Partnership with Rutgers on Project PICKET

Partnership letter for Project PICKET (Parallel Independent Chain, Ka-band Enabled Transceiver).

Today's fielded RF radios generally sit on one band and retune in place when conditions change. Retuning in place means the link goes quiet while the hardware settles, and a quiet link at the wrong moment is the same as no link at all.

Project PICKET takes a different approach. The radio runs several independent chains at once across three bands. One chain carries the traffic while a second chain, with its transmitter muted, tunes to the next candidate band and locks onto it. Only when the second chain is ready does the traffic move across. The settling time still happens, but it happens quietly in the background, so the connection never actually drops.

The second idea behind PICKET is about timing. Conditions may change faster than a radio can measure them, so a radio that waits to detect a problem is answering a question that has already gone stale. This can be approached algorithmically so any idle chain then spends a short moment confirming the best candidate channel rather than searching for it.

A satellite terminal losing signal in a rainstorm losing signal in plasma are, from the radio's point of view, the same request. Both need a way to change bands without losing the conversation. This is why I have partnered with Rutgers' COSMOS lab to find a way of fixing this.

Partnership with Worcester Polytechnic on Project CHIRP

Partnership letter for Project CHIRP (Cognitive Hardware Identification of Radar Pulses).

Most radar sends out a short, powerful pulse and listens for the echo. That is easy to detect, because anyone with a receiver can hear the shout.

A different class of radar was built to avoid exactly that. Instead of shouting, it whispers across a very wide range of frequencies, almost continuously. Any single sliver of the spectrum carries so little energy that it blends into the background hiss that every receiver already lives with. The radar still works, because it knows the pattern it sent and can reassemble the echo. Anyone listening in does not know the pattern, so there is nothing obvious to grab onto. Engineers call these low probability of intercept waveforms.

These signals "chirp" much like a ball rolling around on a stretched surface into a hole, or if you have ever played with spherical magnets you will notice the "chirping" sound they make as they come together. This particular kind of signal is also one that is found when objects rotate and fall into a black hole or as black holes rapidly evaporate as the frequency emission from the object rapidly increases.

CHIRP signals from LIGO.

Finding these signals is possible and using them as a communications channel is the aim of Project CHIRP. You have to listen for a long time and look for structure rather than volume, the way you might pick out a repeating rhythm inside a noisy room. Researchers have gotten good at this. Published work routinely reports accuracy above ninety percent using machine learning on visual representations of the signal.

Here is the part that gets less attention. Almost all of those results come from a workstation running unhurried calculations on stored data. The situation that actually matters is a small circuit board flying on an aircraft at more than twice the speed of sound, with a fraction of a second to identify what is illuminating it and choose a response.

Partnership with Northwestern University on Project GLACIER

Partnership letter for Project GLACIER (Gigabit Link, Athermal Cryogenic Integrated Electro-optic Readout).

Infrared cameras used for space missions work best when they are cold. Very cold. The sensor sits inside a vacuum flask, chilled by a cryocooler, because heat is the enemy of a clean infrared image.

That creates an awkward problem. The sensor produces an enormous amount of data, and all of it has to travel out of the cold box to the electronics outside. Today that data travels over copper wire, and copper wire costs power twice over. It costs power in the circuit that pushes the signal down the wire, and it costs power again in the cryocooler, which now has to remove both the heat from that circuit and the heat leaking in along the wire itself. On a battery powered or satellite mounted system, that is time on station you do not get back.

For a large modern sensor, moving the data can consume as much power as the cooling budget can spare. The bottleneck is no longer the detector. It is the wire.

Project GLACIER is Dreamscape Systems' answer to that problem. The idea is to send the data out on light instead of electricity. Optical fiber carries far more data than copper and conducts almost no heat, so both halves of the power penalty shrink at once. This also fits into previous work we have been doing with Project BLUEBIRD and the GATE Program to read information from optical/spin systems.

An earlier draft proposed to UCLA as a possible partner.

An isometric cross-section view.

Sending light out of a cold box is not itself new. What has been missing is doing it in a way that actually saves power. Earlier attempts bolted an optical chip next to a standard readout chip and connected the two with conventional solder joints. That removed the copper, which helped with heat, but the circuit driving the signal was still built for copper. It still burned the same energy. The saving evaporated before it reached the optical device.

GLACIER takes the approach of designing the driver and the optical device together, so the driver only has to produce the small voltage the optical device needs, and the connection between them is short enough that it does not put the energy back. This first phase is a design and analysis effort. It produces the optical circuit design, a specification of the changes required to an existing readout chip, and a plan for building and testing the result.

Cold optical links matter beyond infrared imaging. Quantum computers face the same problem of getting data out of a refrigerator without heating it, and data centers are already moving to optical connections for similar reasons. Work in one area tends to help the others.