Motortrend Piece on CT issues

[kbolt]

Zero worries, my guy - clearly, *I'm* at least equally triggered.

You're one for whom I can see past the brevity and occasional over-correction to, at bottom, an informed and at balanced degree of curiosity, owing a generous read.

And whom in kind, I also see often affording me the same allowances.

Are you an attorney? Your post reads like you're writing a contract trying to cover as many things as possible. On a forum you can be concise and specific.

Sponsored

 

[cvalue13]

On a forum you can be concise and specific.

Thankfully for you, all my posts are optional reading!

[scrolling works by flicking just your finger on the screen / mouse]

 

[CyberGus]

I was thinking the same. since 5 ms = 0.005 seconds.

• A 0.005 is near-to an electronic switch (I/O) response only,
• There is no such thing as a mechanical switch what-so-ever with that low a ratio response.
• A 0.05 = (or) 1/20th of 1 second, sounds excellent, even for steer by wire.
  • Normally such would be near impossible for anyone other than Tesla.

To a computer, 5ms is an eternity. Here's me pinging my network gateway:

# ping -c5 192.168.1.254​

64 bytes from 192.168.1.254: icmp_seq=0 ttl=64 time=6.665 ms​

64 bytes from 192.168.1.254: icmp_seq=1 ttl=64 time=5.744 ms​

64 bytes from 192.168.1.254: icmp_seq=2 ttl=64 time=7.517 ms​

64 bytes from 192.168.1.254: icmp_seq=3 ttl=64 time=4.041 ms​

64 bytes from 192.168.1.254: icmp_seq=4 ttl=64 time=3.539 ms​

That's taking a packet all the way up the network stack, transmitted wirelessly to the gateway, processed and returned by the gateway, and then received again, all in around 5ms.

Trying a local flood ping, I can average 0.02ms (20 microseconds!) per packet.

root# ping -f -c999 localhost​

999 packets transmitted, 999 packets received, 0.0% packet loss​

round-trip min/avg/max/stddev = 0.007/0.021/0.227/0.019 ms​

So really, the delay between steering input and steering action is induced by the code. With proper optimization, 5ms is an attainable goal.

 

[CyberGus]

Are you an attorney? Your post reads like you're writing a contract trying to cover as many things as possible. On a forum you can be concise and specific.

brevity=wit[soul]

 

[Arctic_White]

No, I haven't owned a Tesla before. For what little it may be worth to you, I have, however, been a close observer of the company and it's products for a long time. I've not owned a Tesla before only because I, for whatever silly reason, have for 30 years driven only full-sized pickup trucks as my daily driver (ignoring my side piece toys).

That said, even if you think only Tesla owners have authority to understand *something* (unclear what exactly, I fail to see what those things have to do with my comments?

Because my comments were about, at bottom, basics of critical thinking/reading. Agnostic to particulars about ownership of a particular vehicle or the company.

But as for particulars about ownership of a particular vehicle or the company, I explicitly invited *you* (the royal 'you') to " I mean, find their names and discredit their expertise, or whatever"

And it does seem that your only report relates to some unclear assertion that the people purported to be experts in the article aren't in fact experts.

However, bizarrely framed as instead telling me that *I'm* not an expert, and in virtue of having not owned a Tesla vehicle?

Sort of, ironically, making exactly my point by example.

For someone who claims you've been a close observer (but not quite an owner nor a TSLA shareholder), you'd know that there are daily hit pieces on Elon, right?

And for someone who claims to have been a close observer, you will note that Tesla's culture is amazing at innovation. Just wait until the Cybertruck comes out. If you are not blown out of your mind, we can then have a discussion. LOL.

 

[PilotPete]

So really, the delay between steering input and steering action is induced by the code. With proper optimization, 5ms is an attainable goal.

The delay time in FBW/SBW is more than just a single code thread. When you add the feedback loop, you add in all the sensors and the servos and their ability to react. And then remember, every 11.7 inches of wire adds a nanosecond! Here is a generic feedback loop for FBW

With that being said, the best hydraulic power steering system on a sports car can’t meet that assistance speed, the valves and pressure wave don’t move that fast. But if that’s their target, expect some dang crisp steering.

You can see in the video (towards the end) where the system delay (intentionally built in) creates what is called PIO, or Pilot Induced OcIllation. It’s caused by either an excessive delay in the system, or gain is too high, meaning the pilot thinks he is moving the control X amount, but the system moves it twice as much, As a result, you create a system where there is just one over correction after another.

 

[CyberGus]

With that being said, the best hydraulic power steering system on a sports car can’t meet that assistance speed, the valves and pressure wave don’t move that fast. But if that’s their target, expect some dang crisp steering.

