TI introduces DLP for headlights | Hacker News Reader

TI introduces DLP for headlights

electronicsweekly.com

26 pointsrbanffy41 comments

I got a demo of this technology close to about 10 years ago from TI. Their "killer app" in the lab was a pairing of the DLP with their DSPs to track rain while driving and reduce illumination of rain drops such that glare to the driver was substantially reduced. It was in a lab, but it worked surprisingly well.

Another thing that many people don't realize is that TI makes a huge number of automotive parts currently with extremely tight reliability controls in place for customers like Toyota, as well as established supply chains from silicon to road via companies like Temic automotive.

Personally, I hope TI continues to develop DLP into new markets. The tech is really cool, their miniaturized projectors using laser sources for embedding into smartphones is another demo I saw that would be really interesting if it ever hits market.

> Another thing that many people don't realize is that TI makes a huge number of automotive parts currently with extremely tight reliability controls...

Any idea how many of those leverage MEMS tech at the scale and complexity of DLP?

DLP complexity + high vibration platform + operating temperature extremes on both ends of the gamut doesn't strike me as reliable by any stretch of imagination.

TI makes some MEMS devices for automotive applications, but I don't know of anything in automotive with the type of complexity involved in a DLP. Their mixed signal integrated controls are pretty nuts but not mechanical.

The reason I'm not to worried about reliability is more or less as follows. The fact that TI makes other automotive parts is important because they know exactly what kind of environment these parts will be subjected to, they know how they will be handled when they are assembled from the chips TI ships into automotive boards, and what corners will be cut when those boards are sold and turned into assemblies which are sold and turned into cars. They have plenty of experience in determining what kind of reliability intervals will be required. TI built their first DLPs back in the 80s. They made their first commercial ones in the mid 90's. They probably never turned a profit on DLP until the mid 2000's which is about when they first demonstrated working prototype DLP headlights. They then spent 10 years refining them before taking them to market. TI isn't Facebook, they don't move fast and break things, they are an old school technology company that moves slow and reliable. If they get a reputation for poor reliability they stand to lose decades of investment and future revenue and they know it.

Another thing a lot of people probably don't know is that every TI part that fails in an automotive application gets returned to TI where a team of engineers meticulously dismantle it until they determine the exact cause of failure. They are legally obligated to do this from contracts with auto manufacturers but the net results for TI has been the development of one of the worlds most sophisticated semi-conductor reverse engineering capabilities.

If we can't have adaptive high beams [https://www.youtube.com/watch?v=wSI-NVD1who] in the US we definitely can't have DLP. DOT for headlights in the US is awful. There should be revamped laws for headlights. When the base model has better output with terrible reflector technology than halogen projector you've screwed up the regulations.

http://www.iihs.org/iihs/news/desktopnews/more-than-half-of-...

Though likely not as sophisticated as those Mercedes ones, the Citroen DS series 3 (1967) had directional headlights[0]. Sadly as with now, this was also prohibited in the US.

[0]: https://en.wikipedia.org/wiki/Citro%C3%ABn_DS#Series_3_-_Nos...

Grandaddy of them all was the Tucker Torpedo...

When you get your retinas burned out by HID bulb in an oncoming car, thank the DOT. Rating is based on watts delivered to lights, not photons delivered to driver. Law shouldn't even have type checked.

About 17 years ago, the NHTSA had a public comment period about glare from headlamps. One response [1] was particularly informative.

[1] https://www.regulations.gov/contentStreamer?documentId=NHTSA...

> Automaker philanthropy cannot be relied upon to control headlamp glare.

This is exceedingly well written.

It would be nice if the US would update their headlamp beam pattern standards to be in line with the rest of the world. They allow low beams with insufficient beam width (leading people to use their fog lamps to address the issue), to much glare for oncoming traffic (lack of an asymmetric cutoff), and high beams that are limited to half the hotspot intensity. Also the fact that they require a manual switch between low and high beam prevents the introduction of more advanced vehicle lighting systems that are available in the rest of the world.

