ePMP unstable Ping

I noticed that on ePMP access points with fixed ratio 75/25, the ping is not so stable: there is 10-15ms jitter in some cases.
I think it’s due to MCS selection and retransmission. Is there something I can do to have a stable ping to my clients? I wouldn’t like to lock uplink MCS on each client…

Here's some screenshots from one AP and one SM of this AP.

This is from AP, as you can see with only 8 clients.

This is from one SM pinging to the main router directly connected to the AP:

And this is from wireless statistics on SM:

This is the cleanest frequency I can choose and allows me to do wireless link test of:


61.583 Mbps


9.196 Mbps


Using 20MHz channel and 75/25 ratio.

I'm satisfied with the speed, but I would like more stability.


How can I achieve that for all clients?

That is to be expected in a real TDD system. 

the frame has a fixed duration and schedule.   no matter the traffic, everything is organized to take place in a specific pattern. 

each frame is 5ms in duration, a ICPM reply cannot be returned within the same frame, it actually ends up almost always missing atleast one turn to return as it goes through the scheduling process.  

its works something like this (its a gross under expectation)  the AP will send the first frame out after the data has arrived (not the actual data, its got the control information for the next frame, and already scheduled data transport)   the next frame up, the AP will transmit the packet in the next frame to the AP that is receiving it, once the SM has gotten the ping, in its next control frame, it requests airtime, the AP assigns it a slot, and it will transmit in its assigned spot. the entire process takes appox 3 frames, or a minimum of 15 MS if your ping is included in the next out frame on both the AP and SM its possible to see right at 10MS pings, but you won't see to many of them. thats a huge chance out. 

from the stand point of 1 or 2 SMs that makes the system seems little shacky, but as you see from your speed test, it is scheduling and moving a lot of data. 

the trade of to forcing this schedule and adhearing to the given order is you'll see stability when the platform is smashed,  with wifi driven system, you'll get 1 ms ping beasue the system doesn't have a fixed schedule, it checks to see if the air is clear and then yells to data, if it makes it... Cool!  if it doesn't, repeat. when you get to have 10 to 15 CPEs randomly starting to yell and are not controlled in a schedule, you end with the wild swings in ping you see on a wifi platform that is struggling.   

on a solid TDD based radio like EPMP, 450, wimax or TDD mode LTE , you can get all of your APs performance from it, and not see much elevation of the pings with dozens of CPEs attached and active. 

in short, you'll see that small amount of jitter fairly consistently as your sub load grows and once you begin to max out a AP, your customers throughput will decline evenly as the AP is out of airtime to feed the need.  BUT you'll still have somewhat low ping and most importantly stable ping.   meaning a good experience still for end users so long as there is enough capacity to fill the need of the web apps thier using like netflix and hulu.  download rates just start to decline instead of utter chaos and mad customers.


Thank you for your super clear answer, that makes sense!
Sometimes some users see this high jitter as a bad point and think it’s a low end internet connection, because fiber or ADSL don’t have such jitter normally, and it’s not simple to explain that the quality is high despite the high jitter.
I use 2.5ms frame to reduce the ping a little bit, and it’s ok, but unfortunately I think the jitter is normal in a TDD system as you explained so good!
Did you see a big increase with 40-50 users in ping?

This ap has 32 subs attached, ap pinged from management vlan, down to the router, then back through itself to the cpe wan interface. it sits around 50% frame usage this time of night. 20mhz wide channel on 5ms frame

I’ll look for a busier AP tomorrow night


@Chris_Bay wrote:
This ap has 32 subs attached, ap pinged from management vlan, down to the router, then back through itself to the cpe wan interface. it sits around 50% frame usage this time of night. 20mhz wide channel on 5ms frame

I was under the impression that the frame utilization stats on ePMP were not correct. Has this changed?

The downlink is correct, the uplink is still bugged