I’ve got the odd Force 300-25 and CSM subscriber with wildly fluctuating download speeds (e.g. Radio set to 50Mbps but speeds 12Mbps - 35Mbps) . Noise isn’t the issue. In some cases, I’m suspecting tree growth causing fresnel issues etc. Regardless, I’m finding that going onto said subscriber radio and forcing the max download MCS to 7 (64QAM) brings their download up to almost 50Mbps with far more stability.
In the image below, the radio marked in purple had the problem. There’s nothing obvious to indicate it has a problem and logging into other radios’ routers (customers all have a Mikrotik router so I can run btest etc) all steady at 50Mbps, even the ones with lower data rates.
I know a certain other equipment provider had an “alternative” data rate algorithm as an option that was more pessimistic in terms of MCS rates. Presumably it backed off radio rates for a longer period of time according to packet loss etc.
I’m wondering if Cambium would consider having a more pessimistic data rate algorithm as an option?
It could be a fluctuating MCS issue on SM side and lowering the Downlink Max MCS to stabilize modulation provides better throughput.
Provide pic of Monitoring > Performance > Downlink/Uplink packets per MCS section of both AP and SM in question.
Also compare that to a known good performing SM. Your target MCS should be the highest percentage, at least +50%. Be sure to reset counters at top of performance page and let run for 5 mins to get the most accurate readings.
EDIT:
I always set the Uplink modulation on SM side to MCS 4. All of our packages are asymmetric (10M is usually the max Upload) so any uplink modulation above MCS 4 is completely unnecessary. This also drops latency considerably.
I have dealt with tree growth before and dropping downlink modulation is a good temp fix. So sounds like you fixed it, Im just explaining the “why”
Appreciate the feedback. I’m just trying to reach out to Cambium developers to point out that this should happen automatically in the software (kinda like BGP back offs). We’ve got ~2500 UBNT radios out there that we’re gradually replacing with ePMP3000 (and hopefully 4000 soon ).
It’s not ideal to have to mess around with max MCS rates after a customer rings up complaining.
I like your logic regarding lower upload modulation and latency.
AP over last 66 days. Affected station over past 21 hours since modulation locked to MCS7
On the AP across all stations (20) retransmissions are about 1.5% on uplink and 2.3% on downlink
Retransmits on this station are high at between 3 and 8% on uplink and downlink is negligible (0.1%).
Link test when max download MCS is 7
Downlink 50.726 Mbps
Uplink 10.612 Mbps
Aggregate 61.338 Mbps
Perhaps I do need to lower the max MCS for upload but again, this isn’t something we should be messing around with by hand on a per station basis. That’s what software algorithms are for.
You can change any setting for all your SMs at one time via CnMaestro. And you can do it with one particular SM.
Radio is always trying to get the best link performance so tries to use highest MCS. In case not good environment it is points to re-transmits and wasted time due to switching between MCSs.
Hi Andrii, IMHO it should monitor the amount of re-transmits and reduce MCS rates more aggressively and/or wait longer before trying to increase MCS rates. This could be a simple tickbox option on the AP. Similar to BGP-backoff.
For all their faults, Ubiqu!ti managed to do this with their Airmax gear. I don’t have to flaff around setting max MCS rates on Ubiqu!ti or Mikrot!k gear to get the best performance. Cambium promote the concept that their stuff “just works”.
On Ubiquti Airmax M gear, under wireless it’s labelled “Alternative” under the Data Rate Module option.
On Ubiquti Airmax AC gear it’s also still there as an option.
I believe after much tweaking and testing (a.k.a. experimenting on end users ) they’ve just integrated aspects of the alternative data rate algorithm into their default wireless software as I’ve not had to use it for a few years.
In general EPMP3000 seems to perform worse in the same situations as EPMP2000, it seems to be the scheduler isn’t as robust and is way too optimistic like OP said.
).
) they’ve just integrated aspects of the alternative data rate algorithm into their default wireless software as I’ve not had to use it for a few years.