I had someone tell me a 20mhz ap having somewhere along 8x-6x mimo from the sm's will have throughput ranging from 90-125 mbps. However, this drops to 60-45mbps when using a 10mhz channel. Looking for documentation to support since other network engineers dont believe us.
What exactly don't they believe? That dropping mhz in half cuts the speed by half actually more (larger channels are more efficient due to less RF overhead such as the HARQ/MAP. Thats why 1x 20mhz is slightly better than 2x10mhz thats just standard rf.
As for the top 90-125mbps remember that's aggregate traffic for the AP and is considering 100% of traffic is at 256qam (unlikely) but i can tell you it seems realistic based on what we've seem, on an AP that's a split of 8x and 6x, about 16 users right now. You can see below, this is 20mhz with 80/20 split so technically should give us 100mbit on downlink based on cambium numbers.
As you will see we'e hitting a reported frame utilization of 60% at 35mbit so in reality we're looking at a saturated sector at 70ish megabit vs the expected max of 100mbit, so as you can see if we keep adding non 8X/8X customers in the same ratio we are now our sector will only handle 70ish megabit. However if we can improve the 1 guy thats at 4X and the other 3-4 people at 6x we stand to improve by about 20-25mbit (Remember an 256 qam to 64qam modulation drop is technically a drop of 75% efficiency on that client... it hurts A LOT and affects the entire sector.
Sorry for the long shpiel but i hate when i come on forums for RF products and ask for information on "real world throughput" and someone throws a single user iperf at me with no explanation (normal of some other vendors, not really cambium) So i figured i'd provide some real world results and graphs.
1 Like
your graph looks very famiiar, what system are you using to generate it?
also, please see below. In fact changing to 10mhz from 20mhz cuts throughput by more than half.
Table 19 Examples of aggregate sector throughput – OFDM MIMO-B (PMP 450 Series) Air Interface |
Rate Adapt |
Ch BW (MHz) |
Cyclic Prefix |
Maximum Aggregate Sector Throughput - RF Link Test (Mbps) |
OFDM (MIMO) (PMP 450 Series) |
2x |
5 |
CP 1/16 |
4 |
OFDM (MIMO) (PMP 450 Series) |
4x |
5 |
CP 1/16 |
8 |
OFDM (MIMO) (PMP 450 Series) |
6x |
5 |
CP 1/16 |
14 |
OFDM (MIMO) (PMP 450 Series) |
8x |
5 |
CP 1/16 |
18 |
OFDM (MIMO) (PMP 450 Series) |
2x |
10 |
CP 1/16 |
13 |
OFDM (MIMO) (PMP 450 Series) |
4x |
10 |
CP 1/16 |
26 |
OFDM (MIMO) (PMP 450 Series) |
6x |
10 |
CP 1/16 |
42 |
OFDM (MIMO) (PMP 450 Series) |
8x |
10 |
CP 1/16 |
55 |
OFDM (MIMO) (PMP 450 Series) |
2x |
20 |
CP 1/16 |
30 |
OFDM (MIMO) (PMP 450 Series) |
4x |
20 |
CP 1/16 |
60 |
OFDM (MIMO) (PMP 450 Series) |
6x |
20 |
CP 1/16 |
98 |
OFDM (MIMO) (PMP 450 Series) |
8x |
20 |
CP 1/16 |
128 |
HEHE :) LibreNMS (though its based on observium, it just tends to be updated more frequently, Â we use mib-polling that was recently added to grab the frame utilization).
Yep that hit of more  than half from 20mhz to 10mhz as cambium explained is the same reason that when they are going to 30mhz we're gaining ~55-60% instead of only 50% more as they only have 1 set of overhead instead of 1.5x overhead in each frame.
1 Like