Gain results observation


#1

Firstly, there are so many gain posts, some of which are dated, I thought I would put this up for comments.

For the hell of it, I thought I would do a 12 hour gain test. Now this was between 6.00 am and 6.00 pm, , with a reasonably large regional airport near by and also a Defence Basic Flight School operating out of the same. My coverage footprint extends down towards Sydney and covers the north west and west approaches to Sydney from Asia and Europe. Average ADS B positions reported during bushiness days between 95,000 and 126,000 positions.

Using the popular original gain scripts, a Prostick Plus, a bias t power injector and a RTL-SDR Blog ADS-B Triple Filtered LNA (Bias Tee Powered, a DPD ADS-B Vertical Outdoor Base Antenna 9db gain), up the top of the pole, utilising Power over Ethernet and a POE Hat, a PI 3B+. 62 seconds over 66 runs,(each run took about 11 minutes to complete) utilising all possible gain settings, including AGC and -10.0 (reason for all possible gain settings, I have found some smaller gain tests resulting generally in the best setting being below 20.0, probably due to the setup,high gain LNA and a high gain antenna. I don’t need much gain to get high figures, SNR reasonably good as well.

The below results are reasonably self explanatory,EXCEPT for the Aircraft result: The results will have Gain, Messages, Positions, Aircraft and largest to smallest including the relevant gain numbers included:

Meassages:

GAIN - MESSAGE - POSITION - A/C
12.5 - 201422 - 45876 - 281
8.7 - 200743 - 45704 - 278
15.7 - 200408 - 45461 - 282
14.4 - 200179 - 45525 - 282
7.7 - 199577 - 45506 - 280
16.6 - 198277 - 44809 - 283
3.7 - 195659 - 44758 - 278
19.7 - 195542 - 44235 - 282
2.7 - 193715 - 44366 - 277
20.7 - 190725 - 43231 - 284
22.9 - 181904 - 41599 - 283
1.4 - 180832 - 41622 - 278
0.9 - 176269 - 40642 - 275
25.4 - 174166 - 40059 - 282
0.0 - 158133 - 36595 - 270
28.0 - 153583 - 35886 - 285

Positions:
GAIN - MESSAGE - POSITION - A/C
12.5 - 201422 - 45876 - 281
8.7 - 200743 - 45704 - 278
14.4 - 200179 - 45525 - 282
7.7 - 199577 - 45506 - 280
15.7 - 200408 - 45461 - 282
16.6 198277 44809 283
3.7 - 195659 - 44758 - 278
2.7 - 193715 - 44366 - 277
19.7 - 195542 - 44235 - 282
20.7 - 190725 - 43231 - 284
1.4 - 180832 - 41622 - 278
22.9 - 181904 - 41599 - 283
0.9 - 176269 - 40642 - 275
25.4 - 174166 - 40059 - 282
0.0 - 158133 - 36595 - 270
28.0 - 153583 - 35886 - 285

Aircraft:

GAIN - MESSAGE - POSITION - A/ C
28.0 - 153583 - 35886 - 285
20.7 - 190725 - 43231 - 284
16.6 - 198277 - 44809 - 283
22.9 - 143247 - 33577 - 283
29.7 - 181904 - 41599 - 283
14.4 - 200179 - 45525 - 282
15.7 - 200408 - 45461 - 282
19.7 - 195542 - 44235 - 282
25.4 - 174166 - 40059 - 282
38.6 - 92118 - 22129 - 282
12.5 - 201422 - 45876 - 281

Now you will note, Messages and Positions both have 12.5 as the winner. Aircraft though has 28.0, double the gain and some, of the other 2 gain fields. A quick observation from my end though, on smaller tests 62 secs and below and over only say 15 to 30 secs, the 3 data fields generally are the same gain level. Long and more extensive checks, generally have the messages and positions the same, but Aircraft different gain level.

Please feel free to comment and may be explain why the Aircraft gain level, over that length of time, would be so different to the message and position. What causes this?


