Man Says CES Lidar's Laser Was So Powerful It Wrecked His Camera (arstechnica.com)

An anonymous reader quotes a report from Ars Technica: A man attending this week's CES show in Las Vegas says that a lidar sensor from startup AEye has permanently damaged the sensor on his $1,998 Sony camera. Earlier this week, roboticist and entrepreneur Jit Ray Chowdhury snapped photos of a car at CES with AEye's lidar units on top. He discovered that every subsequent picture he took was marred by two bright purple spots, with horizontal and vertical lines emanating from them. "I noticed that all my pictures were having that spot," he told Ars by phone on Thursday evening. "I covered up the camera with the lens cap and the spots are there -- it's burned into the sensor." In an email to Ars Technica, AEye CEO Luis Dussan confirmed that AEye's lidars can cause damage to camera sensors -- though he stressed that they pose no danger to human eyes. "Cameras are up to 1000x more sensitive to lasers than eyeballs," Dussan wrote. "Occasionally, this can cause thermal damage to a camera's focal plane array." Chowdhury says that AEye has offered to buy him a new camera. The potential issue is that self-driving cars also rely on conventional cameras. "So if those lidars are not camera-safe, it won't just create a headache for people snapping pictures with handheld camera," reports Ars. "Lidar sensors could also damage the cameras on other self-driving cars."

"It's worth noting that companies like Alphabet's Waymo and GM's Cruise have been testing dozens of vehicles with lidar on public streets for more than a year," adds Ars. "People have taken many pictures of these cars, and as far as we know none of them have suffered camera damage. So most lidars being tested in public today do not seem to pose a significant risk to cameras."

Comments

Lies pushed by big Optometry.

[Anonymous Coward]

"Cameras are up to 1000x more sensitive to lasers than eyeballs"

and what was the shutter speed he was using? Something like one 1/500th of a second, so in 1 second the laser would do 500 times the damage it did to the camera? in 2 seconds it'd do just as much damage.

Bullshit about eye safety.

[Anonymous Coward]

If this is powerful enough to damage a CMOS/CCD sensor then it is most certainly also doing damage to biological tissue in eyeballs.

If this is doing "thermal damage" to CMOS/CCDs, essentially chunks of glass, then it is doing more damage to biological tissues.

Re:Bullshit about eye safety.

[pushing-robot]

The retina is submerged in a water bath while the camera sensor is surrounded by insulating air and plastic. The sensor may have a higher absolute temperature rating but it can't dissipate heat nearly as well as your retina.

Re:Bullshit about eye safety.

[eclectro]

Stop downplaying the dangers of laser technology. Any coherent radiation hitting they eye should be considered very dangerous. Even the cheap laser pointers have a yellow caution sticker on them!

Here is a story about lasers blinding concert goers in Russia.

Re:Bullshit about eye safety.

[tlhIngan]

If this is powerful enough to damage a CMOS/CCD sensor then it is most certainly also doing damage to biological tissue in eyeballs.

If this is doing "thermal damage" to CMOS/CCDs, essentially chunks of glass, then it is doing more damage to biological tissues.

It depends on the laser. Common ones used for LIDAR is 800nm and 1550nm. 800nm can injure the eye, so its use in LIDARs and such where it might hit an eye accidentally mean its power limited to prevent damage.

The thing with 1550nm is that it can't make it to the retina - the fluid in the eyeball is opaque to light at 1550nm, so the power limits are higher because it takes a lot more power to damage the eye when you have to go through the eyeball liquid first (and it's a lot of water so it will take a fair amount of energy before you actually cause real damage).

Of course, camera sensors aren't filled with eyeball fluid and thus the full power of the laser will be able to destroy the sensitive pixels in the array.

"as far as we know"

[SlaveToTheGrind]

"People have taken many pictures of these cars, and as far as we know none of them have suffered camera damage. So most lidars being tested in public today do not seem to pose a significant risk to cameras."

Or maybe, just maybe, this was one of the few instances where (1) camera damage happened; (2) the camera owner realized the damage must have been due to snapping a picture of a self-driving car; and (3) the camera owner knew who owned the self-driving car so they could complain?

Re:"as far as we know"

[Dayze!Confused]

It's pretty obvious which picture it is when 1-100 are good, picture 101 has two dots and a self driving car, and pictures 102 and beyond have that too. Anyone with a camera who detects that problem would go back through their photos and find 101, and self driving cars aren't being sold to the masses yet, it would be easy to track down who owns it; it's not one of millions of individuals, it's one of a handful of companies licensed to operate them.

