Sorry for the lack of updates here! Work has kept me busy, and I've had little time to write detailed blog posts.

I'm now also heavily involved with SatNOGS, where I run a few satellite ground-stations. One aspect of this I've really gotten into is the capture and analysis of radio doppler signatures, which can be used to identify unknown satellites. I've written a guide on how to do this here:Using STRF to fit Doppler observations to a TLE

In a similar vein, I'm also slowly learning about orbital dynamics, and how to generate estimates of orbital elements for upcoming rocket launches, to perform pass predictions. I wrote a short guide on this here:TLE Estimation for an Upcoming Rocket Launch

73, Mark VK5QI

I have recently updated my radiosonde predictions page to include a few more launch site. The updated page is now available here: https://rfhead.net/sondes/

For quite a while now I've been working on the radiosonde_auto_rx project, which is available on Github here: https://github.com/projecthorus/radiosonde_auto_rx

The aim of this project is to make it as easy as possible to track and recover meteorological radiosondes. Recently Michael Wheeler and myself gave a conference talk at Linux.conf.au, discussing this project, and radiosondes in general. This talk is embedded below:

There's a lot of work to be done on the radiosonde_auto_rx project. A few of the major tasks to be done include:

• Add support for iMet radiosondes (commonly used in the US)
• Re-write the signal processing chain to use David Rowe's optimal FSK Demodulator
• Add support for SDRs other than the RTLSDR

Please take a look at the radiosonde_auto_rx issues page and see if there's anything you can help with!

73

Mark VK5QI

Many years ago, I did a radiosonde chase down in Mt Gambier to recover one of the then-new (to us, anyway) Vaisala RS92-SGPA radiosondes. These were being launched by a Vaisala AUTOSONDE station, installed as part of the Bureau of Meteorology's (BOM) effort to make the Mt Gambier site mostly unmanned. Back in 2011, Adelaide Airport was still launching the old 'analog' RS92-K radiosondes, so these newfangled digital radiosondes were a rare thing to us!

Nearly seven years later the RS92-SGPA radiosondes are obsolete, and the new and exciting thing is the Vaisala RS41, which as a friend put it, is 'the STM32 devboard that falls from the sky'. The Bureau of Meteorology are currently in the process of transitioning to the new model, however at the time of writing this post only the Autosonde launch stations (generally in regional areas) had been cut over. Our interest in these sondes is in the potential for re-using them as Project Horus telemetry payloads, particularly if we can reverse engineer the various onboard weather sensors.

So, almost seven years after recovering a shiny new RS92-SGP, a few friends and I decided to head to Mt Gambier over the New-Year break and try and get ourselves a shiny new Vaisala RS41! I was joined by Michael Wheeler and Geordie Millar, both fellow radiosonde hunting enthusiasts, making the trip over from Melbourne. Mt Gambier is about a 5 hour drive away for home for all of us, so this was a fairly serious effort to get one radiosonde!

The Mt Gambier Autosonde Launcher Station

Originally we had planned to spend a week down at Mt Gambier chasing as many sondes as we could. Unfortunately, a discussion with an engineer at the BOM confirmed that regular launches only occurred twice a week, on Mondays and Thursdays. So, we targeted the sonde launch on Monday the 1st of January 2018 - New Years Day.

In the week leading up to the launch the predictions were looking at bleak, with a landing in the Grampians National Park a real possibility. Thankfully by the day before the launch the weather models had settled for a landing somewhere near Dunkeld, just to the South of the Grampians. Still, this was a 160km drive from Mt Gambier, across state boundaries, making this possibly the longest radiosonde chase we'd done in some time.

Chase Car!

We departed Mt Gambier around 9:30AM, and had crossed the border into Victoria by the time we first detected the radiosonde signal on 402.3 MHz - just a few minutes after launch. Using the radiosonde_auto_rx software, the payload position was uploaded into both APRS-IS (visible on aprs.fi), and the Habitat High-Altitude Balloon tracker, allowing members of the local South East Radio Group to follow along with the chase.

