Thousandth of an Inch | Hacker News Reader

Thousandth of an Inch

en.wikipedia.org

35 pointsuserbinator65 comments

What I find amazing is that one thousandth of an inch is not particularly precise when it comes to machining. You can easily make parts to that accuracy with common mills and lathes.

Also surprisingly (to me) is if you have a pin that goes into a hole that is 1-2 thousandths of an inch larger than the pin, it feels like a sloppy loose fit. If you want the pin to feel like a really nice smooth fit you need more like a couple TEN THOUSANDTHS of an inch clearance.

You can do it on common equipment but I dunno if I’d call it “easy.” There’s a reason that +/- 0.05mm costs a lot less than 0.02mm.

This.

For the unindoctrinated, this concise Typical Tolerances of Manufacturing Processes reference[1] may be useful towards establishing pragmatic expectations for manual processes.

[1] https://mae.ufl.edu/designlab/Lab%20Assignments/EML2322L-Tol...

Or as a machinist might say, a few tenths.

Yea, I'm always blown away by the old master machinists I've worked with when they talk about getting to "50 millionths" (ie half a tenth of a thou) on manual lathe and measurable with a bench micrometer [1].

Also the auto body guys whose hands are like profilometers [2] and can just feel various surface quality [3]. This is actually where precision originally came from with the 3 plate lap where you're just using fingers and candle soot to create crazy high surface flatness [4].

[1] https://willrich.com/product/starrett-direct-reading-bench-m...

[2] https://en.wikipedia.org/wiki/Profilometer

[3] https://en.wikipedia.org/wiki/Surface_roughness

[4] https://ericweinhoffer.com/blog/2017/7/30/the-whitworth-thre...

You should watch the documentary on the most famous Card Mechanic (Card Magician, but they like to go by 'mechanic') - Richard Turner.

He fan feel thousands of inches and can note the individual thickness of each card he touches.

He is also 100% blind, and he talks about how sensitive his touch is.

He is unbelievably remarkable

https://en.wikipedia.org/wiki/Richard_Turner_(magician)

the documentary he is in is called "Dealt"

https://www.imdb.com/title/tt3127902/

Very cool - thank you!

Which is why a more common unit in imperial machining is the "tenth", which is one ten thousandth of an inch.

The most annoying thing about the thousandth of an inch is while it’s often called a ‘thou’, it’s also called a ‘mil’ which sounds exactly as the common slang for millimetre (mill) here in Australia at least (used more often in construction, but it’s still common so makes it confusing).

Luckily we don’t really do PCB design in imperial anymore at any of the companies I’ve worked with (I hear even US companies are moving away from it since so much of the manufacturing is offshore)

guess who didnt read the link :-) but agree, i did pcb design w a few old hands who said mills and got me a few times

You're not going to say 20 mil and the context not be clear if you mean 0.508mm or 20mm, are you?

A machined feature could be either.

Now's where one needs the Wikipedia article titled something like "Engineering Failures Caused By American/Metric Confusion".

There's at least https://en.wikipedia.org/wiki/Mars_Climate_Orbiter and TBH I was hoping there'd be a link to such a list on that one.

I seriously hope I'm Poe's Law-ing your post too, and there's an implied /s on there, because if you could see the level of communication effort I see at work … yeah, I'd absolutely believe that someone would use it in a confusable context.

Oh this is my cue to share my favourite story stemming out of this confusion, the story of the Gimli Glider [0]. It is full of incredible events, each of which would be a good story in its own right - and it still has nothing to do with our favourite Dwarf.

[0] https://en.wikipedia.org/wiki/Gimli_Glider

Aaaargh, why do people do such stupid things?? Because they started doing it a long time ago when it wasn't stupid.

When doing PCB layout, you're basically forced to think in thousandths, but work in millimeters... You end up memorizing multiples of 2.54mm (0.100in)

Why is that? Common components use thous but tooling moved to mils?

Essentially, yes. Older components tend to have imperial footprints, newer ones metric.

There's something even worse... 0201, 0402, 0603 are all codes for SMT component sizes... and they mean something different depending on whether they are metric or imperial. 0201 imperial is 0603 metric.

Most parts were imperial IIRC until at some point they started selling parts that were too small (smaller than 0201) and switched to metric. I can understand wanting to switch to metric, but reusing the same goddamn size codes is too chaotic for my taste. Come up with different numbers.

