What Speeds are you guys offering to Customers V.2 (2026 edition)

Products PMP
1

I am curious of what kind of speed-packages other WISPs are offering their customers on Cambium PMP 450 series theese days? Espesially the 450m platform.

There has been a topic on this earlier, What Speeds are you guys offering to Customers , as well as others I think, but I cannot seem to find newer and more recently updated posts regarding this, so I’ll try starting one.

We are currently using PMP450i’s and PMP450m’s througout our network, both with a mix of 20MHz and 30MHz. We have never had any luck running 40MHz because of lack of clean spectrum.

We are offering different packages in different areas, based on the expected number of SM’s connecting to each AP.

Here is our plans:

Scenario 1 (20MHz or 30MHz depending on actual congestion):
450m AP’s expected to have max 80-100 SM’s
450i AP’s expected to have max 20-30 SM’s
Range 5km, all SM’s 450i

  • 10Mbps/4Mbps
  • 25Mbps/5Mbps

Scenario 2 (20MHz or 30MHz):
450m AP’s expected to have max 20-30 SM’s
No 450i AP’s
Range 12km, SM’s mostly 450b HG

  • 15Mbps/4Mbps
  • 25Mbps/4Mbps
  • 50Mbps/10Mbps

Scenario 3 (30Mhz only):
450m AP’s expected to have max 20 SM’s
No 450i AP’s
Range 5km, SM’s a mix of 450b HG and 450i

  • 25Mbps/4Mbps
  • 50Mbps/10Mbps
  • 100Mbps/30Mbps

Scenario 4 (30Mhz only):
450m AP’s expected to have max 50 SM’s, range 4km
450i AP’s expected to have max 15 SM’s, range 2km
SM’s mostly 450b

  • 15Mbps/4Mbps
  • 25Mbps/4Mbps
  • 50Mbps/10Mbps

I would appreciate if anyone would share their own plans for comparison, as well as any thoughts or feedback on what you think about our plans? We would really like to bump up the speeds in scenario 1 to be identical to scenario 4, but are a bit reluctant, as 95% of the customers are cabins, which is used almost only in weekends and holidays - where there is a significant load on the AP’s (especially in the easter- and christmas holidays). It is hard to test how higher speeds will impact the overall performance, as the “real load” only shows during theese few days a year. Maybe someone has experiences from serving more residential areas, where the traffic probably is more continous during the year?

In my experience, higher speed plans doesn’t always result in a huge difference in actual load on the ap’s, as the customers probably uses less time to transfer the data when on higher plans, but haven’t been able to test this on the AP’s with many SM’s because of the beformentioned challenge.

Hopefully someone will share their take on this!?

2

Not taking band or EIRP considerations into account, and we just assume that you have mid to high modulation on all SM’s, then all 4 of your scenarios are easily possible.

As an example, we have a PMP450m 5GHz AP running 30MHz with 40 SM’s on it with tier’s ranging from 20/5 all the way up to 120/20 and it’s not really close to reaching max utilization.

In order to deliver the best service with PMP450:

  • Make sure you’re on the newest firmware release, which is 25.1
  • Make sure that all your SM’s are 450b or 450i’s
  • Enable the “airtime fairness” schedular
  • If you’re really concerned about congestion, use something like Bequant’s traffic shaper to help smooth traffic
3

Thanks for your reply Eric!

I wrote a long reply earlier today, but it is somehow gone now :man_facepalming:

Anyways, I’ll try writing a shorter reply now.

What is the benefits of using firmware 25.1 (cnMaestro recommends 24.2, and it seems to me that new firmware releases theese days maynly focuses on 450v)?

If I set the scheduler to Airtime Fairness, will all the MIR settings on our SM’s be ignored? (Cannot find any good documentation on this feature)

We would really want to bump our scenario 1 up to 15/4, 25/4 and 50/10 plans, but we have had peak frame utilization up to 96% when traffic load is around 220Mbps on a sector with 93 sm’s. All sm’s modulating above 8X/7X except maybe 3 or 4 running at 8X/3X (when checking the link status just now, when there are 79 online sm’s).

4

I’m very sorry - that was my fault.

5

There have been numerous bug fixes and minor improvements for all 450 radios since 24.2. Updates are not just specific to 450v.

No. They will be retained. That being said, using the MIR shaper on the radios is not ideal as when the limits are reached, it typically causes high latency/jitter. It’s better to use a 3rd party upstream shaper, and only use the radio’s MIR shaper as an overflow limit. e.g. If you wanted to deliver 100/20 to a customer, you’d use something like Bequant to set and smooth traffic up to 100/20, and then on the radio, you’d set a hard MIR of 110/25.

Airtime Fairness scheduling aims to have the timeslot resource allocation to each and every SM the same, independent of the operating modulation on those SMs. Operation is the same as selecting the proportional scheduler, setting the same plans for each and every SM, setting all proportional scheduler weights to the same value, setting User Lock Modulation to Enable, and setting Locked Modulation to 8X.

The Airtime Fairness schedular can help reduce frame utilization and help “smooth” out traffic across the AP, typically reducing latency/jitter during periods of high frame utilization.

6

Thank you once again - it is appreciated!

We are not seeing SM’s with high latency/jitter normally, and if we do, we often set up QoS in the customer router, as this reduces bufferbloat.
Implementing a 3rd party shaper might be a good idea, but it introduces quite a lot of challenges I assume. We are bridging all of our SM’s, and using dynamic public ip adresses to our customers. As I understand, a 3rd party shaper is shaping the traffic based on the customers IP-adress - I suppose the shaper does not recognize that the client has changed IP (which happens a lot in our network, as many many customers unplug the power when leaving their cabins to make sure the equipment doesn’t brake in thunderstorms).

So I can just enable Airtime Fairness on the AP, and everything works the same regarding MIR settings - only the SM’s gets equal timeslots resulting in that poorer modulating SM’s cannot reach the same max throughput as the SM’s modulating at 8X? That meaning a poor SM does not “steal” resources from the good ones?

7

If you ever look into it in the future, you can use the CPE’s mac address via option82 to figure out the lease the user’s router has, and use that to map it to the customer for shaping.

8

Yes, this is correct.

9

Thank you jnovak!

But will this solution make the shaping get updated every time the clients router gets a new ip address?

10

Ok. I’ll try enableing it on the ap which I’m connected to at my home, and see if I experience any difference (I assume I don’t, as the AP isn’t congested or saturated).