Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Suspected repeater camera defect that affects FSD performance

This site may earn commission on affiliate links.
The repeater glare seen by many people here has been revealed by the channel below to be a design flaw with the circuit board of the camera. The PCB has three holes which allows the indicator light to leak directly onto the image sensor, bypassing the front of the lens entirely (aka, it's not the external indicator light being seen by the camera).

https://www.youtube.com/watch?v=_BUPsjguqdY

E1874639-3E52-4300-B918-B6A18457AB87.png


Furthermore, they show multiple repeater camera's they've found and documented that Tesla has already acknowledged the design flaw and and began at some point covering these holes with tape, to ultimately be resolved with a new design all-together. This directly conflcits with Tesla's messaging when denying some owners' repair requests by aruging it's a feature of the camera. By definition, light leak is a problem, and in this scenario it's not caused by the itself but the board it's attached to. It's simply not possible this glare was intentionally designed, particularly given the evidence of Tesla trying to correct it.

I believe they're quietly letting this issue fall by the wayside, however, that's not good enough. This will be insufficient for those who bought FSD which we know will almost entirely rely on image feeds from the cameras. It's simple deduction to know that having these blinded by a design flaw at a critical time - lane changes and turning at junctions - will heavily impede FSD performance and potentially be dangerous.

D7695D0A-CA48-4CF9-965F-AFDE3D20C964.jpeg


Tesla is arguing this is simply a characteristic of the camera, and you can pay to have a newer version that is "design enhanced". This is logically inconsistent given their DIY tape solution in production. Also, some owners have had their cameras replaced under warranty, likely before those service centre agents received the memo to downplay this issue. They're arguing these instances were done under 'goodwill'. If anyone has received warranty replacements for the cameras specifically for this glare / light leak issue it would be helpful do feel free to comment here.

BC33557F-DF26-4926-B945-84666F334FA4.png


Long story short, these defective cameras will need to be replaced FOC for FSD owners sooner or later, given FSD entitles us to some degree of hardware replacement to make the feature work if needed. Some owners have had to pay out of pocket to have these replaced for this issue, and will have to likely wrestle a refund out of Tesla at some point in the future. Very reminiscent of the eMMC issue.
 
Last edited by a moderator:
I only have the glare issue with my driver’s side repeater, which I’m assuming means they didn’t take the old version out of inventory after they corrected the flaw. It was pretty funny having the SC tell me this was “as intended.”
 
  • Like
Reactions: K5TRX
I only have the glare issue with my driver’s side repeater, which I’m assuming means they didn’t take the old version out of inventory after they corrected the flaw. It was pretty funny having the SC tell me this was “as intended.”
It's really insulting actually given those working on the cars will know it's BS, and whoever made the high level decision to have the "characteristic" wording sent out will know it's BS. By the time these cameras become the lowest hanging fruit to tackle to improve FSD performance there'll be less of these cameras on the road to replace. Plus the amount of FSD Licensee's that could benefit from a replacement will be reduced from selling cars on or whatnot.

Tesla just refused to fix my cameras under warranty (they’d be happy to take $300 from me to do it). They said it’s not a defect. Does anyone know how to complain to Tesla?
I have already been in the process of doing this. You can contact them here for the UK at least: [email protected]

The more people complaining the better.
 
  • Like
Reactions: K5TRX
I am not convinced the current camera resolution and 14 nm computer are up to the task. My bet is that the next version of hardware with 4x resolution cameras (solve the perception issue and a new fsd computer (designed around the current 3D + time nn methods instead of the old per camera per frame nn methods) will be close.

With the 12k cost of fsd, this could easily pay for a camera and computer replacement when they are readily available. If I were them, I’d wait to replace devices until I had the product that achieved the goal; there is little benefit to this type of incremental upgrade if it has to be done again in 2 years.

All stated, I wish they would fix the issue in the mean time.
 
I am not convinced the current camera resolution and 14 nm computer are up to the task. My bet is that the next version of hardware with 4x resolution cameras (solve the perception issue and a new fsd computer (designed around the current 3D + time nn methods instead of the old per camera per frame nn methods) will be close.

With the 12k cost of fsd, this could easily pay for a camera and computer replacement when they are readily available. If I were them, I’d wait to replace devices until I had the product that achieved the goal; there is little benefit to this type of incremental upgrade if it has to be done again in 2 years.

All stated, I wish they would fix the issue in the mean time.
That's very true, and would be a solid reason why they don't see the benefit in upgrading the cameras now. I would be much more satisfied with Tesla if their response to this issue was "we will replace the cameras as needed for FSD when the time is appropriate if need be" rather than "nothing's wrong".

My only issue with that is that they're actively doing the S & X camera replacement programme to the latest design (read: glare-less) of HW2.5 cams. Why not wait to do that until it's the next generation of cameras?

