From the outside looking in, this is a common thought to have. But adding manpower to programming projects isn't always helpful.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Elon "About to end range anxiety"
- Thread starter Lump
- Start date
From the outside looking in, this is a common thought to have. But adding manpower to programming projects isn't always helpful.
Citizen-T
Active Member
If Tesla is in the same boat as most other software companies, it isn't that we don't want to hire more people, we just can't find enough qualified candidates.
SteveS0353
Member
The same is true for hardware projects too. Believe it or not, the optimum size for a team seems to be 7 or less. On very large projects, I've been successful partitioning the work to teams of 7 or less, but you have to watch ramp-up time and communication channels between teams. Brooks' Law is real.
HankLloydRight
No Roads
Amen, brother!
- - - Updated - - -
I drove to Boston and back this past weekend, and this happened every time I stopped while using the trip planner. The most egregious example was that I got off the Mass Pike at Exit 10A to head towards the Auburn, MA SC. I was about 2 miles away with about 4% or 5% charge left, and I stopped at a Wendy's to get a drink (since the Auburn Mall was closed at this point). I got back in the car, and it had recalculated that I would arrive at the Auburn SC, which was 2 miles away, with -20% charge left. For a second or 20, I was quite worried that I had miscalculated and for some reason either the battery was lower than I thought, or the SC was further away than the Nav indicated. But I just reset the route to the SC and all was good again. This happened two or three more times over the weekend, and yes, each time the "recalculating" would freeze.
Also, the little battery indicator that says you'll arrive with x%, would be nice if that also told you how many miles you would arrive with.
Andyw2100
Well-Known Member
I'm not questioning the validity of Brooks Law. If all of you programmer types agree that it's real, I'll buy that. But it would seem that with good management much of what Tesla is doing could and should be broken up into many small projects, so Brooks' Law wouldn't really apply. There are so many things that Tesla just doesn't get to, or things that are low on the priority list that could be a project unto themselves with enough capable people on staff.
This is another issue. But if you're willing to throw more money at the problem, I expect that's a problem that could be solved.
Maybe I'm wrong. I'm very much on the sidelines, watching this game, and some of you are players in it. At times it's just a really frustrating game to watch. It's kind of like watching the team you know is the better team not play their best game. That's frustrating.
Citizen-T
Active Member
To be fair. Apple did ship the first iPhone without copy and paste.
I don't know why any particular favorite feature or pet bug is not getting high enough priority, but believe me, and I say this from personal experience: Tesla's software engineers are very much about getting **** done. If it isn't done, there's a good reason.
I don't know why any particular favorite feature or pet bug is not getting high enough priority, but believe me, and I say this from personal experience: Tesla's software engineers are very much about getting **** done. If it isn't done, there's a good reason.
breser
AutoPilot Nostradamus
Brooks law is real but you're right about breaking things up to a point. The current trend in a lot of web systems is to use a microservice architecture. Break every feature up into it's own tiny system that's managed by a separate team and has it's own release schedule.
Of course Tesla is not making a web system and with the exception of a few features, the car has to work without online access. Also each deployment results in an outage for the user (i.e. the 1-2 hours you can't drive your car while software in installing). So Tesla is pushed into a monolithic release where everything ships at once in one update.
That doesn't mean you can't apply this way of working to something like Tesla is doing. You just break up various features into libraries or plugins and then when it's time to do a release you take whatever versions each team has flagged as production ready and bundle them up.
However, it doesn't work as well as it does with web systems. These systems almost certainly have some degree of interaction. So you have to do integration testing between the components. If you find a show stopping bug all the other fixes have to wait for it to be fixed or pull the entire change from the team with the critical bug.
However, that team's release might be required for another team to be able to release their latest version because of dependencies. So then that gets pulled. So then you get releases that are disappointing to end users because it seems like Tesla isn't working on the things we care about.
Even worse if you find a critical bug after release then you have to ship and whole new release with (hopefully) a small fix for that critical change. But in order to do that fix you have to do the integration testing all over again.
All of this gets more complicated as you have more and more teams responsible for different things. So even though you arguably could just add more software teams to handle different aspects. Given Tesla's situation it isn't entirely easy to do it that way.
