Is Tesla potentially putting the public at risk with it’s cost-cutting strategy?

I’ve made no attempt to hide the fact that I’m not a fan of Tesla. It was never about the technology, but rather their business practices and the way they treat their customers. However, that may be about to change with what I would consider to be a penny wise, but pound foolish decision to remove radar and ultrasonic sensors from it’s vehicles.

In a report from Electrek, Tesla is basically saying that they’re going all-in on computer vision and dropping other sensors with the laughable justification that it’s how humans navigate. Considering the potential consequences, perhaps laughable isn’t the right word.

One of the features that impressed me the most about Tesla’s early self-driving feature was the integration of radar data. A perfect example of this can be found in this 2016 video of a Tesla detecting a vehicle several cars ahead that was moving slowly/braking and was able to avoid an accident. You can’t see the vehicle ahead as it’s completely obscured, but the Tesla radar could. Can a Tesla camera system do the same? Maybe, but I’m not sure I’d want to trust an autopilot system without radar. (or some other form of beyond visual spectrum object detection)

Tesla’s track record with self-driving cars hasn’t exactly been fool-proof either. (even with the additional sensors) I’ve been in Teslas that have had difficulty in daylight with the painted lines of the road. There have also been numerous incidents reported where autopilot didn’t see a crossing/oncoming vehicle resulting in fatalities. Beta testing software features is fine for the consumer smartphone and computer industries, but it shouldn’t be something that happens in cars on public roads with safety-critical systems.

Tesla does state in their announcement that “at this time, [they] do not plan to remove the functionality of ultrasonic sensors in [their] existing fleet.” Hopefully the same is true for the radar sensors as well. This appears to be targeted at new production models only with models 3 and Y, and later the S and X.

This Reuters article from early June explains some of the details of why a camera-only system could be less safe than one with additional radar sensors. Tesla may think they’re saving (making) money removing the sensors and making development easier for themselves, but not tackling the more difficult problem of integrating different sensors leaves gaps that may put people’s lives at risk.

Update 12/2022: Drivers of new vision only Tesla models already running into problems, some features not yet available

Update 12/16/2022: Tesla appears to be reversing course and bringing radar back based on FCC filing. Elektrek quoted the same tesla filing indicating Tesla would be marketing a new HD radar in mid-January of 2023. This is good news for potential new buyers and potentially for improvements in FSD, but what about those who bought the stripped down versions? If I were one of those customers, I wouldn’t be too happy.

Hey Microsoft! Do Better.

The title might seem a little over the top, but let me explain. This afternoon, I had one thing I wanted to do. I wanted to sit in my racing sim and practice. A seemingly simple thing, right? The thing is, I hadn’t sat down at my sim in about a month. (life has been a bit busy lately) As I expected, I had some AV updates that needed to be applied… no problem. The batteries on my wireless mouse were giving out… annoying, but again, no problem. The system seemed super sluggish after doing the AV updates, so took a look at oh, yep… a bunch of updates from Microsoft. No problem, I’ll let these install and then I’ll play I thought. Oh, how naive of me.

I should have known something was up earlier, but I let my trust in the update process get the better of me. You see, earlier I’d loaded up a piece of software that controls the haptic feedback on my seat and it had a problem because it couldn’t find the output device. The output device is a small usb audio interface that sends analog signals to an amplifier that drives a transducer, which in turn makes the seat vibrate. This sudden change was a result of driver updates, something that I’d specifically disabled. (for reasons that will soon become apparent) That seemed weird, but the device was still there and I just had to re-select it.

The second thing that should’ve tipped me off was an NVidia driver update coming through Windows Update. Strange I thought… I’m sure I disabled that. (I had previously) Still, I thought… what’s the harm. It probably needs it anyway. What I didn’t see was one of the other update packages contained a driver update for the onboard audio. This had already completed, but the worst part was that in the process of updating, it removed the user interface that allows you to configure things like surround sound. I checked the properties for Sonic Studio 3 and found that the application was missing, but there was a handy menu to install it. Awesome! (it wasn’t awesome)

Clicking this menu brought me immediately to the listing for this application on the Microsoft store, but when I clicked it the download immediately failed and told me to log in. I tried to do that and was met with a progress wheel that timed out and failed. This happens when you use a local account with Windows and don’t log in to Microsoft’s services. Ok, annoying, but I’ll just go to the Asus ROG driver page and download them I thought. What I got from the manufacturer support site was a driver-only download. Cue a bunch of searching through forums until I find a link to a slightly older version of the same driver package with the entire suite. (also hosted on the manufacturer site)

What happened became apparent at this point. Asus has moved current & future versions of this application/driver package onto the Microsoft store and removed it from their own support page. This has the effect of tying continuing support of their products to Microsoft’s services. It also revealed that Microsoft is continuing to reset certain settings on your computer to their factory defaults without notice or warning in what are supposed to be security & quality updates. Microsoft, if you’re listening… this is how you erode user trust.