You'll still use metric, but you'll just end up seeing lots of 2.54 or 0.254 everywhere.

My sophisticated wild-ass guess: Film and sheet thicknesses are often specified in mils, and PCBs are essentially fancy sheets of stuff laminated together.

My unsophisticated wild-ass guess about why they persevere: Like Fahrenheit, it so happens that mils as a scale intuitively correspond to sizes of trace we care about. A 5 mil trace is pretty run of the mill for signals. 20 mils for most IC supply traces. 3 mils is small and expensive to make (last I checked! capabilities have changed rapidly in my decade of experience)

People in metric countries just learn to be comfortable with decimals. So you end up with trace widths like 0.5mm or 0.3mm and you just get used to it.

I think the intuition argument is often just a post hoc assumption for things that we are actually just familiar with.

The intuition argument is pretty weak.

Thinking about it a little harder, I think the reason is simply institutional inertia, same as for every other industry that retains inch-pound-second based systems.

> A 100µm trace is pretty run of the mill for signals. 500µm for most IC supply traces. 50µm is small and expensive to make

The numbers get even better on the metric system. I was expecting them to just not get much worse, since they were picked on the imperial on.

A lot of tooling and tolerances still uses thousands of an inch as well, complicating the use of a metric grid.

Curious why we jump from 16ths or 32ths to thous rather than 64ths, 128ths, 256ths, 512ths and 1024ths

I guess they are a mouthful without good abbreviations and there was a tech jump between woodworking and machining so in betweens aren't used

I had an old machinist ruler that actually did have 64ths - barely could read it haha - time to get out the calipers!

One reason these whole even numbers are used is because it's easier to inspect with basic tools - ie just fold/layer it a few times and measure the thickness - like using a rule to measure 64 sheets of 1/64" stock and hoping to get 1". The evenness is about doubling/dividing to achieve this effect.

They also had very primitive measures for bore size that worked the same - ie "gauge" for shotguns was the number of how many balls of the barrel's size added up to 1 pound of lead, so you could have a basic scale and infer everything from there.

I don't think it's usual to make that jump, but to use fractions in some contexts, and thou in others. Around a machine shop the measuring devices go by decimals, and in wood shops they use fractional measures.

It mostly stops around 32nds to at most 64ths at best in woodworking where it gets hard to measure accurately using a tape and the Mark 1 Human Eye. Then in machining there's usually digital or dial readouts for things so it's easier to work in a round number like 1000s.

"instead of our engineers and machinists thinking in eighths, sixteenths and thirty-seconds of an inch, it is desirable that they should think and speak in tenths, hundredths, and thousandths"

Yeah, it's a bit of weird disjoint. Almost a tacit admission, in fact.

I've heard that in some crazy parts of the world - say 95% of the planet - it's base-10 all the way down.

Makes you wonder if they're on to something with that kind of whacky scheme. Hopefully it works out for them.

How do they measure out ⅓ and 1/6 and 1/12 using only base 10?

Here's the trick: They just don't do that.

Same way as you'd measure 1/7, 1/11 etc.

“Only base 10” is already the measure of choice where it matters most for most Americans: gas, money, ammunition. It works just fine.

No, it's because the mil works with decimal fractions.

If a measurement is 3.125 inches, you know that it's 3 inches and 125 mils.

You don't need support for fractional syntax in order to have an input field for inches that goes down to mil precision: just a regular decimal numeric input.

You cannot dismiss the advantage of decimal fractions as casually as you can reject the metric system.

Mils are awesome for wire thicknesses, trace widths, pin spacings on circuit boards, fine drill bit sizes for through holes ...

Guitar strings are, e.g. 9 to 46: .009" high E, .046" low E.

Classic DIP (dual in-line package) integrated circuits and headers have 100 mil spacing: ten pins per inch. IIRC, you drill a 25 mil hole for the pins. This is international.

The metric system doesn't have a nice mil-like unit for those scales; the millimeter is way too crude, and micrometers are too fine.

The mil is just right for all sorts of miniature things that are still large enough to be worked on by human hands and eyes.

> The metric system doesn't have a nice mil-like unit for those scales; the millimetre is way too crude, and micrometres are too fine.

This kind of absurd claim would be more compelling if you hadn't just used 3 decimal places and a quotation mark to define a distance in a format used by at most ~5% of all the people working in such scales:

> ... 0.046"

My point - metric can also use decimals. More adroitly, too, I'd argue, because of the consistency.

