CubeSat Gamma-Ray Burst Timing Payload — Rev A

The intended payload would detect short and long gamma-ray bursts across a target range of roughly 50 keV to 1 MeV, timestamp every accepted pulse against UTC, estimate each photon's energy from its pulse height, and build a light curve showing how the burst brightened and faded. It would recognise a significant rise above its own background, alert the spacecraft, and preserve the record for downlink. On the ground, those events would be compared against Fermi, Swift, GECAM, gravitational-wave observatories and other CubeSats.

Rev A is a schematic and a review. It demonstrates none of this.

One gamma ray produces one flash in the scintillator, one small pulse out of the SiPM, and from that single pulse the instrument has to extract two independent quantities.

The first is when. The pulse goes to P1, a fast comparator, which flips the moment the signal crosses a threshold. That digital edge is what gets timestamped, and it has to be timestamped in hardware rather than by software noticing it afterwards — which is why the review requires the comparator output to reach an STM32 timer-capture input, a connection Rev A does not make.

The second is how big. Amplitude is the estimate of photon energy, and U3 is meant to digitise it to at least 12 bits. But a SiPM pulse is far too brief for a converter to catch at its peak, so something has to hold that peak still. That is the missing shaping or peak-hold stage. Both paths are incomplete in Rev A.

On most instruments a clock is housekeeping. Here it is part of the measurement, because the science question is comparison: does this burst match what another spacecraft saw, and what do the arrival-time differences say about where it came from?

The requirement is absolute event time within 10 microseconds of UTC, and the review breaks it into pieces that each have to be earned. GNSS 1-PPS disciplines the event timer. Electronic timing variation through amplifier, comparator and firmware is allowed under 1 microsecond, which is why the trigger edge has to land on a hardware capture input. Every fixed delay in that path has to be measured and corrected, because a delay you have not measured is an error you cannot remove.

Then the failure cases: firmware noticing that GNSS synchronisation is missing and flagging those records as degraded, and a local clock stable enough to hold time between 1-PPS edges.

Rev A draws two GNSS modules and two crystals. It does not define which does what, connect the comparator to a capture input, or calibrate a single delay.

The support electronics are drawn. The review was about the distance between that and an instrument whose measurements would mean anything.

The detector head is the largest gap and has its own section below. After it comes the analog path: the shaping or peak-hold stage is missing, the comparator does not reach a hardware capture input, and U2, P1 and U3 have no part numbers, so the amplifier noise, comparator speed and converter accuracy the requirements depend on are all open.

The bias supply needs closed-loop feedback, current limiting, voltage measurement and controlled discharge — a SiPM's gain follows its bias closely, so an uncontrolled bias is an uncontrolled energy scale. Gain follows temperature too, which is why U6 has to feed compensation rather than only telemetry.

Then the duplicated parts. Two GNSS modules and two flash devices are on the sheet, and the review's position is that their roles need defining; I am not going to call them a finished redundancy scheme, because the drawing does not say that. Both antenna symbols read 6 GHz, which is not a GNSS band.

Everything on the Rev A sheet processes a pulse. Nothing on it produces one.

The detector head is a scintillator crystal — CsI(Tl), CeBr3 or similar — optically coupled to a SiPM, wrapped reflectively, in a light-tight enclosure with its own temperature sensor. The crystal converts gamma-ray energy into visible light; the SiPM converts that light into the pulse the rest of the board is built around. Neither is drawn, and there is no connector for the SiPM's signal, bias, temperature and ground.

This is not a detail to tidy up later. The crystal sets the energy range and the resolution, the optical coupling has to survive launch without changing the amplitude it delivers, and the enclosure has to keep light out, because a SiPM cannot tell a stray photon from a scintillation flash.

The honest description of Rev A is a payload electronics revision waiting for its instrument.

I drew a board that processes pulses and called it a payload. The review made the distinction clear: the instrument is the crystal, the SiPM and the light-tight enclosure, and what I had drawn is the electronics that serve them.

The second lesson is that a measurement has a shape in time. A SiPM pulse is too short to be measured directly, so the shaping stage I left out is not a refinement — without it the energy number has no defensible meaning. The trigger edge is the same: if software finds out about it late, the timestamp carries that lateness.

The third is about the network. One satellite times a burst; two constrain a direction and three or more do it well. Ten microseconds only matters because somebody else recorded the same photons — and the burst is the one thing you cannot ask to happen again.