A Rail-Mounted Pumping System Developed for Suborbital Rockets

1. Introduction

Suborbital rockets frequently use cooled focal-plane cameras to make their observations during flight, requiring the camera to be operated in a vacuum to prevent forming ice water on the detector surface. Large roughing and turbo pumps are attached to the side of the payloads while the rocket is in the horizontal position, ensuring a high-quality vacuum is achieved in the instrument. Just prior to when the rocket motors are armed for launch, known as category A (Cat A) arming, these pumps must be removed and all current to the rocket must be switched off. Post Cat A arming, no personnel are allowed to return to the launcher. This procedure has been used on several rocket payloads that feature pressure-sensitive conditions (e.g. Rogers et al., 2015; Miles et al., 2019a), and similar pumping solutions have also been used in DEUCE (Erickson et al., 2021) and SISTINE (Behr et al., 2023). In a perfectly designed instrument, the vacuum chamber holding the camera would be virtually leak free, allowing the camera to be initially evacuated to a suitable pressure before cooling the detectors after the pumps have been removed. In practice, it is more likely that the vacuum chamber will have leaked up to a pressure that is higher than the minimum pressure required for cooling. Missions must either be able to tolerate some ice build-up on their detectors, or have a method to further evacuate the chamber before cooling. For payloads that are known to have problems with leak-up rates in their chambers, the time taken for Cat A arming can be a critical factor. If the chamber is allowed to leak up too far, the systems in place to reduce the pressures in the chamber post-arming may not be effective. On WRXR, an ion pump was used to re-establish the detector chamber vacuum post arming, but this ran into problems due to the detector chamber leaking up to pressures beyond the range of the ion pump during arming (Wages et al., 2019; Miles et al., 2019a).

To counter these problems, a rail-mounted pumping system was devised for the Rockets for Extended-source X-ray Spectroscopy (tREXS) (Miles et al., 2019b) that was designed to evacuate the detector chamber from atmospheric pressure to 10${}^{-5}$ Torr through remote operation. This removed the complications associated with power being off during Cat A arming, as any amount of leak-up could be reversed. The pumping system was based around a roughing pump [Pfeiffer ACP 40 Multi-Stage Root Pump (Pfeiffer Vacuum, 2022)] that was attached to the rocket using a 100 ft foreline and a magnetic umbilical. This foreline was attached to the exhaust of an internal turbo pump that flew on the payload [Agilent TwisTorr 74S (Agilent Technologies, 2021)]. This system was also used to monitor the pressures and temperatures in the instrument and manage the detector cooling rate. The focal-plane camera on tREXS must be cooled to suppress the silicon’s dark current, with a target cooling temperature of $-80^{\circ }$C (Tutt et al., 2021).

This paper will detail the rail-mounted pumping system that was designed for tREXS, including the different stages of operation that prepared the instrument on the rail, minimized leak-up during arming, and optimized the camera environment before launch. Section 2 describes the rail-mounted pumping system in detail. Section 3 discusses the umbilicals used in this system, including a unique pull-away magnetic umbilical. Section 4 talks through the procedure for the launch of the payload and Sec. 5 discusses the challenges that arose from operating this system in the field. Finally, the planned improvements to the system for future launches will be detailed in Sec. 6. Though this system was designed for the launch operations of tREXS, the lessons learned from the launch operations, and the planned improvements for future launches make this system adaptable to other suborbital rocket payloads.

2. The tREXS Rail-Mounted Pumping System

To control the rail-mounted pumping system, an Ethernet-controllable relay board was used with solenoid valves and a gate valve to make sure the vacuum system pressure gradient was always away from the detector chamber. The eight analog channels were used to monitor pressures in the vacuum system, the gas pressure at the gate valve, the current draw of the roughing pump, and the temperature of the camera, which was used to control a negative feedback loop for the cooling of the focal-plane. The whole system was controlled with a MATLAB GUI using fiber optic connections to the Ethernet interface from the blockhouse. All hardware used for the rail-mounted pumping system is listed in Table 1.