Flight Path Overview

The radiosonde flight was fairly standard - 5 m/s ascent rate, and a burst around 27km altitude (actually 26.776 km) while over Hamilton, Victoria. The payload descended (much slower than we were used to, likely due to the lighter payload weight) to a gentle landing 10km to the South-East of a small town called Glenthompson. Unfortunately we were not close enough to watch this one land, but did receive data all the way down to the local ground level of 223 metres.

Final Descent and Landing

After a brief chat with the property owners, who were quite surprised that a balloon had landed on their property, we were given a lift to the landing site. A short walk through a field and the payload was found!

Found it!

Remnants of the balloon. Note the plastic parachute, which is contained within the balloon up until burst.

A few stats on the flight:

• First Received Signal: -37.74643, 140.78701, 723m ASL
• Flight Duration: 2 hours, 42 minutes
• Maximum Altitude: 26776 metres ASL
• Landing Location: -37.71748, 142.64814
• Total Distance Driven: 392 km

We now eagerly await the cut-over to these sondes at our home launch locations (Adelaide & Melbourne), apparently occurring sometime early in 2018.

Our recovered Vaisala RS41. DO NOT BURN OR INCINERATE! (Like we would do that!)

If you're a follower of this blog, you'll know I have a bit of a thing for chasing high-altitude balloons, be it Project Horus launches, or Bureau of Meteorology Radiosonde launches. I've developed software (based on Zilog80's RS decoder) to automatically scan for, decode, and upload radiosonde telemetry, available here: https://github.com/projecthorus/radiosonde_auto_rx

In Adelaide, the Bureau of Meteorology are still launching the now-obsolete Vaisala RS92 Digital Radiosondes. I've posted on these previously (7 years ago?! Where did the time go?!!!). Sadly, not much in the radiosonde is re-usable, so all the sondes I've collected have been sitting on a shelf, or have been given away as souvenirs.

A collection of Radiosondes.

What has always been interesting to me (and what most people comment upon when they see one of these sondes), is the nice quadrifilar helix antenna, used by the on-board GPS receiver. Ben Zandstra (PE2BZ) had mentioned to me the possibility of using the antenna and the receive amplifiers as a L-band active antenna. Today, Will Anthony (a fellow high-altitude balloon enthusiast) and I finally gave it a go! Thanks go to Will for writing the following segment:

The methods that we've used may not be the quickest method to re-purpose an RS92 radiosonde into an L-band antenna/LNA front end.  However we wanted to retain the ability to use the battery packs that come (for free!) with the radiosonde while also ensuring that we don't accidentally and unlawfully transmit with the on-board transmitter.  The steps we took don't necessarily have to be performed in the order that we did them, and across the three units we modified we seemed to do it a different way each time.  It kind of boiled down to who was using what tools at the time.

Speaking of tools, you're going to need some if you're following along at home.  Again, you don't have to use all of what we did, or the same types of things we did, but there are few that you aren't going to want to go without.  See if you can spot them as we go along.  We used the following tools in hacking these up:

• 65W temperature controlled soldering iron
• Hot-air rework station (temp and airflow controlled)
• "Soldapullt" manual "solder-sucker" type vacuum de-soldering tool
• Solder wick
• Liquid Rosin flux
• Side cutters
• Wire strippers
• "bell wire" single strand copper wire, commonly used in telecommunications and some network cables
• Thin (0.56mm) Multicore tin/lead/silver solder
• Small "Standard Slot" screwdriver
• Small bag for garbage/discarded parts, wire ends, etc
• RG-316 coaxial cable
• SMA Male RF connector
• Needle-nose pliers
• Fluke 179 multimeter to check for shorts
• Vaisala RS92 Radiosonde (duh!)
• Some kind of SDR for testing! We tried both a RTLSDR v3 and a HackRF

RS92 Modifications

Lots of components had to be removed from the main PCB. The above figure shows what was removed, and where we attached the RG316 coax for signal output. What was removed included (in no particular order):

• The main CPU
• The GPS receiver IC (a uBlox uN8021)
• The 16 MHz crystal (next to the GPS receiver)
• All of the SAW filters
• The balun at the end of the filter/amplifier chain.

Note that you must retain the SOT-23-5 IC that is next to the GPS chip! This supplies regulated 3.3v to the LNAs.

Overview of the RS92 post-modification.

