Combo Feeder Statistic Issue

Nitr0 · October 17, 2020, 5:24am

That’s impressive! I am seeing the same results on my system running an FA Orange on UAT, Airspy R2 @ 24MSPS and the 1.47kb/s is the GPS dongle.

Bus ID 0 (All USB buses)        To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s
  Device ID 2 :                 19.92 kb/s      15861.84 kb/s
Bus ID 1 (USB bus number 1)     To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s
Bus ID 2 (USB bus number 2)     To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s
  Device ID 2 :                 1.06 kb/s       1018.28 kb/s
Bus ID 3 (USB bus number 3)     To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s
  Device ID 2 :                 0.88 kb/s       1.47 kb/s
Bus ID 4 (USB bus number 4)     To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s
Bus ID 5 (USB bus number 5)     To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s
  Device ID 2 :                 15.38 kb/s      14779.07 kb/s
Bus ID 6 (USB bus number 6)     To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s
Bus ID 7 (USB bus number 7)     To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s
Bus ID 8 (USB bus number 8)     To device       From device
  Device ID 1 :                 0.00 kb/s       0.00 kb/s

I hadn’t thought of upping the ante by slapping the Mini or other R2 on to see how far it can be pushed. Thanks for the output - I’m impressed with these RockPi 4’s

I have a shiny new Rock Pi X on order, I couldn’t resist.

wiedehopf · October 17, 2020, 6:15am

Again, that’s not how it works.

If stuff gets lost on the USB bus, there is no indication besides MLAT going wonky.
Lost samples that get lost on the USB bus can’t be logged because the software never sees them and the SDR doesn’t transmit a running number with every sample or something like that.

If you have insufficient CPU, it’s easy to log those lost samples because they occured after we’ve first handled the samples in software.

Without adding -p to the -m 20 that’s not really much difference in bandwidth.

Going to -m 12 and confirming the MLAT issue is gone would be too easy i guess?

ca249 · October 17, 2020, 10:48pm

It was the failure on my part to fully comprehend the outcome of what the
results of that change might be.

I will change the combo feeder’s -m parameter this evening at 0000Z and post the results
at the end of the UTC day tomorrow.

Thank you for your assistance.

ca249 · October 19, 2020, 3:26am

Below are the aircraft stats from FA for feeders rockaware and rock1090 for the UTC day
10/18/20.

On rockaware the R2’s -m parameter was changed from -m24 to -m12 for the entire day of 10/18/20.

This test was set up to possibly expose USB bandwidth issues caused by lost or silently
dropped samples from the R2 on the USB bus. These lost samples, or some other bandwidth
load issue, might be responsible for the difference seen in mode-s aircraft counts and MLAT
counts between these two feeders. The original stats issue is defined in post #1 of this thread.

From the stats below the spread of aircraft counts for mode-s and MLAT between the two
feeders are approximately the same for 10/18/20 as for the previous days.

I think we can assume that the change of the -m parameter from -m24 to -m12 made little change
in the stats.

SITE 108954 rockaware stats - combo feeder
rockaware_stats_2020-10-18

SITE 103073 rock1090 stats - stand-alone feeder
rock1090_stats_2020-10-18

wiedehopf · October 19, 2020, 7:40am

Well looking at the stats, it has worse reception than the other box which can also explain less MLAT.
Why there would be a difference, hard to say.

I’d recommend adding graphs1090 to have better stats, FA stats as previously mentioned can fluctuate.

ca249 · October 19, 2020, 8:16am

Your graphs are loaded on all my feeders. Let me know what time frame you want to see. I will post them up tomorrow. Its late here now.

wiedehopf · October 19, 2020, 8:31am

Whatever you want to compare.
A day with dump978 switched on on both boxes, a day each with dump978 being disabled on one of the two boxes sounds like a good place to start.

ca249 · October 20, 2020, 3:09am

Below are the graphs for both feeders for the 24 hour period ending at
approximately 1955 PDT on Monday, 10/19/20.

SITE 108954 rockaware is in the combo role receiving both 1090MHz and UAT.
SITE 103073 rock1090 is in the stand-alone role receiving only 1090MHz with
dump978-fa stopped and disabled.

Tomorrow at about this time I will post up graphs of both feeders in the combo
role.

ca249 · October 21, 2020, 3:03am

Below are the graphs for both feeders for the 24 hour period ending at
approximately 1953 PDT on Tuesday, 10/20/20.

Both SITE 108954 rockaware and SITE 103073 rock1090 are in combo mode receiving both ES1090 and UAT 978.

wiedehopf · October 21, 2020, 11:16am

It would probably help to make the scale the same:
GitHub - wiedehopf/graphs1090: Graphs for readsb / dump1090-fa / dump1090 (based on dump1090-tools by mutability)

It’s at the bottom of the config, just set a common max y value for the main graph especially.
Also you’ll need to actually right click copy image the graphs, those screenshots are bad resolution.
The message rate and aircraft count and signal and range should be sufficient.

Signal levels are a bit different, i suppose a droop in USB voltage or as you said interference could be the issue.
Not sure how hot the airspys run, could also be the 2nd SDR increasing the airspy temperature.

But really the numbers look very similar.
I wouldn’t worry about the FA stats, as has been mentioned they do fluctuate.
Also not all stations are used equally for MLAT.

sudo journalctl -u piaware -n1000 -e

You can check if you see any out of ordinary messages from the mlat-client or compare both machines.

ca249 · October 23, 2020, 3:03am

I have posted up graphs again for both feeders below for the 24 hour period ending at approximately
1950 PDT, for Thursday, 10/22/20. The graphs are posted up one above the other to make comparing easier.

The FA stats for mode-s aircraft counts are still about 8% lower on the combo feeder verses
the stand-alone feeder. The FA stats for MLAT aircraft counts are still about 40% lower on the
combo feeder verses the stand-alone feeder.

For graphs1090 I set the scale for both feeders the same for message_rate, aircraft, tracks,
range, maxima and rate/aircraft. I posted up the graphs here again grouped for easy comparison.

The signal for rockaware (combo) did look a little ragged. I had moved the R2 about 2’ (as far
as the USB cable would allow) from the other feeders last week to see if there was an RFI
issue with it being close to the other feeders/dongles. That put it outside of the effect of a
common fan for all the feeders. The R2 was noticeably warm to the touch. I moved it back
into the airflow. As far as I can tell the warm R2 had no effect.

The signal level was a little lower on rockaware. This was due to the extra loss of a 6’ length
of RG-316 verses an 8" RG-316 jumper to the splitter. Moving the R2 back near the other feeders allowed me to replace the 6’ RG316 with a 8" RG316 jumper. This brought up the signal to about
the same on both feeders.

The RockPi4Bs came with true USB-C power adapters. They are marked “Output: 5V@3A,
9V@2A, 12V@1.5A”. When first pluged in it puts out 5v for a couple of seconds and then
switches to 9v mode settling at 9.076V @.890A when the feeder is fully booted up and
running in the combo role. It puts out 5.004V to the USB port.

A check of the piaware journals for mlat-client messages did not turn up anything out
of the ordinary, but counts were almost identical between the two feeders.

rockaware 24 hours ending at approx. 1950 PDT 10/22/10

rock1090 24 hours ending at approx. 1950 PDT 10/22/10

rockaware 24 hours ending at approx. 1950 PDT 10/22/10