I do have my old NooElec still. Iâll use it for the test. Is it better to mount the LNA near the antenna or the Pi? Does the bias Tee to be right next to the LNA or can it be on the other end of the cable if you mount the LNA nearer the antenna?
The bias-t needs to be on the receiver side of the LNA not on the antenna side, so the power arrives on the correct end of the LNA.
(If you have >100 ft of cabling the voltage delivered to the LNA will reduces somewhat, in that case iâd feed 4.5 V or 5 V so enough arrives at the LNA)
Antenna â coax â LNA â coax â bias-t â coax â receiver
Those coax pieces are just there to signify a connection, if it is 0 ft or 30 ft of coax doesnât matter.
LNA near the antenna is better, but no matter where it is the performance should improve compared to the prostick+ alone.
If it doesnât improve something isnât working right.
airspy_adsb is actually measuring the average SNR of the frame instead of its average power. The resulting number is then scaled and truncated to fit in one byte for backward compatibility with the existing software. The implication is that once you get âgood enoughâ reception, the value doesnât change much (if at all) because of the limited range available in 1 byte.
Thereâs also another âaggravatingâ parameter: Processing Gain. The filtering/decimation from 12 or 20 MSPS to the final bit decision at 1mbps reduces the noise dramatically which, in turn, increases the dynamic range.
5 Likes
Thanks for the explanation. From that, am I correct to assume that there is no reason to reduce gain unless decodes are being lost at close range? Seems like it would be advantageous to maximise long range reception as far as possible.
1 Like
Correct. The strategy is quite simple: Set the maximum gain that doesnât cause a loss of the local traffic.
4 Likes
prog and caius, thank you both (caius for asking the question and prog for the great reply). this is good information and much appreciated!
2 Likes
I did a quick test earlier. I increased the gain to 21 and observed local aircraft. I didnât really see any data lost at close range due to overloading and tracks were complete (apart from a few low level GA aircraft, but that is typical for here). There was a noticeable impact on received message rate, but no noticeable effect on number of aircraft seen:
I backed off the gain one step at a time (can see at about 1410 and 1430) and at each step the message rate increased. Again there was no obvious change in aircraft seen.
That would appear to imply that losing close range strong signals isnât the only impact of having gain set too high - it would seem that the ability to decode overlapping messages is also hampered. I ended up back with the gain set to 19, and message rate back to normal.
1 Like
Can you show the signal graph for the experiments? 
Here you go:
Though from what prog said, itâs not really indicating signal strength.
It is a mix. After demodulation you compare the âperfectâ full scale signal form you would expect from the bits to the actual received amplitude.
So while itâs SNR, itâs not gonna read 0 dB without the amplitude also being maximum.
Considering that there is always some noise, you only get 0 dB when the signal is already stronger than ADC full scale.
This implies that the source of strong signals is not your local traffic. Try to do some statistics and plot the strongest signals on a heatmap.
That is definitely possible, since Iâm using an antenna that should have about 5dB gain.
I didnât make a separate data set I can plot with the gain set to 21, but this does include that period. The difficulty is that thereâs a lot of signals at the stronger end. Itâs common to get signals at 0dB at quite long ranges:
This was today anyway:
Even plotting with the signal strength palette range constrained to -2dB doesnât show much in the way of a pattern.
Tomorrow Iâll try adjusting the gain and collecting data only from that period, however I think identifying any areas that are losing data due to overloading will be difficult. Iâm trying to think of a way to do it more scientifically than âhmm, that bit looks a bit thinner for data pointsâ. Binning the data and then somehow normalising it to account for the different traffic between settings might work.
can you make this same plot, on the same data, for 0 to -2dB, like rudimentary binning ? The two plots look quite similar indeed. Not sure that there is much to disentangle beyond variability in seasonal air traffic, weather.
The difficulty is that signals showing 0dB are decoded fine, and that those signals can appear at long as well as close range. Even if you just plot those 0dB signals you will get a plot full of points that has a similar pattern - by itself it wonât tell you much.
Take a look at this one scaled from -0.5 to 0:
Itâs almost as if the colours are randomly selected, with very little pattern at all. Thereâs no obvious transition from strong to weaker.
Iâll try binning the data into something like 1 degree by 0.5nm sectors, and then normalising it so that data taken at different times can be compared more easily. What is going to be difficult is if the signals that are lost because they are too strong are spread evenly across the whole area. Spotting that will be hard, as increasing the gain will increase the apparent signal strength of other signals in those areas as well.
Today I made another ground plane antenna. I used some copper flashing to form a cone. The radiator is a piece of superflex 1/2" heliax with a 1/2 wavelength (including connector) of cable. This looks pretty good on the analyser, 1.19:1 VSWR and 47.5 ohms. Probably could have made the cone fractionally shallower than the 45 degrees.
I expect no difference from the previous attempt with the exception that this should be somewhat more rugged and ft for a life outside.
After a few coats of spray lacquer I plan to mount this above the rooftop.
1 Like
