Eagle library editor

(the

second 1234567AB,

everything in

will just be

#1234567A.

versus

board. If

may not

CAD

have a

requirement.

design.

and

a BOM

an

So, if someone in testing prints a schematic, and pencils some private notes on it, the company gets shut down? How about smudges, fingerprints, coffee stains? Post-it notes? Do they get strip-searched by the inspectors?

No, thanks, I won't compete with you for that business.

John

(the

second 1234567AB,

everything in

will just be

#1234567A.

versus

board. If

may not

CAD

have a

requirement.

design.

and

I've never had a customer stop buying from us over quality or procedures issues. We do have a bunch of testimonials and a few physical awards. Most importantly, the purchase orders keep coming in.

One of my lifetime goals has been to design as much good, diverse, challenging, and fun electronics as I can. That goal is directly impeded by getting tangled up in ruts and regulations. So I avoid them. We have, a few times, licensed our designs to people who do want to sell to utilities, or manufacture to stupid ISO quality rules [1,2], nonsense like that. We are willing to take their money.

John

[1] For an extra fee, we'll troubleshoot the units that they build wrong, and don't work. [2] One of our licensees wouldn't buy directly from us because we're not ISO9000. THEY aren't ISO9000! They get testy when I point that out. They send the prints and a rack full of our test set to various contract manufacturers who are ISO. They ship the test rack back to us for annual calibration, at huge expense.

1234567AA (the

second 1234567AB,

everything in

1234567AB will just be

#1234567A.

one versus

the board. If

may not

CAD

have a

requirement.

design.

and

a BOM

for an

In engineering that's ok. In production, service or QC those better be gone the millisecond Inspecteur Clousieau barges in. Once when I helped fix a problem in a production facility they paged me but I couldn't answer right away. Pulled a sticky note out of my briefcase and hung it on the scope to remind me to call back. "Please don't do that in here, might get us into trouble". Had to stick that to my briefcase where it was kosher enough.

No, but inspector might show up unanncounced which can scare the linving daylights out of anyone who doesn't maintain a tight ship in the med devices industry. "Hi, I am from the government and I want to help you" :-)

Hmm, I thought you guys did stuff for MRI machines. That would be medical devices.

--
Regards, Joerg

http://www.analogconsultants.com/

1234567AA (the

second 1234567AB,

everything in

1234567AB will just be

#1234567A.

one versus

the board. If

may not

CAD

have a

requirement.

design.

and

That's good. I did turn down vendors, had to. A typical scenario was a power supply vendor. We got (or wanted to buy) the +/-12V version which was their modified +/-15V version. Came with a schematic that was stamped "uncontrolled". It had a few resistor value changes marked in the feedback and crowbar sections but the title block said "Power Supply

+/-15V". We asked whether they could furnish a released one with a correct title block. Nope. So we looked at each other and saw no choice other than to find a new vendor.

I don't enjoy those rules and regs any more than you do. But I pretty much grew up with them because my first job out of the gate was medical devices. Later when my first aerospace designs came around they were a bit surprised that they didn't have to explain the regulatory rigamaroo to me. Because it wasn't all that different from medical except that theirs is a bit more down to earth style.

Guess it's like the military. People who have served are very used to hierarchies and things like that. For them there's really nothing to it.

ISO9000 contains a lot of hot air :-)

--
Regards, Joerg

http://www.analogconsultants.com/

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

one versus

the board. If

may not

the CAD

have a

requirement.

design.

and

B

It's a BOM

for an

Most of that product line is NMR, which is analytical chemistry. We make two gradient drivers for imaging, but they are not used on humans. One driver is 3-axis at 20 amps, which is used for micro-imaging, essentially 3D microscopy using nuclear resonance. The other is a 120 amp driver, used three at a time for small-bore (like,

8" room temperature bores) imaging, mostly for lab animals. Nowadays a lot of animal drug testing is "non destructive", namely you MRI image the critter to see how its liver is doing, rather than killing it.

John

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

one versus

(a

the board. If

may not

bare

the CAD

have a

requirement.

file

the

that

design.

those

P/N and

rev B

It's a BOM

for an

That may be the reason why they didn't insist on strict adherence to the rules. Normally some of their QC guys would have had to come out and audit you but possibly this stuff is exempt. Although once when I was tasked with a patient interface for animal use they discovered in the middle of the game "Oh dang, it has to be IEC 601 compliant". Luckily I never design them any other way, mine are even all defibrillator-proof.

It greatly depends on where it's used. Scientific lab tests are one things, lab tests in preparation towards a cert for human use (called

501(k) or when totally new a PMA) are a very different ballgame. That's where, in a regulatory sense, the gloves come off.

--
Regards, Joerg

http://www.analogconsultants.com/

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

one versus

the board. If

may not

the CAD

have a

requirement.

design.

and

B

It's a BOM

for an

latest

the

...and I answered. The BOM changes and a new assembly rev/ECO is issued.

