Forums › BB Series Discussions › BB60C, EMC measurements:How to lower the noise floor of my setup ?
- This topic has 5 replies, 3 voices, and was last updated 4 years, 1 month ago by kaiser.
- karlist April 24, 2019 at 5:17 am
Hi. We are trying to replicate EMC measurements made in certified testing house, but we are having some issues.
We have anechoic chamber without apsorbers on the walls.
Our antenna: https://www.aaronia.com/products/antennas/BicoLOG-30100/
Our Spectrum analyzer: BB60C
I added 4 images in attachment:
1. Noise floor without antenna factor table and cable loss.
2. Noise floor with antenna factor table and cable loss.
3. Test house results(quasi peak).
4. Our results.
As you see from images above our noise floor is very high after we enter the antenna factor. Antenna factor is entered correctly, because its datasheet already has antenna factor table values.
Our RBW, frequency range settings are correct, but dwell time is not the same as in test house. That should not affect results that much – I checked it with Meter with dwell time of 10s.
So our 2 major conerns:
1. Why is noise floor so high?
2. What is the reason our results are so much different from test house results?
We think that maybe there are some settings that are entered incorrectly, because if use ref. offest of -20dB we get somehow similar results to test house, but of course there is no reason to randomly put a -20dB offset.
I would really appreciate help.
Attachments:You must be logged in to view attached files.
Justin CrooksModeratorJustin Crooks April 25, 2019 at 9:28 am
With the BB60C, maximum sensitivity is achieved at 55 dBuV maximum input level, or -50 dBm maximum input level.
As far as measurements being 20 dB off… is the antenna the same distance as in your test site, or did you compensate readings for distance to DUT? Did you factor in any preamplifiers or cables used? Using a chamber without absorber will skew your results as well, but not by 20 dB… however, using a test distance of 3 meters instead of 30 meters would give you a 20 dB offset.
kaiserParticipantkaiser April 25, 2019 at 12:16 pm
My guess would be a sensitivity drop. The noise figure of the BB60C is pretty high, and unless you have a LNA before it, you can lose a good amount of SNR that way.
We had a user using the BB60C for test, and saw the 20dB issue. Since they weren’t setting the noise figure of the test using an LNA, the noise floor of the BB60C was determining sensitivity, and it was fairly bad. Then when they changed the Ref level setting, which controls the attenuator, they kicked in 20dB of attenuation up front, it naturally increased the noise floor by 20dB, and they came to me with this exact problem that you’re posing here.
We tossed in a cheapo LNA in front before the BB60C, and it all worked out nominally regardless of BB60C setup.
A good check for monitoring ANY signal is to take your end measurement device, and kick in 10dB or more of attenuation. With spectrum analyzers you can do that by either manually setting the attenuation, or adjust the ref level to make it change the attenuator. Other devices, you put a 10dB pad in front. If the SNR measured changes at all, then you have a bad setup and need to stick more amplification in somewhere to compensate for test losses.
edit: btw, it’s not an anechoic chamber without foam. It’s a faraday cage. It’s the anechoic foam that makes it an anechoic chamber. An anechoic chamber without absorber on the walls (aka a Faraday cage) is typically worse than just doing it in the open unless you’re time-gating to ensure you’re not getting any reflections. I don’t doubt that small changes in the geometry of anything in the chamber can result in 20dB swings due to standing wave interference patterns. You’re building a mode stirring chamber without stirring the modes or the fancy math to back out performance parameters. I understand that you’re trying to shield yourself from external interfering signals, but you need some absorber in there otherwise you’re just in a big echo chamber.
You’re in essentially the worst environment for testing anything that’s not a pulsed, low duty cycle waveform. Get some absorber, mount it to some plywood and build a half-box around your antenna. It’s cheap, easy, and will dramatically improve test repeatability.karlist April 25, 2019 at 10:36 pm
Something is off with the replies in this forum. I edited my reply and it got deleted..
Thanks to both of you for replies!
Reply to Justin Crooks:
is the antenna the same distance as in your test site, or did you compensate readings for distance to DUT?
Antenna is 3m away from DUT in both cases – certified test lab and our lab.
Did you factor in any preamplifiers or cables used?
