4 ms·

And since the car is most often better than you at knowing when you should shift gears, why have the manual transmission in the first place? From my empirical

by Annatar 10y ago

And since the car is most often better than you at knowing when you should shift gears, why have the manual transmission in the first place? From my empirical experiments, and from understanding how modern automatic transmissions work, the assertion above is incorrect, and here is why: a modern automatic transmission is controlled by firmware, a unikernel. The shift points and the shifting logic are based on the developers' very dim understanding of economy and of performance driving, and because the firmware is so shitty, there is often a significant latency involved - by the time the unikernel has managed to process all the inputs from all the sensors, compute a decision and acutally shift the gear, I've long since shifted with a manual. On top of that, the developers' opinions of what's acceptable and what makes sense are truly dim and utterly infuriating: as I drove the manumatic in hypermiling mode in which I normally drive a stick, the idiotic unikernel refused to upshift because the programmer thought that the gear would be too high and the RPM's too low, neither of which was the case. This was a 3 liter V6, so with plenty of low-end torque for the situation I was putting it through. Incensed, I repeated my experiment on other makes and models: the result was always the same. After so many years of automatic frustrations, I will never again own one: the day when I'm no longer able to buy a vehicle with a manual transmission is the day I take a bicycle or walk with the two legs nature gave me, or engineer my own car, and the car manufacturers can go to hell, because they sure won't be seeing any of my money ever again.

eveningcoffee 10y ago
Yeah, downvoted for honest and informed opinion. Unfortunately this is quite common in HN.
- viraptor 10y ago
  I downvoted you for random assumptions and insulting people you never talked to. > a modern automatic transmission is controlled by firmware, a unikernel Have you seen the source? If not, how do you know it's a unikernel? > The shift points and the shifting logic are based on the developers' very dim understanding of economy and of performance driving "dim understanding"? Do you have a reference for any of that? Why do you think it's developers, not mechanical engineers or engine testing that resulted in the shift threshold? > there is often a significant latency involved - by the time the unikernel has managed to process all the inputs from all the sensors Again, that's a heavy assumption unless you have a source for it. Also, from own experience in actually writing control software like that, real world sensors can be read tens/hundreds+ of times per second. But even with tiny 8K microprocessors you can process it more quickly. Mostly you sit in a tight loop and wait for new data. Processing it takes tiny fraction of the wait time. > On top of that, the developers' opinions of what's acceptable and what makes sense are truly dim and utterly infuriating Yeah... you're just throwing ad-hominem around. Please come post some sources which support your opinions.
  - Annatar 10y ago
    I have noticed systemic abuse of ad hominem labeling anyone who doesn't lavish high praise and criticizes instead. Please show some emapthy and think about how the code could be improved instead, so these criticisms can be addressed with a practical solution. In this particular case, all my criticisms can be addressed by offering a manual transmission option with a dry clutch, not even code optimization is required. Also, as a firmware writer, you know that all such source code is under strict non-disclosure agreements, and whatever I know or have seen, I wouldn't be at liberty to discuss or cite, or else I risk legal penalties. We're neither anonymous nor protected here. As for unikernel, any firmware or subsystem concerned with one task - such as running a vehicle - is by definition a unikernel - all firmware is, as its running one single monolithic application image. Developers versus mechanical engineers - in this context a mechanical engineer is also a developer, just of the transmission, and they often write the software themselves, too. Finally, as a firmware writer, please explain to me why there is a lag between depressing an accelerator pedal in a 2011 Volkswagen Golf, why the engine will often konk out unless the accelerator pedal is pressed fully when starting on a hill, and why a 2012 Buick LaCrosse will display "shift denied" on the dashboard when I try to upshift from 5th to 6th going at 1,400 RPM, which would put me at 1,200 RPM, well within the engine's ability to continue running efficiently?
    - striking 10y ago
      > lag between depressing an accelerator pedal Well, doesn't fuel flow need to be increased before acceleration can begin? That causes latency, much more than software would. Perhaps the software is causing a little bit of that, perhaps there's a simple delay loop instead of a sensor there and it's a little too long. We'll never know, but there are plenty of guesses you could make. > why the engine will often konk out unless the accelerator pedal is pressed fully when starting on a hill That doesn't at all sound like a software problem. It sounds like you're conflating "I don't like my car" with "my car software sucks". > upshift from 5th to 6th [...] 1,400 RPM [...] 1,200 RPM Are you certain that upshifting like that would only drop you 200 RPM? Not at least 1000? I'm fairly certain that's not correct. > [...] all my criticisms can be addressed by offering a manual transmission option with a dry clutch [...] Just buy a manual car if you care so much. Or buy a journal and vent into that rather than the HN comment section. Goodness gracious.
henrikschroder 10y ago
I didn't downvote you, but you are on Hacker News, and you are complaining about bad software, and making the claim that said software will never improve, despite car manufacturers having extremely high motivations - to the tune of billions of dollars of market share - to improve this software, to meet stricter fuel efficiency laws and safety laws.
- Nullabillity 10y ago
  Solving good automatic transmission in isolation is basically solving general AI. And then you have the opposite problem of constantly trying to predict how the black box will behave. Much better to have the driver (regardless of human or software) control both transmission and the other concerns.
  - ams6110 10y ago
    Automatics will never shift well until they can consider the road ahead. Upcoming curve, slope up or down, heavy traffic, etc. all affect a driver's decision about when to shift and which gear to select. So you basically need the sensors a self-driving car would have, and at that point, manual vs. automatic transmission is sort of irrelevant.

