Fast steering optical phased array: how to trick a control system

Optical phased arrays are a way to construct a really powerful laser beam, and steer it precisely without mirrors or lenses. Making an optical phased array that can steer millions of times per second would be a huge boon to systems like LIDAR for autonomous cars. But to make something move that fast, you need to trick the control system.

Since their invention in the 1960s, laser have revolutionized technologies from precision measurement to telecommunications. Ways to reliably and rapidly point laser beams (e.g. by deflecting the beam with a mirror or lenses) are needed for applications such as LIDAR for autonomous cars, laser communications links between aircraft or satellites, and directed energy (i.e. Star Wars style laser cannons). You could mount your laser to a gimbal and move the whole thing, or use a mirror or lenses to steer the laser beam around. The problem with all of these systems is that they have inertia, you have to physically move an object. This means that the rate at which you can steer your laser beam (the speed with which you can point it from one direction to another) is limited by how quickly your motor or other actuator can accelerate and decelerate the mass of the mirror/lens it has to move. This means mechanical steering systems are limited to steering the beam only a few thousand times a second. This sounds like a lot, but it means something like a LIDAR scanner for an autonomous car, which might have to scan back and forth hundreds or even thousands of times per second, can only build up a few full images every second. Not fast enough to let vehicles safely travel at 100 km/h.

Mechanical systems also wear out, which is not good when it’s a key safety system in your car, or when that system is on board a satellite where it can’t be fixed or replaced.

Optical Phase Arrays (OPAs) are a solid state (no moving parts) way of steering a laser beam. OPAs comprise several laser beams which are combined into single beam. (Which also means they can be used to construct a very powerful laser from several less powerful lasers.) By controlling the phase of each laser beam in the OPA (that is, emitting the light from different sub-lasers at different times) it is possible to control the combined laser beam without mirrors or lenses. For example, in an array of lasers making up an OPA, if the phases of lasers on the left-hand side of the array are held back, and the phases of the lasers on the right-hand side are advanced, then the combined laser beam will steer to the left. (Insert your own set phasers to stun joke here.)

With no moving parts, an OPA will not wear out like a mechanical system, and can also steer its beam much much faster.

Another benefit is that an OPA like this can rapidly point to random areas in its field of view, instead of having to scan the area line by line like a television building up its picture. It sounds a bit counter-intuitive, but randomly scanning your LIDAR over an area builds up a useful picture faster than TV-style raster scanning does, which is of critical importance for autonomous car sensors.

The problem with OPAs is that the phase of each laser in the array has to be controlled to extremely high precision (equivalent to each laser being aligned within one ten-billionth of a metre of all of the others). There are several ways you could do this, but we used a control system called a phase-locked loop (more precisely, a lock-in amplifier).

A phase-locked loop (PLL) takes an incoming signal (the phase of a chosen laser in the array) and compares it to a reference phase. The result of this comparison is an error signal that is fed to a ‘controller’ within the PLL which, in our case, both updates the phase reference and adjusts the phase of the laser to keep it at the desired offset from its neighbours.

To steer the OPA, we change the phase of the phase reference, which makes the controller change the phase of the laser. We have to do this individually for each laser in the array.

Unfortunately, the PLL has quite a lot of delay in its electronics, which means this method can’t be used to steer the OPA more than a few thousand times per second.

 

By injecting phase control signals after the controller, directly into the laser, we can bypass the PLL and steer the laser as fast as we want, millions, or even billions, of times per second.

But now the problem is that the PLL and its controller will detect these changes in phase of the laser and try to prevent them from happening, thinking that they are due to mechanical vibration in the laser array causing misalignment of the lasers. This doesn’t matter much when these fast steering signals are much faster than the PLL has the ability to sense, but there is a range of steering speeds that are too fast for the PLL to apply itself, but slow enough that the PLL will sense them and try to counteract them.

The overcome this, we can trick the PLL.

By injecting the steering signals in the right combination, we can blind the PLL to the steering signals applied after the controller. To do this, we apply the desired steering signal directly into the laser after the PLL controller, and the inverse of that steering signal into the phase reference. This causes the two steering signals (the original applied to the lasers and the inverse) to cancel out in the phase comparator, meaning the PLL controller doesn’t see any steering signal, but will still be able to see any errors caused by misalignment of the phases of the lasers in the array.

By using this trick, our OPA is able to precisely steer the laser beam at speeds ranging from extremely slowly, to millions of times a second, while still detecting and correcting any phase errors caused by mechanical problems such as vibration of whatever the OPA is mounted on. Correcting these unwanted phase differences is critical to making the laser combine into a single beam efficiently.

The video below shows the beam of the OPA scanning slowly to the right before ‘snapping’ back to the left. Snapping back in this way is only possible thanks to the super fast steering ability of this OPA and its control system.

The reason there are several spots in the video, and fainter spots on the edge that fade in and out as the beam moves, is because our OPA only had 7 lasers and a mediocre fill factor. The more lasers in the OPA, and the better the fill factor, the better they combine into a single central beam.

This OPA design and steering technique was implemented digitally on a field-programmable gate array (FPGA). More information can be found in the published paper here.