WARP Project Forums - Wireless Open-Access Research Platform

You are not logged in.

#1 2009-Feb-06 09:47:13

dpfeil
Member
Registered: 2008-Feb-02
Posts: 47

Reference design v11 performance

We have noticed that v11 of the reference design performs much worse than the previous v9.  We have run tests in both SISO and MIMO using the reference designs downloaded straight from the site for both 11.1 and 11.2, as well as 9, and have noticed that for both SISO and MIMO, the newer versions (11.1 and 11.2) do not perform as well (more packet drops during ping test) as we increase the transmission distance.  We have also tried to recompile the designs, but this did not help, we still get the same performance.  Have you experienced this at all?  Do you have any suggestions for fixing it?

Offline

 

#2 2009-Feb-06 10:06:27

murphpo
Administrator
From: Mango Communications
Registered: 2006-Jul-03
Posts: 5159

Re: Reference design v11 performance

It's likely just that the packet detector is failing to detect lower-SNR packets. Try changing the detection threshold (pktdetthresh in warpphy.c) to something lower (like 3500). This number is used to threshold RSSI summed over 16 samples. RSSI ranges [0,1023] (its noise floor is around 100) and is proportional (in dB) to received power. So a threshold of 9000 corresponds to an average RSSI of 562, which is pretty high. Refer to the MAX2829 datasheet for a plot of RSSI vs. Rx power (top-center plot on pg. 16). The RSSI voltage range of [0.5,2.5]v maps to digital values of [0,1023].

Offline

 

#3 2009-Feb-06 12:11:59

dpfeil
Member
Registered: 2008-Feb-02
Posts: 47

Re: Reference design v11 performance

Thanks.  it looks like we have it set at 9000 for all versions and csma threshold is also consistent at 4000 in all versions.  we could try to drop it down for version 11, but this doesnt explain the difference in performance between the versions.  was something changed between these versions for RSSI calculation or any other factor that would reduce (or increase?) the threshold calculation and cause it to act differently with the same threshold value?

Offline

 

#4 2009-Feb-06 13:34:45

murphpo
Administrator
From: Mango Communications
Registered: 2006-Jul-03
Posts: 5159

Re: Reference design v11 performance

Can you describe your setup in more detail (antenna spacing, modulation rate, packet lengths, changes from the distributed code, etc.)? I've been doing some tests with v11.2 at low SNR (using 80dB of attenuation between two nodes, for example) and still see decent performance. I'd like to try recreating the setup that shows a big disparity between designs.

Offline

 

#5 2009-Feb-06 13:48:50

dpfeil
Member
Registered: 2008-Feb-02
Posts: 47

Re: Reference design v11 performance

4 QAM for all cases, nothing changed at all from posted reference designs
distance between nodes: 7 meters
channel: whatever is set as default in downloaded project
we did then change the channels and were no more successful on any other channel

For MIMO:
antenna spacing of 30cm

so to clarify,  for a set distance and set up, using SISO, version 9 would ping at 100%, where with version 11.1 and 11.2 we could get 50% ping success.  using the exact settings posted in the reference design.

Offline

 

#6 2009-Feb-12 23:35:21

shettysid
Member
Registered: 2009-Feb-05
Posts: 6

Re: Reference design v11 performance

@dpfeil: We are seeing similar problems. We started with 11.2 (QPSK mod) and saw bad performance. Switching to version 9 seems to have solved the problem.

sid

Offline

 

#7 2009-Feb-13 09:25:12

murphpo
Administrator
From: Mango Communications
Registered: 2006-Jul-03
Posts: 5159

Re: Reference design v11 performance

Thanks for the update Sid. Can you describe your setup (like dpfeil did above)? We're trying to understand the scenarios where our changes in v11.2 helped or hurt.

Offline

 

Board footer