#3

L O L, that was quicker than re typing LOL :slight_smile: It’s been a long day at work…


#6

I’ve just done a 32 day performance tracking of gain vs positions/aircraft.
Previously I tested various gain figures and found no advantage running less than 37. In the graph below I have multiplied the gain figure by 2 for visual effect, actual figures are between 37 and max with AGC ( i.e. -10 setting).
The gain was set for the complete 24 hour period.No other changes were made to the system for the thirty day period. From the 9th of April, gain was set -10.
It shows that there must be other factors that are affecting performance, such as weather and atmospheric conditions.


#7

The number of planes in the sky is not always the same. You need to compare the same day of the week with another, at same hour to be more precise.


#8

Nice graph, handy reference. You can see the troughs of each weekend, a dip for the Easter week, and the kind of levels you’re getting mid-week. When you look at those across the different gain settings I think your assessment is right, there’s no signifcant correlation between your different gain settings and the useful data decoded, with auto gain giving similar results and riding out the other variables.


#9

@SoNic67
@chrislfa

Note that the Pos/Acft line in the graph is the product of dividing the number of positions reports received in the 24 hour period by the number of aircraft reported.

Thus the figure should reflect the daily reception conditions regardless of the actual number of aircraft or day of the week.


#10

You can get the number of aircraft in a UTC day from the stats page to compare with your calculations.

I have two systems on the one mast. One is my reference system that is not changed in any way and the other is the experimental system.

Arrow 1 points to aircraft on a very windy wet nasty day when no General Aviation or training was flying (and YMML was delayed about 45 minutes with about 15 aircraft at any time waiting to land.

Arrow 2 points to what could only be described as a beautiful day for flying and there were lots of light planes doing circuit work.

In the early days of the month I was changing antenna, radio, gain, amplifier and filter combinations and it is easy to see when the experiment improved or not compared to the reference.

From about 11 April there has been no change and the relationship between the two systems stays the same regardless of how busy or not the sky is around here.

This method of comparison reasonable removes all other local factors such as traffic levels, types of traffic, weather, day of the week, phase of the moon and performance of your local stock exchange.

As a secondary check I compare my stats day by day with two other local stations to see if they track the same way.

Interestingly, this highlights that whilst aircraft total and aircraft with position may be higher on one system the messages per second is often higher on the other. Range also often tracks counter intuitively. I don’t know why yet but it gives me further experiment on.

I welcome discussion on this methodology for evaluating changes. (Puts on fireproof suit in preparation) :man_firefighter:

S.


#11

I have collected a lot more data than that shown in that graph. As I’ve tried to explain, the data (Pos/Acft) is the product of dividing the number of aircraft positions by the number of aircraft over each 24 hour period. It is from FA and FR24 that I have gathered that data. The assumption is that the calculated figure is largely independent of variations in the actual traffic numbers, so quiet days and busy days should be largely irrelevant.

If the system is working well, it should track aircraft longer outbound and pick up tracks earlier inbound, thus collecting more position reports. If conditions are bad, then the figure will be lower, as aircraft fade earlier or are picked up later and thus produce less position reports.

If that data assumption is valid, it shows that there were other factors than the gain setting that caused the variation.


#12

Pos/Acft is in a direct relation with the distance to those planes. Also is dependent of the number other planes close by the antenna.

Especially when you leave AGC active, a stronger signal from a closer plane will force the receiver to lower the gain, “drowning” the signal from a far-away planes. That “drowning” happens also when the gain is constant, but less prevalent, because of collisions in the transmission band. If the signals from two planes overlap in time, only the stronger one will be decoded. Those collisions are statistically independent, there is no mechanism in ADS-B of adjusting the transmission times to avoid collisions.
You need to remember always that those are not stationary targets with fixed relationships between their levels. they move in respect to the receiver and in respect of each other. So levels and timings are variable.


#13

All true, I told my grandmother the same.


#14

I’ve tested it with my stats and it’s not holding up. I see the same shape for positions/aircraft as I do for positions and aircraft. Which means the data you’re looking at is dependent on variations in traffic numbers and the factors around gain which @SoNic67 mentioned.


#15

Did she also told you that that’s a false assumption?