Antenna cheep vs very expensive for EMI testing

[Bob S]

I am doing some EMI testing and trying to get some idea of my radiated noise levels at a meters or so. I would like to see if I can pickup any of what my near field probes are seeing. My unit has failed in the test labs between 30 and 300MHz so this is the only frequency of interest.

My question is, short of spending thousands on a proper calibrated antenna what might I expect to get out wide band scanner antenna such as this one?

http://www.ebay.com/itm/Taurus-25-1300-mhz-Scanner-Ham-General-Coverage-Discone-antenna-/301908583528?hash=item464b276468:g:Y-cAAOSwB4NWuj9Y

Thanks

[mcinque]

How much your product is exceeding the limits? On what frequency span? 30-300 it's a little bit too generic. The testing lab must provide some accurate report to see where the product is failing.

However, considering that a proper EMI testing require a standardized setup and a proper shielded room/cell, the issue with any uncalibrated antenna is that you will have gain and losses in reception and your readings would be influenced by that.

For example if your product is radiating -let's say- something around 150MHz just a little bit over the limit and that antenna has a loss of a couple of dB just at 150MHz, you would consider the SA reading as "pass" while it's not.

It's better for me to have a proper report from the test lab, knowing what frequency(ies) is/are radiating, so that you can investigate the issue with your NF probes, to find the part of the circuit is radiating that frequency(ies) and try to correct it (shielding, ferrite, loop area etc.)

[nctnico]

The first step is to get a graph or table which shows at which frequencies (or small band) the device has failed. From there you can use cheap (DIY) probes to measure the effect of changes you are making. If the limit is exceeded by -for example- 4dB then you need to make the emissions at least 4dB smaller. You don't need a calibrated antenna for that if you make relative measurements.

[w2aew]

The first step is to get a graph or table which shows at which frequencies (or small band) the device has failed. From there you can use cheap (DIY) probes to measure the effect of changes you are making. If the limit is exceeded by -for example- 4dB then you need to make the emissions at least 4dB smaller. You don't need a calibrated antenna for that if you make relative measurements.

And speaking of making relative near-field measurements - I just completed a video on that...

[acbern]

The problem with near field antennas and using them to reduce some nosue spikes by the 4db as stated above is that just e.g. the way you hold it may make quite a big difference already, hiding such small improvements.

Regarding the scanner antenna mentioned above, one should rather use a (more directional) dipole. That dampens some of the unwanted signals comming from other directions that where the DUT is sitting.

[Karel]

My question is, short of spending thousands on a proper calibrated antenna what might I expect to get out wide band scanner antenna such as this one?

The problem with most antenna's is, you need some distance from the dut in order to let them do their job properly
and because of that you'll need also a farraday cage because otherwise the environmental noise will be too strong.

To overcome this problem, we use this: http://www.tekbox.net/test-equipment/tbtc1-tem-cell

It gives surprisingly good results, although, it's not calibrated and it fits only small boards.
The price is very affordable.

Quote from the manual:

"The TEM cell can be very effectively used to identify potential issues that may result in failing the compliance test.
As a rule of thumb, given that the PCB is positioned not much higher than 1-2cm above the bottom wall,
any spurious with amplitudes higher than 40dBuV are a threat."

[davebb]

Hi when I was working in my last job I went to see a emi test lab and they used log periodic antennas as they are very wide bandwidth,
Dave 2E0DMB

[Richard Head]

I've only ever see them use a bi-conical.

[nctnico]

The problem with near field antennas and using them to reduce some nosue spikes by the 4db as stated above is that just e.g. the way you hold it may make quite a big difference already, hiding such small improvements.

I don't see that as a problem. Just hold them in the same position & location. Been there / done that many times.

[Apollyon25_]

Home made loop antennas (or the nice Tekbox set), pin probes, PCB-based antennas (home made or http://www.wa5vjb.com), TEM cell (Tekbox or home-made http://www.emcs.org/acstrial/newsletters/summer08/pp2.pdf) are all suitable to solve EMC problems.
Oh and there is a really good LISN design here (I've built a 4channel version) (https://www.eevblog.com/forum/projects/5uh-lisn-for-spectrum-analyzer-emcemi-work/30/)

However your 30-300MHz area of interest puts some challenges in terms of a convenient antenna size.

Once you have a list (or plot) of frequencies affected i.e. failing and by how much, you have all you need to solve it.
And you may as well pop a 300-400MHz LPF on the front to get rid of all the other spectral stuff you dont care about.

It is a radiated fail? or a conducted?
http://www.compliance-club.com/archive/old_archive/010422.htm (this is for conducted, though its just one part of a series of excellent articles)

Once you have a probe, start sniffing around, you should be able to account for all the spikes you see.
The frequencies you get from the EMC lab may not be 100% correlated to the frequencies you observe so just look in the vicinity.
Take a peak hold measurement sweep/scan at the same or similar RBW as used by the lab. Repeat with an average reading. Between Average and Peak will be the QP (quasi peak value), and peak is ~= lab 'peak' so you should be able to find the source(s).

Be prepared to hack the board to bits, i.e. cut tracks, swap out values, put in CM chokes etc.

Oh and lastly, I have found frequency decomposition useful - what is a potential fundamental of this frequency of concern. Be prepared to look beyond the fail frequency to its harmonics or fundamental... 

[T3sl4co1l]

Typically, a conical dipole (usually made with a metal wire cage) is used in the 30-300MHz test band.

Polarization is important; it'll almost always be with the antenna parallel to the lengthwise dimension of whatever traces or cables are causing problems.  Give it a spin and check.  Mind reflections from the floor and walls, too; that's why the lab tests at different antenna heights, and uses a fully or partially anechoic chamber.

You should be able to confirm the lab measurements with "at home" equipment, at least qualitatively.  Even if your measurement is off by x dB, the improvement (by shuffling around filtering, caps, ferrite beads, whatever) should be the same relative change.  So if you need to trim away 10dB, take a reference measurement, then trim away 10dB!

Tim

[T3sl4co1l]

I've been wanting to ask you a question for a long time. Is there anything related to EE that you do not know?

Yes, but probably nothing I can't extrapolate from present breadth of knowledge.   Thanks,

Tim