Tuesday, July 14, 2015

Lasers Keep Killing Image Sensors

The unexplained effect of laser causing a dead row or, less often, a dead column in image sensors keep appearing in more and more Youtube videos. Most videos show damage in larger sensor cameras, while smartphone image sensor damage is a rarity. The larger lens probably collects more laser energy. Or the higher quality lens might be able to focus light into a smaller spot. Still, in theory, the on-chip microlens should sway the light away from the metals and pixel transistors, so the damage should be limited to the PD and TG. This is local to the affected pixel and should not affect the full row or column.

So far, the damaged sensors appear to be all FSI. I was not able to find a single BSI sensor affected. This might be an unsung advantage of BSI tech.






Update: As somebody pointed in comments, there is a MSc thesis "Damage Mechanisms in Visible Band Cameras Subjected to Continuous Wave Laser Attack", July 2014 Sacha Casken, Cranfield University. One interesting observation of the thesis is that no damage is made by red lasers, explained by their larger divergence. It remains to be seen whether this explanation is correct or not.

13 comments:

  1. In FSI, circuitry layers are located on top of PD, in BSI, they are located undeneath.
    That's why FSI has more chances to get damaged, I guess.

    ReplyDelete
  2. I guess the row that is currently being addressed/read-out is the one that is being killed. And thats why the damage is always a single row (in most cases). Perhaps this has to do with the dark-sun effect which changes the FD pedastal level so much that the row driver (currently addressed) is damaged.

    ReplyDelete
    Replies
    1. Does no seem a likely reason to me. First, the row driver is connected to gates, and there should be no light-induced gate current, not in green light, anyway. Second, even if we short the output of the row driver to ground, the driver degradation and electromigration are slow processes. They should not kill the row in so short time.

      Delete
    2. Hmm, not exactly light-induced gate current that I was thinking of; I am not a circuit designer but a semiconductor physicist; so I may be wrong. But I was thinking of a pathway through a potential Vdd fluctuation via the reset gate at the moment of row-addressing that ultimately affects the row drivers.

      Delete
    3. Well, the path you are talking about goes through a tiny pixel gate capacitance. Whatever happens there, I doubt this weak path can deliver enough energy to permanently damage the row driver.

      Delete
  3. Are there a lot of BSI cameras out there (not cellphone cameras, obviously)? While I agree BSI should be less susceptible, the lack of laser damage may be related to the number of FSI v. BSI cameras out there. (I have an RX100M2 so I know there are some nice BSI cameras with more to come).

    ReplyDelete
    Replies
    1. I had this happen to me on a BSI smartphone camera, so BSI sensors are certainly not immune to this...

      Delete
  4. There is a masters thesis on this subject by Sacha Casken "Damage Mechanisms in Visible Band Cameras Subjected to Continuous Wave Laser Attack", July 2014, at Cranfield University, Department of Informatics and Sensors, Military Electronic Systems Engineering. It concludes heat to be primary reason for damage.

    ReplyDelete
    Replies
    1. Thanks, interesting reference. Here is the digest of the thesis:

      http://www.laserist.org/camera-sensor-damage-thesis.htm

      Delete
    2. it would be interesting to performs such tests on a rad-hard CIS...

      GC.

      Delete
  5. I think that this is a pure thermal effect on metal lines crossing a pixel. The row signals are connected to gates (RST, SEL & TX) and they are isolated from silicon sub by oxide. Oxide conducts less well heat than silicon, so laser spot are mot likely to burn these wires. The column line is connected to the source of readout transistors that is silicon. So heat can be dissipated more easily. By consequent less chance to have column line burned.

    -yang ni

    ReplyDelete
  6. Doesn't this pose high risk to auto cars when they are moving high speed?

    If we fire waves of high power laser onto Google car and damage the cameras used
    for auto driving etc, the car could technically be out of control. Does it shut down at 70 mph?

    Sure they have sensors redundancy and can tolerate few missing rows, the possibility of operating on defective cameras is there.

    ReplyDelete

All comments are moderated to avoid spam.