That 0.046" is just called 46 mils though, or just 46 when that is understood. Those guitar strings are just "nine to forty-sixes" and that's it.

If you're routing a circuit board, you have a grid that is a multiple of the mil: e.g. 25 mils, so if have a part with 100 mil pin spacing, that is four grid lines. Trace widths are multiples of a mil: 6 mil, 8 mil, ... through holes are mils: 25 mil, 40 mil, ... no decimal point or " mark in sight.

Sure, a tenth of a mm would be a useful unit. Problem is, 0.1 mm has no name. It's either 0.1 mm or 100 microns, which are too small. I've not seen anyone define such a unit in practice; they just plod along with multiples of 0.1.

> Classic DIP (dual in-line package) integrated circuits and headers have 100 mil spacing: ten pins per inch. IIRC, you drill a 25 mil hole for the pins. This is international.

But that's just a historic accident.

Related, the metric system was forced upon trade in Germany not before the early seventies. Only few units previously used survived, e.g. the 'Pfund' (pound, now redefined as 'half kg'). My father continued to use 'Zoll' (inch) and lamented American made machinery using threads on nuts which differed from those used in Germany (and everywhere else). Even I know the folding carpenter's ruler only as 'Zollstock' (no one says 'Gliedermaßstab').

And while I used to be very opinionated about the imperial- vs. metric system, after twenty years in California, I feel comfortable at 75°Fahrenheit and am exhausted, if I have to walk more than a few miles. I even caught myself recently referring to a width as 'about a yard'...

My guess would be that fractions with a powers-of-2 denominator were used in pre- and early-industrial contexts and remain in use today by long tradition, but thousandths started to be used only later in industrialization when the benefits of a base 10 system were more clear

> The units are considerably different: a millimetre is approximately 39 mils.

I've been regularly watching various machinists on YouTube for a few years now, so this feels very normal to me even though I have no actual experience in machining.

Like the article says, we machine parts with tolerances in thousandths of an inch, but certain specifications like paint thickness are specified in “mils” which is also thousandths of an inch.

Another really common one is plastic sheeting. That's probably the item I most often personally come into contact with that is measured in mils. (that is, things I need to know the thickness of, which excludes the paint all around me)

Disposable gloves are also measured in mils. When I bought a micrometer one of the first things I measured was a disposable glove and I was concerned that the measurement was really wrong. Then I searched online and found that the specified thickness is only on the finger pad area of the glove. Measuring there was dead on.

> the specified thickness is only on the finger pad area of the glove

TIL. Thanks for sharing that, this is handy to know.

Pun intended?

I think it was all over once America started using ounces to measure volume.

Ounce / fluid ounce is roughly the same as gram / milliliter. A fluid ounce of water weighs about one ounce (1.04 oz/fl oz or something), like a milliliter of water weighs about one gram (0.997 g/mL or something).

The US customary units are pretty logical, they're just not powers of 10.

> Ounce / fluid ounce is roughly the same as gram / milliliter.

That's how it's used, but the units don't measure the same thing. It's like measuring an engine's power output in square meters, because a kilowatt is about a square meter of sunlight.

> That's how it's used, but the units don't measure the same thing.

I'm having some real trouble parsing this sentence right here. You're saying that the two DIFFERENT UNITS "fluid ounce" and "ounce (weight)" measure DIFFERENT things? Yes, different units measure different things. Or maybe I don't understand what you are trying to say here.

If I said "kilogram force", you'd know exactly what that is.

You are also not confused when I say that water boils at 100 degrees centigrade, and a right angle is 90 degrees.

And don't even get me started on the number of people who argue that 1024 bytes is called a "kilobyte".

I think the GP's point was that "fluid ounce" is a measure of volume, not mass or weight. If you're measuring a liquid whose density is different from water, a fluid ounce won't weigh anywhere near an ounce.

"degrees" is an interesting one, as it's really just another word for "amounts" or "units" or "subdivisions". 100 degrees centigrade is 100 amounts of centigrade. Or actually Celsius, as "centigrade" means "100 subivisions". The name of the unit is just "Celsius".

Really, degrees of arc should probably have been given a "proper" name too, but it's too late for that. Or you could switch to radians ;-p

Language is fuzzy. People know what you mean when you say "degrees centigrade" or "fluid ounces", so it's fulfilled its purpose. But that doesn't mean that some "proper" terms for measurements aren't weirder than others. "Celsius" is straightforward and unambiguous. "fluid ounces" is kinda weird. And yes, "degrees (of arc)" is a poor choice of unit name, worse than "fluid ounces". That still doesn't make "fluid ounces" good though.

