [PSES] Measurement Accuracy and antenna factors I think that's the method I also use when the correction factor is moving over a large range, say, for the low frequency end of a current probe or H-field antenna. And one more thing; all my cal data is swept data (or at least, stepped in really, really tiny increments). I get a continuous plot, plus periodic tabulated points along the curve. As Ken says, you can see the undulation of the curve, and scale intermediate points such as every 1 dB. (Most times, I can visually resolve a half dB, so picking the 1 dB transition points feels pretty accurate to me. Ed Price [email protected] <blocked::mailto:[email protected]> WB6WSN NARTE Certified EMC Engineer Electromagnetic Compatibility Lab Cubic Defense Applications San Diego, CA USA 858-505-2780 Military & Avionics EMC Is Our Specialty ________________________________ From: [email protected] [mailto:[email protected]] On Behalf Of Ken Javor Sent: Monday, June 08, 2009 12:43 PM To: [email protected] Subject: Re: [PSES] Measurement Accuracy and antenna factors That is only true if the calibration is done like in the old days, like my old biconicals tuned at 20, 30, 40, 50, ... MHz. The query of the original post is, given you have received swept data from the cal lab, how many of those points do you import into your AF data correction file, and at what point do you start using interpolation. In that case, what I was referring to was that you could simply eyeball the data and enter data points that are, say, 1 dB apart. Which is what I do. That is a specific answer in terms of the OP and your question. But even granting your premise, that it is somehow an unknown antenna, and I am the antenna calibrator working without the original antenna factor data for the antenna, there is still no need to record volumes of data. All I need do is set up the two antenna test, run a sweep, and observe where the isolation is relatively constant, and where it changes rapidly, and once again I can concentrate my efforts so that across the entire antenna range, I take data at points roughly 1 dB apart. Ken Javor Phone: (256) 650-5261 ________________________________ From: "Grasso, Charles" <[email protected]> Date: Mon, 8 Jun 2009 12:29:43 -0600 To: Brent G DeWitt <[email protected]>, Ken Javor <[email protected]>, <[email protected]> Conversation: [PSES] Measurement Accuracy and antenna factors Subject: RE: [PSES] Measurement Accuracy and antenna factors But Ken – You don’t know a priori where the AF slope changes. ________________________________ From: [email protected] [mailto:[email protected]] On Behalf Of Brent G DeWitt Sent: Sunday, June 07, 2009 8:35 PM To: 'Ken Javor'; [email protected] Subject: RE: [PSES] Measurement Accuracy and antenna factors Well said Ken. I have also dealt with the absurd number of points taken by automated systems. Many years ago I wrote an QuickBasic program (yeah, that many years ago), that decimated the data based on exactly the same thought. It did a simple piecewise first derivative as well as looking for total changes around .5 to 1 dB, depending entirely on how skeptical I was. It resulted in a huge reduction in frivolous data. Brent DeWitt Westborough, MA From: Ken Javor [mailto:[email protected]] Sent: Sunday, June 07, 2009 1:43 PM To: [email protected] Subject: Re: [PSES] Measurement Accuracy and antenna factors I’m not a fan of all this tenth of a dB concern with uncertainty. I also disagree that antenna factors are selected randomly to be entered into a data file. It seems obvious to me that intelligent data entry would use an analog simulation of what the questioner is after: Lots of data points (high density) where the factors change rapidly with frequency, fewer points where the factor is relatively constant. I once had a commercial facility calibrate an antenna, and they did so at hundreds of frequencies, with the values bouncing around hundredths or tenths of a dB from data point to data point. I’m sorry, but that seems a moronic waste of time and money. All they were plotting was the error bounds of their measurement system, not the actual performance of my antenna. It’s time a for a little common sense to be displayed on this topic. Apologies in advance if I have hurt anyone’s feelings! Ken Javor Phone: (256) 650-5261 ________________________________ From: "ce-test, qualified testing bv - Gert Gremmen" <[email protected]> Date: Sun, 7 Jun 2009 18:47:36 +0200 To: <[email protected]> Conversation: Measurement Accuracy and antenna factors Subject: Measurement Accuracy and antenna factors A lot of effort has been put into specification of measurement accuracies in radiated emissions. CISPR 16-4-2 has a number of uncertainty budgets listed. One factor that I have not seen in any budget is the error introduced by interpolation between antenna factor calibration points by the measuring receiver. In general the characteristics of a calibrated antenna are entered into the measuring receiver as a number of F/AF pairs, more or less randomly selected from the calibration graph. Then the AF values for frequencies in between those pairs a quadratic spline function is used to interpolate. The function requires 4 calibration pairs to operate correctly of which 2 must be lower and 2 must be higher then the interpolated frequency. Especially near 30 MHz, where modern antennas have steep AF graphs, a calibration point below 30 MHz is not always available and I assume the software duplicates the 30 MHz pair to say 25 MHz to complete the function’s