We have a PMM 7000 precompliance receiver that we use for all kind of EMC testing. However we recently had a CE failure at a test house that surprised us as we didn't see it on our setup. That made us start poking around and we came up with two potential issues:
1) In the 150kHz - 30 MHz region the minimum step size for the receiver is 10kHz. The RBW is set to 9kHz as per CISPR. Is that not an issue? If the filter was perfect then it would leave narrow bands between the "bins". Perhaps the steepness of the cutoff of the filter is not that steep? Anybody got any thoughts?
2) We have seen significant differences in results sometimes with different dwell times. I wondered if this could be the issue. I struggled to get a straight answer out of the test house as to what we should use. I guess as long as it needs to be to see worst case? Any advice?

Looks like I am talking to myself and (partially) answering my own question - no change there!! Re (1) it is clear that the roll off of the filter in the receiver is not that steep as if for instance I set the receiver to 200 KHz and use a sig gen to put in 195 KHz at 80dBuV I see 76dBuV from the peak detector. So no hiding at the fringes of the bins, just a slight error in amplitude reported.

This is a significant part of what I called Murphy's corollary "It takes 98% of the time to do the design and it takes 98% of the time to do the last 2% of the project". Spectrum analyzers are mysterious beasts. To have it set correctly and to have a proper screen room, is a week's job in itself.
You have my sympathy and some advice: barely anyone passes the first EMI test.

Marty - I agree your 98% rule is not far from the truth. It is a shame customers won't believe you when you tell them that though.....

Marty, you're right there: rarely does a new project meet EMI on the first test. Charlie, I'm going through a similar exercise right now, trying to determine the differences between a test lab and our home-built EMI lab. All I can say from experience is that the spectrum analyzer often seems to have been set up by another group for some other project, resulting in something like the bandwidth setting being different. My first 2 cents of advice is to make sure your step-size is narrower than your switching frequency so that the harmonics of that frequency which you expect to see appear at their actual narrowband levels.

Darrell - Fortunately we have a receiver where we can save the setup files but even with that ..... We have the set size set to 10 kHz which is OK for most things we do but not for higher power motor drives.

So a couple of things to look at that can cause the behavior you are seeing.

First any receiver has a definite and limited out of band rejection. That is out of band signals are mixed with signal before the receivers frequency rejection stage. Some receivers have little to no filtering before the modulation to base-band part of the receiver. This can cause the inaccurate measurement of some signals. We found this while measuring some power supply emissions near and airport. To fix this we needed controllable notch filters before the receiver. Large VFD drives can often cause the same effect.

Second if the emission is not continuous and the dwell time of the measurement is long, which it often is in pre-compliance receivers, then you can miss it or badly measure the spectral power.

Hope this helps.

Hamish - reading you comment again I am wondering if I have misunderstood something and you hare it correct. I would have thought the longer the dwell time the more likely you will "see" a discontinuous emission whereas you are implying the opposite. Could you clarify please?

Charlie - This is a good question. The detector or receiver in an EMC measurement system has a transient response to non steady EM signals. Typically there is some sort of low pass filter in the receiver hat determines the signal amplitude and so sets the transient response. So when the EM signal is changing the detector output is not always a valid representation of the signal strength.

Long dwell time receivers often use the entire time to do signal acquisition and often use the long dwell time to maximize signal sensitivity. If the EM discontinuous emission is not present for the entire dwell time then some receivers can report lower signal levels.