this post was submitted on 26 Aug 2023
652 points (93.9% liked)
Technology
59467 readers
4510 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
LOL
Yeah ok.
Tell me you know nothing about manufacturing, without telling me you know nothing about manufacturing.
That one quote - assuming it is accurate - explains that Musk is even more of an idiot than everyone already knew he was. You don't make things at those tight tolerances. A couple of dimensions on a part might be (for instance the bore on a press fit sleeve), but you'd almost never, ever hold an entire part to that tight of a tolerance.
In imperial units, 10 microns is .00039". A human hair is roughly .001 to .005" thick. So he is asking for a tolerance that is 3 to 10x smaller than the thickness of a human hair. To put the absurdity of Musk's demand into perspective, most parts that go into a car are roughly an order of magnitude looser in tolerance with some dimensions being 2 orders of magnitude looser.
That difference might not sound like a lot, but holding something to +/-.0039 versus +/-.00039" could easily triple the price of an item or more. Easy. You use a tight tolerance only when you need to - that's engineering 101. Some parts could easily be +/- .039" and not affect their performance on bit. Close tolerance engine parts might be held at what Musk is demanding, but never "ALL PARTS" would be held to that.
As someone who knows almost nothing about the topic, wouldn't some (most?) of these parts be big enough that a small change in temperature or air pressure alone would cause these parts to expand/shrink enough to go over the tolerance limit?
Yes, and different materials will have different rates of thermal expansion. That's probably why the pixel 7 camera glass was cracking for no apparent reason when winter hit. Imagine coming out in the morning and finding all the glass in your car shattered because it got cold overnight. Or even worse you take it out of a heated garage on a cold day and the glass shatters while you're driving.
Thermal expansion of steel is .0000072" per degree F. All you would need is a 100 degree F temp change to blow that tolerance out of the water. And 100 degrees is not that much when it comes to cars. A freezing cold day versus a boiling hot day in the summer is a temperature swing more than 100 degrees. A ICE engine runs at roughly 250 degrees F so that right there would easily expand parts way more than the tolerance he is calling for. On an EV, I don't think anything gets that hot, but the motors still get warm.
Is that per inch or per foot?
We compensate for thermal expansion. The standard temperature things are measured at is 68F/20C. So if it's 72 degrees we'll compensate it back to 68 in software for the material we're measuring. We use scale bars of known length and similar material type to verify scale. (I run laser trackers and laser radars.)
For measurement equipment that's stationary, like CMMs, you just control the environment.
I work in semiconductor industry where machines need to have sub-micron positioning accuracy and even we don't generally design parts with 10 micron tolerances, unless it really needs to.
No customer would ever pay for that accuracy on all parts.
No customer would NEED that accuracy on all parts. Just shows what kind of clown Elon is.
When I was young I got a job at a manufacturing place that made all sorts of parts for sensitive equipment. Younger people, or people with steady hands would debur and smooth. We would have these huge magnifiers and friggin microscopes and be working with what looked like a really long tiny exacto knives that needed to be replaced every 5 minutes or a couple dozen uses to get that stuff to spec. You can spend 20 minutes on a piece, think it's perfect and then QC would send it right back because they somehow found some tiny inconsistency or groove you didn't or couldn't notice.
There is no way you can expect that level of accuracy, unless your willing to pay for clean room level stuff. Even we weren't always quite that accurate depending on the end use and they charged like almost $50 for something that looked quite like something you get a hardware store for 50 cents.
I've told many (usually new) design engineers that they're stupid for asking for 0.001" tolerance on parts when they only need 0.005 "or 0.010". The difference between 0.010" and sub-10 micron is easily a factor of 100 in most parts, ESPECIALLY when you're talking larger steel components like panels on a freaking car.
not just that, he said "all parts". The stitching on the seats, the floor carpets, USB ports, cupholders and the A/C vents have to be more accurate than the width of a human hair too
Unless you are talking about a press fit location or some kind of high precision alignment issue, almost nothing needs anything tighter than .001" and .005 or .010" is perfectly fine for most things. I work with a lot of weldments so if we're within .030" we consider that good enough.
+/- .030 good enough for most skin panels on airplanes.
Gear teeth I believe need to be precision ground to less than a thou for best efficiency and life. But that's not for most things.
hah historically Teslas body panels are closer to +/- .39"
I don't see anywhere in the article where Musk says "tolerance". He specifically says "accuracy" and goes on the talk about listing more decimal places instead of rounding. Any mention of tolerance is done by the author of the article. If certain dimensions are not naturally rounded to one, or even two decimals, there is no reason not to list it to three or more on modern drawings. GD&T can specify whatever tolerance is necessary without relying on a decimal-based block tolerance. I'd be interested in seeing the original email but it seems like there is a misunderstanding by the author given the context being discussed.
I default to three decimal places for all my basic dimensions on both in and mm drawings. One of the benefits of GD&T is that you can give provide additional dimensional accuracy, completely independent of the tolerance being specified.
That's a really good point. If this is actually for the intervening calculations, that'd make a lot more sense.
The linguistics of metrology is not exactly a topic I'm particularly passionate about, but if yiu look at the technical definition of accuracy, it essentially is the same as a tolerance.
Accuracy: the degree to which the result of a measurement, calculation, or specification conforms to the correct value or a standard.
And when it comes to decimal places, you'd never display more than you really need. If a dimension is +/- .010" there is absolutely no reason to display it to 4 decimal places. That doesn't win you anything. More importantly, I'm sure a company like Tesla doesn't even use drawings at all. I'm sure they are paperless and send out their models for machining.
I think there is a very important distinction between accuracy and tolerance in engineering. +/- .010" is not a dimension, but a tolerance that can be applied to a dimension. However if your example was changed to a .010" dimension, I would agree with you as I stated in my last comment. There is no need to give any further accuracy to that dimension if you are just adding zeros to the end (unless you are using block tolerances that rely on a specific number of digits to correspond with a standard tolerance). Unfortunately, not everything is designed using the same units and you will inevitably end up with a part designed in mm that uses a bolt-on component using a hole span in inches (for example, a nice round 1-in span). If you want a +/-1 mm tolerance on that part, you wouldn't want to round every dimension to the nearest mm because you may end up with a tolerance of 24-26 mm when you really wanted 24.4 to 26.4 mm. I like to provide true dimensional accuracy (to microns or .0001" if I'm not just adding zeros) and then apply a suitable tolerance independently, using GD&T.
Regarding paperless manufacturing, I agree that many components are made straight from the models these days and imported directly into a CNC machine. However, there should always be a drawing or a digital equivalent a drawing. This is the contract that specifies acceptable tolerances to the manufacturer, and it will be used during QA inspection to determine if an acceptable part has been delivered.
I think there is an important distinction between accuracy and precision in engineering. I'm having flashbacks of sitting in class when the professor was going over this stuff. I honestly always found it some of the most boring topics in the curriculum.
One of my physics profs had a story about this. He needed two resistors to be very similar in performance for a circuit he was making, so he asked for a couple of the super-high-quality ones.
His advisor said "fuck that, get the 1% bin, they'll be bimodal at 1% above and below rating, sort em and find two that match to the degree you need"
That's kind of analogous; do you need to try to hit a particular value (accuracy) or do you need things in a consistent relation to one another (precision)?
Just coming in to say you are completely and absolutely correct. 🍻
Engine parts? Tesla's don't have engines, they have electric motors which shouldn't need this level of precision. Electric motors they have today work pretty well already.