requirements. This must introduce interpolation errors near 30 MHz. I do now know the error that might be introduced by this Type of function. I know that Taylor series have alternating sign In their expansion, and that the values diminish each term, so the error of approximation remain smaller as the last term used to interpolate. But Taylor does not suit itself for approximation of non computable data (such as AF). My questions for the group are: What requirements are to be met for the F/AF pairs to minimize errors? What are the errors introduced by interpolation? How do YOU handle this additional uncertainty…? Gert Gremmen Ce-test qualified testing bv Van: [email protected] [mailto:[email protected]] <mailto:[email protected]%5d> <mailto:[email protected]%5d> Namens Bill Owsley Verzonden: zondag 7 juni 2009 4:36 Aan: [email protected]; [email protected]; GheryPettit Onderwerp: RE: CISPR 22-2005: testing on interconnecting DC cables? I routinely measure the same, but I have not been able to establish that there is any requirement for a direct measurement. In general, if the EMI >from the DC cables causes a problem it will show in the usual required tests. A test on the DC cables just focuses on the problem area and helps with debug efforts, but I have not been able to claim that it is required by CISPR 22 (or related standards) ps. Some of the DC cables are much longer than any standard one normally used and so come fall under some of the immunity tests, so by quantum leaps in logic, we apply the emissions test to them. But when it comes time to ship, no problem... - Bill Indecision may or may not be the problem. --- On Fri, 6/5/09, Pettit, Ghery <[email protected]> wrote: From: Pettit, Ghery <[email protected]> Subject: RE: CISPR 22-2005: testing on interconnecting DC cables? To: "[email protected]" <[email protected]>, "[email protected]" <[email protected]> Date: Friday, June 5, 2009, 2:25 PM Pat, Annex C deals exclusively with telecommunication ports. This is clear in the first sentence of the annex. If a port isn't used for telecommunications (see article 3.6 in CISPR 22:2008 for the definition) then Annex C doesn't apply. And while the term "mains" isn't defined in the standard, it commonly is taken to mean the low voltage distribution network in a building that is supplied from the public power supply. Thus, the mains port is the port that plugs into the wall socket. I don't see how the DC output port on your power supply is either a telecommunications port or a mains port, so this test by your customer doesn't make sense to me, at least not as a 'requirement' in CISPR 22. I hope this helps. Ghery S. Pettit -----Original Message----- From: [email protected] </mc/[email protected]> [mailto:[email protected] </mc/[email protected]> ] On Behalf Of [email protected] </mc/[email protected]> Sent: Friday, June 05, 2009 10:48 AM To: [email protected] </mc/[email protected]> Subject: CISPR 22-2005: testing on interconnecting DC cables? Good Friday morning all, We have a customer who is measuring conducted emissions on the DC output of our external switching power supply (laptop-style power supply), claiming it is required by CISPR 22. As I read through CISPR 22-2005 for rebuttal material, the phrase telecom port was defined and the measurement details looked clear. Until I got to Annex C. Clause C.1.5 is titled 'Flowchart for selecting test method', and says the flowchart in Figure C.6 is applied to different ports. The flowchart has a decision block at the top based on whether the port is a telecom port. If not, no testing is necessary. If the port is a telecom port, you choose between 4 methods: - Unscreened pairs - Screened or coaxial - Mains - Other Certainly, Mains ports need testing regardless of whether the EUT has telecom ports, so the flowchart has logic errors. But does the port choice 'Other' mean you must test any port not already covered? Can a single statement in a flowchart define testing requirements not detailed elsewhere? BTW, the flowchart says 'Other' ports must meet the telecom test limits. Pat Lawler EMC Engineer SL Power Electronics Corp. - - This message is from the IEEE Product Safety Engineering Society emc-pstc discussion list. To post a message to the list, send your e-mail to <[email protected]> All emc-pstc postings are archived and searchable on the web at http://www.ieeecommunities.org/emc-pstc Graphics (in well-used formats), large files, etc. can be posted to that URL. Website: http://www.ieee-pses.org/ Instructions: http://listserv.ieee.org/request/user-guide.html List rules: http://www.ieee-pses.org/listrules.html For help, send mail to the list administrators: Scott Douglas <[email protected]> Mike Cantwell <[email protected]> For policy questions, send mail to: Jim Bacher <[email protected]> David Heald <[email protected]> - This message is from the IEEE Product Safety Engineering Society emc-pstc discussion list. To post a message to the list, send your e-mail to <[email protected]> All emc-pstc postings are archived and searchable on the web at http://www.ieeecommunities.org/emc-pstc Graphics (in well-used formats), large files, etc. can be posted to that URL. Website: http://www.ieee-pses.org/ Instructions: http://listserv.ieee.org/request/user-guide.html List rules: http://www.ieee-pses.org/listrules.html For help, send mail to the list administrators: Scott Douglas <[email protected]> Mike Cantwell <[email protected]> For policy questions, send mail to: Jim Bacher <[email protected]> David Heald <[email protected]>