We really should give these guys a lot of credit. They have brought software updates to a car. To the point that they have pushed updates that have improved the driving dynamics of a car. Nobody else is doing this. So sure they aren't necessarily able to scale their team and deliver as fast as we might like. But they're delivering a hell of a lot more than any other car manufacturer. Because effectively most other manufacturers are delivering nothing (or maybe a release every year that requires a dealer to install). They sure as hell aren't going to put out a release that makes existing cars quicker, they'll just put that on the next model year. This is a remarkable achievement in and of itself.
Andyw2100
Well-Known Member
I definitely do give the software guys already at Tesla a lot of credit, for all the reasons you've so eloquently identified. The gist of my argument was that I just wish there were more of them. From what a lot of you are saying, it sounds like perhaps there are compelling reasons why there aren't.
As I see it they have a big job ahead of them. The built-in navigation clearly has a suboptimal route determination routine, which bases its choices on a simple bigger-highway-is-better heuristic. Google Maps is far superior - it recognizes when the lesser highway is better.
(Actually the built-in navigation is pretty dumb... on my recent trip through upstate NY there was a tiny little region showing heavy traffic - completely bogus - and it worked out a complex route to avoid it. I ignored the directions and drove straight through. As expected there was NO traffic jam there whatsoever. I saved myself a big detour.)
Calculating the optimum route is an NP-complete problem analogous to the infamous traveling salesman problem. A perfect solution is impossible for a computer. There are are of course very good approximate algorithms these days, but it's not trivial stuff. They have to recreate Google's far-better algorithm themselves (or buy it) and implement it in the touchscreen.
Add in to the above the necessity of figuring in Supercharger stops and it's even more complicated.
(Actually the built-in navigation is pretty dumb... on my recent trip through upstate NY there was a tiny little region showing heavy traffic - completely bogus - and it worked out a complex route to avoid it. I ignored the directions and drove straight through. As expected there was NO traffic jam there whatsoever. I saved myself a big detour.)
Calculating the optimum route is an NP-complete problem analogous to the infamous traveling salesman problem. A perfect solution is impossible for a computer. There are are of course very good approximate algorithms these days, but it's not trivial stuff. They have to recreate Google's far-better algorithm themselves (or buy it) and implement it in the touchscreen.
Add in to the above the necessity of figuring in Supercharger stops and it's even more complicated.
I had something less than 90 and strong headwinds.Still made it, but it was crazy close.
I'm guessing you didn't have super strong headwinds, or else you were driving extremely conservatively (like 60 mph). With 200 rated miles leaving Atascadero in my MS 60, in strong headwinds I only have about 20 rated miles when I get to Gilroy, and that was keeping my speed down around 65 driving up the Salinas Valley. Twice I've done this with basically identical results. Figure an MS 85 has maybe 40 miles of actual additional range, with really strong headwinds through the Salinas Valley, I would think you'd be able to get to Palo Alto but SFO would be really pushing it
I don't own an HPWC and have never read the manual. However, twice I've come across 32A ones in the wild.
My impression is that the list is just HPWCs and they multiply the planned amperage by 200V. Power is overstated quite a bit. There are several 40A sites that are posted as 16 kW. On the other hand, there is a 19.2 kW station that only gets credit for 16 kW.
breser
AutoPilot Nostradamus
The current HPWC has jumper positions for the following breaker sizes: 15A, 20A, 30A, 40A, 50A, 80A, 100A.
Older HPWCs had jumper positions for the following breaker sizes: 40A, 50A, 60A, 70A, 80A, 90A, 100A.
The dip switch settings match up as follows between the two versions (breaker amps and bolded lines where the same switch settings result in different amperages):
| Old | New |
| Test Mode | Test Mode |
| 40A | 40A |
| 50A | 50A |
| 60A | 15A |
| 70A | 20A |
| 80A | 80A |
| 90A | 30A |
| 100A | 100A |
Curiously the 4 switches allow for 16 combinations but they only use 8 and as already noted they reused some for the new lower possible values.
Old manual here (there's probably a link on Tesla's site but this is what I found): http://download.waidy.com/EV/HWPCinstallation.pdf
New manual here: https://my.teslamotors.com/sites/de...wall_connector_install_guide_northamerica.pdf
6 kW could be a 240V/24A HPWC rounded up from 5.7kW. Tesla rounds up the 19.2kW on the max amperage of the HPWC to 20kW. So that's what my guess is on what that charger is.
Also I should point out that the car only sees 80% of whatever the breaker setting is. So a 40A setup shows up on the car as 32A. So that probably accounts for some discrepancy between what was discussed by TexasEV and what you've seen. There may be reporting issues of people reporting breaker amps rather than max usable amps accounting for some discrepancy with what is listed.