The sensor and radio daughterboards were also removed, using pin-by-pin de-soldering of the pin headers connecting these to the main PCB.

Once all of the parts that need to be removed are removed, it's a good idea to check the preamp-chain rail voltage to make sure you haven't accidentally short-circuited the 3.3V rail to ground. We managed to accidentally do this when soldering on the coax - a few of the resistors surrounding where the GPS chip was are connected to the 3.3V rail - it's very easy to short circuit them with the coax shield!

Once this was all done, we plugged in a battery pack, checked all the rail voltages were correct and plugged our RTL-SDR into the output.  A few minutes and some free open-source SDR software later and we were able to see signals on the first attempt.  This modification method verified, we then did several more because we enjoy suffering with fiddly things and burning our fingers.  So now we've got the most eco-friendly L-Band receive antenna in the neighbourhood, 100% recycled!

Looking at some L-band signals using a RTLSDR v3

These antennas have been successfully used to decode Iridium signals, and observe other L-Band signals. Some screen-grabs from GQRX are below:

Various Inmarsat Downlinks

Iridium signals around 1624 MHz.

Update:

I was able to hack on a SMA connector to the input of the amplifier chain and get a measure of the gain. The Spectrum Analyser I have access to tops out at 1.5 GHz, but around this point I was seeing > 40dB gain.

RS92 L-Band Amplifier Gain

Test setup for gain measurements.

In my never-ending quest to develop the perfect high-altitude ballooning chase car, I've been playing with preamplifiers mounted close to the receive antennas. This has some nice benefits in that the receive system noise figure can be very low (<1 dB), and the cable loss from a few metres of RG316 is lessened.

I've also added a new antenna - a commercial 70cm turnstile antenna purchased off eBay. With a few modifications and some bracket work, it now mounts nicely to the roof rails of the Rav4:

70cm Turnstile Antenna

This nicely fills in a null that's existed in my overall car 'antenna pattern' for a while - in the direction of 'up'. Using vertical antennas on both balloon payloads and cars means that when the balloon is at a high elevation relative to the receive antenna, the end-fire nulls of both antennas meet, resulting in poor signal. The turnstile antenna has a roughly hemispherical pattern which removes the null on the receive side, improving performance.

The pattern isn't perfectly hemispherical though (a result of the reflector). I've noticed a deep null at about 20 degrees elevation, and performance below 10 degrees when receiving vertically-polarised payloads is very poor.

So, obviously, I need more antennas!

I've always wanted to have the ability to switch antennas quickly while on a balloon hunt, and with the purchase of a MiniKits Relay kit and a bit of enclosure and bracket work (with help from Dennis VK5FDEN and Peter VK5KX), I can now switch between the new turnstile, and a vertical antenna.

The above diagram provides an overview of the new receive system. I now have two receive antennas (the turnstile, and a generic SO239 base), both with PSA4-5043 preamplifiers right next to the antenna feed-point:

More Antennas! (The antennas at the rear are for 2/70 comms, HF, and 4G)

A coaxial relay can be used to select which antenna is to be used, with the controls for this within the cars glovebox. A bias tee just after the switch provides power to the selected antenna preamp.

A splitter is used to drive two SDRs - one, an AirSpy is connected to the Car-PC (an i3 NUC), and is generally used with SDR Console v3 for general purpose receive activities, such as tracking RTTY payloads, or radiosondes. The second SDR, a NooElec RTLSDR (R820T2), is used for Wenet reception, or other Linux-based SDR RX activities. If nothing else is going on, I connect it up to a Raspberry Pi mounted in the back of the car and use it to automatically decode radiosonde launches.

The relay, bias-tee, splitter, and SDRs are all mounted in a semi-temporary manner to the cargo barrier in the rear of the car:

Switching, Splitting, and Receiving

From left-to-right, you can see the relay, bias-tee, splitter, AirSpy, and RTLSDR-in-box. The Raspberry Pi is mounted within an enclosure just visible at the top-right of the image.

Overload Issues...

The preamps being mounted right next to the antenna does pose a few problems. Whenever I key up on HF,  2m, or 70cm, the amplifiers are immediately driven into saturation, and there's bugger-all I can do about it.