The rail-mounted pumping system was split into five different regions (Fig. 1):

2.1. Initial pumpdown

The MATLAB GUI^(a) (Fig. 3) was used to control almost every aspect of the rail-mounted pumping system and so is an excellent conduit to explain how the system as a whole works. Figures 4 and 5 show a simplified layout of the rail-mounted pumping system.

In the initial operation of the system, the detector chamber was at atmospheric pressure ($\approx 750$ Torr), and the internal door between the telescope section and the detector chamber was closed. The solenoids and gate valves were also closed (this means there was no pneumatic pressure reaching the gate valve) and the rail-mounted pumping system was not powered. The rail-mounted pumping system used three dedicated 110 VAC power lines that were switched on at the blockhouse and they powered the three AC–DC power supplies and turbo pump controller (Fig. 2(a)), the vacuum sentry, and the roughing pump. The three AC–DC power supplies served different parts of the rail-mounted pumping system. The first, a 12 VDC power supply, provided power to the relay board (Fig. 2(b)). The second, a 24 VDC power supply, was the input voltage for the relays on the relay board. Finally, the third, a 24 VDC power supply, powered non-switching hardware such as the temperature monitor and the vacuum gauges.

Vacuum gauge information was returned to the MATLAB GUI, allowing the pressure of the foreline, the turbo pump, and the detector chamber to be monitored. The GUI was used to switch on the roughing pump and a current monitor (NCD Veris Hawkeye H822-20 Solid Core Current Sensor 0–20 Amp AC to 0–5 VDC) measured the current draw of the pump and returned the valve to the GUI. The reading on the foreline vacuum gauge did not start to drop at this point as it was placed on the upstream side of the foreline solenoid (Fig. 4) so that the vacuum could be maintained at the magnetic umbilical interface. The gradual drop in current draw from the roughing pump showed that the foreline was being evacuated. Once the current stabilized, the foreline solenoid was opened. As soon as the foreline vacuum gauge reading started to fall, the turbo solenoid was opened. Then, when the turbo and the foreline vacuum gauge readings started to fall, the gate valve was opened. The gate valve is pneumatic, so the gaseous nitrogen (GN2) solenoid needed to be open. A pressure of at least 65 psi is required to actuate the gate valve, and a reading of the pressure from a transducer on the gas line could be read in the MATLAB GUI. If the pressure was above 85 psi, a pressure release valve opened in the electronics section to start a nitrogen purge (this was to minimize ice build-up during cooling (Fig. 5)). This pressure was selected so the gate valve could be operated without running the purge for lab testing, but the pressure used during launch operations of 99 psi (a high-pressure threshold for NASA safety operations) would cause the purge to run during launch operations. As the detector chamber was at atmospheric pressure initially, it was important not to create a large pressure gradient between the turbo side and the detector side of the gate valve as the acoustic shock of opening the gate valve could damage the optical blocking filters in front of the focal-plane. It is a good practice to maintain a small pressure gradient away from the detector chamber to avoid back-streaming dirty air into the chamber; therefore, the timing of opening the gate valve is important to protect the filter while ensuring a small pressure gradient exists. During the operation of tREXS, the gate valve was opened seconds after the pressure at the turbo vacuum gauge dropped below the pressure in the vacuum chamber.

With all gate valves and solenoids opened, the pressure on each of the vacuum gauges (foreline, turbo, detector chamber) started dropping. Once the pressure in the detector chamber reached cross-over pressure (10${}^{-2}$ Torr) the turbo pump was started using the A-Plus software, the standard Agilent pump control software. The turbo spun up and the pressure measurement from the detector and turbo vacuum gauges fell rapidly. Once the detector chamber was at 10${}^{-5}$ Torr the detector cooling could begin; however, on launch night the detectors were not cooled until post-arming to make sure the vacuum level was low enough to prevent detector icing during rocket motor arming.