AFAIK, Tesla abandoned hydraulic power steering long ago for an electric motor, which should be very responsive.

 

[PilotPete]

AFAIK, Tesla abandoned hydraulic power steering long ago for an electric motor, which should be very responsive.

So did Porsche. They went “Electric Assist”, which is very different than SBW. They did it to keep the steering crisp and reduce the engine load of a hydraulic pump.

 

[Gojuryu]

To a computer, 5ms is an eternity. Here's me pinging my network gateway:

# ping -c5 192.168.1.254​

64 bytes from 192.168.1.254: icmp_seq=0 ttl=64 time=6.665 ms​

64 bytes from 192.168.1.254: icmp_seq=1 ttl=64 time=5.744 ms​

64 bytes from 192.168.1.254: icmp_seq=2 ttl=64 time=7.517 ms​

64 bytes from 192.168.1.254: icmp_seq=3 ttl=64 time=4.041 ms​

64 bytes from 192.168.1.254: icmp_seq=4 ttl=64 time=3.539 ms​

That's taking a packet all the way up the network stack, transmitted wirelessly to the gateway, processed and returned by the gateway, and then received again, all in around 5ms.

Trying a local flood ping, I can average 0.02ms (20 microseconds!) per packet.

root# ping -f -c999 localhost​

999 packets transmitted, 999 packets received, 0.0% packet loss​

round-trip min/avg/max/stddev = 0.007/0.021/0.227/0.019 ms​

So really, the delay between steering input and steering action is induced by the code. With proper optimization, 5ms is an attainable goal.

Attainable to mechanical action (triggered response at the wheel) or attainable at the compute level.... Drastically different time sets.

I believe it's better to query the writer of the original article to clarify their point, regarding what is or was an actionable response and where or at what level they set their expectations.

 

[cvalue13]

For someone who claims you've been a close observer (but not quite an owner nor a TSLA shareholder), you'd know that there are daily hit pieces on Elon, right?

sure

there are also other pieces that are fair critiques

none of which appears relevant to the discussion at hand

except if you are suggesting that the existence of hit pieces on Musk requires the conclusion that all pieces on either Musk or Tesla are by definition hit pieces

but such a bizarre conclusion would be reached only by someone who isn’t an observer, but instead a level-4 fanboi whose judgment can’t be relied upon for the exact same reason that “hit pieces” can’t be

rest of your message seems non sequitur

 

[ricinro]

Wow! So you are saying Telsa's target is to match the control surface response time of modern aircraft controls?

What is the speed of mechanical steering? At some latency a driver will perceive a loss of steering. Much like the servo needs of Optimus precision requires fast encoders and a dedicated control loop so human input, much slower, would not notice a lag.

 

[Diehard]

This review of the Lexus steer-by-wire shows a rather substantial delay between steering input and the wheels. It's completely unacceptable IMHO. Hell, if I had a 50ms lag in Counter-Strike, I'd ragequit

I wonder if some of that delay is intentional to reduce the chance of flipping the car if a bee get trapped in the car with you. I definitely don't like any delay.

This whole thing makes me wonder if FSD reaction time (from the time light from a falling tree branch hit the sensor to the wheel turning) is less than human reaction time (250 ms eyeball to hand plus mechanical transmission steering to the wheel).

 

[CyberGus]

I wonder if some of that delay is intentional to reduce the chance of flipping the car if a bee get trapped in the car with you. I definitely don't like any delay.

This whole thing makes me wonder if FSD reaction time (from the time light from a falling tree branch hit the sensor to the wheel turning) is less than human reaction time (250 ms eyeball to hand plus mechanical transmission steering to the wheel).

 

[Kahpernicus]

I wonder if some of that delay is intentional to reduce the chance of flipping the car if a bee get trapped in the car with you. I definitely don't like any delay.

This whole thing makes me wonder if FSD reaction time (from the time light from a falling tree branch hit the sensor to the wheel turning) is less than human reaction time (250 ms eyeball to hand plus mechanical transmission steering to the wheel).

There is a safety feature for over correction at high speeds built in.

That's different from low sleep latency that still seems too laggy.

 

[PilotPete]

There is a safety feature for over correction at high speeds built in.

That's different from low sleep latency that still seems too laggy.

Typically, there is a gain variable that changes based on speed. It’s not a delay (nor did you way it was). Same thing on planes. You don’t want the same control deflections at 360kts that you get at 120kts. You would overstress the plane real dang quick (and tick off quite a few passengers!). And if you used the 360kt gain values, the thing would be virtually impossible to fly in the smallest of winds or wind shift during takeoff and landing.

They also (typically) provide “firmer” controls as the speed increases, to artificially give you the natural feeling of direct connections/

Sponsored