HN Theater @HNTheaterMonth

A YouTube channel primarily dedicated to precision: https://www.youtube.com/c/machinethinking

Example video: https://www.youtube.com/watch?v=gNRnrn5DE58

I think it's fair to call these mini-documentaries.

"Origins of Precision" Best video from Machine Thinking: https://www.youtube.com/watch?v=gNRnrn5DE58

(Hand) scraping refers to https://www.youtube.com/watch?v=REeGn4hN1Bg .

With two surfaces you can still have some bias (convexity/concavity) even though they appear to be relatively flat, you need three surfaces for absolute flatness. See the Whithworth Method described earlier in this topic.

If you're interested in this sort of stuff and want a video introduction, see: https://www.youtube.com/watch?v=gNRnrn5DE58

Obligatory machine thinking: https://www.youtube.com/watch?v=gNRnrn5DE58

Altough, I disagree with the "random fashion". Alternating the pairs AB -> BC -> CA seems more logical to me.

⬐ etrautmann

or a long-form textbook on the foundations of mechanical accuracy. Someone linked to this a year ago and I found it a fascinating deep dive:

https://pearl-hifi.com/06_Lit_Archive/15_Mfrs_Publications/M...

⬐ reportingsjr

It is necessary to grind in a random fashion (OP is talking about the method of grinding, not the order), or you will end up with imperfections. Lots of info about this if you look in to grinding mirrors for telescopes.

Also, the order being random wouldn't effect the end result.

⬐ platz

When grinding the concave plate why didn't that make then other flat plate convex again?

⬐ jlokier

As the sibling comment explains, it does make the other flat plate convex, but by a smaller amount. Their shapes sort of average out.

I'll add that this depends on the materials having similar properties, so when ground together, they are grinding each other. This means whatever residual shape you have in one of them, it won't be transferred completely to the other.

If the materials had very different hardnesses, one of them would dominate over the other when they're ground together.

⬐ wizzwizz4

It did – but less so than the original was concave. By rubbing them together, you're effectively “averaging out” the convex / concavenesses.

⬐ openandshut

If it's a machine doing the grinding, what is the source of randomness?

⬐ ruined

input

⬐ dylan604

function setPostion(float x=random(), float y=random()){}

⬐ datameta

I imagine the PRNG they use for this is a rather thorough implementation. Unless perhaps it is even necessary to used a cryptographic chip to generate true randomness.

⬐ dylan604

pseudocode my man, pseudocode. ignite the imagination, and let the reader run wild.

however, now my imagination is now running, and thinking about how overkill "true" random might be for this application. something "random" like an orbital sander would probably be enough. you're just trying to get away from side-to-side, left-circle, right-circle, up/down patterns. you're doing this to 3 different surfaces, so they would grind out any slight patterns which is the point. seems like a try crypto random would be as "effecient" as my roomba appears not to be.

⬐ datameta

I suppose I went a bit overboard (no pun intended). 1 KB implementation should be more than enough for their precision.

Maybe some code then, instead: https://www.dwitter.net/d/20446

You might enjoy this video, 'Origins of Precision'. https://youtu.be/gNRnrn5DE58

From what I've seen as an amateur that is yet fascinated by mechanical engineering: iteratively work up to the precision limits with your current methods then try to find/research a (slightly) better way of measuring and/or manufacturing. Since humanity has worked up to this point over centuries/millenia, you wouldn't need to "bootstrap" it from the beginning anymore, just choose the appropriate level of (manufacturing/measuring) precision for your usecase. Otherwise start with a surface plate.

Nice summary:

https://www.youtube.com/watch?v=gNRnrn5DE58 ("Origins of Precision" by "Machine Thinking")

⬐ _pmf_

This is a beautiful video!

How the screw of the wine press was the central part of Gensfleisch/Gutenbergs invention which allowed all the other previously existing ones to finally come together (though I also read that certain advances in metallurgy were also essential). Also see the previous video on screws, a much underappreciated invention.

It then describes how the availability of books made propagation of existing technology and newer developments possible and faster.

The only other comparable moments in history are probably the evolution of spoken language, the invention of writing and most recently the internet.

Also much recommended from that Chanel are the videos on the "Origins of Precision"[1], and "The 1751 Machine that Made Everything"[2].

1: https://www.youtube.com/watch?v=gNRnrn5DE58 2: https://www.youtube.com/watch?v=djB9oK6pkbA

Also see:

https://www.youtube.com/watch?v=gNRnrn5DE58

⬐ DavidPeiffer

Thank you. As an industrial engineer, I geek out about things like this. The video is immensely interesting and well presented.

⬐ Lev1a

> I geek out about things like this

Same.