ECO the assembly. The raw board doesn't change at all. The schematic controls the raw board.

The one in my desk drawer (actually in a notebook above it). If service wants a copy, I'll give them one. The *controlled* one in Agile doesn't because that would mean a change to the board rev/ECO. Can't do that!

The controlled document isn't current but that's the price one pays for keeping it under control. We *knew* which version to update for the next board rev. Before we did this, it was impossible to know which rev to use and there were too many chances to have an uncontrolled schematic change. Now there is _only_one_ schematic that controls the board. There may be many BOMs. If there was a schematic with each BOM rev who guarantees they're the same?

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

one versus

(a

the board. If

may not

bare

the CAD

have a

requirement.

file

the

that

design.

those

P/N and

rev B

It's a BOM

for an

latest

the

Ok, that really answers the question. If you guys did med stuff or things like that you'd likely run into major problems.

In my line of work the respective ECO does. There is a current schematic for each kind of unit ever produced and it is properly ECO'd. Having an uncontrolled schematic out there in the service or production areas in the kinds of businesses I often consult for would have a similar effect as blowing past a cop at 100mph, on a motorcycle with a sawed-off exhaust.

--
Regards, Joerg

http://www.analogconsultants.com/

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

one versus

(a

the board. If

may not

bare

the CAD

have a

requirement.

file

the

that

design.

those

P/N and

rev B

It's a BOM

for an

latest

the

That's almost exactly what we do. The values on the schematic are advisory, as there can be multiple BOMs that call out different stuffing options, or ECOs that change things with varying effectivity.

Our schematics have a note "ECOs may apply" that gets us off the hook on values.

In Joerg's case, he releases a different schematic for every possible board configuration, every resistor value change. I assume that, if some resistor or IC isn't stuffed, it's disappeared from that rev of the schematic. Whole sheets could disappear. I suppose blue wires show up as regular connections. I can't imagine how he controls the schematic tree, namely which schematic you pick up as the baseline for a new one, or which one you keep aligned with the bare board, if any.

John

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

one versus

(a

the board. If

or may not

bare

the CAD

have a

requirement.

file

the

that

design.

those

needed.

P/N and

rev B

It's a BOM

for an

latest

the

documentation

That's exactly the mess we got into before we tied the schematic to the board. What if that schematic (the one with the disappeared components/pages) was used for the next board ECO? You're now completely out of control. Not to mention simple mistakes that can happen when changing components on "live" schematics. The *only* way to do it was to have the schematic control the board and BOMs control the assemblies. That's not to say that a schematic doesn't have BOMs for several variations encoded in it. They're not valid after the first ECO to the assembly, though.

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

one versus

(a

the board. If

or may not

bare

the CAD

have a

requirement.

file

the

that

design.

those

needed.

P/N and

rev B

It's a BOM

for an

latest

the

documentation

And we do this too. The purpose of the schematic is a method of design entry to generate the netlist for the bare board.

In our case what is *released* is the gerber file set (for the bare board) and a BOM (for the assembly). In our case the first release process also generates a "base" BOM to populate our BOM / parts database. We typically then have to add a few things by hand which do not appear on the schematic, fasteners, heatsinks and so forth. This then forms the released BOM, which defines that release of that particular *assembly*.

But in our case the assembly is a different part number from the bare board and they have separate revision control. It has to be since one appears in the parts list of the other!

Part value changes result in a new BOM revision and perhaps a new assembly part number (depending on interchangeability with existing parts). But the schematic is not changed and the bare PCB revision is not changed.

Our database generates the parts list for the PCB assemblers, they never use a schematic.

If an accurate schematic were ever to be required for a particular

*assembly* then this would be a separate document, in the same class as illustrations for manuals perhaps depending on the intended purpose.

Well I suppose there are procedural ways to do that.

I think all he is really doing is the extra work to always generate a schematic for each shipped variant. The one that generated a particular gerber set could be identified easily enough, giving it same part number as the bare board perhaps, and the variants the part number of the assembly. Or some other way.

I'm not dogmatic on any of this stuff yet - still wondering myself if we are doing this the best way, which is why I have been following this thread. We don't work in a medical/aerospace environment, so can adopt any reasonable system.

--

John Devereux

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

different one versus

BOM (a

documenting the board. If

or may not

bare

the CAD

do) have a

requirement.

file

the

that

design.

those

needed.

P/N and

rev B

It's a BOM

document for an

latest

of the

first

documentation

wants

and

No, the resistor will still be on the schematic because it has a location on the board. It will just says "N.I." or something for "not installed".

How? That has never happened in my whole career.

