Graph artifact or significant boundary condition?

I’ve been tuning the gain on my system for a while, initially using gut feeling and seeing the impact on daily positions and aircraft reported.

More recently I’ve been using the excellent graphs1090 and during a spell of relatively flat propagation conditions I’ve more or less settled on my optimum gain setting.
In my case, at my particular location, --gain 20 gives the best results. I’m sure every other site will get a different optimum value.

However what I didn’t expect, is that a 1 db change on my gain manifests in such a noticeable CPU utilization pattern difference.

Before ~ 11.21 I was running with --gain 21 (notice the finegrain cpu utilisation up/down)
After ~ 11.21 my “optimal” --gain 20 was set (notice the much smoother CPU utilization!)…

dump1090-localhost-cpu-8h

Now I have no idea if this is just a graphing artifact or if the slightly higher messages > 3dBFS causes more processing for the CPU, but I do find this interesting nonetheless!

The graph for the message rate also appears smoother after the gain reduction and I certainly don’t suffer lower number of positions reported…

…or degraded weak ADS-B Signal Level performance:

dump1090-localhost-signal-8h

To me it looks like the dynamic range is slightly better and cpu usage smoother.

Any thoughts?

Change it back and see if you can reproduce (i would bet against it).

My guess: just some strange timing between stats.json being written and then read (both actions are on independent clocks around 60 seconds).

I see exactly the same effects. Higher and lower gain results in more and less CPU utilisation in the graphs.

So you see the saw tooth pattern visible on the left of his graph?

lower/higher CPU is indeed expected when changing the gain.
But he was talking about the smooth vs sawtooth line.

Yes. I have had the same effects with other metrics too. I suspect it’s an ‘interference pattern’ caused by sampling vs rendering, and gain adjustment is just one way to push the alignment of those into different ‘phases’.

I did qualify that with “Higher and lower gain results in more and less CPU utilisation in the graphs.”

Try as I might, I was unable to reproduce the effect today :frowning:
It happens now and again when I change the gain around my prefered optimal setting.

dump1090-localhost-cpu-8h (1)

While the CPU utilization does go up and down with the gain changes, I cannot reproduce the sawtooth pattern seen previously (much to my frustration) - so I’ll put this down to nothing to worry about.

Thanks guys for your input. At least I’m not the only person to have witnessed it.

Why would you say that?
The graphs clearly show 20 is the right ballpark.

If someone doesn’t want to miss close by flying aircraft, the gain can’t be higher than it currently is as can be seen for the >-3 dB percentage.

25 might produce some more range, but even that completely depends on the setup and situation.
Might not give any benefits.

Well you obviously don’t have the same setup.

For the graphs he gets with a gain of 20, he has an additional external LNA.

Examples:

  • rtl-sdr LNA + rtl-sdr v3 dongle: optimal gain usually between 20 and 35
  • uputronics LNA + Fa prostick+: between 15 and 30 (i think)

Have you looked at his >-3dB percentage? It’s clear he’s at least in the correct ballpark.
That’s why i’m curious why you would suggest a gain that’s somewhere else.
I get the impression you assume everyone is using the Fa Prostick+.

1 Like

@wiedehopf you are correct to think I have uputronics LNA (actually its the SAW version)!
The other items for the record are:
ADS-B Antenna 1.2m, 8.5dBi (Pimoroni)
~5.5 m Aircom Premium (The DX Shop)
FA Pro Stick Plus (Moonraker [uk] Ltd)
RPi 4B with passive cooling in loft https://smile.amazon.co.uk/dp/B07XL17RQ4
Antenna is mounted on chimney stack on a 3m pole as per my user picture at https://uk.flightaware.com/adsb/stats/user/hphillip

@hdtvspace The choice of gain 20 suits me and it’s right that any more gain will cause an excessive amount of >3dBFS messages. I seem to do well compared with nearby sites

Although I’m listed as nearest EGDT (Wroughton) I see plenty of light aircraft activity from Cotswold Airport EGBP especially on fine days, which is only 10 NM away. Gain above 20 means I loose considerable local traffic :frowning:

My average max range is about 210 NM with a peak of 248 NM in the last 24 hours.
The predicted 1090Mhz horizon for my location and elevation using HeyWhatsThat is remarkably consistent with the results I see in practice. To the South and South-East towards France I am limited by geography to a range under 200 NM.
I get better results to the North and through the M4 corridor East towards The Netherlands.

The site seems to be working well, I get plenty of ADS-B positions and my rank is creeping up - I’m hoping to get in Top 50 UK before the end of the month, weather permitting :wink:

2 Likes