We do not use external preamplifiers so we did not have to factor them in. Regarding cables- yes, we added cable path loss table.
But can you tell why the noise floor is so high? And what result should we even expect in empty chamber? 30dbuV/m seems a lot.
Reply to kaiser:
So if I understand correctly you advise us to add a LNA to our setup right?
Can you give me one example of cheap LNA? I only found this one and it does not seem cheap to me :
I will do the test you described about ref level change and SNR.
Yes, I know we do not have anechoic chamber, we are gradually upgrading our chamber, but it should not affect the measurement if we are measuring empty chamber. We were surprised that we had 30dbuv/m noise floor in empty chamber.
BTW, we can even see the standing waves if we move the antenna 😀karlist April 25, 2019 at 11:03 pm
I just compared results between with/without attenuator.
I added a 20dB attenuator infront of BB60C.
I also added a 20dB offset in Spike software because of this attenuator.
Graph is almost the same, but now I have increased noise floor, so SNR is worse. From what you wrote I understand that this is sensitivity issue and an amplifier might fix this.
Also it seems that increasing the Max Input parameter increases the noise floor. I have to increase it because otherwise I get the IF overload warning. I hope amplifier wont make this worse.
kaiser April 26, 2019 at 8:50 pm
- This reply was modified 4 years, 1 month ago by kaiser.
Yea, I understand where you’re at regarding the noise floor. The thing is that you’re just continually bouncing waves in the chamber, which will build up (you’re in a microwave oven) and increase the noise floor.
Think of the difference in background noise in a small library or church that’s all tile, granite and glass. You can hear everyone doing everything, and there’s a small “din” as all the sound bounces around with nowhere to go. You’re noise floor is higher. Take that same building with the same people, hang some tapestries, put that carpet down and it will be tens of dB quieter and without that “din”. You don’t need to upgrade the whole chamber, but some basic foam like I said can help a lot.
For the LNA, you don’t need anything fancy. I don’t know your frequency range and necessary specs, but we typically would use something like this, maybe even two of them depending upon line loss and the gain budget:
Put it as close to the receiving antenna as you can (right on the connector if possible), so it drives the signal down the cable and into the BB60. This should increase your SNR.
If you’re already near max input power level, the LNA could cause some issues there. You may want to toss a filter in in front of the LNA to knock down whatever else is causing the issue. But some red warning light in my head is going off because what’s being described isn’t totally matching with what I’ve seen in the field. The BB60C should have more than enough dynamic range for nearly any test. If you’re driving it near IF overload, but you’re still having sensitivity problems as evidenced by the attenuator test, then it feels like something else is going on. Have you checked for other interfering signals, zooming way out on the Spec An? Maybe your chamber isn’t as tight as you thought, or there’s some injection leakage. Is your test equipment *outside* the chamber, and just the antennas inside? Is the chamber appropriately sized for your lower frequencies (too small and it will act more like a waveguide)? Or maybe the chamber is just bouncing the signals around too much and you’re getting to much of that echo bouncing around nailing you. Have you tried it outside the chamber with the power tuned down some?
A box of anechoic foam (4 pieces) isn’t too expensive, and clever placing of it can help a lot. We don’t always use a chamber (in fact we often don’t), like I said slap a couple of pieces on plywood and make something that looks like an open box laid sideways that the antenna fits in and you’ll knock out the vast majority of interference. Add in a height differential or a piece of foam at the halfway point on the ground, and you’re pretty good to go.
So, try the LNA, that will definitely give you more sensitivity for better measurements. But my guess is that there is another gremlin or two out there that you’ll need to slay to get good results. It probably will be something you would never have thought of. Keep at it, RF test is hard!
edit: You also could always have a bad device; but that’s one of the gremlins to slay that I was talking about. Get some of the other stuff figured out, and see where it leads you.
Bypassing open air and just connecting Tx (or an appropriate surrogate) to Rx with an appropriately large attenuator (start with lots for safety, then dial it back) should give you the best signal characteristics and can help determine if you have a bad device or not. Could also be a bad Tx that’s putting out broadband noise also… Then you can also add in components walking the chain to see where things change. And antenna / atmosphere is a component in your link budget, so eventually you add that back in also 🙂
You must be logged in to reply to this topic.