2.2. During arming

No current can flow during arming; therefore, the rail-mounted pumping system had to be shut down. Evacuating a 100-ft foreline (between the roughing pump and the magnetic umbilical) took $\approx 10$min, so to make the post-arming process more efficient a vacuum sentry was used. When un-powered, a vacuum sentry is a valve that closes when air flows in the wrong direction (in our case, through the vacuum pump and into the detector chamber). By turning off the vacuum sentry before the roughing pump, we were able to keep a 10${}^{-2}$ Torr vacuum in the foreline. The vacuum sentry was needed as the vacuum pump was modified to let atmosphere into the foreline when it was powered down so we can more easily detach the magnetic umbilical before launch.

To prepare the rail-mounted pumping system for arming, the vacuum sentry was powered down using one of the 110 VAC breakers in the blockhouse. The gate valve was then closed, followed by the turbo and foreline solenoids. The turbo pump and roughing pumps were then powered off and the GN2 solenoid was closed. The rail-mounted pumping system was then powered down by turning off the relay board and roughing pump 110 VAC breakers. The motors could now be armed. During the arming process, the detector chamber was sealed and at vacuum. The foreline solenoid caused a pocket of vacuum to be trapped across the magnetic umbilical interface, which helped the magnets maintain the vacuum connection during arming.

2.3. Post arming

Post arming, it was important to make sure the roughing pump was powered before the vacuum sentinel to maintain the vacuum in the foreline. Using the 110 VAC breakers, power was restored to the relay board and the roughing pump. Once the current draw in the pump had stabilized the 110 VAC breaker to the vacuum solenoid was switched on.

The detector chamber leaked up to a higher pressure than 10${}^{-5}$ Torr during arming, but was still lower than the crossover pressure used to turn on the turbo pump (10${}^{-2}$ Torr). With the whole vacuum system at cross-over pressure, the acoustic load in the system was minimized, protecting the filters. However, it is good practice to open the gate valve when the turbo and detector vacuum gauges are reading approximately the same values. Based on the pressure in the detector chamber, it was necessary to start the turbo to bring the turbo and detector vacuum gauges into the same range as the detector chamber before opening the gate valve. Once all gauges were equal, pneumatic pressure was restored to the gate valve by opening the GN2 solenoid, and the gate valve was opened.

Once the detector pressure was at 10${}^{-5}$ Torr, the cooling was able to begin without risk to the detectors. The cooling system was designed so that the LN2 solenoid opened and closed with a cadence that caused the focal-plane detectors to cool at $\approx 3$K/min. A temperature sensor gave information to one of the analog channels on the relay board, so the temperature of the focal-plane could be monitored (Fig. 5). Once the temperature reached the target temperature ($-80^{\circ }$C), the cooling loop transitioned into a steady state mode. If the temperature reached 10^∘C below the target temperature, the LN2 solenoids closed and if the temperature reached 10^∘C above the target temperature, the LN2 solenoid opened. It took $\approx 90$min to cool the focal-plane, at which point the rocket was ready to launch. The cooling system used on tREXS was an open loop system where the LN2 flows through an aluminum block and then was exhausted from scuppers on the rocket skin. The aluminum block was thermally strapped to the focal-plane using copper plates, allowing the detectors to be cooled (Fig. 6). The scupper exhausts were symmetrical placed on opposing sides of the transition skin so that any force from the exhaust is canceled out (Fig. 7) without impacting the attitude-control system (ACS).

2.4. Pre-launch

Prior to the rocket launch, several components of the pumping system had to be transferred to a safe-for-launch state. The magnetic umbilical needed to be detached and the foreline was vented to atmospheric pressure by closing the turbo solenoid and stopping the roughing pump. This vented atmosphere into the foreline as the vacuum sentry was still powered and open (it took 20 s to vent the foreline). The next step was to detach the magnets; however, once the magnets were released they could not be re-attached without bringing the rocket back to a horizontal position, hence, the magnets were released as late as possible. Powering the magnets caused them to be de-magnetized and they were pulled-away from the rocket using bungees. A spring loaded slammer door then closed over the hole left by the magnets to improve the aerodynamics of the payload during flight, Fig. 8.

With the magnets detached, the rest of the rail-mounted pumping system can be shut down. All internal solenoids will close on launch due to the disconnection of the electrical umbilical, but the gate valve is left open as late as possible so that the turbo can continue to pump on the chamber. The GN2 umbilical will only release if it is at atmospheric pressure, so this gas line is also vented just before launch.

