In the TT7B project, I attempted to design a low power tracker aimed at collecting data during long duration flights under custom made superpressure balloon envelopes. The design and development of the tracker are described in detail in the following two blog posts:
While working on the hardware and firmware, a couple of tools used mainly during testing were developed:
Rather than using commercial machinery, I put together a couple of tools designed solely for the purpose of making balloon envelopes, both described in the following blog posts:
A detailed description of the manufacturing process of the superpressure envelopes along with a general discussion about the used material can then be found in this blog post:
The details, progression and data of all individual flights done as part of this project can be found in the following blog posts:
The following paragraphs contain a summary of the experiences with launching the balloons, collected data during the flights, and a conclusion of the project.
Three trackers were built, and due to two successful recoveries, a total of five balloons were launched within a period of 30 days in the beginning of fall 2019. A rough overview of the flights can be found in the following table.
| designation |
launch date |
envelope type |
altitude [m] |
distance [km] |
duration [day] |
| TT7B1 Flight |
09/13/2019 |
Circle 1.35 |
454 |
0.36 |
0.0 |
| TT7B2 Flight |
09/15/2019 |
Obround 2.35 |
13,500 |
790 |
0.9 |
| TT7B1 Relaunch |
09/21/2019 |
Circle 1.35 |
10,800 |
526 |
0.5 |
| TT7B2 Relaunch |
10/01/2019 |
Obround 2.35 |
13,400 |
57,355 |
27.8 |
| TT7B3 Flight |
10/13/2019 |
Circle 2.04 |
9,898 |
123 |
0.1 |
The first flight, or rather attempted flight, paid the price for over-reliance on theoretical calculations with little margin for error. Although the method and ascent rate targets used in sizing the lift were identical to the TT7F flights from two years ago, I didn't take into account the fact that those balloons were launched later in the day and always spent some time in direct sunlight, heating up the lifting gas, prior to the actual launch. TT7B1, although ascending initially, climbed only several dozen meters before it encountered some downdraft or nonlinearity in air density beyond its lifting capability and descended into treetops a few hundred meters away. The tracker was eventually recovered later in the day.
The second flight was launched two days later with higher free lift and a couple more hours after sunrise to avoid the same issue. It was the first flight to provide the desired data on superpressure and supertemperature. However, as the balloon settled to float at its equilibrium altitude, the data began to show decreasing pressure inside the envelope. The pressure kept decreasing suggesting the hydrogen was continually leaking from the envelope until after 17 hours of flight, the balloon lost lift, started to descend and landed in northern Croatia. Luckily, it was recovered by Kruno 9A3SWO who contacted me and sent the tracker back to me in the mail.
The third flight was a relaunch of the TT7B1 tracker that ended up in treetops under a new envelope. It was released with a higher free lift and later in the day than the failed attempt. Shortly after stabilizing at its float altitude, it became obvious from the pressure data that the balloon was once again slowly leaking hydrogen, and that it would descend later that day. The inevitable happened several hours later, and the balloon landed in western Romania.
The fourth flight was a relaunch of the TT7B2 tracker recovered in Croatia. The identical failures of the two previous flights led me to suspect the glue that was used to seal the balloon's gas inlet with the pressure and temperature sensors inside the envelope. As a result, this balloon's inlet was heat sealed and the sensors were left outside the envelope. The suspicion was proven correct by this flight as it floated and reported its position for 28 days during which it circumnavigated the Northern Hemisphere.
The fifth flight wasn't planned originally. The TT7B3 tracker differed from the previous two and was constructed relatively last-minute without extensive testing. It was a solar powered, combined APRS and WSPR version of the tracker which used an unideal antenna to transmit in the 2 and 20m bands. The signal reception was significantly worse than on the previous flights, and the last packet was received over Poland after just a couple of hours in flight while the balloon was still ascending. There was no further information about its fate.
The trajectories, and receivers in case of the shorter flights, can be seen in the maps above. The trajectory of the longest lasting flight was reconstructed from packets received live and backlogged historical data the tracker stored transmitted with every live packet.
