Dump1090-fa adaptive gain and wiedehopf autogain

FlightAware General
1

1.Does wiedehopf autogain work while dump1090-fa is running or just at 0245am in the morning ??
2.Can it run along side the adaptive gain in dump1090-fa or without ??
3.Or can adaptive gain cope with auto adjusting the gain on its own ??
4. Is wiedehopf autogain more suited to readsb were it does not have adaptive gain ??

Value feedback on this

Hope this makes sense ?

setup Flightaware Pro Stick Plus v1.0
Raspberry pi 4b + RPI Buster OS

2
  • readsb doesn’t have adaptive gain, not a feature i deem necessary.
  • adaptive-gain will be off if you install autogain
  • dump1090-fa having adaptive gain … i didn’t make it work again after the config change.
  • still works on piaware images

Change the decoder, see how you like readsb … you can always change back to dump1090-fa easy enough.
After changing decoder, you’ll need to reinstall graphs1090 if you have that.

3

Few days ago on one of my Pi, I started using readsb instead of dump1090-fa (reimaged microSD card).
I have noted that the CPU and Memory usages both have gone down somewhat.

 

image

 

image

 

image
image1258Ă—327 59 KB

 

4

That is something i have realized as well.

5

dump1090-fa
dump1090-fa1842Ă—664 181 KB

Dump1090-fa1
Dump1090-fa11302Ă—696 191 KB

dump1090-fa2
dump1090-fa21302Ă—778 173 KB

dump1090-fa3
dump1090-fa31310Ă—706 171 KB

That is interesting all of you mention this .
I have enclosed screenshots of same .
The Raspberry pi4B and small screen is my testing unit ( no room to install the fan) which had also READSB installed and other software.
As you see the temp is extremely high .
About 20 minutes i installed READSB and all of the software relevant to Wiedehopf also the offline maps for Tar1090.
There is an old rtlsdr connected without an antenna but looking at the graphs1090 core temp with fan is running at Max 43c.
Wiedehopf software installation using auto scripts went well without any errors.
Away to connect the Flightaware pro plus v1.0 to the other setup with the 10inch screen and monitor the gain etc till tomorrow .
Will update next few days.

6

I have now slipped out microSD card with readsb, and slipped in the old microSD card with dump1090-fa. The comparison is below:

CPU (Down from 22% to 12%)

DUMP1090-FA
image1

READSB
image2

 

MEMORY (Almost same)

DUMP1090-FA
image3

READSB
image4

1 Like
7

That doesn’t seem right.

Are you running that on a single core?
It detects that and assumes rpi zero … and reduces the CPU usage (which also reduces performance a bit).

8

No, I did NOT make any changes in settings to run it on single core. It is running on default settings on Pi model 4 with 1 GB RAM

9

I don’t suppose you have the graphs for message rate?
That seems a lot of CPU usage for dump1090-fa, not sure it’s the norm.
Oh well … not important. Glad readsb is using CPU conservatively.

10

Message Rate

image
image1258Ă—327 107 KB

11

Yesterday I wrote 64-bit RaspberryPi OS on a spare microSD card, and booted RPi Model 4 with it. Next installed following (in the sequence shown below):

  • Readsb
  • Graphs1090
  • Flightaware feeder (piaware)
  • Flightradar24 feeder
  • Planefinder feeder
  • Radarbox24 feeder
  • Adsbexchange feeder

Let it run for 24 hrs. Today evening when I checked the graphs, I found “Messages > -3dBFS = 29.4%” and ADSB CPU usage 12%

I checked the gain setting of readsb, found it was at default -10. Changed it to 32.8.

Checked after about 1.5 hrs, found
(1) The bad red Messages > -3 dBFS have become very low
(2) The ADSB CPU usage has dropped from 12% to 6%

image

 

image

 

image

 

12

Probably the single biggest thing that affects CPU use is the input gain and local noise environment (i.e. how much noise gets thrown at the preamble detector) as abcd’s later graphs show. For an apples-to-apples comparison you’d need to ensure the same gain settings.

2 Likes
13

Congratulations
It works as expected

Exactly the same i see when i forget to change gain

14

I have now slipped out the microSD card with readsb, and have slipped in another microSD card with Piaware microSD card image. Now the decoder is dump1090-fa instead of readsb.

I have set “adaptive gain = no” and manually adjusted gain to 60 (which is same as readsb’s -10). I will leave it running for a day, and tomorrow evening change gain to 32.8, as I have done with readsb, and will let it run for new gain settings for say 2 hrs and check again.

There is a problem with graphs with dump1090-fa. If adaptive gain is made “no”, the gain graph becomes blank, I dont know why. Anyway I have embedded the gain button in graphs to see the “Current Gain” along with the graphs.

image

 

image

 

Traffic is very low here from 2330 hrs to 0630 hrs

image

 

 

15

dump1090-fa does not provide the gain in the stats.json file when adaptive gain is off.
Thus it’s not available in the graphs.

16

Yeah, this is just an unfortunate implementation accident where I put the current gain within the adaptive gain stats. no adaptive gain → no current gain. I’ve been meaning to fix that.

3 Likes
17

As the 64-bit readsb was installed on 64-bit OS, while 32-bit dump1090-fa on 32-bit OS, their performance comparison cannot be considered exact.

I will now prepare 3rd microSD card with 32-bit Raspi OS Lite, and install 32-bit readsb on it and run it to see how much is it’s ADS-B CPU Usage, first with gain -10 and after a while reducing gain to 32.8.

Following are results with Piaware SD Card image (32-bit OS + dump1090-fa)
Up to 1415 EDT:
Gain = 60
ADS-B CPU Utilization = 25%

From 14:15 EDT to !6:15 EDT
Gain = 32.8
ADS-B CPU Utilization = 20%

image

 

image

18

You could also post the overall CPU use … maybe i screwed something up with the CPU accounting.

19

@wiedehopf

microSD card #3 in RPi Model 4

OS = 32-bit RaspberryPi OS lite (Bullseye)

Decoder = 32-bit readsb

Initial gain = -10
Final gain = 32.8 (gain changed at 14:50 EDT)

 

image

 

image

 

image

 

image

 

 

20

I guess it’s less CPU. Probably mostly because i’m only trying to fix DF17 for flipped bits in the DF type.
That’s pretty significant.

Rpi4 power usage depends very little on CPU usage though … so doesn’t make too much difference.

1 Like