GPS sync tests drastically reduce bandwidth

I've been running 3 APs in flexible mode for many months. As we get ready to add a 4th AP, I wanted to see what GPS sync did to throughput.  We tested with two clients and the results surprised me.

On 5780 Ghz with a 20Mhz channel, we tested two SMs with flexible, 75/25 and 50/50.  To start off, the clients have a RSSI of: -56/-60 each and a SNR of 28/24 each. Not the cleanest RF environment but not bad. Clients are each about 3 miles from the AP.

One SM has a MIR of 26000/26000 and the other is set at 53000/53000

I am using medium packet with a 10 second duration for the radio speed test. No other SMs on this AP during testing.  Both SMs are Force 200's.

In flexible mode, I see 27/18 and 53/21 respectively.

In 75/25 mode, I see 27/5 and 52/7 respectively.

In 50/50 mode, I see 27/9 and 42/11 respectively.

I was a bit surprised to see the upload speed drop so dramatically with GPS sync on. The download speeds seemed to be about what I expected. We repeated the tests with 2.5 ande 5.0ms frame sizes - that had no effect on speed but watned to make sure.

 Is this normal for upload to take such a hit?  With a 20Mhz channel, I should see ~100 Mbps theoretical in throughput. in a 50/50 mode, that would leave me ~50 Mbps for upload minus around 10% for overhead.  I see that in download but the upload takes a much bigger hit.

It's normal! 

In Flexible mode you don't have 100 + 100 (about the theorical max throughput at 20MHz), but 100 aggregate.

If you do your tests separately, you'll have ~100 in download and ~100 in upload in ideal conditions in Flexible mode, because your frames will be used all for download when you'll use download, and all for upload when you'll use upload.

If you set GPS sync mode with 75/25, you use 75% of the time for download and 25% for upload, so you can't go higher than about 85Mbps in download and 25Mbps in upload.

With 5ms you have about 10% more throughput than 2.5ms, but you'll have bigger latency (and a lower TCP throughput in some cases).

Take a look at Understand ePMP throughput capacity at a detailed level in the support section, and it'll be all more clear!


Hi AU Wireless,

Giuseppe4 totally right about understanding of ePMP throughput capacity and differences between flexible and defined ratio modes.


Also please note that you have not perfect RF conditions.

SNR 28 it is close to MCS13 and SMR 24 it is close to MCS 12.

In that case expected throughput is less than 100 Mbps.

I can assume that Max MCS are not fixed for DL and UL and it causes additional Throughput reduction.

Try to fix UL Max MCS - 12 and DL Max MCS - 13.

Thank you.

1 Like

Thanks for the reply.  I did go back and re-read that white paper on capacity.  What is confusing me still is the reported numbers for my upload.  The download speeds are somewhat in line with what I would expect to see moving from flexible to TDD.  The upload seems wrong.

If a station can achieve 20 Mbps upload in flexible mode, should it not be able to achieve 20 Mbps in say 50/50?  Or, is it going to cut my upload in half and if I can do 20 in flexible, I get 10 in 50/50 and ~5 in 75/25?

With my download, I have a station that (with MIR on and limiting at 53000) can achieve 53 download in flexible.  By going to 50/50, I would expect 50 as well - or very close to it. However, in reality, it tops out at 42.

Those real world examples don't match up with how my mind is interpreting the white paper's description of what I should see.

If station can achieve 20 Mbps upload in flexible mode, it should achieve 20 Mbps in 50/50 with your conditions (DL - MCS13 and UL - MCS12).

Try to use longer time periods for tests it can be related.

Thank you.