MEB rooftop mMIMO experiment error for script 'rl_ofdm_simo.m

[sayaz...@gmail.com]

Hello Powder team,

I am trying to do an experiment using MEB rooftop mMIMO nodes and got stuck with an error.

Profile: renewlab

Nodes: pc19-meb, mmimo1-meb, irisclient2-meb

I open Matlab (in pc19-meb) with the script 'rl_ofdm_simo.m' (provided by renewlab  Github link ) for uplink to do initial test. The simulation runs w/o any problem. 

To test OTA, i've first checked the BS_ID, UE_ID with python3 -m pyfaros.discover and the active HUB_IDs as below. The center frequency, tx/rx gain were also changed accordingly.

Without changing any other parameters I run the experiment, while the below error pops up:

As I understand it is something to do with the interface between Matlab and Iris python driver. Can you please point me on where to look for a fix ?

Thank you,

Ayaz

[sayaz...@gmail.com]

Also, please ignore the last error part image saying HUB_ID "FH4B000021" (image was captured before). It is actually "FH4B000003".

Thank you,

Ayaz

[Oscar Bejarano]

Hi Ayaz,

Have you attended office hours to discuss the mMIMO setup? 

There’s a few things it might be worth going over before you continue running experiments.

A few comments:

1) Make sure you stay between 3.5450GHz and 3.6GHz. 

2) Avoid increasing the TX GAIN at the base station beyond 81.

3) You are using extremely old code. Please switch to the “develop” branch and start from those scripts we provide there. Let us know if you face any problems.

Best,

-Oscar

On Dec 10, 2022, at 2:01 PM, sayaz...@gmail.com <sayaz...@gmail.com> wrote:

Also, please ignore the last error part image saying HUB_ID "FH4B000021" (image was captured before). It is actually "FH4B000003".

--
You received this message because you are subscribed to the Google Groups "Powder Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to powder-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/powder-users/621dc51d-a80e-408f-b2d7-1116aaa1c63bn%40googlegroups.com.

[Oscar Bejarano]



[sayaz...@gmail.com]

Hi Oscar,

Thanks for the heads up. I was actually using an older script, now pulling from "develop" branch. I am also using frequencies/tx-rx gains in the range.

I am trying to run 'rl_ofdm_ul_mimo.m' and 'rl_ofdm_siso.m' and getting a new error:

Let me know if you have any comment on this. I will also set up an office hour meeting and understand the flow. 

Thank you,

Ayaz

[Oscar Bejarano]

Hi Ayaz,

Something changed in one of the libraries that broke this.
Seems like matlab is loading the python modules using a very old version.
The fix is very simple but we need to test it on POWDER.

If you have a reservation in the coming days let's meet so I can walk you through it.
I'll send you an invite shortly.

-Oscar

To view this discussion on the web visit https://groups.google.com/d/msgid/powder-users/594ed205-a520-42fd-bd3b-11e0dd1480bcn%40googlegroups.com.

[sayaz...@gmail.com]

Hi Oscar,

I have a reservation that starts from Dec 15, 2022 10:00 AM. We can meet at that time if you are free.

I'm also available tomorrow 10am-4pm to meet incase you are unavailable on 15th Dec. 

Thank you,

Ayaz

[sayaz...@gmail.com]

Hi Oscar,

As you suggested changing the way we load the python module did the trick. The experiment ran fine w/o any error.

I did encounter two more problems:

1. While running 'rl_ofdm_ul_mimo.m', with [1x8] uplink setup, most of the times the BS antennas are unable to detect the frame start (either 1 rx antenna or sometimes all of the antennas). The error that pops up "No correlation peak at antenna X sched sym idx: 1", drops the frame, and continues with the next frame. Sometimes (50% or more) the correlation is unsuccesdfull. May be very low power on the receive signal / interference . The transmit gain is already at 81. Is there any way to improve this scenario from HW perspective (by some gain control) ? Or its just the nature of setup we have ?

2. The MATLAB (GUI) crashes randomly, and I need to restart every now and then. X11 is upto-date. Anything to stop the crash so frequently?

Thanks

Ayaz

[Oscar Bejarano]

Hi Ayaz,

Great. Glad to hear that worked. Since you managed to test it we'll push those changes to the repo.

About your issues:

1) Gain is usually the one to blame when that happens. In this case it would be the UE's TX gain. You can play a bit with it. You can go up to 109. Try to increment it gradually. I would keep RX gain at about 65, 70 at the highest. Once you go any further the signal quality degrades a lot. Also if you notice that the error is always with one specific radio, please try to use another one, in case something might've gone wrong with that radio. Most of the time we can fix the issue by rebooting it. If you are still struggling, ping me and we can meet.

2) I suggest you avoid using the Matlab GUI when working on the server. It's very unstable and extremely annoying. Try running Matlab like this: "matlab -nodesktop" and just work from the terminal.

Hope this helps,

-Oscar

To view this discussion on the web visit https://groups.google.com/d/msgid/powder-users/69729a9f-cdcf-48ec-be39-afd83e67957dn%40googlegroups.com.

[sayaz...@gmail.com]

Hi Oscar,

Its better now. Thanks for the help.

Thanks

Ayaz