| designation |
VB [m3] |
mB [g] |
mP [g] |
mFL [g] |
vE [m/s] |
v2k [m/s] |
altE [m] |
altF [m] |
altd [m] |
| TT7B1 Flight |
0.434 |
116.0 |
11.0 |
5.1 |
0.78 |
- |
12,410 |
- |
- |
| TT7B2 Flight |
1.045 |
224.0 |
18.5 |
10.8 |
0.91 |
0.65 |
13,810 |
13,510 |
300 |
| TT7B1 Relaunch |
~0.430 |
116.0 |
18.6 |
7.2 |
0.90 |
0.69 |
11,930 |
10,796 |
1,134 |
| TT7B2 Relaunch |
1.099 |
224.0 |
19.1 |
13.2 |
1.00 |
0.89 |
13,880 |
13,416 |
464 |
| TT7B3 Flight |
1.497 |
274.0 |
8.2 |
14.5 |
1.00 |
0.62 |
15,190 |
- |
- |
The table above contains detailed parameters of each flight, while the resulting launch trajectories are recorded in the ascent rate and altitude profile charts. The $V_{B}$ parameter represents the measured envelope volume - in case of the third flight the value is only a pre-flight estimate as I wasn't able to actually measure the envelope that time. Data from the actual flight suggest rather smaller internal volume of just around 0.380m
3. Parameters $m_{B}$ and $m_{P}$ represent the masses of the envelope and the payload, respectively. The values in the $m_{FL}$ column represent the final measured free lift on a sealed and prepared balloon (cut excess film) with payload attached and thus slightly differ from the intended target values. The $v_{E}$ ascent rates then represent expected initial ascent rate for the system, while $v_{2k}$ column contains actual average rates from launch altitude to 2,000 meters. The last three columns show the expected float altitude $alt_{E}$ for the system and atmospheric conditions on the day of launch (data from closest weather station's sounding), the actual initial float altitude $alt_{F}$, and the difference $alt_{d}$ between the two.
It is apparent that specific atmospheric conditions on the day of launch play important role in the initial ascent rate at low lifts such as these. Especially noticeable is the varying ascent rate on the fifth flight despite relatively high free lift compared to the other flights. A difference between envelope types is observable in the data as well. The obround shaped envelopes peak at significantly higher ascent rate when they are fully inflated than the circular envelopes due to different drag coefficient. The data also show that except for the fourth flight, the ascent rate all the way to roughly 10,000m was very similar on all the other three flight. The difference between the expected and actual float altitude in case of the second and fourth flights (300 and 464m) is, in my opinion, due to some inaccuracy in the volume measurements and due to some remaining air taking up space in the sealed envelope as it is not easy to get it all out after pre-stretching. The large difference between expected and actual float altitude on the third flight (1134m) is mainly due to the poor estimate of the unknown envelope volume. There were some issues with the pumps during its pre-stretching, so I made a guess about its volume purely based on pressure levels achieved during pre-stretching. The in-flight data suggest I was way off.
The six charts above show a 28 day long, detailed recording of TT7B2's (relaunched) altitude progression. Unexplained issues with the GNSS module, however, led to erroneous altitude reports on a number of occasions during that time. The obvious errors were removed from the data set. Other suspicious data points were left in it. When lone outliers such as on the 2nd of October and a two hour long slump in the middle of the night on the 19th/20th of October (likely a GNSS module error) were removed, the variation in the balloon's altitude over the 28 day period was 900m (maximum: 13750m, minimum: 12850m). Upon a closer look, there is no clear, regular day to night variation in altitude distinguishable from development of the air mass in which the balloon floated. The balloon's altitude would vary within a range of less than 200m on a typical day with a few exceptions where the balloon's altitude changed by up to 500m.
This chart captures TT7B2's altitude with respect to the balloon's latitude. Generally speaking, closer to the poles a balloon is, lower in altitude it will be, while closer to the equator it is, higher in altitude it will float. Latitude, however, is only one factor contributing to the height of the air density level at which the balloon found equilibrium. Large scale weather systems and consequent evolution of the air mass surrounding the balloon disrupt the linear relationship. Thus the general tendencies can be observed in the data, but so can exceptions.
The collected altitude vs. latitude data inspired me to look into the phenomenon more closely. The three charts above contain a whole year of sounding data from three weather stations at different latitudes. First from Alert in Nunavut at 82.5°N (blue), then from Omaha in Nebraska at 41.3°N (green), and from Belem in Brazil at 1.4°S (red). The original data can be found on the University of Wyoming
website. The first chart captures how high in altitude an air density level is throughout a year above the station. Specifically a level of 0.2664kg/m
3 which is where the relaunched TT7B2 found equilibrium. It is apparent at first sight that the equatorial levels are the highest throughout the whole year and also the most stable with a difference between the minimum and the maximum of only 300m. At the pole, the density levels are the lowest throughout the year regardless of weather systems underneath, and they undergo much more prominent swings with minima and maxima difference of 1100m. At mid latitudes, the density levels experience the largest swings throughout a year and also the most prominent differences between the summer and winter months. With altitudes approaching the equatorial values during the summer, the minima differ from the maxima by 1600m. The data basically represent the float altitude the relaunched TT7B2 would have had had it flown over the specific station at the specific time. The second and third charts, then, provide the values of pressure and temperature, respectively, at the density level. The relationship between the three properties is described by the following equation: $$\rho = \frac{p}{R_{specific} \cdot T}$$ where $\rho$ is the air density, $p$ the air pressure, $T$ the air temperature in Kelvin (K) and $R_{specific}$ the specific gas constant for air (287.058J/kgK). For a balloon floating at a constant density level, the relationship means that if either of the two properties - air temperature or pressure - increases the other increases as well, and vice versa.