Though I "blame" my father ("Schiffbauingenieur" ^= "ship/naval engineer") for my interest in engineering in general, my late grandfather for my interest in electrical engineering and my discovery of "This Old Tony" on YouTube for my recent interest in (manual) machining.

> the precision metal tools

I was wondering about how those came about not long ago, and stumbled over this[1] nice video, "Origins of Precision".

[1]: https://www.youtube.com/watch?v=gNRnrn5DE58

⬐ dredmorbius

Excellent video.

See also Simon Winchester's The Perfectionists

https://www.worldcat.org/title/perfectionists-how-precision-...

Indeed, precision was key: https://www.youtube.com/watch?v=gNRnrn5DE58

⬐ jeffreyrogers

There is a really great book called "Foundations of Mechanical Accuracy" that shows in detail (and with a lot of very good photographs/technical illustrations) how extremely precise measuring tools can be built up from a simple flat plate. One of the most interesting things I've read recently.

⬐ mvidal01

It's on Archive.org - https://archive.org/details/FoundationsOfMechanicalAccuracy

⬐ CamperBob2

Archive.org is so weird. Henry Ford died in 1947, so Moving Forward is out of copyright (and consequently available on Project Gutenberg, as another post points out). Yet on archive.org I can only "borrow" an encrypted copy.

Meanwhile, Foundations of Mechanical Accuracy was published in 1970 by Wayne R. Moore, who was presumably still alive at the time, so its copyright status is still very much in force... but archive.org says "Here, have a nicely-OCR'ed .PDF."

I hope they're able to survive their present legal difficulties, but I'll be darned if I can see how.

⬐ thechao

Currently priced at 2349.56$ with 3.99$ in shipping, on Amazon.

⬐ SAI_Peregrinus

Or $150 from Moore. http://mooretool.com/publications.html

⬐ petertodd

While that's an excellent book, one of the things that I find fascinating about it is what it doesn't cover all that well: how do you make an accurate screw thread? The usual way you make a screw thread is by cutting it on a lathe. But the heart of a lathe is an accurate screw thread...

How to making flat plates from scratch with the 3 plate method is relatively easy to understand. But screw threads have a much more complex geometry, with no obvious way of making one from scratch. They also have many more parameters to consider, including relative pitch accuracy, and absolute accuracy.

As far as I can tell achieving accurate screw threads it something that took a huge amount of work by many different people effectively working together to get successively better and better threads, using each others' screw threads made using multiple techniques to average out errors. But I've never actually seen anyone describe from start to finish out to go from nothing to an accurate screw thread. It's not even clear to me that it's possible to do alone in a reasonable amount of time.

Relevant: https://freechaptersinbooks.wordpress.com/2012/09/18/screw-t...

⬐ dylan604

I learned this lesson with the difference in an all-thread rod compared to a lead screw. even though both were at a typical 1/4"-20 pitch, the all-thread had too much slop in it to make for an accurate, repeatable motion. I went down the same rabbit hole of reading about how accurately turn a screw.

⬐ YZF

Slop can be taken out with preload... or always approaching a point from the same direction. Though if this is something like a milling machine then you have forces acting on your nut to deal with. But if all you need is repeatable motion you could do better than your slop.

⬐ dylan604

Lots of ways to ultimately handle slop. Stepper motor vs servo motor makes a big difference. I didn't know if I wanted speed or accuracy when I stared. Also had to make the decision of belt driven vs screw. Ultimately, accuracy was most important, so lead screw with a stepper. It's just a pain to carry around as the lead screw dictates the size of the rig.

⬐ aaronblohowiak

I imagine you can do this through gear reduction -- reducing the ratio of your leadscrew to the tool movement essentially reduces the inaccuracy of your leadscrew, iiuc. the percent error is the same, but you care more about absolute error than percent.

⬐ petertodd

That's a good idea. But I don't think it actually works, at least with the obvious way to do it.

Suppose you have a lathe whose screw thread has errors in pitch such that the position of the middle thread is incorrect. With the right gears, you can use that lathe to cut a second thread with a different pitch. But regardless of what pitch you choose, the middle of that second thread will have the exact same absolute error in that position as the original thread. So you haven't actually improved anything.

The best I can think of to improve the accuracy of a thread is to measure the thread against multiple length standards, and grind/lap away material by hand to bring the actual position along the thread length as close as possible to your length standards. However, as the number of length standards you can practically measure against is limited, you'll need to already be at a point where relative pitch error is small. Averaging out errors with a long follower nut is probably one way to do that.

⬐ aaronblohowiak

Why would the error the same in absolute sense instead of relative to the gear ratio?

⬐ aaronblohowiak

I will write up my idea later today.

⬐ aaronblohowiak

After thinking about it, you’re right!

I also found this: https://en.m.wikisource.org/wiki/Encyclopædia_Britannica,_Ni...

⬐ petertodd

Thanks! Interesting link.