Not just cameras

[ColaMan]

When I use multiple LIDARs on a machine their beam sweep has to be synchronised otherwise the reflections of one beam can interfere with the other.

I'm waiting to see what happens with a freeway full of cars with LIDARs, all flinging their beams at each other willy-nilly with direct beams and reflections all over the place. If you're unlucky you'll get a beam from another vehicle just after yours has sent a pulse out - resulting in a false return showing something right in front of you.

I'm guessing that most of the time with enough units around you all you'd get is the equivalent of "static" on your laser sweeps, where you briefly get invalid results for a few degrees of sweep. If you're really unlucky, you blind your sensor, temporarily (bad), or permanently damage it (bad and expensive).

Re:Not just cameras

[somenickname]

Your "static" scenario is exactly what will happen with AM pulsed lidar. It's (one of many) dirty secrets of the lidar industry: AM lidar doesn't scale. There are alternative ways to do lidar (FMCW for one) that can scale much, much better but, these kinds of systems are still a bit expensive and so won't see widespread adoption for a few years.

Re:Not just cameras

[Solandri]

Radar and sonar overcome this problem by constantly varying the frequency in a series of chirps. It's highly unlikely that there's another radar/sonar transmitting at the same frequency at near the same time. And even if there is, it's unlikely to be varying the frequency at the same rate/range.

Another advantage of this is that you don't need as strong a sweep signal. With a single frequency, you're emitting a pulse, then waiting for the reflections of the pulse. In order to avoid the possibility of spurious noise from another source being interpreted as a reflection, your pulse has to be high-power (basically make the reflected signal stronger in strength than any noise). 1000 to 5000 Watts was typical for boat radars using pulse beams. But when you use a varying frequency, you can compare reflections at one frequency with subsequent reflections at a different frequency (there's no need to wait for return reflections - subsequent pulses will not interfere with previous pulses, so can be sent before reflections from previous pulses arrive). Noise will show up at just one frequency, making it easy to spot and trivial to filter out. Consequently newer frequency sweeping boat radars only need to emit at a few tens of Watts.

That said, the parking sensors in your car use this frequency varying sonar. And I've noticed other cars' parking sensors trigger mine about once a day. So some more work needs to be done on standardizing frequency sweeps and noise filtering to reduce signal collisions. But the problem is not as insurmountable as you'd think from your LIDAR experience.

Re:Not just cameras

[guruevi]

People in the industry know. Right now self-driving cars are an easy way to raise funds. Much like the .com era, everyone wants a piece of the pie. Eventually after about 20 years we'll laughing at the time we were awed by the self-driving car equivalent of Badger badger and lament the situation with Pewdiepie while everyone just uses self-driving cars to get delivery of hookers and blow.

There is less and less reason to keep driving around like fools every day for work. Cars, self driving or not, will feel that pressure as the next generation will demand more to flex home and work. So cars will once again become a thing for leisure and the powerful who will either be driven around by someone experienced or doing it themselves for the thrill of it.

Self-driving cars for the 'common folk' will be around but will be very simple automatons without all the high-tech gadgets driven in tandem with a skilled delivery driver much like today's truckers, taxi and UPS drivers.

Re:Not just cameras

[ColaMan]

I didn't say it was insurmountable, but as another poster has pointed out, there's very few LIDARs on the market right now that modulate their beam. Unlike radar, it's difficult to vary the actual frequency as such as they're diode lasers and generally fixed. So we're kind of stuck with just beam modulation unless we want to do something fancy like driving multiple lasers, which gets tricky when you're all sharing the same optical path.

LIDARs also have the difficulty (or advantage, depending on which way you look at it) of having a much tighter beamwidth. So it's much more difficult to rotate onwards in your scan and listen to previous reflections as you're using the same optics to both transmit and receive your signal. Even at one foot per nanosecond, you start to have trouble out at 300 feet with the typical half or quarter degree resolution and 50-100Hz scan rate of recent LIDARs. With the telescope-type lensing on them, it would be fairly easy to blind/overload the sensor assembly with a beam (not a reflection) from another unit - they are already significantly blinded by the setting sun on the horizon or other brightly lit environments, for example.

Anyway, it just seems that every autonomous vehicle manufacturer right now is all "oh wow, look at what our vehicle can do!", with very little info on how they'll go when there's 20 of them at a stop light. Hopefully technology will catch up to them before they're at scale, or full-scale production will force the solution, one or the other I guess.