These three charts show the same sounding data. This time in full scope (pressure and density only up to 15km in altitude for better discernity). During the year worth of data, it was the coldest, perhaps unintuitively, above the equator, but only in a range of altitudes a few kilometers above typical float levels. At float altitudes, it was the warmest during local summer close to the pole. The data also suggest that temperatures above the typical minimum rating of -40°C for electronic components are a rarity anywhere around the globe at the altitudes the TT7B balloons floated. In case of air pressure, the higher surface temperatures at the equator cause air to rise consequently leading to higher pressure than close to the pole when considering the same high altitude. In case of air density, a range of altitudes between 6500 to 7000m seems to be a turning point. Above, a specific density level can be found higher in altitude during local summer than during winter, while below this range a specific density level can be found higher during local winter than during summer. Similarly, the same density level can be found higher in altitude at the equator than at the pole above this range, and vice versa below this range.
For completion, these are altitude profiles of the relaunched TT7B1 flight and of TT7B2's first flight. Both exhibited a similar range in daily float altitude to the relaunched TT7B2 throughout the single day they floated. Notice that as the balloon slowly leaks, the float altitude stays more or less the same initially, only after the buoyant force of the remaining volume of gas becomes less than the weight of the system, the balloon descends fairly rapidly.
As the sounding data suggest, the TT7B balloons that managed to float did so in very cold air masses outside the typical operating range of electronic components. During the day, the electronics get warmed up by the Sun to within the operating range, however, during the night the temperature of the components eventually falls to the ambient air temperature. This typically led to some of the components not working properly, or to a complete end of the tracker's operation until the next morning suggesting either the MCU or the transmitter were affected by the cold. In case of the GNSS module (u-blox ZOE-M8B), the average time required to produce a valid positional solution would increase, but it would stay operational if the ambient air temperature stayed above about -58°C. When the ambient air temperature was between roughly -58°C to sometimes down to -63°C the tracker would still transmit packets, but the GNSS module would no longer respond to communication, and the attempt to poll data would time out in about 2 seconds.
The tracker on the first TT7B2 flight contained an instruction to do a system reset with a fixed five minute period for satellite acquisition in case it failed to acquire a valid solution five times in a row which can be observed in the first chart above. This instruction was removed for the latter flights along with shortening the maximum duration during which the GNSS module attempts to get a valid solution. Since the 16th day of the relaunched TT7B2's flight, its GNSS module began to experience difficulties maintaining lock onto satellites for some unknown reason. This lead to a significant increase in average active time, and erroneous positional and altitude measurements started to show in the data. Note that these sometimes significant errors were passed as valid solutions by the GNSS module. This unexplained behavior drastically decreased the expected lifetime of the battery as the average GNSS active time was the main contributor to the tracker's consumption.
The average active time on the first TT7B2's flight was 6.4s when the system reset periods were omitted from the data, and only 2.4s when only data from the day were considered. The average number of satellites used in a solution was 6.2 with a maximum at 11 satellites. The relaunched TT7B1, on the other hand, averaged only 5.2 satellites in a solution with a maximum of 7. This probably contributed to a higher active time of 5.6s during the day it managed to float. The previously discussed TT7B2's second flight (relaunch) saw initially average daily active times in the range of 3.9 to 6.3s (7.2s when the GNSS module operated throughout the night). After the GNSS module began suffering from the unknown issue, the range increased to 10.8 to 17.5s. However, data from the last day show a return to the initial range with an average of 3.4s. The average number of satellites used in a valid solution was within a range of 5.2 to 6.5 with maxima of up to 11 satellites throughout the whole flight.
Another issue with GNSS became apparent on the TT7B3 tracker. This tracker transmitted 2 minute long WSPR transmissions which when ended saw the GNSS module having lost fix to all satellites and requiring several seconds, sometimes tens of seconds, to reacquire it. Something in the transmission, or the Si5351B's chain of frequency reference manipulation must have been interfering with the GNSS reception. The earlier APRS transmissions were probably too short to cause a noticeable issues. Later, I encountered this interference problem separately with a different u-blox module and Si5351B in a different configuration as well.