⬐ Cerium

One way I have heard of is starting with a lead screw composed of a rod wrapped in wire.

⬐ jbay808

I believe they said that their screw threads were lapped by hand while being checked by interferometer readings.

An absurdly difficult process, but the result is a full-area bearing screw that never wears out.

⬐ petertodd

Yes, that's how the Moore Tool Company did it. However the first screw-cutting lathe actually dates back all the way to 1800. While as far as I can tell the first use of interferometry for distance measurement was the Michelson interferometer, in 1887; it took until 1960 for the meter to finally be defined in terms of wavelengths of light.

⬐ jbay808

Oh, I see what you mean now; I misunderstood your question!

Here's a screw cutting machine from Da Vinci's notebook:

https://www.ssplprints.com/image/100676/leonardo-da-vinci-sc...

But as you observe, it already has a screw in it. Two screws actually, one on either side, which advance a carriage that cuts a new screw in the center from a blank. It can either duplicate the pitch, or using a gear ratio, it can cut a different pitch than the two master screws.

The central screw averages the error of the two master screws, so with certain assumptions, the new can end up more accurate than either of the masters; the master screws can then be swapped out for new screws made this way, and so on. But there are limits to that approach, because common errors won't get averaged out. However you could, for example, assume that two screws made this way with the same masters are accurate duplicates of each other, then when you replace the masters with those two new screws, you can mount one of them flipped end-over-end, or with a rotational offset of eg. 90 or 180 degrees (offsetting the leadnut to compensate), and gradually average out more errors that way.

Another way as you noted in a different comment is to use a long, soft leather lead-nut to average multiple threads, and use that as a master to create a new thread (assuming a low cutting force -- for example with force amplification, or for checking a thread, and so on).

There are some first principles approaches you could use, like making a master cylinder and then wrapping a wire of constant-diameter around it to make a literal 'thread', and bonding it in place. It wouldn't be the most robust threadform but that's another way.

⬐ jeffreyrogers

Haha, I wondered the exact same thing when reading it. Same thing about the spindle. It talked about how you could measure how accurate a spindle was but not how one was constructed.

⬐ mauvehaus

With a screw origination machine, of course ;-)

Henry Maudslay invented the first one that saw widespread use.

⬐ petertodd

...and notice how his screw-cutting-lathe has an accurate lead screw in it? :)

https://www.ssplprints.com/image/100286/henry-maudslays-orig...

https://www.gracesguide.co.uk/Henry_Maudslay:_Machine_Tools

I've never actually seen a detailed explanation of how he cut that first accurate lead screw! Like I say, as far as I can tell it took an iterative variety of processes that made a variety of screw threads, with no one method alone being enough by itself.

⬐ Accujack

Wrap a triangle around a cylinder. Adjust the angles of the triangle to govern the pitch of the resulting spiral. Use the resulting spiral as a master guide for screw threads being cut on a metal cylinder.

⬐ mauvehaus

The top two pictures of the gracesguide.co.uk page show a screw originating machine, not a screw cutting lathe. And that's the machine you need.

AIUI, it's used to cut a lead screw in relatively soft material by holding a cutting tool to it at an angle. Once you have that, you can generate a screw in harder material from it (carefully, one has to assume).

⬐ petertodd

Ah, I see what you mean. I was aware of those machines. But IIUC they by themselves don't generate accurate enough screws, as the blade doesn't track the material exactly, so errors build up over the length of the screw. Getting those errors out in subsequent generations of screws is the hard part.

⬐ yetihehe

You can average out errors over thread by repeately sliding a long nut over your screw while using lapping compound. Each part of nut will contact only highest (in direction of travel) parts of thread and slightly cut them. You will need to cut off about the same length as nut from both ends of your screw. Nut will have to be made from two parts so that it can be slightly tightened on rod.

⬐ YZF

One observation is that if you can measure your position accurately then you can correct your inaccurate screw to machine a more accurate screw. So an inaccurate, but repeatable lathe, can make a more accurate screw. I know that when you order a ball screw you will typically get a compensation curve with it, which lets you position yourself to better accuracy than the thread on the screw .. though presumably that's not enough to machine a better screw, otherwise they'd do that ;)

⬐ petertodd

The trick there is figuring out how to actually do that compensation in the pre-CNC world. I've seen references to cams used in certain screw-cutting devices. But not a detailed description of how exactly those cams worked or how people were actually measuring the error in the first place back then.

> though presumably that's not enough to machine a better screw, otherwise they'd do that ;)

These days it's probably cheaper to just program the compensation curve into your CNC controller than to actually build a better screw! The really accurate machines often use separate distance encoders anyway, rather than rely on a ball screw that's being stretched and compressed by both load and temperature effects during use.