With the PSA4-5043, this results in almost 100mW of power being produced by the preamp, going straight into my receivers. Ouch. A BAV99 dual-diode on the RF output of the Bias-Tee helps limit that to something a bit more manageable, which the SDRs own input limiting can handle.

Also, the preamps are wide-open. No input filtering. Yes - this results in a LOT of intermod in the presence of other strong transmitters. However, once out of the city area, they seem to work fairly well. Since install I've performed a few radiosonde hunts with good success.

I'm currently working on a re-designed preamp PCB with a bypass path. This will let me switch out the preamp when I'm closer to the transmitter, and will also allow me to use the antennas for transmit.

Satellites!

When I heard about the LilacSat-1 cube-sat and its Codec2 transponder, I knew I'd *have* to give it a go, if only to make sure David Rowe got a chance to talk through it (after all, he did write the codec!).

LilacSat-1 has a 2m FM uplink, but the downlink is 9600 baud BPSK on 70cm and multiplexer telemetry information with a 1300bit/s Codec2-encoded version of the received FM signal. The LilacSat devs provide their own GNURadio out-of-tree package to decode the downlink, but I ended up using Daniel Estevez's lower-latency decoder. I added a QT Constellation sink to his decoder so I could get an indication of receive quality.

A few hours messing about on my ThinkPad, and I had gpredict, gqrx, and Daniel's decoder all talking nicely to each other, and I was ready to try it out on a real signal! I organised to catch up with David for a coffee and some nerding out at an AREG working bee, which happened to be on right when a nice high-angle pass was going to happen.

Remember my discussion above about intermodulation issues? The working bee was right in the middle of Adelaide's eastern suburbs - lots of interference. However, I had a secret weapon:

No, the tea-towel is not the secret weapon.

Cavity filters are awesome. AREG, also being awesome, have a lot of cavity filters.

I tuned a band-pass cavity up for 436.510 MHz, the LilacSat-1 downlink frequency, and voila - no more issues with out-of-band interference. The insertion loss of the cavity was about 0.5 dB, and the 20dB bandwidth was about 10 MHz. Instead of using the turnstile on the bracket, I sat it on the roof and angled it closer towards the horizon, covering the sector of sky LilacSat-1 was expected to be in.

For uplink we used a Kenwood TM-D710G in another car about 20m away. We ran about 25W into a short Diamond dual-band antenna.

I hooked my ThinkPad into the RTLSDR, and once the sat got to about 20 degrees elevation, I could clearly see the downlink signal!

Main Screen Turn On!

Keying the uplink radio resulted in the downlink turning on, and after some fixing of settings, we could hear audio from the downlink. Andy VK5AKH was also at the working bee and managed to capture part of the pass on video:

The pass only lasted a few minutes, but it was great to see all the software working smoothly, and to see David use his own codec via a satellite!

LilacSat-1 also has an onboard camera which can be controlled via the 2M uplink, and will downlink images via the 9600 baud stream. It's currently not enabled as the cube-sat is currently oriented to optimise GPS reception. Once it's enabled, I look forward to getting out mobile and trying to capture an image of Australia from space!

... and you all thought this blog was all about cars!

Recently we conducted a high-altitude balloon launch for LaunchBox, Horus 37. On this launch I was able to include a project I've been working on with David Rowe for some time, a high speed imagery downlink making use of SSDV.

Unlike analog SSTV, SSDV sends down compressed JPEG images via some form of data link. Written by Philip Heron, the SSDV software converts a JPEG image into a set of packets which can be transmitted via a radio link and then re-assembled on the ground. Unlike regular JPEG images, if a packet is lost, SSDV will still produce a full image, albeit with some portions missing.

Image received via SSDV on Horus 37, with some packets missing.

SSDV has been used in the UK for many years now, though is generally transmitted via slow RTTY or LoRa links, with fairly low resolution images. We decided to turn the speed knob up to 11 and try and send down images that were a bit bigger...

In our payload, we use SSDV to compress images captured by a PiCam, then transmit them via 70cm FSK at 115kbaud. I'm currently using a HopeRF RFM22B as the FSK modulator, running in 'direct-asynchronous' mode, driven from the Raspberry Pi's UART. There were no real design choices with this modem chip, it's just what I had lying around.

