We have a backhaul link comprised of two 2.4GHz Force200 radios. Running firmware 3.2.2, 3.3 or 3.5 in any combinations we see a problem in ePTP mode where the received signal level (in the uplink direction particularly) is dipping by up to 25dB for brief moments, and the modulation is dropping to lower capacity accordingly.
Attached is a screenshot of the eAlign screen of the BH Master, showing the 10dB-25dB dips in signal level. We first discovered the problem when reviewing the throughput capacity graphs generated via SNMP polling every 60 seconds, which indicated fluctuations as high as 123Mbps and as low as 20Mbps.
We also see similar dips at the BH slave end, though not as frequent and not synchronized with when they appear at the Master end.
When the link is switched back to TDD-PTP mode the fluctuations vanish.
We tried various software versions, and replaced both radios and their PoE injectors, no change. I've never seen this on 5GHz links, just this lone 2.4GHz link. Clear LoS (tower to water tank, 130ft mounting height both ends, 6 miles apart.
That’s odd but the only explanation I can come up with is that CSMA is causing this in ePTP mode. This mode uses a hybrid/highly optimized version of CSMA and I’m wondering if intermittent high noise in the area is causing this. I can see the transmitter backing off due to this. Also, this will be done independently by each transmitter as a reaction to localized interference which may explain why you don’t see it on both sides simultaneously.
Now, all that said, it doesn’t explain why the RSSI dips at varying levels. With the transmitter backing off, it should drop to -90 every time. Weird. How’s the 2.4 GHz spectrum around the ends? Any luck when you change channels?
We’ll give this a try in the lab to reproduce the problem.
Attached is the past 48hrs throughput graph for that link - you can see when the change was made from ePTP to TDD, about 2pm thursday. (I don't know what caused the hole in the dataset last night) The thin lines are representative of the capacity at that time, based on modulation. Data is polled every 60 seconds and fed into RRD. Radio is running 40MHz wide channel, have tried all available channels with essentially the same result.
Interference would of course not surprise me, what surprised me was that it was actual received signal being impacted, not the noise floor. If the radios back off transmit power in response to high interference (why??) then what we're seeing would make sense.
Also attached is our signal level graph for the same 48hr period, same 60-second polling cycle. To clarify, between 1pm and 2pm we replaced both radios, leading to the holes/spikes midafternoon. The evening holes are currently unexplained.
I wanted to mention that the daily usage peak on this link had been 50Mbps-60Mbps in ePTP mode, but since changing it to TDD-PTP mode the same client load on that tank has been pulling 110Mbps peak most days. So the ePTP mode was most definitely restricting the available bandwidth.