This topic has been brought up several times, but I want to confirm if there are any current issues with this configuration with either the CBRS or UNII PMP450 products. Currently I am using this config with CBRS and saw an approximate 20% increase in throughput / AP utilization and happier customers, but Cambium support expressed some concern with this config and I have delayed implementing the same configuration on my 5GHz UNII APs.
My CBRS APs have been humming along for a couple months now without issue, so barring any known issues, I would like to increase my 5GHz capacity with the same config.
Any feedback is appreciated.
Are you referring to the original PMP450, PMP450i, or PMP450m AP’s?
The original PMP450 AP’s don’t have the processing capabilities to properly use 30 or 40MHz carriers. 30/40MHz carriers are supported on the 450i and 450m platforms.
Basically, there are some issues with the PMP450 platform that cause 5ms 30MHz and 40MHz to not perform optimally. This is called “the deep VC queue” issue. The throughput doesn’t scale up properly when using 5ms frames and those larger channel widths, the way for instance it does using 2.5ms frames. This doesn’t cause any serious issues… it’s just leaving quite a bit of performance on the table. Running 2.5ms frames and 30/40MHz is the recommended configuration, but there are instances where you have to run 5ms, like if you need to sync with LTE, or legacy Canopy or ePMP radios.
The deep VC queue issue is scheduled to be fixed around August of 2024.
1 Like
Thank you for the clarification, I am referring specifically to the PMP450i and the PMP450m.
Is the concern then only for the PMP450 opposed to the other two?
I have several other CBRS and 5GHz technologies in the area and the intent is to better match the 5ms frame period of those technologies, but I am also seeing +95% DL frame utilization during busy hour (6pm-midnight) hence the need to increase capacity.
The deep VC queue issue effects all PMP450 AP’s (450/450i/450m), but due to the lack of processing capabilities, it’s not recommended that you run 30 or 40MHz channel widths on the original PMP450 AP.
Again, running 30/40MHz 5ms is fine on 450i/450m and you’ll see some speed increases over 20MHz, but it’s not going to scale up properly.
How many clients are on that AP?
With a 450i or 450m, how does 5ms @ 30MHz vs 2.5ms @ 40 MHz compare? I’m wondering which config to use until they come out with the update…
Cambium recommends that you stick to 2.5ms frames @ 30 & 40MHz until there’s a fix. I’d only run 5ms frames @ 30 & 40MHz if you need to sync with another platform that supports 5ms frames, like LTE on 3GHz.
Is there a fix coming in the next firmware? Seen a few posts on other platforms as well saying a few things are needing fixing on the 450/450m platform.
Like I said earlier, planned fix around August of 2024.
Ah shoot, I missed that lol, my bad. Thank you.
I had a ticket open with Cambium support on this issue and I was notified that this has been pushed up to June 2024. Whether that’s true… I can only speculate.
Ok question,
does anyone actually have the benchmarks on what 450m to X looks like on 5ms vs 2.5ms on 20,30,40 mhz, i’ve been wondering i understand its a bug on the 450’s but does the 2.5ms to 5ms on 40mhz set 450’s lower than say 30mhz throughput? I ask because i’d imagine most us run mixed 450/450b networks so begs the question what the impact actually is.
And the deep queue bug, is it impacting the full sector because of the 450 sm’s or just the 450sm themselves throughput? If we’re going for max performance on the overall 450m … especially given MU-MIMO
@Eric_Ozrelic question, i take it this turned out to be a more difficult fix than originally intended? I was looking back through old webinars and didn’t even realize this was mentioned in the 30-30-30 presentation that it was supposed to be in 23.1 (Dec 2023) for the Rearchitecture of the MAC Layer specifically 40mhz+5ms mentioned.
Also wondered, did cambium ever roll out the rest of the first 30 of the 30-30-30 the “Downlink Slot Optimization” that was going to see us get another 7-10% improvement on top of the Multiple Groups per frame?
I performed some tests awhile back and found that 2.5ms @ 30 and 40MHz often times would produce higher aggregate speed test results on 450m then 5ms @ 30 and 40MHz. At 20MHz, IIRC, 5ms would produce slightly higher aggregate throughput results then 2.5ms.
Maybe @Charlie can add a bit more information on this?
Yes, this is a very difficult issue to fix, it’s buried deep in the firmware. Delays to the deep vc queue fix were partially because of development work on the Downlink Slot Optimization, work on the interference cancellation features, and 450v development.
And yes, the Downlink Slot Optimization feature was [added in R22.0].(PMP450x Release 22.0 is now available)