SNR of 1 - firmware bug?


#1

Just installed 2 B5’s as a backhaul. I am using 2 40mhz channels, both in the DFS band (5275 and 5665). They link up and I have around -58 signal. I chose those 2 channels manually as they were totally clear on both ends of the link.

However, when I light up the link and look at the SNR, it is show an SNR of 1 for both chains. It reports the noise floor to be 1 dB lower than the signal - and it will track that remaining 1 dB lower as the signal strength changes.

Is this a firmware bug? I’m running 1.2.3. I would expect to see a SNR of ~30 based on the noise floor I can verify with multiple radios…


#2

Is the distance, perhaps, between 1.5 and 3 miles?

We’ve had a serious bug in the 1.2 stream where it gets the SNR wrong and, unlike yours, shows the other side’s signal in the spectrum analyzer. It seems to be a timing issue at distances near 2 miles. They almost fixed it in 1.2.3 but that had a serious memory leak in its SNMP so we are waiting for 1.3.


#3

Hi Chadwick,

Yes, there is a confirmed bug on links which are near 3.2 km (2 miles) in length. A timing problem causes the Spectrum Analyzer to see the link peer as noise. This causes the Rx Noise to rise, and the SNR to fall as a result.

In the background, the rate adaptation algorithm uses PER for making modulation decisions, so you shouldn’t see any performance impact as a result. You can verify this on the Dashboard > MIMO Status panel (Streams table) by observing that a higher MCS is maintained than the SNR would support.

This bug is addressed in the upcoming firmware version 1.3.0 which also addresses the unrelated SNMP memory leak that Fred mentioned.


#4

Well, that seems to be exactly the issue. I am about 2 miles and it does look like it reports the link as noise. Excellent. Does not seem to effect performance, just thought it looked odd.


#5

Fred, my signal is in the spectrum analyzer shot. I took that picture after shutting the link down to verify the noise floor. When the link is up, I do see it in the spectrum. Sounds exactly like the bug.


#6

By the way your shot confirms another bug in the spectrum analyzer, though not as serious. Notice the little peaks at 5000 and 5400? Mine have that too – even when indoors away from a window, with no signals. It appears to be either a narrow spur inside the radio or a bug in the analyzer.

The funny thing is that while those are out of band (in the US) and thus shouldn’t matter, I do pick up real signals on those frequencies in certain locations… there are certain vendors whose radios are not limited to US frequencies (15C approvals only), and who put them on unapproved frequencies for customers who don’t know the difference (not WISPs). Very annoying. So out-of-band analyzer coverage can be revealing.