The HW3 replacement programme was at least somewhat more understandable as that was some time ago. Although that's also worrying seeing as they're giving up the redundancy with their dual chips in favour of maximum processing potential. They seem to have boxed themselves in and that entire replacement programme might amount to a waste of money if that computer isn't sufficient.
 
Currently they need to train all with one camera setup, so the replacement makes some sense to me. I see the most recent move to unprocessed raw images as a last ditch effort to see if they can avoid new cameras ( imho it won’t be: the cameras are low resolution and don’t have enough light sensitivity).

For the light leakage issue, I see this as super sun optimal, but not critical yet. Once they are using object persistence through long short term memory nn (hopefully soon? Crazy it isn’t baked in yet), they should be able to take the info from when the blinker is off and project object positions through time. It won’t be perfect, as they would loose a lot of data when the blinker is on, but it could work in a lot of cases.

I don’t see the point of dual fsd chips for redundancy when they don’t have redundancy or wipers in front of the cameras. The probability of failure (obstruction) of a camera is way higher than the failure of a chip. Maybe I am missing something about the redundancy plan that is clear to others.

In all, this is an interesting convo. Thanks for engaging :)
 
Currently they need to train all with one camera setup, so the replacement makes some sense to me. I see the most recent move to unprocessed raw images as a last ditch effort to see if they can avoid new cameras ( imho it won’t be: the cameras are low resolution and don’t have enough light sensitivity).

For the light leakage issue, I see this as super sun optimal, but not critical yet. Once they are using object persistence through long short term memory nn (hopefully soon? Crazy it isn’t baked in yet), they should be able to take the info from when the blinker is off and project object positions through time. It won’t be perfect, as they would loose a lot of data when the blinker is on, but it could work in a lot of cases.

I don’t see the point of dual fsd chips for redundancy when they don’t have redundancy or wipers in front of the cameras. The probability of failure (obstruction) of a camera is way higher than the failure of a chip. Maybe I am missing something about the redundancy plan that is clear to others.

In all, this is an interesting convo. Thanks for engaging :)
Ah no problem! It's definitely interesting. I would be surprised if these conversations weren't being had by FSD devs at Tesla. I posted a reply in another thread which touches on some of the same points you did. I'll paste it below:

I strongly suspect they will have to replace these cameras at some point for FSD. I have been told at a service centre that if needed, Tesla will replace cameras for FSD to function (and have evidence of such).

In a couple of discussions people defending Tesla on this issue claim this won't affect FSD performance - I strongly disagree. I work in VFX / video production and do motion tracking fairly regularly. The process in which I do that will be somewhat different to Tesla's on-the-fly solution, but the underlying principle will be the same:

1. Load the data (i.e - the input from the camera, or in my case a video file from a camera)
2. Analyse and identify points of high contrast
3. Track movement over time of those points
4. Build a solve (output a virtual camera that attempts to mimic the movements of the original perspective)

You can then go on to use that solve. For me, CGI placed into a video will appear to be 'in' the environment thanks to an accurate solve. For Tesla, the virtual car in vector space should be accurately placed relative to the virtual environment around it. The virtual environment and virtual car should closely align to the real environment and real car, and therefore the solve needs to be as accurate as possible.

Going back to the blinded camera, that's affecting step 2 in the above process - Analyse and identify points of high contrast. If you're blinding a camera with indicator light 50% of the time during a lane change or junction turn you're screwing up 50% of your data to base decisions on. Everything in the visual area of the glare is having its perceived colours, contrast and brightness dramatically shifted. As far as the computer is concerned - everything in that area is now completely different. You can either develop some kind of advanced algorithm to try and account for the glare in every scenario (which isn't going to work well because depending on what's behind the glare the area can be different in colour or contrast AND will require you to spend additional processing power), or completely disregard the frames with the glare on and extrapolate what's happening using object permanence. Either way, you're spending a ton of development time compensating for a manufacturing defect, both of which are going to strongly affect the decisions FSD is making in strange ways.

Let's not forget a repeater will likely be the only camera in many scenarios to base decisions off of during manoeuvres/lane changes/turnings, particularly when there's another vehicle directly behind you blocking the rear camera.

Over time there'll be more vehicles with corrected cameras - that's a problem. FSD is supposed to be trained on data from, and deployed on cars using, the same sensor suite. There's now a proportionally shrinking number of cars using defective cameras. Will It make sense to spend more dev resources compensating for this shrinking group? Someone's eventually going to say 'screw it' and just make the argument to upper management these rubbish cameras need to be replaced because it's straining FSD development.

