SRS good performance?


#1

Hi guys! First of thing, sorry for my english, but I dont want use Google translator to be sure that my comunication is good enought.

A few days ago I did an installation of A5c with MTI 4x4 MIMO 65º 17 db sectorial anntena, so here is the picture of the performance on a costumer house, beta-tester of course :slight_smile:

Did you think is normal that poor performance on speed with the parameters of picture? I have 40/5 Mb. from my provider over VDSL and is receiving about 40-45% less, about 26 mbps/ 3.8 mbps.

All comments are welcome!

Thank you!


#2

Hi Enrique,

We would normally expect higher throughput for the PHY rates in the screen capture you provided, but cloud data shows that you have intermittent interference (up to -57.8 dBm) on the current channel, leading to thousands of short preamble failures. Much of this appears to be due to a channel change from 6025 to 4965 MHz.

What throughput do you get when you switch back to the cleaner channel with the same Tx power and channel width as before?


#3

Hi Cris, yes the reason was to try with lower frequency to now if the anntena work better than the higher one.
The throughtput was the same, even with C5 PHY at 300 mbps.

Another think very strange is that if I choose 40 hz. the throughtput is reduce at 14-16 mbps. download. same on 6025.

Thank you!


#4

Where do you see this intermittent noise on cloud ? It could be useful to me to understand where I need to read in similar case.

Thanks


#5

Can you post a picture about your spectrum. Combined spectrum of course.


#6

Yes, here is:


#7

What happen if you use 80 Mhz in same channel you choose ?

I think you only could choose frequency you actually set. No other channel seems to be available here.

Another think is have you set the AGC to right value as described here: http://ap.help.mimosa.co/srs-tuning-agc


#8

I think is right setting on AGC. Here you have the pictures from working at 40 to 20. 80 is even worst.

At 40 band, with more PHY speed 300/300 the real speed from A5c to C5 is very low.


#9

Hi Massimiliano,

On the Device Page, the Spectrum, Noise, and Wireless Errors graphs provide this information - like in the screen capture that I posted above.


#10

The wireless statistics look very good otherwise. It could be that the speed test above may be throttling TCP traffic, or that there is another bottleneck. What results do you get when running an iperf test across the C5 and A5 in isolation?


#11

Hi Chris, here you have downloading&uploading iperf3 test. I think a bottleneck is not possible because just few days ago I replace mi PtP Ubiquiti with M5-400 25 dbi for the A5c with C5. At this moment only two C5 are connected, this a field test before to deploy 2 nodes with 2 A5c on each one. So if that is the performance Im serously to move to another solution from another brand. Mimosa in that case is no cheap and I think dont offer me the step ahead that I want to have.

Even uploading is better than downloading. That C5 is an clear LOS.

Thanks.


#12

I’ve seen.

What it could be the cause of that problem ? It’s reall strange. I have
seen in particular noise environment, it’s really difficult to have a good
bandwidth traffic for real, but with a such clean channel, I have never
seen following strange behaviour.

Massimiliano Iavazzo
Responsabile tecnico
Telecomunicazioni srl con socio unico
sede legale: Centro Direzionale di Napoli is A7 - 80143 - Napoli NA
sede operativa: via G. Verdi 48/A - 81031 - Aversa CE
P.IVA: 07136161218
t: +39.081.19309465
f: +39.081.19758072
www.connecting-peoples.com
supporto@connecting-peoples.com
Rispetta l’ambiente. Stampa questa email soltanto se è estremamente
necessario.


#13

I have seen on oher my device’s on cloud. Thanks. It was really useful.


#14

Sorry. In the first post you told this is a VDSL link.

Well. VDSL is a shared link not a dedicated link like fiber or other. So it could be a moment where the shared link is particularly stressed by other VDSL’s users.

Have you tried with a serious dedicated line ? Or better, have you got some speedtest without A5, but directly connected to VDSL router.

I think it’s too strange to be an A5 problem here.


#15

Hi, here you have. PC connected on same VDSL router. Beleive me, my problem isn´t my VDSL, router or whatever, 3 days ago change my PtP with Ubiquiti gear and after 2 years using, 0 problems. :slight_smile:

Sometimes we try to find the problem in a lot of variables forgetting the main change, the AP.

Anyway in the next days I´ll try to do more test with another C5 but more close, maybe 1 or 2 km away. But to be honest, if i need to invest a lor of money on gear to bring 20-30 or even 50 mbps. to customers and they should be maximum 1-2 km. from AP, maybe is better to move ahead A5-14.


#16

I think it’s better to clarify some concepts here.

First point is A5-14 is done to be used in a maximum of 600 meters of diameter, so 300 meters in radius. You can surely find many help about that on this forum, talking with @Kent about his experience for example. This concept is called Micropop. Like in cellular network, many micropops connected togheter, will probably offer a real way to compete with main incumbent provider.

If you need to cover distance more than I depicted earlier, you need to turn on A5c with KP Performance antennas or even MTI 4x4 MIMO Sector antenna.

In your case, I can hint you to follow these steps:

  1. To be absolutely sure your VDSL line, goes always maximum buyed profile value, executing regular speedtest or test always at same time of the day, or better in time not of traffic peak.

  2. Execute such test even in WiFi interop mode to see if it change something.

  3. Compare the test with GPS synced mode.

  4. Check A5’s ethernet statistics, duplex, speed or other to try to exclude compatibility problem with your router.

  5. Check A5’s log to find something of strange.


#17

Hi again guys, so finally I have my full speed ( nearly full almost ) on C5:

But was very strange, after spending a lot time doing some field tests, even with another brand new C5 at1 km. from A5c in clear LOS& similar high, a hill in front of AP, the results was very frustated:

Speedtest to 2 servers:

iPerf3 between 2 PCs:

1 km at 20 hz with MCS8 and 177/177 PHY Mbps. all ports 1 Gbps.

from the other side:

And the speed on that PC connected to the same switch Gigabyte like the A5c: ( same servers like before )

So, very frustated with the performance of that unit even at 1 km, in the night I was decided just to try a littlebit more and I change the C5s AGC from AUTO to manually ( was at 23 db setting up by remote A5c ) I let at 15 db and increase step by step up 20 db, after a few minutes to let algorithm to work well i feel that when I open new web sites was faster than before. So I tried another speedtest and here was:

With iPerf3, now at 3 km. is about 39-43 Mbps. maximum.

Today I repeat many speeds test with similar result, so that is great.

But, for resume Mimosa´s staff:

what kind of black magic is that ? :astonished:

Thanks.

P.D. Massimiliano, as I know the problem wasn´t my VDSL speed, router,etc. Anyway thank you so much to try to help me :wink:


#18

In SRS (TDMA), the client AGC value is determined automatically, and the A5 AGC value is entered manually. Normally, the AGC value entered on the A5 is close to the lowest client RSSI. I wonder if we are confusing AGC and Tx Power in the description above.

Enrique, would you please post a screen capture of the Channel & Power page showing the power and AGC configuration?


#19

Hi Chris, here you have the screen captures:

A5c

1st C5

2nd C5

On that C5, AGC is on in AUTO from the first day.

Thanks.


#20

Yes, AGC should be Auto on both C5s. You can’t change it in SRS.

It looks like you clicked the Power Override button on the C5 which allowed upward adjustment of Tx Power within your regulatory domain. The relationship between AGC on the A5 and the C5 RSSI is important to consider when tuning.

  • If the AGC threshold value is set too high, it can cause problems for low-RSSI (i.e. long distance) clients.
  • If the AGC threshold value is set too low, it can allow more noise into the system and reduce SNR.