FlightAware Discussions

Pushing the envelope

Firstly, I am experimenting with my units and the antennas available. No need for any “why would you do that” replies. I have 3 antennas available up a mast. 2 x DPD adsb omnis and a Jetvision 1090 yagi. All amps to be used are ceramic uptronics. As you will see in the diagrams, I am running odroid N2’s and XU4’s and Airspys. The crucial part of this is a SMA t piece. At the moment I have 2 units mounted on the mast in their cosy little enclosure, a N2 and dual antenna Radarcape, one of the DPD’s and yagi into the Radarcape and the other DPD omni onto the N2. I am planning to move my XU4 to the enclosure. But what I am going to try is running 2 antennas, ommi and yagi, into the XU4, utilising the t piece. the radarcape will go onto another antenna system. The question is the placement of the amps, as diagrammed. I will be utilising the gain feature of the Airspy, to make adjustments as required, to compensate.

First setup:

Second set up:

Pending the results, I would transfer the best setup onto the N2. Thoughts please :slight_smile:

My first thought doesn’t take the amps into consideration at all. I’m merely thinking about what I surmise is a passive splitter (tee) that will essentially be halving the signals coming in from each antenna.

For Setup #1 – I think it will all work, but I’m concerned that the amp will not be able to compensate for the halving of signals from each antenna (induced by using the tee). It should work, but I’d not expect the antenna system to have anywhere near the end-to-end gain that a single ant/amp setup has. And if the amp is something like a 14-15 db amp, you are going to be pulling in less planes on each antenna – even if you happen to pull more in combined by virtue of one antenna covering an area the other doesn’t. I’m thinking that if you were going to use a passive tee like that, you probably really want to use an amp with significantly more gain than that which you would normally need for a single antenna setup. We already know the Airspy needs gain by default. Prog (Youseff) one stated how much gain is needed from an amp just to break even at 978 / 1090. And when you’re halving signals, that suggests to me (maybe incorrectly) that you are going to need an amp with more gain.

For Setup #2 – i don’t know if you are going to be able to power two amps with the Airspy. Even if the Airspy can provide the juice, I’m not sure if the electrical component of that setup is going to work and actually power two amps. I’m not currently capable of understanding how the circuits between the two amps are completed (and if they are) in a way that ensures you have full DC power to each of the two amps by when feeding it up one feedline to a tee.

I’ll certainly be interested in hearing about your experimenting though. I might end up being very pleasantly suprised. Sounds like fun either way!

Solid T, or Passive?
Insertion Loss (dB) ?
Isolation between 2 antennas (dB) ?

SMA Solid T



Insertion Loss = 0 dB

Isolation = 0 dB
image 225x225
image 225x225
SMA Passive
Frequency Range: 100-2700MHz
VSWR: ≤1.4
Isolation: ≥20dB
Insertion Loss: ≤1
Passband Ripple: ≤0.3dB
Phase Imbalance: ≤3°
Impedance: 50Ω
Maximum Input Power: 25W
image 321x251

If the cable lengths are such that one is shorter with lambda/2, then the signals will be 180 degree in phase and will cancel each other. In air lambda is 275mm, in cable is approx 340mm (depends of the actual cable)… So 1/2 of that is like 170mm.

Even shorter differences in length will lead to some cancellation.


1 Like

A resistor, possibly.


1 Like



The diagrams below are from the article linked by @eaglemark in his above post and clearly show that in Wilkinson’s combiner/splitter, the elements are λ/4. This makes it suitable for only one frequency (or a narrow band of frequencies).

On the other hand, the ebay seller claims a very wide-band (Frequency Range: 100-2700MHz). How ebay combiner could achieve such a wide band?




1 Like

Both amps are externally powered from power packs

Why would you do that? :wink:

But seriously - what’s the expectation here? I’d expect that the overall system would perform something like the average of the gain pattern of the two antennas - is that what you’re aiming for?

You may also see some unexpected directional effects depending on the separation of the antennas and any phase difference in the feed lines.


It’s very very likely not going to improve your performance, probably even get worse performance.


Well, the OP is ignoring my previous post and will be for a rude awakening.
You just don’t mix RF signals like DC ones. No matter what a Chinese seller might tell you, the relative phase of the AC signals matters.

What I need to add to the previous post is that even the relative position of the antennas, in report to the direction of the signal, will matter because it will introduce additional delays (propagation time in air).
Since planes can be virtually anywhere in a 360 degree circle, two combined omnidirectional antennas (even with exactly equal cable lenghts) will have different “gain”, dending of the actual angle of the airplanes to the line between the two antennas (like any directive antenna plot).

PS: That’s a whole theory around that simple effect and steerable arrays of antennas are deployed in a multitude of places, based on adjusting the delay time at each antenna. From military radars to WiFi routers. But that’s beyond this simple discussion.


Yep. What @jlb56 built is a “random phased array” with “nulls” at “some unknown frequency” and some extra gain at the axis orthogonal to the nulls.
That said, this could be useful in some situations when properly built.


I wonder what the circuitry of one of these is like?

1 Like
1 Like

Very much like the picture in this post:


I understand why you think this, but there are several reasons why you might want to do it and more why you shouldn’t.

First point of course is that it is so simple and obvious, why are you the first to try?

Second point (as above) a “CoAx T”, is not the same as a splitter/combiner (except at DC).

Third is that (as above) the two antennas need to be mounted with the same “phase center”. This is the same as being at the same electrical position in space (not to be confused with physical position).
Another way of saying the same thing:

  • What you expect: 1 + 1 = 2 (you signal has doubled)
  • What you may get 1 - 1 = 0 (no signals)

Try it by all means, but be prepared to admit it doesn’t perform as you expect.

How about beam sharpening via a sidelobe suppression algorithm?

Two antennas feed a pair of SDR 1090 MHz receivers. One antenna is omnidirectional, say a collinear vertical of +6 dBi gain. The other is directional, a Yagi with +12 dBi in the main beam.

The SDR outputs are processed in parallel. The preamble of each decoded transponder reply are amplitude compared, and only those replies that are stronger in the Yagi channel than the Omni channel are forwarded for full decoding.

This “beam sharpening” should result in replies being limited to the direction of the yagi antenna main beam.

The 6 dB beam width of a 12 dBi Yagi is about +/- 30 degrees. The Yagi response is actual, the omni is idealized.

Only replies from about 60 degrees of the 360 horizon are processed.

Anyone conversant enough with the preamble processing of replies care to comment on the difficulty?

You would need synchronized sampling on the two receivers (a two-channel SDR would probably be simplest; you might be able to do something with a pair of receivers driven from a common clock, but aligning the sample streams could get messy)

A simpler approach might be to just decode independently and then look at signal strength after decoding.

1 Like

The OP wants to feed one receiver with two antennas.

You probably could do the DSP, but it seems like a lot of work for no benefit.

The Pi3 has enough CPU to run two receivers feeding a copy of Dump1090 each. Why filter out data when you can just plot it? It doesn’t matter if a position is plotted twice (it’s the same position after all) - if it is a problem, discarding a redundant position would be easier than the complexity of trying to sync two receiver chains.

Edit: when en I say “it doesn’t matter…” I mean in terms of displaying for your own interest. If you target is point scoring, then I guess it does.