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

HW4 leak: 1 fewer camera

This site may earn commission on affiliate links.
Not clear how reliable this info is, but tl; dr; front 3-camera unit becomes 2 cameras, cameras themselves become higher-res, B-pillars get heating elements.

 
Heaters in the B-pillar cameras are definitely needed. The B-pillar cameras do suffer from condensation buildup in certain weather conditions.
I, like most, have this issue. I always thought the solution would be to drill a hole in the bottom of the black felt lining, venting the camera to the inside cabin environment. I read somewhere that's already the case, but when I examine my cameras closely with a light, I don't see any vent holes in the black felt lining.
 
Heaters in the B-pillar cameras are definitely needed. The B-pillar cameras do suffer from condensation buildup in certain weather conditions.
Yeah I almost consistently can't use NoA after hitting the sun on the freeway during some parts of the year. Interestingly FSDb is usually unfazed in identical conditions (eg exit the freeway manually because the camera is considered blocked, enable FSDb as soon as the visualization appears).
 
  • Like
Reactions: enemji
Yeah I almost consistently can't use NoA after hitting the sun on the freeway during some parts of the year. Interestingly FSDb is usually unfazed in identical conditions (eg exit the freeway manually because the camera is considered blocked, enable FSDb as soon as the visualization appears).
The reason is that the old Driver Assist stack uses just a single front camera instead of all 3. Advanced Driver Assist uses all 3 cameras, including the wide-angle camera, which can compensate for an obscured B-Pillar. There is a video I've posted awhile back that showed how amazing the system is when covering the various cameras and observing what the system can still see.
 
I sure hope they can retrofit heating elements in the B pillars
The existing camera module has video and power, so the new module should be plug-n-pray assuming the wiring harness can take the extra power to supply the heater elements, my guess is yes since it wouldn't need much power to resistively heat such a small area.
 
A higher resolution camera (which has been rumoured for a while) could replace 2 of the forward facing cameras which are at different focal lengths - ie it’s just a digital zoom. So it makes perfect sense, easier to integrate the images as they’re already integrated, but no real increase in resolution, at least on the one image.
 
  • Like
Reactions: gsmith123
Sounds like quite good news for everyone on HW3 in terms of those vehicle’s ability to support everything Tesla are currently expecting to be able to deliver in software.

Most of what’s mentioned in that article seems like an incremental upgrade rather than massive new capabilities. Only possible exception would be new radar but that seems still unconfirmed at this point.
 
  • Funny
Reactions: AlanSubie4Life
Sounds like quite good news for everyone on HW3 in terms of those vehicle’s ability to support everything Tesla are currently expecting to be able to deliver in software.

Most of what’s mentioned in that article seems like an incremental upgrade rather than massive new capabilities. Only possible exception would be new radar but that seems still unconfirmed at this point.
I think I view that as bad news. I still don’t see how the current system has a good enough view to turn perpendicular onto an uncontrolled road. I feel like that means the band aids continue. Right now it’s pretty scary.
 
  • Like
Reactions: DanCar and Jcl007
I think I view that as bad news. I still don’t see how the current system has a good enough view to turn perpendicular onto an uncontrolled road. I feel like that means the band aids continue. Right now it’s pretty scary.
Yeah, Tesla might feel like FSD is good enough with some band aids. But, even if they can get it to be mostly safe, and it still drives like a student driver, that's not finished IMO. I'm sure this is where the disagreements will start. Some will say it's fine as long as it is safe, who cares if it drives slow. But a lot of us paid too much money for an unfinished system.
 
Isn't 2x 5k > 3x 720 ?
Careful in comparing resolution numbers. I believe the old cameras are 1280 x 960 (960p not 720p, though roughly in that class). I don't see numbers for the new cameras, but a good guess would be double the linear resolution i.e. 2560x1920, hence the 5MP description.

I think the 5K number is 2x too high and the 720p number is a bit low. Nonetheless, I think it's still true that the improved resolution could justify the removal of one of the cameras - however the net result may not be nearly as good as if all three cameras were kept an upgraded.

For resolving and classifying distant objects, the most important Improvement is the ratio of linear resolution which appears to be double, or close to that. And we assume that the optics support that, not limiting the sensor.

It's probably a safe assumption, if the two cameras are set with different focal lengths, that the wide front camera will remain - because they need that basic field of view. The doubled resolution could, very roughly, double the claimed distance it's capable of, from "60m" to perhaps "120m".

The next question is whether they will keep the FOV of the present medium camera, and use its doubled resolution to justify an improvement from"150m" to perhaps "300m", slightly better than the claimed "250m" of the existing narrow front camera. Alternatively, they could keep the narrow FOV camera, leveraging the doubled resolution to go from existing "250m" to perhaps "500m". This is a more significant Improvement for high-speed highway driving, and the justification would be that the medium camera is no longer required because it's role can now largely be taken by the wide-F9V camera.
( sorry for the tedious use of quote marks around all the distance numbers; I'm doing that because these are all approximate claims based on debatable blur standards etc.)​

Of course, they could be doing some in-between FOV for the second camera, not reaching quite as far but doing a better job to make up for the lost medium-view camera.

Another possibility is that they use two symmetrical equal-FOV cameras, a left and a right that together cover the wide view but with better resolution then before, because of the doubled sensor resolution combined with a narrower view for each. There would be some overlap of the two views directly forward, and signal-processing principles would result in a further virtual-resolution improvement (but less than a complete doubling) due to the merging of two images in the central overlap area. For @brkaus : this would maybe justify the use of the word "binocular", but I wouldn't assume. too strongly that the word was used properly or conventionally, in this leaked internal (translated Chinese?) document. In any case, too much is made here on TMC of the binocular i.e. stereoscopic range-finding advantage of two closely-spaced cameras. This is important for humans in the near field, but not too helpful at longer distances on the road. The triangulation effect is insufficient unless the cameras were to be moved farther apart (increased "baselength").\

It will be interesting to see the revealed setup, and like many others I hope there will be an upgrade path (whether "free" for FSD purchasers or not).