Other FaceID analysis

Discussion in 'iPhone' started by 537635, Nov 8, 2018.

  1. 537635 macrumors 6502a

    Joined:
    Mar 7, 2009
    Location:
    Slovenia, EU
    #1
    A user on this forum "stafil" has started a couple of topics in regards to eye safety with iPhone X and the new XS and XS MAX phones. He asked if somebody could measure what is really going on, so here goes.


    We used a device to measure the power with which the two infrared sources are emitting. Unfortunately there were other light sources in the room so it was difficult to get really exact measurements. Still we repeated the measurements a couple of times and they were consistent to about 0.05 mW.

    4.JPG



    Additionally we also analysed how the sequence looks like in terms of pulse durations. Dot projector appears to work with about 5ms pulses.

    5.JPG



    There appear to be three distinct sequences.



    1. Upon waking, the phone is using the flood illuminator to check for surroundings, most probably looking for a face to identify. A this point the flood illuminator is emitting about 0.2 mW of power.

    1.gif




    2. When the IR camera detects a face, dot projector (short flash on the right side) emits a much more powerful beam. It doesn't look like it from the video, but we measured it to about 1.5 mW. Still this is just a single shot and if successful, the phone unlocks.

    2.gif




    3. During normal operation flood illuminator fires up every couple of seconds, most probably to check if the user is looking at the screen. The power stays the same, about 0.2 mW.

    3.gif



    From eye safety standpoint it is still difficult for me to comment. Before I dismissed these concerns as utterly irrelevant and paranoid. Now I'm leaning more towards "most probably OK for normal use, but don't put your phone next to your eye". Flood projector is most probably completely safe, because it is just an infrared flashlight. Yet the dot projector is actually quite focused and apparently ten times more powerful.

    I still believe Apple did quite some testing and deemed the dot projector perfectly safe. Yet in a way it conforms to class 2 laser standard, if we speculate our measurements were on the high side (class 2 is limited to 1mW). Dot projector being near infrared also makes it a bit more dangerous, because there is no blink reflex.

    https://en.wikipedia.org/wiki/Laser_safety#Class_2

    Additionally dot projector is seldom activated, it is always at a considerable distance from the eye and it is dispersed into numerous small dots, from which most probably only one or two hit the eye.
     
  2. bbrks, Nov 8, 2018
    Last edited: Nov 8, 2018

    bbrks macrumors 65816

    bbrks

    Joined:
    Dec 17, 2013
    #2
    Thanks for the effort. I sincerely hope that now there will be no more paranoia when discussing faceID. :)
     
  3. sddabrow macrumors 6502

    Joined:
    Oct 7, 2011
    #3
    This is very informative, thank you. I would imagine from a laser safety standpoint, they would have to be Class 1 or Class 1M, as it is intentionally looking for open eyes.
     
  4. 537635 thread starter macrumors 6502a

    Joined:
    Mar 7, 2009
    Location:
    Slovenia, EU
    #4
    ... and they most probably are. It is impossible to measure the power of an individual dot, I haven't found any information about the number of dots altogether. Yet the sheer power could be potentially problematic if one was to put the dot projector aperture right on top of the eye.

    But I guess this falls into the category that iPhone can be dangerous it somebody throws it at you.
     
  5. sddabrow macrumors 6502

    Joined:
    Oct 7, 2011
    #5
    Yes, interesting. Do we know the laser wavelength of the dot projector?
     
  6. 537635 thread starter macrumors 6502a

    Joined:
    Mar 7, 2009
    Location:
    Slovenia, EU
    #6
    I read somewhere that is is between 950 and 1000 nanometers. I'm also looking into measuring this, will post here.
     
  7. BugeyeSTI macrumors 68030

    BugeyeSTI

    Joined:
    Aug 19, 2017
    Location:
    Arizona
    #7
    Thank you for posting your findings and future results as well..
     
  8. stafil macrumors 6502

    Joined:
    Sep 15, 2017
    #8
    Amazing work. Thanks so much for taking time to do this!
     
  9. Tijdelijk macrumors regular

    Joined:
    Jul 8, 2017
    #9
    So what is the conclusion ?
    Is it harmless or ??
     
  10. pdp1 macrumors regular

    Joined:
    Oct 15, 2018
    #10
    So can we assume the 1.5mW from the dot projector is a sum of all the dots? Which means if only 1 or 2 of the dots would only make up a small fraction of the 1.5mW?
     
  11. 537635 thread starter macrumors 6502a

    Joined:
    Mar 7, 2009
    Location:
    Slovenia, EU
    #11
    Yes. Sensor aperture is about 2x2cm and we held it just next to the iPhone screen.
     
