Eagle library editor

BOM

There is nothing in your system to show the release. You don't know what you have. That's "out of control", in my book.

So you have two types of ECOs. An original and changes. Why?

the

But it's not tracked by the ECO system. Problem Reports are and can be entered by anyone (even sales, if they're so inclined).

No, I'm saying that the engineer's writing should make sense before anyone signs off on the ECO. If not, kick it back.

So you have two systems. Why?

Making something out of nothing is a change. Engineering did it. ;-)

You also said that there is no ECO for lettered revs. A rev from A to B certainly *is* a change.

writing home

design is

defined by

schematic he

ECOed

separate

BOMs,

There

the

calls

It certainly generates the netlist but not necessarily the BOM. ...and not necessarily the complete BOM (non-electrical widgets).

The old one doesn't get "processed". Once anything is changed it's the "new" one.

Usually because one doesn't understand it well enough. I was certainly in that category when I worked for Big Blue. It made *absolutely* no sense. The EC number didn't do anything, there was one for every P/N and a P/N for every EC number.

I agree with you here.

writing home

chip design is

defined by

the

chips

schematic he

PCB

matches

be

wasn't ECOed

separate

in BOMs,

were

the

parts

There

using the

calls

the

There is no way to guarantee the schematic hasn't been changed if components on it have been changed. It can't be guaranteed correct so don't bother. The schematic forever stays with the first instance of the rev.

The schematic stays with Rev A because it's really documenting the board. If PCB gets rev'd for one product, the others it represents may or may not follow.

writing home

chip design is

are defined by

the

chips

schematic he

PCB

matches

not be

wasn't ECOed

a separate

in BOMs,

were

the

parts

bare

There

using the

calls

are

the

controls

Sure there is a guarantee: If the assy rev level hasn't changed then the schematic linked to it cannot have changed. Because those are married. But the schematic is not married to the rev level of the bare board.

The schematic does not document the bare board. It documents the finished board (a.k.a. the assembly). Again, if you change one particular resistor in value there is no need whatsoever to make a new layout. But the assy and schematic rev levels must be bumped. Even those do not necessarily have to be the same, but if not then there must be an ECO-released document that says which goes with which.

Usually the module spec is master. This contains the schematic (with rev level), the BOM, and so on. The BOM in turn contains the part number and rev level of the bare board because that's one of the many parts needed for assembly. It doesn't have to be that way but makes most sense because anyone who pulls this can learn almost everything they need about this assembled board.

Let me give you an example where that was clearly not the case and would have made no sense to do:

A few years ago I designed a fairly large board. Nothing special about it except for one circuit area. That area could be configured in several ways, making this design useful for more than one of their products. The differences were the use of different footprint transformers and the stuffing or leaving off of versious parts. The transformers even have a different number of pins. So here we have the exact same rev level bare PCB but numerous assemblies with different schematics and different netlists.

We use dash numbers for assembly variants, sort of an old aerospace tradition.

We can have 22A230-1A VME thingie, 32 channels 22A230-2A VME thingie, 32 channels, with BIST 22A230-3A VME thingie, 64 channels 22A230-4A VME thingie, 64 channels, with BIST

and so on. Each dash version has its own BOM, calling out the bare rev A board (22D230-A) and all the parts. If we roll the PCB to rev B, we'd have

and not totally trash the manuals and price lists and BOMs. Simple edits.

But getting back to PCB CAD, when we rev a board we start with the existing released PADS schematic, say rev B, and the released rev B PC board. We edit the schematic and save as rev C. PADS makes a forward ECO file by comparing the schematics. We pour that into the old PCB and route the changes, calling that pcb rev C. The rev C schematic and PCB get netlist crosschecked and released. So we're all synchronized, and we have the baseline files to make rev D some day.

If the changes were non-trivial, we will often resequence the reference designators to keep production and test folks happy.

So if you are spinning off schematic rev letters without changing the PCB, how do you do a PCB rev? It sounds like you have to have one thread of schematic revs for PCB CAD use, and all sorts of asynchronous offshoots, assembly letter revs with associated schematics, unconnected from the PCB control thread.

When you do rev the PC board, how does that affect all those other spinoff schematics? What if you resequence?

John

writing home

chip design is

are defined by

the

chips

schematic he

to

The PCB

matches

Rev

not be

wasn't ECOed

a separate

done in BOMs,

were

denote the

parts

bare

number. There

using the

calls

are

in the

controls

products

in

be

The way CAD systems work, *some* schamatic has to document the bare board, because it generates the netlist. Which one? How is it controlled?