The next component noticeably affected by the cold was the pressure sensor (MS5607), specifically its pressure measurement as it provides temperature data as well. There were two pressure sensors on each tracker. One directly on the PCB (onboard), the other (external) on an extended cable glued inside the balloon envelope in case of the first TT7B2 flight and the relaunched TT7B1. In case of the relaunched TT7B2 the cable was left hanging outside and shielded by a piece of silver rescue blanket. The sensors would typically produce sensible pressure readings as long as the tracker was illuminated by the Sun. After the Sun set, the pressure readings would either begin to drift significantly or max out completely. The TT7B2's first flight, because it descended during the night, provided insight into the air temperature at which the sensors began to output adequate readings again, around -34°C. This behavior unfortunately meant that the pressure readings couldn't be used to verify whether the relaunched TT7B2 indeed descended during the night from the 19th to 20th of October by 1000m in an hour and a half, or it was just a GNSS module error.
Based on the multiday data from the relaunched TT7B2, the MS5607 would return to outputting sensible pressure values when it measured temperatures above -40°C, its datasheet stated operating range. The chart above shows one such sunrise where the onboard sensor warmed up enough about an hour earlier than the external sensor.
A couple of times the external pressure sensor didn't work properly the entire day. A bit of a mystery about the relaunched TT7B2 flight was that most of the time the onboard sensor would measure higher temperature and work even when the external no longer did, but there were occasions where it was the other way around including the measured temperature. I am wondering whether the external sensor's shield perhaps kept moving or unwrapping throughout the flight influencing the situation.
The temperature data were obtained from five temperature sensors on each tracker. There were two thermistors, one onboard the other external, each of the two pressure sensors output a temperature reading as well, and the fifth temperature measurement came from the MCU. All the individual sensors were affected by self-heating and absorption of solar radiation to a degree. On the relaunched TT7B2, the least affected was the unshielded onboard thermistor, but even that one measured 10-15°C higher air temperatures than during the night. Its night-time measurements, then, were about 0-1.5°C off when compared to the closest sounding data on a couple of occasions I checked. In contrast, the external thermistor was shielded by a reflective rescue blanket canopy, and it typically measured temperatures about 25-35°C higher during the day and about 5°C higher during the night. The also shielded, external MS5607 measured comparable temperatures to the external thermistor during the day. During the night, it measured temperatures about 5°C higher. The onboard MS5607 and the MCU measured similar temperatures and also the highest of all sensors during the day, while outputting comparable data to the external thermistor during the night.
| sensor |
Tmin [°C] |
Tmax [°C] |
| MCU |
+44 |
+57 |
| Thermistor (onboard) |
+9 |
+18 |
| Thermistor (external) |
+27 |
+35 |
| MS5607 (onboard) |
+43 |
+57 |
| MS5607 (external) |
+27 |
+35 |
This table is a result of comparing an hour of temperature measurements from the sensors to the closest sounding data on a few days the relaunched TT7B2 tracker was in a relative proximity to a weather station around local noon. The data represent the difference between measurements by specific sensors and the sounding data in a form of an interval.
This chart shows the temperature variation as measured by the onboard thermistor throughout the whole relaunched TT7B2 flight. The data from the ambient light sensor show several periods where the tracker at least partially operated during night-time.
In the cases of the TT7B2's first flight and the relaunched TT7B1, the external sensors were sealed inside the balloon envelopes measuring temperature of the hydrogen. While the balloons were floating, the external MS5607 measured hydrogen temperatures within 5°C of the external thermistor measurements.
One of the main goals of the TT7B flights was to obtain measurements of the hydrogen pressure and temperature inside the balloons with respect to ambient air pressure and temperature, the superpressure and supertemperature quantities. Unfortunately, due to my inability to seal the sensors inside the envelope airtight, only partial data were collected on the first flight of TT7B2 and on the relaunched TT7B1 flight. Despite the two envelopes being of different type, the measured pressure inside peaked at between 1.8 and 1.9kPa in both cases suggesting that that was the limit of the sealing method. The peak in the balloon's initial altitude came about 4 minutes later in case of TT7B2 and about 2 minutes later in case of TT7B1. Since then the pressure gradually decreased until the balloons lost lift and started to descend.
