I always refused to do dumb things, and that resulted a couple of times with my departure. Most often, the boss just slinked away and pretended that the confrontation never happened. Try it.
But bosses are less likely to know engineering lore than the engineers are. So we get back to the mystery of standards and practice.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
That's cool. I just ordered a copy, used, $4.98 including shipping.
I've worked for/with operations that just assigned the next sequential number to anything... schematics, parts, PCBs, manuals, forms, even people. That has to be backed up with product trees to remember the relationships. It's horrible. The geniuses who design such systems also tend to use long strings of unpuncuated numbers, so a schematic might be 45931653 and the bare board might be 46795532 and...
One of my customers uses 12 digit numbers for everything. They often email us to ask us what their 12NC actually is. Usually, we don't know either.
We might sell a thing called a "V545"
Our schematic might be 22S545 rev A
The PCB design is 22D545 A
The PCB assembly drawing is 22A545 A
The assembled physical board is 22A545-1A. A different version, stuffed differently somehow, might be 22A545-2A.
So everything is obvious. The BOM for the -2 assembly is file
22A545.2A, and it begins with some plain English comments telling what it is.(In the aircraft industry, -2 is the mirror image of -1, so one drawing can define both wings of an airplane. We don't make planes, so we just use -1, -2 etc for assembly variants.)
We don't allow different things to have the same assembly number, even if someone thinks they are interchangeable.
Our purchased parts have a stock number, and that number must always work. If we buy a part that is not an exact drop-in-equivalent of the original qualified part, it gets a new stock number and a new bin.
These are the sorts of issues that I'm interested in, how to really control electronic design and manufacture. This thread alone shows how much variation there is between organizations, and how unscientific the process is. It's immensely complex.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Hadn't you suggested, once upon a time, that part numbers should simply be hashed?
Or perhaps, derided... :)
Some people try to build a toxonomy-of-everything*. Inevitably, trouble strikes when they try to squeeze everything into, say, five digits.
*Ooh, that's a good typo. I'm leaving it in.That's what happened at a PPoE. They used five digits for everything, with the first two digits being for part class.
They had never done circuit board assembly before, and it was insisted that ALL parts on the PCBs be received, stocked and managed.
Nevermind that they had no intention of buying anything but fully assembled PCBs. I suppose the motivation was having repair stock, but geez, DK will send ten 0603s next-day, no problem.
After designing a few boards, and going through a few spins thereof (oh, and the boards needed numbers reserved, too, even protos), I was already dangerously close to filling up my meager 1000-long inventory slot.
That's not sustainable. But clearly, neither is the desire to be employed by such a backwards company...
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
Love it! Origin?
?Bodge?.
I suggested that we might hash the part number and serial number of a box to make a key that allows field firmware upgrades. If the customer sends us money, we could email them the upgrade key to enable some feature. Most everybody does this lately.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Even when he asks you to tell him how something should be done.
He's probably speaking from direct experience. QA doesn't get shirty at one here in the US of A, but I've seen them get bent out of shape at too many board mods.
-- Tim Wescott Control systems, embedded software and circuit design I'm looking for work! See my website if you're interested http://www.wescottdesign.com
Kluges, done right, don't make a board unreliable.
When you have to kluge 16 data acquisition channels on a board, it takes 16*N hacks.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Sequential systems ought to be the shortest. We basically do that, and our bought-in part numbers are only 4 digits. You do need a database to know that 1446 is a RES 0805 1% 10k.
OK. The top-level product numbers tend to look like Vnnn.
S for schematic. Is the "22" significant?
OK, so you order the board under that reference ("22D545 A").
Say there is a single-variant product and you find you need to change a part value on it, to improve performance. Does that generate a new assembly revision B? Or is that a new "stuffing variant" (2A say).
Do the revision letters appear on each part in the BOM? In particular parts that you control, e.g. the bare PCB and any sub-assemblies like your surf-mount adapter thingies.
What about a system with more than two levels, does a PCB change propagate a change to the top-level assembly revision?
Yes that book uses the "dash number" scheme too for that. Where distinct parts are controlled from a single drawing.
So any change in any of the sub-assemblies - even at the lowest component level - changes all the upper-level numbers too?
Yes that is basically what that book proposes, although I guess you need to be careful how you define "exact drop-in equivalent".
Yes it is
-- John Devereux
Because people don't. You do need to write down enough to let somebody else reconstruct your train of thought.