Re:Not just cameras

[somenickname]

You are 100% correct but, you are describing frequency modulation. The vast majority of lidar companies (on the order of 99%) are producing amplitude modulation systems that shoot out strong amplitude signals and cross their fingers hoping they can see and distinguish it when it comes back. These systems often use "avalanche detectors" to help their probability of return detection (look it up, it's insane). Driving directly towards a sunset can literally cause these systems to emit their magic smoke. Accidentally hit the wrong angle on a retro-reflector? Smoke. Having another system pointed in their direction means "something made a peak" and there is no way distinguish your light from another sensors light.

These problems are solved in the radar space but, the solutions are taking a long time to trickle down into lidar space. Mostly because it's very easy and very cheap to produce an AM lidar.

Re:Responsibility.

[Anonymous Coward]

Eyes don't withstand being pointed at the sun for very long. Neither do most cameras.

What's good for the goose...

[Arzaboa]

All I hear is that there will now be some "smart" people trying to outsmart new tech and cameras by firing lasers, not strong enough to poke your eyes out, but strong enough to burn the retina of the machine.

--
When you have confidence, you can have a lot of fun. And when you have fun, you can do amazing things. - Joe Namath

1550 nm wavelength is (relatively) eye-safe

[Vadim Makarov]

As the original article duly explains, the laser light at the wavelength of 1550 nm used by this lidar scanner does NOT reach the retina of the eye. At this wavelength, it is fully absorbed in outer parts of the eye (cornea, lens, etc.) before it could get focused into a tight spot on the retina. This makes this wavelength (relatively) eye-safe, comparing to visible and some other wavelength ranges. There is no such protection for the camera however, whose glass optics happily focuses 1550 nm into a small spot... so the sensor damage may happen.

Laser safety regulations are primarily concerned with (a) no damage to humans, especially their eyes, and (b) laser beams not setting things on fire. Neither of this has happened in this incident. So we are good.

If you are interested in technical details of laser safety, read ANSI Z136.1 standard. Warning: it requires technical expertise.

Re: 1550 nm wavelength is (relatively) eye-safe

[Vadim Makarov]

While I don't know the actual filter construction, a couple possibilities come to mind. First, the filter may be bonded or integrated or be a deposited layer on the sensor itself, and a physical crack in the filter may propagate into the sensor chip. Second, any filter has a finite suppression, and I don't see why the one in the camera has to have it high. So some fraction of light can still get through it, and that could be enough to damage the sensor.

Re:1550 nm wavelength is (relatively) eye-safe

[Swave An deBwoner]

Retinal damage is not the only ocular risk.

https://velodynelidar.com/newsroom/guide-to-lidar-wavelengths/

1550 nm systems use a wavelength that is allowed to run more power compared to 905 nm. However, under certain conditions, the 1550 nm wavelength of light can still cause corneal damage and potential damage to the eye lens.

Re:1550 nm wavelength is (relatively) eye-safe

[Vadim Makarov]

Maximum permitted exposure (MPE) in 1500-1800 nm band is the same for the eye and the skin. For continuous-wave light it is 0.1 W/cm2, for pulsed light it is 1 J/cm2. Reference: ANSI Z136.1, see Tables 5a and 7.

In other words, if the 1550 nm laser beam is not burning your skin, it is safe for your eye.

This is remarkably untrue at other wavelengths, where light is dramatically more dangerous to the eye than it is to the skin.

There go the police ticketing cameras

[Flexagon]

So, how long will the various municipalities' automatic red light ticketing cameras last with this?

Re:Responsibility.

[thegarbz]

A camera should withstand being pointed at the sun.

a) why? It's not what they are designed for.
b) sun happily damages cameras too, it all depends on exposure.

Broadcasting light interference is no different than broadcasting radio interference (in terms of responsibility, not physics).

Which is why we have part 15 of the FCC rules. You will accept the interference and you will like it.

Shutter speed doesn't matter for mirrorless camera

[JoeyRox]

The camera in question is a Sony A7rII, which is a mirrorless camera. Such cameras constantly expose the sensor to light in the scene [while not taking photos], which is necessary to provide the video-like image stream used for the electronic viewfinder and LCD display.

Sony and security...

[zmooc]

Sony does not have a good security track record so this does not come as a surprise to me.

OWASP Secure Coding Practices Checklist section 1 about input validation was clearly not applied at all. Specifically, they failed to implement "Validate data range" :p

Re:Responsibility.

[thogard]

Lasers are controlled in the US by 21 CFR 1040 and have been for a very long time. There is a odd loop hole in that if you if have your laser hit a lens that spreads out the beam so that at the end of the laser is larger than an eyeball, it can deliver far more power even if the beam is much smaller far away. Some early traffic speed lasers took advantage of that. The ANSI standard has the same problem. They should require the test at the end of the laser and 100 meters away.