Manufacturing Process of a Graw Radiosonde

From 16.09. – 25.09.2019 I had the opportunity to get an impression of the production of radiosondes at Graw Radiosondes GmbH & Co. KG in Nuremberg in the form of an short-term internship. I experienced almost every step in the manufacturing process of a radiosonde myself and was able to gain first-hand impressions.

DFM-17 also already entered regular production

Two things surprised me the most. On one hand how much manual work is in the manufacturing process of each radiosonde. Each sonde goes through many individual steps from the pre-check of the circuit board to the assembly and calibration all the way to the packaging and, by the time it is finally on its way to the customer in the packaging, it has probably been touched two dozen times. The other thing is that what we mainly deal with, GPS, microcontrollers and radio, is hardly important compared to the know-how that goes into the development, production and calibration of the sensors.

Most sonde hunter would get envy about such tracking equipment on the roof

Graw currently produces three different types of radiosondes: the DFM-09, the DFM-17, its successor, and the PS-15, a pilot sonde without TU sensors. Apart from the calibration of the sensors, which of course is not necessary for the pilot sonde, the production is relatively similar for all sondes. The production of the sensor booms is also done in-house and is very similar for DFM-09 and -17. The ground stations are assembled in the production facility of the sister company Noris, which is located in the same building complex.

Pilot sondes have no sensor boom containing TU sensors and can only report wind speed and direction

First the production of a sensor boom will be described. The basic material is a flexible PCB that is supplied externally. The PCBs are supplied individually punched out. Ten PCBs are each placed and clamped in a rail-mounted frame. Solder paste is applied to the pads for the thermistor of the humidity sensor (which will not be fitted until much later) using a time-pressure dispenser, followed by manual fitting of the thermistors measuring size 0402. Under the microscope, the thermistors are soldered with hot air and aligned manually.

In the next step, the thermistors for the temperature measurement, which are THT components, are placed in a similar jig over the corresponding pads on the sensor booms, solder paste is applied and the thermistors are soldered under hot air. The sensor booms are then cleaned from flux residues in an ultrasonic bath with rubbing alcohol.

A thermosetting insulating paste is then applied to the solder joints under the fume hood and then cured on a hot plate. In addition to mechanical protection, this also serves as insulation against the reflective coating that will be applied on the sensor boom later. In this step, a piece of kapton tape is also put over the footprint of the humidity sensor so that it will not be coated. The reflective coating of the sensor carrier and the cap covering the humidity sensor is performed by an external specialist company. This process includes not only aluminum deposition, but also the additional application of a protective lacquer. Finally, the coated sensor booms are tested in a test jig for insulation/conductivity of the respective contacts and the (coated) kapton tape is removed.

For the sondes, the manufacturing process begins with a pre-test. The assembled PCBs arrive at the factory unflashed, but separated from the panel and with a soldered switch/pushbutton in reusable boxes from two differnt PCB assembly companies. The PCBs themselves come from two different manufacturers in China.

During the pre-test, the sondes are electrically tested, flashed and in part adjusted and/or calibrated. In addition, each sonde is assigned its serial number and the first small label at this stage.

For the DFM-09, both PCBs, which are still connected with perforated bars, are first electrically connected with a ribbon cable. The sonde is then placed in a test jig that automatically tests basic parameters such as power consumption. The quartz and oscillator frequencies are set via two trimmer capacitors, which are operated via two stepper motors with shafts and screwdriver heads attached to them. In addition, the sonde is flashed and the GPS and 403-MHz-TX are contacted and tested via coaxial test connectors. For the DFM-17 the programming is done via the groundcheck connector by bootloader in temporary lack of such a jig.

The sensor booms are then inserted into the connector located on the PCB and the sondes are moved onto a production cart. The temperature calibration can then be performed. To do this, the sondes are inserted into a carrier frame and then calibrated in temperature baths at different temperatures. The liquid used is ethanol at the higher temperatures and a special coolant at the lower temperatures. During these calibrations, the sondes are connected to the control system via pogo pins, and if calibration is successful, the control system stores the correction data on the EEPROM of the sonde.

After this step, the humidity sensor can be soldered on. A reflow process is also used here, but due to the sensitive nature of the sensor, it is placed with a vacuum gripper and soldered on a hotplate. The humidity sensor can then be calibrated. Therefore, the sensor booms of sixteen sondes are positioned at the same time in a duct through which air of variable humidity is circulating. In this channel, air can be introduced which can be controlled in its water vapour content by drying agents or by passing over a water surface. The calibration takes place at different humidity contents, but only at room temperature. Again, the correction data for each sonde is calculated and stored on the EEPROM after successful calibration.

Now the manual soldering can take place. In the case of the DFM-09, this includes the antenna, the batteries and the coupling of both circuit boards through the RF shield and the GPS antenna, the latter can be omitted in the case of PS-15 and DFM-17. This is followed by the final inspection, in which the sondes are tested in a second test station similar to the one used for humidity calibration. Here various parameters such as plausibility of PTU values, telemetry transmitter and GPS reception are verified. For this purpose a GPS repeater system is in use in the factory, which receives the GPS signal through a roof antenna and distributes it to transmitters on each test stand. Each sonde that successfully passes the final inspection is visually inspected and the protective cap is placed on the humidity sensor.

