I’m just thrilled to see all of the new ‘Blue (or is it Purple)’ planes on my PiAware / Dump1090! Certainly I undrstand how ADS-B is very accurate, as the plane is reporting it’s exact location. **What is the accuracy of MLAT, especially since you don’t know precisely where all of the receiving stations antennas are located? ** Thanks again for all of your hard work and it is certainly fun watching your product improve (although my little Pi B+ is now using about 50% of it’s CPU cycles trying to keep up with all of the plans and MLAT…).
The more receivers, the more accurate the plot.
The better configured each receiver, the more accurate the plot.
The closer to accurate time each receiver has, the more accurate the plot.
And so on, and so on.
There are some pretty funny looking tracks when there are only 4 or 5 receivers.
As time goes on obj and the FA crew work to eliminate erroneous results and throw out plots that are ‘obviously’ wrong, it will get better over time.
I suspect you have tried this, but I would be curious to see how close an MLAT location would be to the actual ADS-B location. Should be possible to ‘ignore’ the location information on ADS-B equiped planes and ‘pretend’ they are only Mode S. Then you could use the MLAT calculations and compare it to the actual location (received from the ADS-B)? Just wondering how close we think MLAT is (in an area with enough receivers). Are we talking 100 meters? More? Less?
In theory, MLAT can tell you exactly where the plane is, as long as:
The altitude reported on Mode S is correct
The height of the antenna on all the receivers is configured properly
The height of the ground at the receiver is configured properly
All receivers are synced on time
I don’t know if FA would ever use MLAT over and ADS-B report, I highly doubt that. GPS (again in theory if programmed correctly) should always be the most accurate.
But I have a plane that lands here nightly on the taxiway, so you tell me how accurate anything can be in real life 
If you trust what the plane transmits, anyway 
**What is the accuracy of MLAT, especially since you don’t know precisely where all of the receiving stations antennas are located? ** Thanks again for all of your hard work and it is certainly fun watching your product improve (although my little Pi B+ is now using about 50% of it’s CPU cycles trying to keep up with all of the plans and MLAT…).
Receivers with wrong locations get ignored pretty fast - they have large jitter in their clock synchronization. Similarly for receivers with an unstable clock, which is the other common problem (CPU overload / dropping samples manifests as an unstable clock). Somewhere on my list is feeding this information back to the stats pages so you can be aware there’s a problem rather than just wondering why you have no results.
The clock errors for a properly located/synchronized receiver typically add up to maybe 200-300m. The overall error depends on the individual receiver errors and goes down as you get more receivers seeing the target. After filtering I would estimate the usual position accuracy at +/- 500m or better. It depends quite a lot on the positions of the receivers relative to the aircraft. The altitude accuracy is pretty terrible if there’s no transponder altitude available (it’s because most receivers are on the horizon, not below/above(!) the aircraft). The filtering process doesn’t produce results at if it estimates the current error at >4km.
Currently there’s nothing done about things like multipath, because (a) it’s hard without a custom receiver and (b) having lots of receivers is a simpler way to filter that
Comparing back to ADS-B positions was something that I did very early on and the accuracy was good, but it’s a bit of an unfair comparison because you’re using the transmitted position to synchronize the clocks, then using that synchronization info to multilaterate the position … of course they’re going to be close. It was a good way to test the mlat solver if nothing else It might be interesting to do this for aircraft that transmit a low position accuracy category that aren’t used for sync.
I hate when that happens…
Hahah - I’ve got one from today. Looks like the plane mis-aimed and tried again, mis-aiming again.
http://victorspictures.com/img/s7/v152/p718304115-4.jpg