  12. Ralfi macrumors 68030

    Ralfi

    Joined:
    Dec 22, 2016
    Location:
    Australia
    #12
    I'd be shocked if there's anything to worry about - a company like Apple would've tested it extensively at normal range & the closest of distances....there are kids who use these things & if there was even a 0.5% chance of eye damage, Apple would've axed it.

    Imagine the law suites...Apple have history in court for many insignificant matters (in comparison), but to have ok'd a biometric system which could damage eyes takes things to another level.
     
  13. 537635 thread starter macrumors 6502a

    Joined:
    Mar 7, 2009
    Location:
    Slovenia, EU
    #13
    This is an excellent summary! On the other hand I wanted to offer some concrete numbers in terms of light intensity in light of Apple's secrecy in this aspect.
     
  14. serialiphoneuser, Nov 9, 2018
    Last edited: Nov 9, 2018

    serialiphoneuser macrumors regular

    Joined:
    Sep 21, 2016
    #14
    The flood illuminator does not activate to its peak output when there's enough visible light. Try it in the dark. If the measurement tool is too close, the proximity sensor might limit the output of the flood illuminator. The dot projector must be evaluated in direct sunlight conditions, the output could spike by several multiples. The machine learning model may have been already trained to detect special circumstances and play 'nice' (dieselgate). The results may also be skewed by observer effect in other unknown ways, due to other factors. Using Face ID = Rolling the dice and hoping that it's safe in the long term (it's gotta be!, right?! How would they release a product if it wasn't? it's not like they had any other options - Touch ID under the display is a total myth!!).
     
  15. 537635 thread starter macrumors 6502a

    Joined:
    Mar 7, 2009
    Location:
    Slovenia, EU
    #15
    Not true. It works in daylight also and with same power as in total darkness.


    True. But again, it either works or not. No amplitude modulation as far as we've seen.


    Again, this doesn't appear to be true.
     
  16. serialiphoneuser macrumors regular

    Joined:
    Sep 21, 2016
    #16
    How would you explain the varying levels of intensity of the flood illuminator as seen in the original two GIFs (bottom left flashing dot). In the first one, it's very dim and rapidly pulsing. Where as in the second one it flashes very faintly for a split second and then it's a bit brighter?

    Was the testing rig taken outside to take the measurements on a bright sunny day?
     
  17. user1234 macrumors 6502a

    user1234

    Joined:
    Mar 3, 2009
    Location:
    Sweden
    #17
    Great!
    This is exactly what I have been looking for, and what I would have done myself if I had access to the tools necessary. Thanks for your effort!

    On another note, no one seem to have measured the PWM frequency of an Apple Watch, at least none that I can find. This would be interesting too if you feel up to it
     
  18. 537635 thread starter macrumors 6502a

    Joined:
    Mar 7, 2009
    Location:
    Slovenia, EU
    #18
    You guys are really incredible. Yes, we'll take the whole lab with expensive IR laser measuring equipment outside and do the tests. Sure. We didn't see any difference between all lights on and total darkness. I doubt it goes into "retina burn mode" when outside.
     
  19. Tijdelijk macrumors regular

    Joined:
    Jul 8, 2017
    #19
    Retina burn mode :D:D:D:D
     
  20. stafil macrumors 6502

    Joined:
    Sep 15, 2017
    #20
    Just to say that not everyone is like this. Some of us really, *really*, appreciate you taking time to do this and post the results here.
     
  21. BugeyeSTI macrumors 68030

    BugeyeSTI

    Joined:
    Aug 19, 2017
    Location:
    Arizona
    #21
    His signature is “fail id”, what do you expect! Most appreciate the time you’ve taken to try and provide some actual data on the subject..
     
  22. serialiphoneuser macrumors regular

    Joined:
    Sep 21, 2016
    #22
    I asked because you gave an almost conclusive answer.

    It's good that we have some insight into how the TrueDepth sensors work. That's a great start!

    And, why the blackouts in the first picture?
     
  23. matthijst macrumors regular

    Joined:
    Jun 8, 2009
    #23
    Thanks for doing these measurements : )
    Think the dot projector has 30.000 dots, so they should be very low powered.

    You can see how many hit your eyr in this photo:
    [​IMG]


    Don't think it's dangerous, but could be annoying.
     
  24. 537635 thread starter macrumors 6502a

    Joined:
    Mar 7, 2009
    Location:
    Slovenia, EU
    #24



    In this specific photo, zero (0) dots hit the eye (as in through the pupil), all of them stop on the iris.
     
  25. SteelHeart macrumors newbie

    Joined:
    Oct 8, 2018
    #25
    A question , during the third sequence , does the flood illiminator operates during our normal usage of phone even if we turn off the face id??
     

Share This Page

37 November 8, 2018