The test stand for the final inspection resembles the one used for humidity calibration

The finished electronics are then mounted in the housing. With the DFM-09, this is done in two steps. First, the sondes label is printed out and attached to the enclosure, the sonde is then placed in the enclosure. In the second step, the cover is glued to the enclosure, again using a time/pressure dispenser. With the DFM-17, the batteries are inserted and the sonde is checked for function (by switching it on). The maximum run time of the sondes is one minute when leaving the factory. The sonde is placed in the polystyrene housing, the lid is put on and the sonde is locked with a cable tie, before the sondes label is stuck on.

The last step is to pack the sondes. In general, customers can choose between vacuum and cardboard packaging, which has hardly any disadvantages under normal storage conditions. With the DFM-09, in this case ten sondes are packed in one cardboard box. Before this, however, the sondes are inserted into the plastic sleeves, which are assembled in an adjacent workshop for the disabled. The unwinder, which is delivered fully assembled from a workshop for the disabled in Belgium, is inserted into the part of the plastic sleeve that positions the sensor boom in flight. The DFM-17 is placed in a box with a cardboard inlay that keeps the sonde and unwinder in place. With the DFM-17, the stick is already pre-assembled on the unwinder and only needs to be clicked into position by the user inside the cover of the sonde around the cable tie. Furthermore, the unwinders no longer have a plastic sleeve, but a cardboard disc on the upper side and are exclusively equipped with green balloons.

During my internship I gained a lot of insights into production and development and probably held more radiosondes in my hands than I can collect in the next years and decades. For this I would like to thank the whole team of Graw and Noris, especially Mrs. Christina Flohry, Mr. Florian Schmidmer and Mr. Norbert Traeger!

Launch of an Ozone Sonde in Uccle

On 29.07.2019, on my way back from the Tomorrowland Festival in Belgium, I organized a trip to the KMI in Uccle to bring back my two Belgian ozone sondes and visit the institute and attend an ozone sonde launch.

The KMI has now had a continuous ozone measurement series of radiosonde ascents for almost exactly 50 years. Launches are always Mon, Wed and Fri at 11:30 UTC. ECC ozone sondes are used, since the middle of the 90s, before Brewer-Mast sondes were used.

In general a normal radiosonde, here a Vaisala RS41, is used for ozone sounding. The ozone measuring device is connected to this sonde, which consists of an electrochemical reaction cell, a sampling pump, a control board and the necessary batteries, and is connected in a polystyrene box. If you don’t already know it, take a look at the Vaisala Ozone Sounding Guide, which is very good and detailed.

The preparations on the launch day start with the electrochemical testing of the cell starting at 7:30 UTC. We reached the institute at 9:00 UTC and were immediately taken into the balloon hangar to inflate the balloon. 1200 g balloons from Totex are used, which are filled with hydrogen to a counterweight of 2400 g. The hydrogen storage is located about 30 m away from the balloon hangar and is connected to it via underground piping. When the takeoff volume has been reached (the balloon is a  considerable bit larger than a normal helium balloon from the automatic launch system despite the hydrogen filling), the balloon is tied up with cord and attached to the parachute. The balloon is removed from the filling device and attached to a weight, the cord between parachute and probe is measured and attached to the parachute.

the filled ballon can stay in the hangar for a few hours, while the other preparations take place

Then we went on to the laboratory where the electrochemical preparations take place and which is located in the basement of the institute. This is where the ozone sonde supplies and the test station are located. During the preliminary examination for the next but one ascent, my sondes were handed over. For the staff in Uccle, it is okay if the finder opens the sonde to reduce the packing volume and to remove the batteries. The air intake hose is also replaced each time the sonde is launched.

The tour continued in the attic, where the receiving station and the ground check station are located. Here I got the opportunity to program the sondes I brought with me to a different frequency.

About 1 hour before the launch the ozone sonde is brought up for the groundcheck. The heating battery is attached and the motor battery is fixed in the battery compartment. Since Vaisala no longer distributes water activated batteries, Uccle switched to lithium batteries in spring as one of the last stations in western europe. Vaisala now supplies 2 9V lithium batteries instead of the water activated battery, which are already connected in series, fitted with the correct plug and glued shut.

A radiosonde is unpacked and placed on the groundcheck device. After the ground check has been completed and the correction values for the ozone sonde have been entered in MW41, the radiosonde is attached to the ozone sonde. The motor is switched on and the motor current is controlled. Now ozone is flown into the sonde again and the measurement is checked before the motor is switched off again. The probe is then placed on the roof to obtain the GPS fix. The paper documentation of the sonde is entered into an online form and stapled together. The sonde is completely glued shut using tape.

15 minutes prior to launch, the values of the ground-based weather station are entered in MW41 and supplemented in the documentation. The sonde is brought into the balloon hangar and fastened with both cords (one serves for the improvement of the flight stability) to the long cord leading to the parachute.