I produced a set of 'glue' code in Python which handled the capturing and transmission of images from the payload, and also the reception of images on the ground, using SDR. You can find the source code for the project on Github: https://github.com/projecthorus/wenet

While we've had the system working on the ground for some time now (you can read about some early work on the project here), on the 18th of September Horus 37 was launched, giving us the first real-world test of the system. You can find a writeup of the launch on the AREG website here, and some more technical information on how the imagery system performed here.

I added a few new 'features' to my Rav4 to support reception of this payload. First, I built an antenna mount with a preamplifier installed near the antenna feed-point. This provides a very low noise figure (1 dB) for the receive system. I also built a new cross-dipole antenna, tuned to cover the downlink frequencies of both the imagery system, and the regular RTTY downlink.

Crossed-Dipole antenna, on antenna mount with built-in preamplifier.

The preamp was powered via a bias tee, and the signal fed into a RTLSDR mounted in the boot of the car. As the Wenet software currently only runs on Linux, and my Car PC (a Gen-4 i3 NUC) runs Windows 7 (for annoying legacy reasons), I plumbed the RTLSDR into a Virtual Machine running Ubuntu.

Wenet RX software running within a Ubuntu Virtual Machine.

This VM image was also used in Matt VK5ZM's chase car for the hunt, and also at a stationary receiving location run by Andy VK5AKH.

We're hoping to perform another launch in later October, by which time we should have some improvements (forward error correction) to the imagery downlink ready.

See you then!

73

Mark VK5QI

As many would know my Hilux was recently towed home with a busted harmonic balancer that resulted in an engine rebuild.

I finally got around to creating my own blog so I've moved the original post and we can continue the discussion here (ZedM.net).

Hope to see you there !

A few months ago I purchased a new (secondhand) car, a 2011 Toyota RAV4. Of course, my first priority was to kit it out with amateur radio gear! Fast-forward to the present, and I've now got an auxiliary battery system, a Kenwood TM-D710G, a Codan NGT SRx, and a Codan 9350 auto-tuner antenna installed.

RAV4 + Radios!

Now, I've been following the development of Codec2 and FreeDV (Open-source digital voice for HF) for quite some time, and have been involved in testing of FreeDV over-the-air for a while. One of the early tests involved using FreeDV while mobile, on the Hume Highway near Yass, NSW. This required a fair bit of infrastructure to get going (laptop, rig interface, headset), and so the release of the "Smart-Mic" SM1000  promised to allow easy use of FreeDV in a mobile situation. So, at the last AREG meeting, I bought a SM1000 straight from the author of Codec2 and FreeDV himself - David Rowe.

Over the last few days I've been working on interfacing the SM1000 with the NGT SRx, and David suggested I document the process, so here we are!

One of my aims for using the SM1000 in the car was being able to really use it like a microphone. One cable, low clutter.

The SM1000 has an isolated rig interface via a RJ45 socket with a configurable jumper block, and has a DC supply on a separate connector. I didn't really want to use the DC supply connector - this meant soldering wires to the DC connector pads, connected to the patch panel. The final wiring diagram is below:

SM1000 to NGT SRx Interface Cable

A few notes on the connections:

• The audio input on the NGT is balanced. I experimented with grounding one side of the input and using it single-ended, but this resulted in feedback issues. With the MIC GND separated from the other grounds, it can be used in a balanced mode. I used one of the CAT5E pairs to carry the balanced signal.
• All the grounds on the patch panel are separate from the main SM1000 DC ground, so to power the SM1000 and have the PTT and Speaker output work, they must be connected together. Again, more soldered wires.
• The case of the SM1000 is nicely connected to the DC ground, which is now connected to the ground of the NGT. If the NGT isn't bonded to the car body well (i.e. braid to an lug or similar), you'll get an interesting tingle whenever the antenna is tuning!

The patch panel ended up looking like this:

SM1000 Patch Panel

We could definitely use a bigger patch panel in SM1000 revision F! Also breaking out the supply and ground rails to the header would definitely make what I was trying to do a lot easier!

Finally, the output level of the SM1000 is only 50mVp-p. This is fine if you are driving a microphone input, but the NGT requires an input a bit closer to line level (300mVp-p).