While I’d managed to turn off driver & application updates via Windows Update (again) and had fixed the sound issues… I was done. This whole debacle cost me over an hour of my time, by which point I was so disgusted I didn’t even want to do the one thing I’d sat down to do in the first place.

Building custom boards for retro PCs

In my last post I went through the process of bringing an old 286/12 back to life. At the end of that post, I left off with the machine unable to boot from it’s own hard disk. This meant I would need to find another way to allow an old machine like this to boot from more modern media. Thankfully I found a solution in an open source hardware kit called the XT-IDE. The XT-IDE kit is a very simple 8-bit ISA hard disk controller for IDE drives that coupled with some firmware allows XT/AT & clone computers to boot from larger drives. Since this board is only a hard disk controller, I would also need a floppy controller. (more on that later) I ordered up a kit and when it arrived, I immediately tore into the package and built it up.

This being the 4th revision of the board, it was an incredibly well polished design. I did actually have to add some pin headers for the jumpers which were not included with the kit, so just keep that in mind if you build one of these. I read over the instructions for assembly just to be sure, but the board has such a nice silkscreen it almost isn’t necessary. Assembly went relatively quickly with the board completed in about 1.5hrs.

With the board assembly complete, I put it under a crafting lamp with a magnifying glass and inspected all of the solder joints to look for any bridges and touched up anything that didn’t look quite good enough. After that, it was time for the moment of truth. I plugged it into the 286, connected up the hard disk and it actually booted into DOS from the hard drive! For this test, I had actually kept the multi-i/o board installed, but with the IDE section disabled. (just in case I needed the floppy)

Speaking of floppy disks… I didn’t want to keep the multi-i/o controller in the 286 as I planned to return it to the 486SLC where it belongs. (and it’s my only surviving multi-i/o board of this type) Having had so much success with the XT-IDE and given the prices of original floppy controllers from this era, I decided to see if there was another open source controller option for floppy disks. Sure enough, I found that someone named Sergey Kiselev had designed just such a board. I ordered a complete kit from eBay and built it up with the same enthusiasm as the XT-IDE.

This kit was based primarily on a pair of PLCC chips and was a more compact design than the XT-IDE. This made assembly a bit more difficult as there were more connections closer together. I initially tacked the parts in place by soldering just 2 connections each and then went back later to fill in all of the rest. The sockets that came with this kit were the cheaper single-wipe type. I don’t mind these for the smaller DIP chips, but I replaced the one for the ROM with a nicer round hole socket that I had on hand. I also made a small mistake you can see above in that I forgot to socket the larger crystal oscillator, so it’s permanently on the board.

Overall, I’m pretty happy with how the Floppy/Serial controller turned out. I did end up having to change the configuration to make this card play nice with the XT-IDE, but thankfully this is well documented online. All of the chips fit in their sockets really well and the board aligns decently with the ISA slots on the motherboard.

I added a Teac 1.2mb floppy drive (also from eBay) to this system for additional authenticity. (and also because my existing Matsushita drive is finicky when it comes to reading 360k floppies) The open source controller works great so far and it performs just as you’d expect any other floppy controller to function. I’m not currently using the built-in serial features as the DTK 286/12 motherboard actually includes onboard serial and parallel ports. It was really nice to be able to read some of my older floppies that were previously unreadable. At some point I’ll have to see if I can source some DS/HD floppy disks to format at 1.2mb as I never really tried them as a kid. I pretty much jumped straight from 360k floppies to 1.44m 3.5in disks.

These were both fun little projects that made my DTK 286 restoration project complete. If you’re in the process of bringing an old XT or AT machine back from the dead and have similar needs, I highly recommend these kits as they’re inexpensive, fun to make and work really well.