A second employee now exits the hangar with the sonde, and the assembly as a whole is brought to a nearby meadow. The balloon is risen gradually. Now was my moment: I was allowed to hold the sonde and wait until the time for the launch had come. Then I released the sonde on its journey, which took it west of Antwerp.

I would like to thank Roger Ameloot, Roeland van Malderen and the whole KMI team for giving me this great insight into their work!

My first Hat-trick

I had been waiting for my first hat-trick for quite some time, but until now there had only been two consecutive sondes. That changed when I was awakened on Wednesday morning by manoeuvre sonde from Baumholder who were just approaching my QTH. I quickly got out of bed and took the next train to Marienheide, where the first 0630Z sonde had just landed in the forest near Listringhausen. While I was waiting for the bus there, the next sonde already landed behind Marienheide.

#21 Baumholder – Listringhausen
Graw DFM-09 / 17007958 / 05.06.2019 0630Z
🔗 in storage
🔗 public transport

After the position of the first sonde had been decoded at the bus stop, I went 2 km into the forest in the best weather, only to find out that the chute was hanging on the edge of the forest at a height of about 7.5 m, and the sonde was not visible. So the parachute was quickly caught and pulled down, and lo and behold – the robust DFM could be pulled over the treetops, and eventually abseiled.

#22 Baumholder – Marienheide
Graw DFM-09 / 17050161 / 05.06.2019 0800Z
🔗 in storage
🔗 public transport

Further on to the second sonde, this time by bus to the Bruchertalsperre, there across the dam wall and this time about 3 km into the forest, beforehand the position was decoded. This time the parachute was hanging on a high spruce, which bordered on a spruce conservation, and I had to climb along the slope to the landing site. Again the parachute was just within reach of my pole, again the sonde could be abseiled. But this time I had to extend it’s string with the string of the previous one. New tool in the sonde backpack: DFM string.

#23 Baumholder – Nümbrecht
Graw DFM-09 / 17050614 / 05.06.2019 0930Z
🔗 in storage
🔗 public transport

In the meantime the following sonde had gone down faster and had only made it to Nümbrecht. Short investigation on the way back revealed that the landing site could be reached by foot from the central “bus station” aka road with many branches and bus shelters, Homburg-Bröl in the southern county past Wiehl. Therefore quickly into the TH in Gummersbach to empty the now full backpack in my lab and refill my water bottle. The new “express bus” 302 took me there. The sonde had already stopped transmitting and was not visible in the clearing where the forecast saw it. I stomped a few steps into the high undergrowth, which was full of ferns, and saw the sonde as ready to be picked up on the ground. The hat trick was complete. So quickly back to the bus stop, where I could also stop three ticks from getting too comfortable on my calves… and another tool for the sonde backpack: tick card.

Gulaschprogrammiernacht, the second one

NP4 Idar-Oberstein – Boppard
Vaisala RS41-SGP / P2050587 / 03.06.2019 1200Z
🔗 given away
🔗 individual transport

When the way home was waiting for us on Monday, the forecast for the midday probe from Idar gave us reason to take a look over Koblenz, especially as we wanted to see the stream of the WWDC keynote at flynamic in Bonn in the evening.

While the sonde was on its way, the forecast predicted a landing south of Koblenz, we stationed at Vapiano and then took a short walk to the Deutsches Eck. Meanwhile the sonde fell very fast and didn’t even make it across the Rhine and landed near the A61 at Boppard.

The last position was reached quickly, but the signal of the sonde could hardly be heard, let alone decoded. An apparently nearby BOS-Funk station also littered the spectrum with TETRA, which didn’t make things any easier. Only with a mobile Yagi increased to six directors it was possible to achieve anything. When the sonde was found, it was clear why: the antenna had buried itself in the soft forest soil and the sonde body was on top of it. The small red parachute and the big balloon remains hung in a nearby tree out of reach of the pole and could not be pulled down either.

Souvenir from the Gulaschprogrammiernacht

#20 Stuttgart – Bietigheim-Bissingen
Vaisala RS41-SGP / P3620018 / 02.06.2019 1200Z
🔗 in storage
🔗 individual transport

Flynamic, nikcani and myself drove to the GPN19 in Karlsruhe over Ascension Day holidays. Of course the sonde hunting equipment was aloon board, one doesn’t know what might happens. And indeed, the Sunday midday sonde from Stuttgart landed at a reasonable distance southeast of Bietigheim-Bissingen, on a paddock behind a farm.

I was afraid that the on-time recovery of the sonde would be harder in the more fought-over Baden-Württemberg area than at my regular QTH, and therefore also registered the planned recovery at radiosondy. Unfortunately we didn’t met anybody spontaneously at the sonde. When we arrived about 90 minutes after landing, the sonde was active and was quite easy to salvage.

The sonde was on a path behind a farm, the big balloon remainings and the unopened parachute were situated behind an electric fence on a paddock, but could be pulled over.