The requirement for the pressure in the chamber for tREXS is $<10^{-5}$ Torr; however, the system was able to achieve a lower pressure of $5\times 10^{-6}$ Torr during lab testing. This pressure could have been lower, but due to a leak in our detector chamber, described in Sec. 5.3, our lowest possible pressure was limited.

3. Umbilicals

To make the rail-mounted pumping system possible, the rocket payload required four umbilicals: magnetic, GN2, LN2 and electrical.

3.1. Magnetic umbilical

Roughing pumps are too heavy and too large to fly on a sounding rocket; therefore, the decision to use a turbo pump to generate the low vacuum required when cooling the detector required a foreline that could be remotely detached from the rocket skin. Pull-away foreline umbilicals have been used on previous sounding rockets, in particular on FORTIS (McCandliss et al., 2010; Fleming et al., 2011, 2013) and EVE (Wieman et al., 2016), but details of these systems are not well documented. A system is also under development at the NASA Wallops Flight Facility to develop a launch vehicle vacuum system based around a cam and groove clamp on the side of the rocket. As no clear consensus has been developed on the best way to detach a foreline from a suborbital rocket skin before launch, the tREXS system was developed using magnets.

The umbilical designed for tREXS was based around four electromagnets (Eclipse M52178/24VDC). These magnets were chosen for this application as they are magnetic when powered off and demagnetized when power is applied, which is ideal for sounding rockets as all power has to be off during arming. The Energize-to-release electromagnets are based on an electro-permanent concept where a permanent magnet is surrounded by an electromagnet. When un-energized, the magnet acts as a permanent magnet, but when a potential is applied to the electromagnet, the permanent magnet and electromagnet cancel each other out, causing the overall magnetism to be zero.

Each magnet has a magnetic pull of 50kg and four are used to attach the roughing foreline to the rocket skin. A special port on the skin has been made to house four armature plates to which the magnets are attached. The foreline connection goes through a bellows between the magnets and then seals to the rocket skin via a centering ring. This ring is fixed onto the rocket skin using Scotch-Weld 2216 (Ellsworth-Adhesives, 2022) so it releases in a controlled way at launch. Internal to the rocket, a foreline runs from the skin to the exhaust of the turbo pump (Fig. 7).

3.2. GN2 umbilical

The GN2 umbilical is a standard NASA Sounding Rocket Operations Contract (NSROC) high pressure umbilical that is used for boost guidance (S-19), and the ACS. Gas pressure was provided to the rocket at 99 psi using LN2 boil-off and was used for a nitrogen purge in the electronics section (to minimize ice build-up during cooling) and to provide pneumatic pressure for the gate valve. This umbilical had to be at atmospheric pressure during launch or it would not release properly. To vent this line, a 110 AC solenoid was connected to the gas line, and it was operated just before launch using a breaker in the blockhouse. Due to the loss of gas pressure at launch, the gate valve had to be closed just prior to launch.

3.3. LN2 umbilical

The LN2 umbilical consists of a PTFE hose that was fitted onto a barb on the rocket skin. This slip-fit released when the rocket launched. The dewar, which was mounted on the rail, has a remotely operated cryo-solenoid to isolate the dewar from the rocket (Fig. 1). A vacuum jacketed line was used to transfer LN2 from the dewar to the PTFE hose to minimize boil-off. The dewar provided 22 psi LN2 into our camera cooling system, allowing the detectors to be cooled to their operating temperature.

3.4. Electrical umbilical

The electrical umbilical is a standard NSROC umbilical. A 37-socket, D-sub connector is mounted internally to the transition skin of the payload electronics section. Harnessing from the launch rail is then mated to this connector in the form of an electrical umbilical that releases at launch. This umbilical allowed control of the hardware internal to the rocket skin up until launch. This included cycling the LN2 solenoid, control of the gate valve and turbo solenoid, power and analog output from vacuum gauges, and power to the turbo pump. Once the rocket launched, all power would be lost to the internal systems of the rail-mounted pumping system, with all hardware defaulting to a closed state to protect the detector chamber.