I always assumed Tesla was stating output power based on maximum rated input voltage of 250V - i.e., 250V * 80A = 20kW
They state in the new HPWC manual that :
"The maximum current for charging the vehicle is 80A or 20 kW.
At 240V, this is 19 kW maximum."
Last edited:
breser
AutoPilot Nostradamus
Yeah I was more referring to the website where (at least in the past) they have rounded up to 20kW. I suppose it's possible they used 250 * 80A but that seems really strange to me. As I understand it 250V is the limit of the equipment. Nobody would actually be getting 250V on a regular basis.
When I'm pumping out the solar (and thus reducing the load on the line from the street/transformer) I can hover in the mid 240 range (245-247) under load while charging. Never have seen 250V though...
I wonder how much the design team for navigation may be hamstrung by having to integrate with the on-board Navigon system in order to provide navigation outside of cell coverage.
I suspect the possibility that the Navigon system cannot accept a "route" as an input, but only a destination. It then calculates it's own route, and the on-board center console is then able to overlay that route on the center console.
In other words, I don't know that it's possible to allow the center console to use Google's routing algorithms (or Waze integration, etc...) and then force the dash to use that for turn-by-turn. This may also be why removing individual superchargers from a route is problematic.
Note that this is all conjecture... but given that Tesla apparently licensed Navigon for the offline/dash/turn-by-turn portion of the system, they may be at the mercy of whatever integration that product allows...
You make a good observation here, Doug. But that's old skool thinking, in a way.
In today's interconnected world, routing can also by done by cloud computing and what I'd call 'the wisdom of the fleet'. Just observe the most succesful routes travelled by the fleet members in the past and update that information with the events of the last few hours and the car's SOC. My guess is that these kinds of algorithms are part of the magic behind Google Maps and Waze. Tesla has the drawback of smaller data sets and less map coverage, being limited to their 'own' cars.
This is why truely integrating Waze is the way to go for Tesla Motors. (Yes, what to do if there is no internet connection? Just follow the road signs, I'd suggest. Where there is no internet, there isn't much routing to be done either.)
E.g. if you try an alternative route for your daily commute (with Waze active) and it turns out better than Waze's suggestion, after a few days Waze will suggest you the better route. Not only to you, but also to others!
My hope is that Tesla Motors doesn't persist in their current, futile efforts due to top brass' ego clash with Google. (If you can't beat them, join them.)
Being a software engineer myself, I would say that many of us would be more than happy to work for Tesla if we didn't already feel a strong sense of commitment to existing projects and/or want to stay put and not have to move. In other words, if I were prepared to make a clean break and Tesla would let me work at home, I would love to do so, but I don't see that happening for multiple reasons.
After our last trip, It's quite clear to me that with the new "Range Anxiety" preventing software, you still have to use your head while on a trip. We stopped at the Ft Myers SC a few days ago on the way home to the East Coast of Florida. It was our first stop at a SC on a trip since the "Anxiety Elimination" Software download. After 17 min charging the touch screen said we had enough juice for the 150 mile Highway trip home. Rated range was at 175 mi. Quite a bit lower than I would have stopped at in the past. But wanted to give it a test so unplugged and got back in the car and immediately got a "Drive Slowly" warning. Plugged back in and charged a few minutes to 185 mi thinking it would be a good test but knowing I might have to drive slower than I normally do to make it. Sure enough, about 15 minutes into the drive the % remaining at destination started dropping and the "Drove Slowly" warning came up. Still no problem as driving right at at the Speed Limit (and being passed by nearly every car on the road) held the % remaining constant at 8%. Then the unexpected (which in S Florida should always be expected I suppose). A very small rain cell over the Everglades passage blew up into a monster. Intense rain and wind and its power sucking by-product forced us to drive as slowly as 45 mph and even at that speed our % remaining after 20 min in the storm was at 6% with 100 mi to drive and 120 rated miles. Much smaller buffer than I would ever want at highway speeds 100 mi from home. Had to drive at or slightly below the speed limit & pulled into garage with 14 mi rated range remaining. Was an interesting test of the new software and it certainly underscored the fact that one should never trust the software alone. Will be taking a much longer trip in a couple of weeks and will most definitely go back to doing my normal calculating and reasoning.
Similar threads