As for the temperature of the hydrogen, the initial expansion of the gas due to decreasing ambient air pressure steadily decreased the gas temperature until the envelope was filled completely. At that point, the hydrogen temperature stopped decreasing, while the ambient air temperature continued decreasing until the balloon ascended to its float altitude. The TT7B2 flight provided some insight in to the day to night variation in the hydrogen temperature and recorder a decrease of about 2.5°C. However, since the gas was continually escaping from the envelope, the difference between initial amount of hydrogen and amount not able to provide sufficient buoyancy to sustain float was about 5%, I am not sure to what extent this gas loss affected the temperature of the remaining gas as it had more space to expand into.
| designation |
mg [g] |
pa [Pa] |
Ta [°C] |
Vb [m3] |
st [°C] |
sp [Pa] |
| TT7B2 Flight |
18.95 |
16,176 |
-67.9 |
0.952 |
13.5 |
1,776 |
| TT7B1 Relaunch |
10.61 |
24,446 |
-52.6 |
0.376 |
11.0 |
2,496 |
| TT7B2 Relaunch |
19.17 |
16,047 |
-63.3 |
0.985 |
13.5 |
1,892 |
This table contains the masses of gas $m_{g}$ with which the individual balloons were filled. Then sounding data from the closest weather stations at the balloon's float altitude $p_{a}$ and $T_{a}$. This data is used to calculate density of air $\rho_{a}$ at the float altitude which equals the density of the balloon/tracker/gas system $\rho_{s}$. By utilizing the known balloon, payload and gas masses, the inner volume of the balloon $V_{b}$ is calculated. The measured gas temperature $T_{g}=T_{a}+st$ is used to calculate the expected superpressure $sp$ exerted by the amount of gas $m_{g}$ on the envelope of volume $V_{b}$ at ambient air pressure $p_{a}$. $$\rho_{a}=\frac{p_{a}}{R_{a}\cdot T_{a}}\;\;\;\;\;\; V_{b}=\frac{m_{b}+m_{p}+m_{g}}{\rho_{s}}\;\;\;\;\;\; sp=\frac{m_{g}\cdot R_{g}\cdot T_{g}}{V_{b}}-p_{a}$$ where the specific gas constant for air $R_{a}$ equals 287.058J/kgK, the specific gas constant for hydrogen $R_{g}$ equals 4142J/kgK, and the temperatures $T_{a}$ and $T_{g}$ are in Kelvin. The envelope volumes $V_{b}$ in the table differ from the earlier measured values, because they represent the actual volume into which the lifting gas could expand. The difference being due to a portion of the volume being occupied by expanded air that remained in the envelope and due to measurement error. When compared to the measured superpressure values, the pressure inside the larger TT7B2 envelope peaked at around the calculated expected value, while inside the smaller TT7B1 envelope, the pressure peaked significantly lower than the theoretical value suggested it would. In case of the relaunched TT7B2, the superpressure figure is only an estimate based on the gas temperature of the previous TT7B2 flight.
In this chart, data from the onboard thermistor and pressure sensor on the relaunched TT7B2 were used to calculate air density and then envelope volume over the course of the whole 28 day flight. Unfortunately, since the thermistor data were noticeably affected by absorbed solar radiation, the calculated volumes suffer from it as well, and can't be considered accurate. Nevertheless, the absence of any long-term trend in the data suggests the envelope didn't undergo any additional expansion due to creep. At least not any noticeable.
This chart shows the voltage of a AA battery (Energizer Ultimate Lithium) as measured over the course of its 28 day lifetime on the relaunched TT7B2 tracker. The voltage was noticeably influenced by temperature of the battery in day-night cycles. The peaks and troughs of the cycle dropped only marginally over the course of the 28 day period. As there was no clear slump on the last day of activity, it is unknown whether the battery died, or a different unrelated issue led to the end of the tracker's transmissions. Although the last day was the day with the lowest recorded temperature peak during the day, and even sounding data suggest temperatures around -70°C in that area.
Conclusions.
- Unfortunately, the trackers didn't manage to collect superpressure and supertemperature data from long duration flights. Only several hours worth of data on two leaking envelopes.
- On the other hand, the relaunched TT7B2 tracker managed to circumnavigate the Northern Hemisphere and lasted at least 28 days in the air proving the custom made balloons are in principle capable of long duration flights.
- Out of five launched balloons, though, only three managed to float, and only one survived for more than a day. Not a good success rate.
- Despite the relaunched TT7B2's perfect positioning in terms of available receivers on its last day, the cause of its failure remained unknown.
- Unfortunately, light weight trackers such as TT7B are not able to operate reliably during the night at such high float altitudes even when powered by lithium based prime batteries. The temperatures simply stay too low.
- Gluing the PE surface of the gas inlet with Loctite All Plastics didn't manage to create reliable seam that would withstand pressure from the lifting gas at float altitude temperatures.
Data for download.