⬐ mauvehaus

If you're into this sort of thing and in the Northeast, the American Precision Museum in Windsor, VT is worth a visit. Among other things, they have Bridgeport milking machine number 1.

A lot of the original work in precision manufacturing was in the arms industry. Interchangeable parts in rifles were a huge deal when they came about: no longer did you need a skilled gunsmith to individually fettle replacement parts when a gun needed repair, you could swap one worn part for another in less time and closer to the action.

⬐ tejtm

"Bridgeport milking machine" -> Bridgeport milling machine"

in case your internal spell check does not fix it for you.

⬐ canadian_tired

If you prefer a book, take a gander at "The Perfectionists" by Simon Winchester. If you enjoy the pedantry of accuracy vs precision (or wish you did), this book is for you.

⬐ aesthesia

I was disappointed with the book. There was not enough actual technical detail about achieving high levels of precision and too much journalistic fluff.

⬐ yetihehe

Yes, that book is more about people who made precision, not about technology (I didn't finish it yet, bought it after last article about precision on hn). But I'm not picky, it's nice read.

⬐ seiferteric

He mentions that all bricks in a building are relative to the first brick... I am not an expert at all, but aren't bricks laid using a leveling line? And the mortar will provide some "wiggle" room to fix small mistakes?

⬐ dylan604

Also known as the cornerstone. Even though the mortar might wiggle, if the first stone is not correct,the building will not be in the correct spot.

⬐ petertodd

You're absolutely right: https://www.youtube.com/watch?v=lORIZ1shRIM

Given that mortar is semi-solid, there's no way you could ever reliably get the same thickness of mortar without an external reference.

Older tall chimneys are often visibly crooked in parts. I'm no chimney builder, but it's easy to imagine it being difficult to get the alignment perfect when you're trying to make an angled structure, high up in the air, using primitive tools. (remember that tall chimneys are narrower at the top than the bottom to save money, so you can't directly compare it to a vertical plumb-bob line: you have to offset each layer slightly from vertical)

Equally, once the error builds up to the point where you can see it from the ground, getting the rest of the chimney back to true would be easy: just lay the next layers of bricks slightly offset until it looks right again. Results in an slightly ugly looking chimney. But a lot cheaper and faster than demolishing the crooked bit and starting over.

⬐ mikewarot

One of the things he mentions is a chapter in Henry Ford's Book "Moving Forward" about the very practical need to be able to measure to a millionth of an inch... I highly recommend reading that chapter

http://gutenberg.net.au/ebooks17/1700321h.html#ch14

the lathe is basically the one machine required to kickstart the industrial revolution. (and a way to drive it ofcourse).

Also, three surfaces that can be ground truely flat is also vital.

i can highly recommend the youtube series "machine thinking" if you have interest in this[0], aswell as the book "the perfectionist"

[0] https://www.youtube.com/watch?v=gNRnrn5DE58 [1] https://www.amazon.com/dp/0062652559

⬐ zrobotics

Although truly flat surfaces can be easily generated (relative to building a screw cutting lathe from scratch). It just takes a lot of time and labor, but lapping two granite plates together would produce a good-enough surface plate; considering the accuracy likely to be achieved with a one-off homemade lathe.

Realistically, the absolute hardest part would be the castings. Cast iron is no simple feat, as I'm sure the future posts in this series will show.

⬐ 082349872349872

The Gingery books' path is via cast aluminium.

⬐ skykooler

Aluminum is going to be much harder to manufacture on Mars than iron, at least initially, because it requires so much more energy to process.

⬐ imtringued

Here is a video series of someone building the Gingery Lathe: https://www.youtube.com/watch?v=zPGZg45dGXA

⬐ jojobas

Cast iron is much easier than steel. You won't get to the lathe without cast iron (and what will you lathe anyway?)

⬐ bluGill

You can make a lathe from other materials. Most people building one in their garage these days use aluminum. Cast iron is better in a lot of ways, but the temperatures of aluminum are much easier to deal with, and today we can get good aluminum alloys to work with. You can also use bronze if you wanted. Granite is probably better than cast iron if you design for it. Concrete has been used commercially.

⬐ tejtm

Lapping three surfaces together is the path to flats.

Lapping just two together is how we make concave/convex hemispherical pairs for telescope mirrors.

https://en.wikipedia.org/wiki/Flatness_(manufacturing)

Related:

Jacques de Vaucanson

https://en.wikipedia.org/wiki/Jacques_de_Vaucanson

Origins of Precision and first project introduction

https://www.youtube.com/watch?v=gNRnrn5DE58

(The last video, incidentally, is related to the problem of being stuck on a desert island, and how to rebuild tools that produce other tools...)

This is a really nice video if you are interested in measurement precision for manufacturing:

Machine Thinking: Origins of Precision and first project introduction https://youtu.be/gNRnrn5DE58