John

That makes sense.

You have to, because one part in the BOM has changed (the bare board).

That doesn't always make sense to me and I've never seen that as a hard requirement anywhere. But, in my case the customer is king, I always ask them what numbers and rev levels need to be where and make sure it gets done that way. When I was da boss at a division I wouldn't have allowed it as standard procedure, unless it was really necessary to do a new layout. That is often not the case and then we had no PCB rev level bump.

You leave the bare PCB rev where it is, only the assy rev level gets bumped. Unless you need a new layout for technical reasons.

Correct, the bare PCB is always lower on the pecking order, it is only a part. We often use the same PCB for totally different assemblies. A classic example are filters. On one ultrasound machine we had four different plug-in filter modules, each had their own rev level which was marked on the empty field. The bare boards all had the same rev A. Why should these filter modules be controlled by a PCB contrl thread?

That's what systems such as Agile are for. It is the same scenario as if you changed a part. One of the engineers must sit down and investigate the impact, then write a plan of action. The computer will tell you that this part (PCB, transformer, metal piece, whatever) is used on assemblies 322, 768 and 983. The action plan will then spell out that all those must be checked and possibly the documentation must be re-done, requireing one of more additional ECOs.

Those are the more boring aspects of engineering in a highly regulated field, but must be done. It's probably about 10% of my work here, impact studies and finding solutions if a bind shows up.

writing home

chip design is

are defined by

with the

chips

schematic he

to

The PCB

matches

Rev

not be

wasn't ECOed

and a separate

done in BOMs,

products were

denote the

parts

That

bare

number. There

using the

(which calls

are

in the

linked.

controls

products

The

in

be

No, the bare board is only documented by the Gerber file set and the CAD file set that generated the gerbers. Those can (and typically do) have a generic schematic from which the netlist came but that isn't a requirement.

When you have a filter you can have one PCB and several different sets of schematic. On one assembly L3 is 100nH, on the next it is 150nH, and on a third L3 is not stuffed. Yet they are all built on the same board.

worth writing home

Most chip design is

are defined by

with the

chips

schematic he

at

Fremont, to

The PCB

matches

Rev

not be

wasn't ECOed

and a separate

done in BOMs,

products were

denote the

parts

That

bare

number. There

using the

(which calls

there are

in the

obviously

...a

schematic?

linked.

controls

products

forever

The

everything in

just be

If

That terrifies me. Sounds like you can't be sure which schematic file generated any specific Gerber set, and the Gerbers themselves are the only way to figure out the connectivity, the part footprints, all that. PCB revs could get out of hand over time, if the schematic that creates a PCB rev is not always released with the PCB file and the gerbers.

We release the schematic, PCB, and Gerbers together, at the same letter rev level, rigidly cross-checked. After release, none of those rev "X" files are ever changed. They will all go to rev Y if needed.

We create assembly variants with dash numbers, where each dash number has its own BOM. For small changes and specials, an ECO can create a dash number, too.

Manufacturing will often take our BOMs and resection them. They may create their own -99 version of a PCB assembly, and use that as the starting subassembly for our intended -1, -2, -3 versions. We don't care, as long as the final assembly is what we specified.

We use sequential dash numbers. The aerospace boys traditionally assigned -1, -3, etc to assembly variants, and reserved -2 and such for mirror images.

They wouldn't even document this on drawings. The machinists just knew how to mirror things.

Note that 12345A is a drawing, rev A, but 12345A-1 is a thing. If not specified otherwise, a drawing is assumed to define the -1 part.

John

everything in

just be

If

PCBs always have to be ECO'ed. This ECO contains all the back-tracking info you need. Which set of Gerbers it is based on, which schematic set, who dunnit, who was the Republican candidate back then, who won the SuperBowl that year, and so on.

So what could possibly get out of hand?

No client or employer I ever worked with does that.

If you follow your scheme to the letter that means you must scrap remaining circuit boards. If you only changed R92 to a different value in the schematic that would mean throwing away perfectly good and useful boards, wouldn't it?

Small changes? That would be watering down your rigid scheme already. One must bump the rev level if a change in design happens. I doubt dash-numbers for real design changes would fly with the FDA.

No EE sign-off required on that? We usually require full ECO release for versions.