I had this idea more than a decade ago but for another reason. If you're driving in heavy snow, the headlights light up large flakes near the car and make it hard to see beyond them. If you add another light at a non-visible wavelength and an image sensor whose pixels can be mapped to the DPL light at very high frame rates, you could turn off the visible lighting that illuminates the big bright snow flakes. Suddenly the stuff you want to see in the distance would be more visible. When I first thought of it, resolution seemed OK but frame rate would probably not have been high enough. I recently revisited the idea and it seems entirely possible today.

Snow Lights.

That's such a good idea, and it's just the sort of inventive thinking that I really like hearing: the problem is obvious and anybody that knew anything about DLP's should realize that they can very rapidly control the illumination of any pixel, but who would come up with the idea of controlling light like that!

TI's DLP technology, used in 85% of digital cinema displays, utilizes a matrix of tiny articulated mirrors, one for each pixel. Each of these mirrors has two controllable positions (for example +17 degrees and -17 degrees). A separate light source (LED, laser, etc.) shines on the matrix, and the mirrors within the matrix at one of the orientations reflect light at the target. These devices have been used in projection systems since 1997. See [1] and [2].

The advantage in automotive applications is that the light can be bright enough for automotive applications while allowing the illuminated area to be controlled through software. Illumination levels and regions can be modulated in response to the car's speed, steering, location, recognition of oncoming cars, etc. [3]

[1] http://www.ti.com/dlp-chip/getting-started.html

[2] https://en.wikipedia.org/wiki/Digital_Light_Processing

[3] http://www.ti.com/dlp-chip/automotive/overview.html

I noticed on the TI DLP product web page for automotive [1] that heads up displays are another intended use. I've driven my daughter's car and really like color heads up display in her vehicle--It took a few minutes to get used. The ability to see the essential GPS information floating ahead in my field of view was far less distracting than having to look down at the instrument panel to read it.

[1] http://www.ti.com/dlp-chip/automotive/applications/applicati...

This sounds expensive and unreliable and will just result in very expensive headlight assemblies that have to be replaced entirely. Also inability to get the parts to fix them a decade down the line.

This is already a problem with some vehicles with relatively simple lamp assemblies.

They will certainly be more complex in the sense that software will control the areas in the drivers view to be illuminated, but headlight assemblies that respond to steering position and car pitch angle are already expensive and burdened with electromechanical mechanisms to direct the light. They make a large difference in safety though.

TI DLP will allow this to happen without all of the electromechanical mechanisms. The DLP technology has been around for 20 years.

> TI DLP will allow this to happen without all of the electromechanical mechanisms.

Do we not count a DLP as an electromechanical device?

Well, I did hesitate when I was writing that statement, and technically you are right that microscopic mirrors positioned by electrostatic forces are electromechanical. :)

My original point, and what my complete sentence said, is that the this one large solid state chip is able to serve as in advanced automotive lighting functions without all of the (currently used) electromechanical mechanisms. It really is quite different and has unique advantages. Here's a YouTube video that explains the operation of a DLP:

https://www.youtube.com/watch?v=9nb8mM3uEIc

DLP projectors have been around for decades and have proven reliable.

Yes but do you stick them to the front of your car?

I would do this in a heartbeat if it weren’t so expensive, just for the magical day when I can print “YOUR LEFT BRAKE LIGHT IS BROKEN” on someone’s rear view window (and then get pulled over while the cops try to find a law against it).

DLPs are sensitive enough for shock and alignment issues when in a projector - how are they planning to do this with a car?!

Great, now you can pay for your car to project ads onto the road!

I imagine instead of making headlights easier to replace it's going to skyrocket in price and run off of closed source software. Why would this be a good addition for cars?

These headlights will seldom require replacement during the life of the vehicle.

I can see these headlights going for an easy $500 via non-standard proprietary tech, let alone imagining what service labor will charge...because you know they're going to pull some bullshit like, "due to inherent system complexity, a factory certified technician is required to ensure proper service and calibration".

The keyword seldom is a red herring in that if you have to replace them even once--at $500/ea material + who knows how much in labor--you would have exceeded the lifecycle maintenance cost of a traditional HID setup by roughly an order of magnitude.

Furthmore, headlight housing these days are largely manufactured using polymers which are susceptible to "fogging up" over time, so replacement may not necessarily be driven by DLP failure mode. However, given DLP is essentially MEMS tech and target platforms are subjected to significant vibration and wide operating temperature gamut, I question how MTBF of these chips will be impacted vs. their traditional fixed consumer application over commercial temperature ranges.

