What explains the unintuitive numbering of chip pins? | Hacker News Reader

What explains the unintuitive numbering of chip pins?

spinellis.gr

38 pointsDSpinellis32 comments

I think this is wrong. Pins really pads that are numbered on the die. The die is square (or was back then). Pads are numbered in sequence around the pad ring. So the concept of columns doesn't exist in the context of the die. As to why number counterclockwise vs clockwise, possibly that's to do with existing convention from valves. However note that in the beginning dual inline packaging wasn't the norm. There were flip chip packages, die on board, circular TO-xx metal can packages etc. So whatever led to the counterclockwise numbering convention certainly predated DIL packages.

That's a really good point. So if you wanted a single numbering system for 4-sided and 2-sided chips, the current approach is quite a sensible compromise.

If you have the little key symbol in a corner, you can just "walk the chip" from there, regardless of sidedness

These days of SMT chips the trend is towards more QFN and QFP so columns make less sense as well.

I'd say the trend is towards BGA on the active components.

Also column wise counting would be a recipe for disaster. It's way to easy to "slip" by a column/row if you move across the chip, so everyone would still be counting up and down the sides just maybe in steps of 2. Might as well do it right then

>> There were flip chip packages, die on board, circular TO-xx metal can packages etc.

Do you have a good reference on those predating DIPs? Whenever someone balks at such "advanced" manufacturing methods due to price, I ask if they ever opened an Atari 2600 cartridge from the 1970s. IIRC some were die wire bonded to the board under a blob of epoxy.

One reference to early IC packaging is: https://www.computerhistory.org/siliconengine/package-is-the...

Note that the Apollo Guidance Computer for the moon flights used flat packs, flip chips, and circular cans.

The only pin numbering I ever perceived as unintuitive is on PLCC packages where pin 1 is in the middle of one of the 4 sides. Especially since there's still a marker dot in one of the corners generally.

Luckily, PLCCs are rare these days.

Example: https://www.zilog.com/manage_directlink.php?filepath=docs/z8... (Note that the orientation on the PLCC package is indicated by one of the corners being different, yet that's not where pin 1 is.)

Yeah, I've seen those. Do you have any idea why they did that? I had twenty or thirty on a recent board and they took so much extra work to double check because they didn't follow convention. I probably read that pinout in disbelief for an hour before I called them to check.

I think that's why people number them the counterclockwise way, if they didn't they'd get furious engineers calling them and yelling (I didn't yell, I was just very confused) that they aren't following the standard. There are so many ways to make a mistake in a circuit board already, once someone decided on a numbering scheme momentum probably just kept it going to be honest. Why would you change it when there is no functional difference and it makes mistakes less likely?

> Do you have any idea why they did that?

No, not the slightest clue. If you find an answer, I'm interested too...

All I can find is the definition, which states that pin one is in the middle of the top at least in the 44 pin package.

http://www.interfacebus.com/ic-package-plcc-dimension-outlin...

No idea on the history though, hopefully someone can enlighten us, Google did not.

> Why are pins counter-intuitively numbered in a rotating fashion rather than by columns as one would expect for a rectangular package?

I am not sure why this is counter intuitive, on a 8 pin package, it makes sense to keep pin 5 close to pin 4 to simplify the layout of the wires, and in general, it makes sense to keep consecutive pins close to one another. This applies to vacuum tubes as well.

Is there something particularly five-y about pin 5? Surely if you called it 8 nothing else would change. No one is suggesting redesigning the whole thing. Just move a few labels.

When I look at pin 5 on a schematic, it makes it easier to find on a PCB if it is physically close to pin 4 and 6.

This is true in programming too. Spatially close data should be stored in memory adjacent to each other.

I am definitely open to the idea that the current labeling is actually more intuitive, but note that this is not at all the argument I replied to.

Yes, consecutive pins are often related in some way. Let say pin 3,4,5 and 6 belong to the same port, and must be connected to a connector on the other side of the PCB.

If pins were numbered by columns, you would have to connect two tracks to one end of the chip and two other tracks to the other end of the chip. Once you've done that, routing the remaining tracks is basically impossible on a single PCB layer.

In other words, the standard numbering has the nice property that two pins with a consecutive id number are always relatively close to eachother (whatever the package).

> If pins were numbered by columns, you would have to connect two tracks to one end of the chip and two other tracks to the other end of the chip.

You have missed my point entirely. All you have to do is re-label them. There is absolutely no re-wiring or re-routing implied by changing the pin numbers.

No I haven't, but I've failed to convince you, it seems.

It does make a difference because the chip designer doesn't know or doesn't care how the chip is going to be packaged and where the pins are going to be placed on the package.

With the standard numbering, two consecutive pins are always placed next to eachother no matter the package, and the chip designer can use this information to optimize the placement of related pins.

This is not true with the alternative numbering "by column", where pin 4 and 5 are close on most packages, but far away on a DIP 8 package.

Well, we now have packages where pins are on 4 sides (not just 2 sides). For example [1]. So here it certainly makes sense that numbering goes around the package instead of in columns/rows.

[1] https://en.wikipedia.org/wiki/Quad_Flat_Package

Why would a convention be considered unintuitive? You learn it and it's just a design fact. I'm not sure I agree with the article - it seems like a case of human-pattern matching.

What's unintuitive is that it doesn't follow other conventions that could reasonably be applied.

Which conventions would that be?

Reasonable candidates, per the article:

- https://en.wikipedia.org/wiki/Left_to_right

- https://en.wikipedia.org/wiki/Clockwise

By that logic:

- https://en.wikipedia.org/wiki/Counterclockwise

Also, it's clockwise when viewed from what used to be the wiring side, i.e. bottom.

Also-also, mathematical/engineering sense of rotation for positive angles is counterclockwise. ("The mathematician works against the clock.")

Also-also-also, you've listed 2 conventions with no preference between them. If either would have been picked, we would be having the same discussion right now on why it wasn't the other one. The only thing that would give any advantage here is if there were established context or logic giving significant preference to any of them.

Yes, you understand correctly. In this case, assume the life and cultural context of the author.

If you consider those reasonable, then it's pretty hard to argue that the one they picked isn't reasonable.

But "unintuitive" isn't the right word for picking one out of three reasonable options. It's not like you can avoid picking one.

The argument comes as a thought experiment: If you handed me an arbitrary object and asked me to number its parts, I'd probably default to one the above conventions. This outcome is incidental, but a predictable consequence of my cultural background and individual experiences. It would be surprising for me to make up something different.

The insight offered by the article is that they way we ended up with something different is that, amending the hypothetical, the object was actually introduced upside-down.

So you're arguing that clockwise is intuitive but counterclockwise isn't? (In that case it's confusing that you said "other" conventions that could reasonably be applied. If you say "here's two other cats they could have picked", that makes it sound like what they picked was a cat.)

I don't think I agree with that. Especially since if you hand me something and I start counting left to right, oops I probably went counterclockwise.

Also, when creating wire wrap circuits (https://en.wikipedia.org/wiki/Wire_wrap) one works from the underside of the board, so the underside pin numbering image shown in the webpage is the pin numbering one follows while wire wrapping a circuit.

It seems like the author just wanted to write about something, and he chose something he's not familiar with but what appears as quirky to him. I expected the article to talk about IC design and peripherals (the placement of which is what is actually something that might seem unintuitive at first to many, well perhaps more annoying than unintuitive).

It seems like the author just wanted to write about something, and he chose something he's not familiar with but what appears as quirky to him.

Welcome to Hacker News, you must be new here. =)

Reply on news.ycombinator.com