Ouch! Maybe they just haven't been caught yet :-)

Yeah, but if 12345A-2 isn't documented in the drawing that would be a violation in the markets I usually deal with.

worth writing home

Most chip design is

shapes are defined by

with the

mere

potato chips

schematic he

at

Fremont, to

The PCB

matches

one. Rev

should not be

wasn't ECOed

and a separate

done in BOMs,

products were

denote the

the parts

That

good bare

number. There

each using the

(which calls

there are

instructions in the

obviously

...a

schematic?

linked.

controls

products

forever

components

The

everything in

just be

If

Not to mention that the Gerbers can't get you back to the original design.

I agree 100%. Any assemblies made from the board have a separate P/N and release chain in fact.

We used different P/Ns, since they *are* different.

everything in

just be

If

So then what do you guys do if R92 gets changed to 14.3k and a new rev B set schematic is released? Throw the remaining 137 rev A bare boards into the dumpster?

everything in

just be

If

Well, you need a serious database to keep track of all the related files, and hope that people will enter all the data properly. I've seen lots of big organizations that don't.

Next you'll be telling me that you use a VCS to document what goes with what.

Maybe you have led a sheltered life?

We don't ever change a released document. The BOM controls which parts get put on a board.

I don't do anything that requires regulatory approval. Too much hassle, slows things down.

But our dash number system evolved from MIL and aerospace conventions. I once did flight hardware for the S1-B moon rocket, the C5A, and tons of Navy and Coast Guard ships, inspected to the hilt, and all our document controls passed.

No. If they build things that conform to our drawings, we don't care how they stage the process.

We can write an ECO, or release a new BOM, to create a dash number. Easy.

John

everything in

just be

board. If

Of course not. R92 doesn't change and "14.3K" isn't on the board. It's a BOM change. No problem.

worth writing home

Most chip design is

shapes are defined by

with the

mere

potato chips

schematic he

unravel at

Fremont, to

G. The PCB

matches

one. Rev

should not be

schematic wasn't ECOed

and a separate

done in BOMs,

products were

denote the

the parts

That

good bare

number. There

each using the

(which calls

there are

instructions in the

obviously

...a

schematic?

linked.

board

controls

products

forever

components

The

everything in

just be

If

We can do that, if needed. We'd just create a new BOM that calls out some bare board and creates some new assembly. But if the product *is* the board (which our VME products are) we almost always keep the etch rev and the assembly rev the same.

If a box contains boards, the box rev and the subassembly revs need not match up. Again, the BOM controls what goes inside.

If we make different things from one bare board, we have the option of creating multiple dash numbers, or (rarely, for marketing reasons) calling different versions entirely different products.

What matters is that we always know what's been built and what its state is, and that we can reproduce anything we've ever built. And that everything is backed up unambiguously.

I've seen lots of startups create horrible messes because they had no document control.

One big machine-tool N/C company had no controlled assembly drawings or schematics. Everything was built from a "gold" model. Change control was some engineer hacking the model.

John

everything in

just be

board. If

We don't do that. We only change the schematic when we rev the PC board. An ECO tells manufacturing (or test, or a customer) to change the resistor, but we don't issue a new schematic. If it's a permanent change, the BOM is updated and "ECOxx incorporated" noted.

Every schematic says "Engineering Change Orders May Apply."

John

everything in

just be

board. If

Even in the days of the paper computer this worked very well. As I said, the document number and rev level of the underlying schematic is written _into_ the very file that releases the bare board. So all I ever had to do in older organizations is go to the doc center, ask for the 1045-7853 Rev A file, look up the underlying schematic, ask them whetehr they could please pull that for me as well, and whether I could have one of those mint cookies from the bowl.

In more modern organizations this can all be done from the screen with programs such as Agile. That method has less calories but also less exercise.

Ok, >50% of my jobs are in highly regulated industries where what I described is pretty much mandatory. But sheltered I am not, due to being a consultant I get around a lot. Start-ups often need some help in that domain so I give them hints how to build up their doc system.

So let me repeat the question. Say one of your guys finds out that 12.1k is too low and R92 needs to be 14.3k. What happens next? Do you not change the schematic to reflect that change?

That's a bit surprising. Normally they control all the document numbers and such.

Ok, just don't build any med gear that way because that won't fly there.

But above you wrote "They wouldn't even document this on drawings". To me that sounds like the respective ECO wasn't written, because there is no reference on the drawings.

everything in

just be

versus

board. If

Do you not update the schematic?