The output level is set by the trimpot R47, and fixed resistors R48 (10K) and R49 (4R7):

SM1000 Microphone Output Circuit

Switching out R48 with a 1K resistor results in a maximum output level of around 350mVp-p, which can easily be adjusted down with the trimpot. R48 is an 0402 SMT resistor, so a hot air rework station is recommended:

R48 - 10K 0402 Resistor

Unfortunately I only had 0603 resistors in stock, but I managed to shoe-horn one into position:

R48 - Now a 1K 0603 Resistor

The day after modifying my SM1000, I headed over to Andy VK5AKH's QTH. Andy also owns a RAV4, and also has a NGT SRx, so we modified his SM1000 and made up a suitable interface cable.

To configure the SM1000's levels we set up our cars near each other, with the NGTs set to transmit on a clear frequency. We also took the whip antennas off the Codan 9350s to limit propagation (the 9350 will quite happily tune the antenna stub on 40m and above). First we tried a basic SM1000 to SM1000 contact, but it quickly became clear that setting levels would be easier with a laptop running the FreeDV software at one end. Luckily I had brought my Signalink USB. A quick modification to the Signalink's internal jumper block and it was made compatible with the NGT to SM1000 cable, and I was able to run FreeDV and look at SNRs.

Testing QI Mobile (TX) to AKH Mobile (RX) (approx 15dB SNR)

Testing AKH Mobile (TX) to QI Mobile (RX) (Approx 20dB SNR)

A bit of tweaking of trimpots later, and we were able to get at least 15dB SNR between the cars. For some reason I was able to get better SNR from AKH->QI than QI->AKH. More investigation required!

After all this we still had pretty significant garbled voice issues, with conversations somewhat intelligible, but nowhere near as good quality as we had encountered with the FreeDV software. We decided to hit the road and headed over to David Rowe's house for a coffee and some more testing.

On the way we were able to maintain a conversation on 40m for a while via ground-wave, though still with the garbled voice issues. We arrived at David's place, where we started performing a few tests:

• Carrier TX using SM1000 test mode: Worked fine. Clean RX tone, no distortion evident on either transceiver.
• Analog audio TX using SM1000: Pretty bad distortion on the audio, as received by the other car. Tweaking the mic gain didn't seem to help... (hmmmm)
• Transmission of FreeDV test signal from David's home station: Decoded audio sounded great coming out of the SM1000. This means the receive chain on the transceiver + SM1000 checks out... must be something in the transmit chain.
• FreeDV transmission using test audio into SM1000 External Mic socket: Sounded great! At this point we figured something must be wrong with the SM1000's internal microphone.

After some further adjustments, we had both SM1000's operating nicely with the Codan NGTs, giving good quality audio!

Two operational FreeDV mobile stations!

Over the next few weeks we'll be using the SM1000's during drive-time (probably on 80m or 40m), getting a feel for how usable they are in a noisy mobile situation. Stay tuned!

Update #1

HF conditions have been pretty poor, to the point that even SSB comms aren't working on either 80m or 40m in the morning or afternoon (bring on a 60m allocation!!). So, we haven't been able to really give the SM1000s a decent test (apart from some short ground-wave contacts).

I also had the need to interface my laptop with the NGT again, via my Signalink USB. This meant modifying the Signalink's internal jumper to be compatible with the NGT-SM1000 cable:

NGT-Compatible Signalink Jumper Block

This jumper block should also be compatible with the 6-pin Mini-DIN cable which Tigertronics supply.

A fairly straightforward chase and recovery, taking a scenic route through the Adelaide hills towards Mannum. Gary VK5FGRY came along for the chase.

Tracking the balloon mid-flight.

We stopped up on the top of the Adelaide hills to experiment with some different receiver options. My trusty Icom IC-R10 has developed a BNC connector fault, so the HackRF had a chance to work as a receiver. I used GQRX to demodulate the FM signal, then passed that audio into a Windows 7 VM where Sondemonitor and my mapping software was running.

Landed!

We made it to the landing area just as the balloon did, but we overshot the landing site by a few hundred metres and missed watching the landing.

A short bit of DFing later, and the payload was recovered!

Payload recovered.