-- Bill Sloman, Sydney
European bosses can't fire you at will. If you are told to do something really stupid - as sometimes happens - the fact that the boss told you to do it is no excuse for going ahead and creating a disaster.
And bad ones get fired if they make a habit of insisting on people doing stupid things.
-- Bill Sloman, Sydney > >
Such boss' stupidity is still a good reason to look for employment elsewhere.
We've all seen cases of the Peter Principle. We've all seen case where, to be somewhat euphemistic, cream isn't the only thing that floats to the top.
It's certainly part of the design process.
NT
At that point it becomes a financial decision. Does it cost more to hack or spin. Of course, other business issues (opportunity/lost business/time) matter too.
There are always assumptions made in the design process. It's good to document assumptions (if even possible) before the design starts, if for no better reason that everyone starts out on the same page. I've had product managers say "that's not what I wanted" after the product was complete, when what he really meant was "I never thought of that problem". That was a result of an assumption that I probably wouldn't have even thought to document.
Throwing away $4000 boards is painful, too.
We spin pcb revs to eliminate kluges, or to improve performance, or to make them easier to build/inspect/test, but we rarely throw away a built board. One of our best selling products is now rev K.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
I designed a nice NMR gradient driver and was proudly presenting it to the customer engineers. A marketing type accidentally popped into the conference room, looked at it, and declared it unacceptable, in about
10 seconds. She was a big cheese, so I had to spin the board.Well, I'm still in business, and they're not.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
But sometimes there's also a reason to stay.
Nod.
NT
We use a 7-digit stock number for purchased parts (or for subassemblies, or finished products... anything that can be put into stock.) There are algorithms for assigning numbers to things like resistors and caps that have value. So the parts wind up listed and stored in order of value.
22 is the VME product line.
Yes. But it also has a stock number and a bin in the stock room.
22D545B is in a separate bin from rev A.If the bare board is rev A, the assembly is always A. An ECO can produce a variant, even a one-off, and that gets a new dash number.
Purchased parts don't have rev letters. If a part is really different, it gets a new stock number, and BOMs can reference that if they want to.
Assemblies and subassemblies of different revs get different stock numbers.
That's a judgement call. If changing a subassembly doean't affect the customer, we don't change the top assembly rev letter or dash number.
Except some customers demand that we change absolutely nothing (the Intel "Build Exact" pathology) so we're supposed to tell them if we add a lockwasher, which would change our assembly number, and they would requalify the box and change their 12-digit number. In that environment, broken things can stay broken for literally years.
Right, that's an ancient aerospace/military convention. Part, defined by a drawing, defaults to -1. A drawing note could say something like "add this hole for -2 only" or something. We handle board stuffing options with dash numbers.
Not necessarily; see above. This is messy, but the entire problem is messy.
An ECO can change things too, like a resistor value or something. We keep records of the state of every serialized item, so we know the history of every product or subassembly.
Well, nobody is perfect. And sometimes one manufacturer's LM123 doesn't work like someone else's LM123, and we find out the hard way.
We once bought a gigantic hydraulic punch press with a Westinghouse n/c controller. The service engineer said that they didn't have schematics or assembly drawings; their production people had one unit that had been tweaked to work, and they made copies of that.
I wrote the compiler that made punched tapes; one little bug could have torn the 30-ton thing apart. Fun.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
[...]
(hope you don't mind more questions, it is interesting for me and I think you probably have a better system than us!)
Is the "stock number" the same thing as your "part number"? 22V545A is the stock number for the product? And 22D545A is the stock number for the bare board?
Is the stock number just "22D545 A" or is there some parallel system
OK I see now.
That is, purchased parts that you don't control the drawings for?
The "Engineering Document Control" book seems to assume a possibly deeper heirarchy of assemblies where a lot more of the parts are built to controlled drawings. So all parts have drawings with revision letters, and there is then the issue is the revision letter part of the part number. The author insists that they are not. Different revisions of the same part number are "interchangeable" (otherwise they have a new part number). They can e.g. be mixed in the same picking bin and are not generally identified as being of a different revision. The revision letter does *not* appear on the BOM so that revisions to parts do not change all the higher level documents.
Probably wise from your point of view.
Yep. Conversely with consumer computer boards it was common to have totally different PCBs, different chipsets, with the same "part number".
The "Engineering Document Control" book uses a lot of mechanical examples, so you could have a single drawing controlling dozens of different lengths and diameters of struts or something. With a table listing all the dash numbers.
Yes that is the sort of thing I was thinking of.
-- John Devereux
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.