Long story short, at some point, cameras being blinded will become one of the lower hanging fruits to improve FSD performance compared to say, chasing diminishing returns for the march of 9's. I think if you go ahead and pay the fee now you might find yourself fighting Tesla to get a refund on it when this replacement programme ends up happening.
 
Last edited:
Does anyone know if this issue is related to the mysterious error message "Auto park unavailable" that I get in my MX from time to time. It only seems to happen when there is bright sun on the side of the car.

I am guessing the error is due to the repeater being blinded. And yes, the message should be inhibited if the car is in 'Drive'.
 
  • Like
Reactions: Jaro73
Does anyone know if this issue is related to the mysterious error message "Auto park unavailable" that I get in my MX from time to time. It only seems to happen when there is bright sun on the side of the car.

I am guessing the error is due to the repeater being blinded. And yes, the message should be inhibited if the car is in 'Drive'.
Wow so this is bigger than just me. I get that on my 2018 X as well. Not often but it comes up even in highway driving.
 
warranty vs goodwill aside, I feel like this particular flaw won't materially affect FSD performance, and if it does, it's in very limited situations.

The problem only exists when the turn signal is active and at night. The repeater cameras are most useful for lane changes/merges, less so for turns. If there is a car approaching in an adjacent lane, the headlights of that car trigger the camera to adjust for the significant increase in light coming through the aperture and reduces the sensitivity of the exposure. This causes the glare issue to go away because it simply isn't bright enough to register in the new exposure setting. Same concept of background stars not showing up when you are exposing for something much brighter, like the moon. Also why you don't see any of this glare during daytime.

The other scenario where this might be a problem is a left turn where the turn angle is very sharp (way past 90 degrees), so the car is essentially facing away from the direction of the turn. Here, the left repeater camera will be needed to spot for oncoming traffic on the left. Since you're making a left turn, the left signal will be on, and the problem will manifest. However, like the scenario above, if there are lots of cars coming from the left, their headlights will cause the repeater to reduce its exposure setting, which significantly reduces the internal glare problem.

Where FSD might struggle is when there is no approaching traffic, and the flashing glare confuses FSD into thinking it needs to yield for something. Tesla can likely just add some software logic to ignore or filter out the blinks, since they know what hardware is in what car, and I assume they know the real-time blinker interval as well.

Anyway, I'm not trying to defend Tesla for pushing back on fixing this. I agree it's a design defect that "work as designed," which is never an excuse when it works poorly. It is pretty annoying at night to have this magenta rectangle flashing on the screen. But based on my observations, it doesn't seem like an FSD dealbreaker.
 
I don't think the leaking turn signal light affects FSD. If it did, we would have already seen the effects, and Tesla should be able to fix them in software. Also, higher resolution cameras are not needed when the limitation is processing all the pixels in the current setup and the power consumption of the FSD computer.

That said, I'm disappointed that Tesla didn't have the foresight to see future possible issues with the older camera circuit board design with the 3 holes.
 
I wonder if this is an easy DIY fix. I've never popped out my repeater assembly, but seems like it would be easy to take apart.
We'll see eventually how big of an issue this is I suppose. I'm just speaking from my experience with motion tracking that this issue raises major red flags to me. Maybe there's a regulatory loophole that allows them to reduce the blinker rate for additional data, maybe the issue gets superseded altogether with camera upgrades, maybe they figure something out on the software side. I just don't want to be taken advantage of given I paid for FSD which has delivered zero value for me so far as a UK resident. I can only really hope they deliver what they promise, and get defensive when it looks as if they're trying to get away with delivering less.

If you watch the video at the top of my OP someone from these forums found a way to fix it, but it requires you to drill the repeater housing which'll deffo void your warranty if Tesla ever ends up seeing it. Maybe they'll be noting any logs of anyone disconnecting and reconnecting repeater cams.
 
We'll see eventually how big of an issue this is I suppose. I'm just speaking from my experience with motion tracking that this issue raises major red flags to me. Maybe there's a regulatory loophole that allows them to reduce the blinker rate for additional data, maybe the issue gets superseded altogether with camera upgrades, maybe they figure something out on the software side. I just don't want to be taken advantage of given I paid for FSD which has delivered zero value for me so far as a UK resident. I can only really hope they deliver what they promise, and get defensive when it looks as if they're trying to get away with delivering less.

If you watch the video at the top of my OP someone from these forums found a way to fix it, but it requires you to drill the repeater housing which'll deffo void your warranty if Tesla ever ends up seeing it. Maybe they'll be noting any logs of anyone disconnecting and reconnecting repeater cams.
Speaking from my experience using FSD I have not seen any effect that is concerning or different in It’s performance. The slightly orange sky effect (beyond dusk on unlit streets) has also not ruined my experience much either. All said granted I’m not hunting for something to complain about.