Check out these very interesting spectrum charts now posted under the Docs->KnowledgeBase area:
https://www.arednmesh.org/content/spectrum-charts
Thanks to Ken KE2N for capturing this data on a Rocket M2, Rocket M5, and a Mikrotik Basebox 2. We were discussing what conclusions we might draw from these charts, but seemed to have more questions. These charts were produced on the bench with nodes in proximity to each other, so you can assume the highest MCS rate was in use to capture the data.
Here's some of the questions below. Take a look and if able, do some testing on live nodes and longer distance links to help determine if we can come to some general guidelines and conclusions based on everyone's results.
When we review the Ubiquiti device specs the power is automatically reduced as MCS link rates go up in AirOS. The theory says, the ave power and peak power delta of the signal becomes greater when the follow occurs: A) channel bandwidth is cut in half; B) increase in MCS link rate (when the modulation changes, not error correction coding changes); C) when the xmit power is increased. Did I miss any factors? At some point, with a combination of these factors, the range of the PA is exceeded and the signal starts to get dirty.
1) Suspect that the open source driver is not lowering power as link rates go up like AirOS. The charts indicate as such.
2) How much is thoughput affected by lowering power -3dBm or -6dBm from max on a tower node and remote node with cleaner signals?
3) Are we a better neighbor by lowering power -3dBm or -6dBM? The charts show the signal to be cleaner and a better neighbor by lowering power.
Joe AE6XE
https://www.arednmesh.org/content/spectrum-charts
Thanks to Ken KE2N for capturing this data on a Rocket M2, Rocket M5, and a Mikrotik Basebox 2. We were discussing what conclusions we might draw from these charts, but seemed to have more questions. These charts were produced on the bench with nodes in proximity to each other, so you can assume the highest MCS rate was in use to capture the data.
Here's some of the questions below. Take a look and if able, do some testing on live nodes and longer distance links to help determine if we can come to some general guidelines and conclusions based on everyone's results.
When we review the Ubiquiti device specs the power is automatically reduced as MCS link rates go up in AirOS. The theory says, the ave power and peak power delta of the signal becomes greater when the follow occurs: A) channel bandwidth is cut in half; B) increase in MCS link rate (when the modulation changes, not error correction coding changes); C) when the xmit power is increased. Did I miss any factors? At some point, with a combination of these factors, the range of the PA is exceeded and the signal starts to get dirty.
1) Suspect that the open source driver is not lowering power as link rates go up like AirOS. The charts indicate as such.
2) How much is thoughput affected by lowering power -3dBm or -6dBm from max on a tower node and remote node with cleaner signals?
3) Are we a better neighbor by lowering power -3dBm or -6dBM? The charts show the signal to be cleaner and a better neighbor by lowering power.
Joe AE6XE
joe,
Since your post I have been doing some testing of the theory. I have a common M5 Rocket with a 120 degree sector that has a selection of Nanostations, Nanobridge, and a few PowerBeam dishes pointed back at it all at distances between .5 and approximately 8 miles. I have consistently seen lower LQ on the links to the PowerBeam devices while their NLQ has always maintained at 100. Originally thinking this was an aiming issue the alignments have been checked many times. SNR on these links is in the 23-29 range which I always thought was low given the relatively short distances. After initial deployment I had dropped the power to anywhere from 2-4 points below their max ratings.
Over the last few days I have played with this some dropping power on the remote nodes 2 Dbm at a time. Initially I did see a small gain however after driving the node with traffic, the gain quickly faded back to the original LQ. I dropped power continually until seeing an overall drop in performance and then came back up.
Of note in this test pool, the Nanobridges are performing better than either the Nanostation and PowerBeam. The PowerBeams were put in for testing as they becam e useable with AREDN and because they had the memory to support future upgrades. I plan to replace one of the PowerBeams with a Nanobridge today to see what the LQ does as it used to be a solid 100x100 pre PowerBeam.
The Rocket sector is currently running 88-014217a as is one of the PowerBeams in question. Everything else is running the 20180119 release. All LQ and NLQ readings are taken from the Rocket Sector. The other factor I note is impact from the number of nodes and places where OLSR is evaluating the path efficiency. The bench test was P2P and field is P2M.