4. Launch Operations

For the launch of tREXS in September 2022, the payload was mounted onto the launch rail 5 days before launch. The payload is mounted horizontally, as shown in Fig. 1, and the telescope section pumps were attached to the outside of the external skins. The payload was kept at a stable temperature of 20^∘C using a Styrofoam box and air conditioning unit. The external and internal pumps were left running throughout the 5 days before launch, apart from when vertical tests on the launcher were performed and the external pumps had to be removed.

Three hours before launch and just prior to Cat A arming, the external pumps on the telescope section of the payload were removed. The detector chamber was isolated from the telescope section by a door, and the rail-mounted pumping system was used to shut down with the gate valve being closed first to maintain the vacuum in the detector chamber. During arming, which took $\approx 45$min, all power was off.

Post arming, the vacuum in the chamber was re-established as described in Sec. 2.3, and the detectors were cooled. Cooling took $\approx 2$h, at which point the rocket was ready to launch. The steps that were completed prior to launch are as follows:

The connection of the exhaust of the turbo pump to the roughing pump is removed at T-120 s, but lab testing has shown that the turbo exhaust pressure took more than 15min to build to a level that impacted the pump’s operation. Due to this, the turbo was left powered and spinning, pumping on the chamber to help maintain the vacuum until the last possible moment (T-15 s when the gate valve was closed). At T-15 s the turbo was switched off, but there was no plan to break the blades with a nitrogen backfill — the blades would still be spinning at launch. Catastrophic damage to the turbo pump during launch was an accepted risk, but the pump did survive vibration testing, launch and recovery. However, due to the failure of the rail system described in Sec. 5.1, the turbo was not spinning at launch as the failure acted to break the blades. This will have aided the survival of the pump.

The foreline took $\approx 20$s to vent to atmospheric pressure, at which point the magnetic umbilical was no longer being vacuum attached to the rocket. Just before launch (T-15 s), the gate valve was closed to protect the detectors from the turbo solenoid if the turbo blades were damaged during launch. The cooling loop stopped at T-0 s as the payload was no longer connected to the LN2 umbilical and the power had been severed to the LN2 cryogenic solenoid.

5. Implementation Challenges

The rail-mounted pumping system was used during flight operations and was able to get the tREXS instrument in the required state for launch; however, some problems were encountered during the first field testing of the system.

5.1. Relay board operation

The cooling system caused problems with our relay board operation. A second cooling MATLAB script had to be initialized within the main rail-mounted pumping cooling GUI to allow for controlled cooling of the camera. The nested coding caused a regular communication issue between the relay board and MATLAB interface that closed the turbo solenoid every few minutes. This caused pressure to build at the exhaust of the turbo pump, causing the pump to overheat. Several attempts were made to correct this communication error and subsequent closing of the turbo solenoid, but with the launch date approaching, a work-around solution was developed. To counter the turbo solenoid closing during cooling, a command to re-open the turbo solenoid whenever it closed was added to the cooling control loop. Unfortunately, this led to a failure in the pre-launch shutdown sequence due to the turbo solenoid being re-opened when it should have remained closed. Post-flight data showed that there was a failure during this pre-launch procedure. The pressure in the detector chamber and telescope section increased 90 s before the rocket launched. Opening this solenoid vented the chamber to the atmosphere, iced the focal-plane camera, and the pressure differential on the detector door allowed pressure to enter the telescope section.

5.2. MATLAB

MATLAB is an effective way of controlling hardware and the app designer was an easy way to make a GUI. The problem with using MATLAB was due to the linear way in which the code operates, and the challenges in running sub-scripts within the main GUI code. Software interlocks would have added additional safety to the rail-mounted pumping system in case of solenoid or pump failure, but this would have required separate scripts to run concurrently with the main code. This had already been shown to be problematic with the cooling script; therefore, software interlocks were not able to be developed.

The goal was to have all control loops and interlocks integrated into the main GUI. However, it became clear that the additional scripts that adding these features required caused the GUI to slow to a level that made it unusable. To realize the ultimate goal of a fully integrated control system, a different control platform will have to be investigated that could take advantage of parallel processing.