by Annatar 10y ago

eveningcoffee 10y ago

Yeah, downvoted for honest and informed opinion. Unfortunately this is quite common in HN.

viraptor 10y ago
I downvoted you for random assumptions and insulting people you never talked to. > a modern automatic transmission is controlled by firmware, a unikernel Have you seen the source? If not, how do you know it's a unikernel? > The shift points and the shifting logic are based on the developers' very dim understanding of economy and of performance driving "dim understanding"? Do you have a reference for any of that? Why do you think it's developers, not mechanical engineers or engine testing that resulted in the shift threshold? > there is often a significant latency involved - by the time the unikernel has managed to process all the inputs from all the sensors Again, that's a heavy assumption unless you have a source for it. Also, from own experience in actually writing control software like that, real world sensors can be read tens/hundreds+ of times per second. But even with tiny 8K microprocessors you can process it more quickly. Mostly you sit in a tight loop and wait for new data. Processing it takes tiny fraction of the wait time. > On top of that, the developers' opinions of what's acceptable and what makes sense are truly dim and utterly infuriating Yeah... you're just throwing ad-hominem around. Please come post some sources which support your opinions.
- Annatar 10y ago
  I have noticed systemic abuse of ad hominem labeling anyone who doesn't lavish high praise and criticizes instead. Please show some emapthy and think about how the code could be improved instead, so these criticisms can be addressed with a practical solution. In this particular case, all my criticisms can be addressed by offering a manual transmission option with a dry clutch, not even code optimization is required. Also, as a firmware writer, you know that all such source code is under strict non-disclosure agreements, and whatever I know or have seen, I wouldn't be at liberty to discuss or cite, or else I risk legal penalties. We're neither anonymous nor protected here. As for unikernel, any firmware or subsystem concerned with one task - such as running a vehicle - is by definition a unikernel - all firmware is, as its running one single monolithic application image. Developers versus mechanical engineers - in this context a mechanical engineer is also a developer, just of the transmission, and they often write the software themselves, too. Finally, as a firmware writer, please explain to me why there is a lag between depressing an accelerator pedal in a 2011 Volkswagen Golf, why the engine will often konk out unless the accelerator pedal is pressed fully when starting on a hill, and why a 2012 Buick LaCrosse will display "shift denied" on the dashboard when I try to upshift from 5th to 6th going at 1,400 RPM, which would put me at 1,200 RPM, well within the engine's ability to continue running efficiently?
  - striking 10y ago
    > lag between depressing an accelerator pedal Well, doesn't fuel flow need to be increased before acceleration can begin? That causes latency, much more than software would. Perhaps the software is causing a little bit of that, perhaps there's a simple delay loop instead of a sensor there and it's a little too long. We'll never know, but there are plenty of guesses you could make. > why the engine will often konk out unless the accelerator pedal is pressed fully when starting on a hill That doesn't at all sound like a software problem. It sounds like you're conflating "I don't like my car" with "my car software sucks". > upshift from 5th to 6th [...] 1,400 RPM [...] 1,200 RPM Are you certain that upshifting like that would only drop you 200 RPM? Not at least 1000? I'm fairly certain that's not correct. > [...] all my criticisms can be addressed by offering a manual transmission option with a dry clutch [...] Just buy a manual car if you care so much. Or buy a journal and vent into that rather than the HN comment section. Goodness gracious.

henrikschroder 10y ago

I didn't downvote you, but you are on Hacker News, and you are complaining about bad software, and making the claim that said software will never improve, despite car manufacturers having extremely high motivations - to the tune of billions of dollars of market share - to improve this software, to meet stricter fuel efficiency laws and safety laws.

Nullabillity 10y ago
Solving good automatic transmission in isolation is basically solving general AI. And then you have the opposite problem of constantly trying to predict how the black box will behave. Much better to have the driver (regardless of human or software) control both transmission and the other concerns.
- ams6110 10y ago
  Automatics will never shift well until they can consider the road ahead. Upcoming curve, slope up or down, heavy traffic, etc. all affect a driver's decision about when to shift and which gear to select. So you basically need the sensors a self-driving car would have, and at that point, manual vs. automatic transmission is sort of irrelevant.