This isn't about performance. It's about suckering a market into gold plating a system to achieve IP lockdown--solving a problem that simply doesn't exist. OEMs are already refusing to sell factory service manuals to vehicle owners. This is yet another step in the wrong direction towards a completely unmaintainable vehicle.

I had the same thought - headlights are a very weak link in a minor accident such as a collision with a deer or low-speed rear ending. If the cost of replacement goes from $100 to $1000, that's a big deal!

Exactly - the adaptive headlights on my 2014 grand cherokee at $1200 a pop greatly contributed to the decision to total the vehicle in a very minor collision when it was only 3 years old... (not to mention the $3000 radar sensor for adaptive cruise.)

I'm currently in the annoying situation where I need to replace the "Adaptive" LED headlights on my BMW, these units cost ~£3000 which is significantly more than the alternative (bulbs)....

> BMW

You signed up for this.

why would you want a specific shape to be projected by your headlights? Having a sharp level of light beam defined, I agree.

Why would you want shape? You want everything to be lit up in front of the vehicle. This would hold true even for autonomous vehicles, where you blast the area in front of you with some wavelength of light.

I understand your comment as the details of the usefulness are buried pretty deep. I had to dig down to the 2nd half of sentence #2 to get the answer: "while minimizing the glare to oncoming traffic or reflections from retroreflective traffic signs."

>reflections from retroreflective traffic signs.

So you mask out the stop sign in front of you, and that diminishes the visibility of the sign. Or the light masks the stop sign and lowers the intensity of light within that octagon?

Then you have a camera from the perspective of each headlight, then what? You flood the scene with IR light? I guess it could work - It will require a lot of supporting technologies that TI is probably not in the position to develop.

> TI is probably not in the position to develop.

Once again, they've buried the details way deep down in this article, nearly impossible to find. In this outlandish case, I had to reach as far down as the 1st half of the second sentence which subtly discloses (emphasis mine):

"Automakers and Tier-1 suppliers can use this new programmable ADB solution to design headlight systems"

Redisplay the stop sign in the (TI DLP) HUD.

Outline pedestrians, similarly.

Give me a distance, stopping reaction countdown, and alert me to children doubly so lest they suddenly break from adults. I would think that indication that pedestrians are facing towards or back to traffic is important for me to get quickly as well.

This kind of information would greatly reduce my fatigue when driving long journeys through England's west country and any area with narrow twisting, often tree covered, roads.

>I understand your comment as the details of the usefulness are buried pretty deep. I had to dig down to the 2nd half of sentence #2 to get the answer

Now, now...

https://news.ycombinator.com/newsguidelines.html

Please don't insinuate that someone hasn't read an article.

One obvious application, you want to poke holes in the image where faces (or other optical sensors) are to avoid reducing the visibility of others. Also communicating pedestrian locations to the driver.

However you need to see pedestrians, other vehicles as clearly as possible.

You don't need to blind other road users. That's a bug with the headlight design, not a feature.

Lets start with the fact that you already have a specific shape in your headlights. Typically, driver and passenger side headlights have different reflector and beam shapes and are aimed differently.

Think about the need you have for each headlight, they aren't really the same. The outside (shoulder side/passenger side) realistically has a greater need for 'spread' to light the side of the road and see things that could potentially quickly intrude on your space. It also is less likely to blind oncoming traffic because it is pointed slightly outward so the beam can be higher and broader. The inside (drivers side) headlight has pretty much the opposite needs. It is still important to an extent to light the other shoulder but you also want to avoid blinding oncoming drivers and will probably use that to illuminate further down the road. With this you could, theoretically, do something like create a hole in the beam that is 'off' to not blind oncoming drivers and tracks their location while leaving the headlights at a higher and safer level that increases driver vision.

Add in that as you speed up and slow down and corner as you drive in both directions. That is why many cars already have headlights that shift as you turn, this is just the next logical step

@post_break posted this link which is a good explanation (try to ignore the overly dramatic music and narrator's voice):

https://www.youtube.com/watch?v=wSI-NVD1who

Reply on news.ycombinator.com