Saturday, 17 September 2016

ADS-B Testing Notes

There is a huge community out there who love anything connected with airplanes. Yours truly included, who thinks that leaning forward in St. Maarten into a departing 747's jetwash is great fun, and a good night out is listening to Approach, Ground and Tower from a hill. If you snigger, stop reading, this post is not for you.
Because ADS-B performance relies on equipment, and I do ADS-B performance comparisons, it is necessary to elaborate on setup details. This is a technical, and somewhat boring post, but required for full and absolute transparency. Furthermore, individuals (and manufacturers) might be interested in how I reached a particular result.




Back-end computing



I already had a Pi 2 and a Pi 3, on different Amperage power supplies, in different cases, with different lengths of Ethernet cable, and system variance was below 2 percent.
ModMyPi came on board and agreed to supply two identical packages (many thanks, review here), so same Pi 3 and power supply and MicroSD card and Ethernet cable could be used.


Two percent rule



Results, pro and contra, should be viewed knowing that system variance is maximum two percent.
Comparative values showing less than 2 % difference could be down to setup, manufacturing tolerances, and a host of other variables which are impossible to eliminate in a real-world environment with more than 20 individual components.
2 % figure has been achieved with two factory-fresh premium dongle sets from two different manufacturers.
Retested the main (red) ModMyPi setup with the latest v.3 dongle from rtl-sdr.com before the Pro Stick review went online, and variance was less than 1 %.


Data aggregators



All data from feeding FlightAware, because:
1) most popular with like-minded community,
2) ease of use,
3) verifiable and easy to present findings (printscreens).
PiAware version 2.1.5 used, as the latest version with WiFi support just came out, and the previous iteration works just as well. Wifi would have ben easier to setup, but
1) extra variable, and
2) placing a 2.4 GHz transmitter right next to a sensitive receiver might influence end results.
Simultaneously feeding Planefinder for range polar plots, which are much more detailed than FlightAware's.


Baselining


I planned to use chinese generic dongles as didn't want to abuse the goodwill of premium dongle manufacturers, but variance was in the neighborhood of 3-4 %, unacceptable for reliable results. Initial testing has been performed with indoor antennas and two SMArts.
These two SMArts used for baselining had to be opened for reviews and photos, so Nooelec kindly sent two factory-fresh dongles.
Rtl-sdr.com also sent two new v.2 dongles (previous generation), so I could do back-to-back comparative testing.
Neither dongle sets were opened, modified or tinkered with. Two dongle sets enabled me to double-check results and to cut time in half.
On both setups, figures differed ranging from 0.2 % to 1.7 % with same brand same model dongles over 12 hours minimum.
Dongle pair from rtl-sdr.com had 0.3-0.5 % less variance, and lowest value observed was with them, plus the SMArts are used for other experiments, consequently two previous generation rtl-sdr.com dongles will be used as comparison baselines, one in the Pi3/Pi3 and one in the Pi 2/Pi3 setup.


Splitters



Two-way splitters used, so same antenna can be utilised.
Tried and tested four-way splitters with four SMArts, and results were unreliable due to four-way splitter.


Connectors and adaptors


The bane of my existence. Imagine setting up, checking reception, and finding that no data is coming in. Power fault? Card corrupted? Antenna shorted? LNA died? PiAware reinstallation?
9.9 times out of 10, the source is a faulty connector, an F-type or SMA not fully screwed on. Or no contact because SMA Male pin is too short. Splitter is F-type, but dongles are SMA or MCX, so F-Type to SMA pigtails and adaptors from Nooelec after splitter, because they have never failed so far, and tried to use pro gear as much as possible.
Had to use homemade gear with RG6 coax and F-connectors, same length down to millimeter level.
Impedance mismatches exist, so antenna comparisons won't be 100 % accurate. I can live with that, and this fact will be noted in any upcoming review.


Coax, mounts, antennas



To keep results real and to approximate a general user's profile, two identical cables and mounts supplied with the Nooelec SMArt, RG59 50 Ohm, length 6.5 feet / 2 metres.
Metal telescopic antennas supplied with RTL-SDR dongles, extended to the first joint. Half-wave on a ground plane, which is a sweet tin, antenna bases as close as possible.
Bought and tested a FlightAware antenna, but performance improvement in range and received signals does not justify the $45 price tag compared to stock antennas on a magnetic mount.


Timeframe



FlightAware counts a new day at 0100 local time, but I can't be always up to change receivers; so I log results and do printscreens close to bedtime, double-check stations and connections afterwards.
20 or 23 hours of data collection period did not change percentage numbers due to local flight patterns.


Your results won't be the same, but will be very close



Location, antenna, positioning, weather, plus a great number of other variables mean that your ADS-B results differ. If you see 157 mile range with a receiver, don't expect that buying the same receiver and components will get you 157 mile range.
This is a comparison setup, with one singular purpose: to find out ADS-B performance of various dongles and / or antennas.
Compared findings with other enthusiast living literally on the other side of the globe, and my results for SMArt vs Mini 2 + were very close, despite significantly less position reports and shorter timeframe.


I'll always listen when you got something to teach



More dongles than I can locate, four Raspberry Pi microcomputers, professional antennas, connectors, preamps and the joint goodwill of Nooelec, rtl-sdr.com, ModMyPi, Adam, plus my purchases and time.
The more I know, the more I realise I know nothing, therefore please share your opinion, tips, tricks, any observation, here as a comment, on reddit, on rtl-sdr.com, on a FlightAware forum, or by sending an email (address in Manifesto).
Many thanks in advance, and I can't express my gratitude to all contributors who made this comparison possible, from manifacturers supplying equipment all the way to enthusiast Joes and Janes buying my books.

2 comments:

  1. do you know the power consumption of the nesdr smart? This is critical in my stratux build and I know that the nano's are generally around 280ma and the low power sdr is 180ma. But I had 3 of the low power sdr's fail, the several nano's fail, and so I'm interested in the better heat dissipation of the smart, but I can't find good details as to the power consumption.

    ReplyDelete
    Replies
    1. I don't know the official figure for power consumption, ask Nooelec for an official figure. I'm running SMArts for monts in a terrestrial setup and had no heat related issues. If you had Nanos fail it's due to their small size, and because the IR receiver is directly on top of the chipset. Try removing the IR receiver and adding holes to the plastic case if you need the small size of the Nano, but the SMArt is much more advanced dongle.

      Delete