> I think the GP's point was that "fluid ounce" is a measure of volume, not mass or weight. If you're measuring a liquid whose density is different from water, a fluid ounce won't weigh anywhere near an ounce.

Yes, and equally so, if you are measuring a substance other than water, the milliliter will have a different mass than a gram.

But "millilitre" isn't named after a unit of weight.

A fluid ounce of mercury doesn't weigh an ounce. But a millilitre of mercury still takes up one one-thousandth of a litre, which is itself one one-thousandth of a cubic metre.

Funny thing, "millimeter of mercury" is a unit of pressure, but "millimeter" is a unit of distance.

"Fluid ounce" is a unit of volume, but "ounce" is a unit of mass.

"Pound" (or "kilogram") is a unit of mass, but "pound force" (or "kilogram force") is a unit of force.

I understand the argument, it's just not a very compelling argument. Explaining the argument one more time is not really convincing. If I'm going to embark on a mission to remove ambiguity from measurement systems, my first choice would be to abolish the use of KB=1024 bytes. The "fluid ounces" and "ounces" thing just does not seem like it poses as much of a problem.

Addendum: I'm also not trying to be contrarian here, and I'm not trying to advocate for some kind of supremacy of US customary units over SI. Most countries have some form of non-SI customary units still in use and some amount of SI adoption. The SI units are preferred for lots of reasons, but it's not like the various systems of customary units are illogical or nonsensical. There's an explanation and a logic behind them, like how US customary volumetric units follow powers of 2, or how a "mile" is a thousand paces (nice round power of 10). The ounce / fluid ounce relationship is has the same physical explanation as the gram / milliliter relationship, because they are both approximations for the density of water, as a reference point. Neither relationship is exactly the density of water, because the respective units have been redefined, and the density of water is now something you look up which happens to have a value close to 1.0 with those particular choices of units. Just as the meter was originally defined as one millionth of the Earth's meridian, the mile was originally one thousand paces, and after standardization, neither fact is true.

Again, I can understand why people want to advocate for SI, but that doesn't mean that customary units are illogical or nonsensical.

> Explaining the argument one more time is not really convincing.

Sorry, I wasn't explaining again because I thought you weren't convinced of my point. Based on your previous reply, I thought you hadn't understood my point. Because if you had, the "well, a millilitre of other liquids doesn't weigh the same either" seems like a complete non-sequitur. I couldn't figure out what point you would be trying to make with that comment if you had understood me. (Knowing that you did, I'm still not sure? Like... so what?)

I guess my response is also "so what?"

If you measure a liquid other than water, then a fluid ounce does not weigh an ounce... so what?

Even water, the density is 1.0 oz/fl oz, but that's as many significant figures as you get. You would often want more significant figures, even in non-scientific contexts.

> If you measure a liquid other than water, then a fluid ounce does not weigh an ounce... so what?

So that makes "fluid ounce" a bad name. Because it doesn't measure ounces of fluid.

Ok, I had assumed you knew what fluid ounces were. Yeah, that assumption doesn’t make sense since it’s not used outside the US and a couple other countries.

A fluid ounce is 29.5735 mL.

The “fluid ounce”, being a measure of volume, can be used to measure fluids.

"mmHg" is also on my list of units that are stupid. :P

Likewise I understand your argument (yes, of course I know "fluid ounce" is a unit of volume and "ounce" is a unit of weight), I just think naming a unit of volume after a unit of weight is dumb.

> The US customary units are pretty logical, they're just not powers of 10.

It somewhat slides off the cliff when you take into account the existences of the imperial ounce and the US food labelling ounce, though.

The US uses the US customary units and metric, not the imperial units.

Not that it matters much, since the US customary ounce and the imperial ounce are the same mass. Maybe you're thinking of the fact that an imperial pint is different from a US customary pint?

More fun with units, a "tenth" is a tenth of a mil.

i know a few machinists who can think in thous.

my dad's a machinist; I am from exposure. Every time I express anything technical to him there is a translation that has to take where we stumble around until we convert what i'm saying (metric) into thou.

i've been doing it for years and i'm still shite at it.

Reply on news.ycombinator.com