5.3. Detector chamber leak

To be able to get the electronics for the 12 detectors that make up the focal-plane into the vacuum chamber, 2 Vacuum Transition Boards (VTBs) were sealed into a flange with an Arathane plug. This plug formed a seal that was good enough to reach the tREXS required pressure of 10${}^{-5}$ Torr, but limited the ultimate lowest possible pressure. The design of this plug can be seen in Fig. 9. If the chamber were being redesigned, the extent of the Arathane plug would be minimized by using flex cables as the transition hardware, not full PCBs. A leak-up test was performed in the lab to investigate the size of the leak by shutting the gate valve between the chamber and the detector chamber and measuring the pressure increase over time. This result can be seen in Fig. 10. This test was completed with the detector warm so as not to risk damaging the camera performance and the pressure was not as low at 0 s as we achieved on launch night.

6. Future Improvements

While MATLAB did an excellent job in the operation of the GUI, it was not able to run fast enough to make effective software interlocks based on vacuum gauge data, and it had problems running the cooling system at the same time as the main GUI. Programmable Logic Controllers (PLC) are designed for this type of operation; therefore, revision two will move to using a PLC to operate the system. It was decided to not use PLCs in the first revision as we were planning to use MATLAB to control the turbo pump; however, it proved easier to control the turbo pump using Agilent’s native software, so we are moving back towards using PLCs.

The PLC revision has many benefits over MATLAB. Automation Direct’s BRX series of PLCs provides dedicated hardware for analog and digital signal processing, relays, and RS-232 serial communications. Additionally, the C-MORE^(b) platform of human–machine interfaces (HMI) provides integrated graphics processing when paired with Do-More!^(c) PLCs. These attributes enable precise, repeatable control of rail-mounted pumping at minimal computational expense. The inexpensive nature of PLC computations allows ample room for the development of interlocks.

The interlock system prioritizes protecting the detector chamber and detectors from icing and acoustic shock in the event of failure. During automated operation, the PLC utilizes RS-232 serial communication with the turbo pump contained within the rocket skin. The pump returns temperature and state data. If the pump or analog signals return data outside expected operating conditions for the current stage, the PLC attempts to automatically correct the discrepancy by quickly cycling the turbo and foreline solenoids, opening and closing the gate valve, and terminating and reestablishing communications with the turbo pump. If the prior error persists, the system returns to a fail state, similar to the arming stage, and allows for manual override.

The harnessing will be re-designed to make it easier to rig on a rocket launcher. This will include the use of junction boxes so that individual pieces of hardware can be isolated to check that they are operating correctly.

The NCD relay board was limited to measuring 0–5V analog signals. Choosing PLC modules that can measure 0–10V analog signals will remove the need to use potential dividers to bring analog signals into the measurement range of the hardware. This will help improve the accuracy of the readings from vacuum gauges and pressure transducers.

Finally, the use of PTFE or Delrin centering rings and plastic clamps will further allow the isolation of the focal-plane camera and the rail-mounted pumping system. This could help reduce interference further; therefore, leading to improvement in detector noise performance.

Acknowledgments

The authors would like to thank the NASA Sounding Rockets Program Office and the management, engineers and technicians of the NASA Sounding Rocket Operations Contract for supporting this program and collaborating on interfaces and implementation. In particular, Fellissa Selby and Keith Foster were instrumental in getting the magnetic umbilical designed and tested. The tREXS project was supported by the NASA Astrophysics Research and Analysis Program under Grant Number 80NSSC18K0282.

ORCID

James H. Tutt  https://orcid.org/0000-0002-1613-0796

Vincent A. Smedile  https://orcid.org/0009-0008-8819-0481

Jessica Mondoskin  https://orcid.org/0009-0008-6867-9212

Garrett Brady  https://orcid.org/0009-0003-0256-1119

Ross McCurdy  https://orcid.org/0000-0002-4483-5363

Randall L. McEntaffer  https://orcid.org/0000-0002-3719-8212

Drew M. Miles  https://orcid.org/0000-0001-5982-0060

Notes