Not 100% regular, they are marked as rework. The main purpose of a schematic is to enable someone skilled in the art to understand the inner working of the board. Service, final testers in production, new engineers, QC employees, and so on. If the schematic is wrong because it is old and tied to a board rev level that doesn't reflect today's reality then mistakes can be made and auditor will (rightfully) flag this as deficient.

It does not matter which one you pick to create a new rev, regardless of whether that will be based on the existing layout or it is planned to do a new one. Because _all_ existing schematics are properly released documents and have a full design history trail. Usually the best avenue is to take the latest. Which, in your scenario, you don't even have unless it's an uncontrolled document.

Think of it as a family tree that is 100% complete and goes back all the way to Abraham and his wife. That is what US government agencies typically expect to find.

Take a few schematic from modules you guys buy. Power supplies and such. You will see that they (usually, at least the med grade stuff) document the same way I have outlined it. On older schematics there is typically a field in the upper right or lower left corner that outlines the various revisions this schematic has had, often with a reason summary line. On new stuff this is stored electronically, tied to a specific ECO, and can be printed out.

That usually points to a major problem in the documentation system of a company. Not that unusual, and can be fixed. And should be fixed.

Nothing gets out of control if done correctly.

IMHO that is not the correct way. If there is no live schematic for the exact configuration of a produced unit one can only hope that it is not a product in a mission critical application. That's one reason why auditors roam the lands, to make sure corrective action happens in case they find this situation. Their visits are not always announced.

Not valid typically means big red flags and big problems after the Federales were there. Unless you are not in a highly regulated market _and_ nothing happens. But if something bad does happen ... oh boy ...

--
Regards, Joerg

http://www.analogconsultants.com/

1234567AA (the

the second 1234567AB,

Gerbers, and everything in

1234567AB will just be

#1234567A.

different one versus

BOM (a

documenting the board. If

or may not

bare

and the CAD

do) have a

requirement.

file

are the

that

the

original design.

those

needed.

P/N and

rev B

boards

It's a BOM

document for an

in

latest

of the

first

documentation

wants

and

Yes, but mistakes happen. That mistake will be carried forward to the next board spin and there will be no "location on the board".

When you delete them because that feature isn't used.

Even if that one has a mistake that doesn't matter and can't be detected in that revision.

We have that. For assemblies, it's the BOM. I would argue that you *don't* have that because you don't know from which tree the schematic came from. Not to mention the "mistake" problem, above.

Medical stuff uses medical processes. Ain't that unusual? ;-) I think your "medical processes" are quite lacking.

It was, with the ECO system John and I have described.

Mistakes happen. If it's caught, no harm. You can't catch such errors and eventually they *will* get you.

Why? The schematic is the input for the layout. It only has to be absolutely correct for that operation. The BOM controls different manufacturing processes and, likewise, has to be absolutely correct for that operation. Other operations are much looser and some latitude is OK.

Why? A value error on the schematic doesn't cause anything bad to happen, other than *maybe* to inconvenience an incompetent tech. The documentation that matters is in perfect order.

1234567AA (the

second 1234567AB,

and everything in

1234567AB will just be

#1234567A.

different one versus

BOM (a

documenting the board. If

or may not

bare

the CAD

do) have a

requirement.

file

the

that

design.

those

needed.

P/N and

rev B

It's a BOM

document for an

latest

of the

first

documentation

Correct. It also conveys an understanding of the board to others but that's an advisory sort of thing.

We got around this by putting such things into the schematic. If you look in the bottom corner next to the title block you'll see things that look like a hex nut and a crude stick drawing of the board itself. These insert the P/Ns into the BOM automatically. It works out really well.

Yes, that's what we did, too.

Hey, Joerg! Make that three of us that do it that way. ;-)

Yes, and it can be an uncontrolled document because it isn't used to build the product.

How do you *guarantee* that the schematic is absolutely correct? You're saying that it is correct, so anyone wanting to ECO the widget can pick it up and run with it. What if it contains a mistake introduced when you created it?

Same here. We came to the above process because of problems *we* had with the previous process. Using this process solved those problems with the minor issue that the *released* schematic may not accurately reflect the current BOM.

1234567AA (the

the second 1234567AB,

Gerbers, and everything in

1234567AB will just be

#1234567A.

different one versus

BOM (a

documenting the board. If

may or may not

the bare

and the CAD

do) have a

requirement.

schematic file

are the

all

schematic that

the

original design.

those

needed.

separate P/N and

new rev B

boards

It's a BOM

document for an

in

latest

of the

first

documentation

wants

be

and

the

That's why it is important to check the netlist against the schematic. Yes, originally by hand, because netlist routines do glitch. Some schematic programs (we know which ones ...) have a habit of not connecting somthing that is connected on the screen and even in the print-out.

The least one can do is vet the netlist electronically for consecutive board version to make sure nobody accidentally deleted R92 or something. Easy enough to do.

If a mistake happens and isn't caught, the board is spun and you haven't used R92 in the last 15 years, it would probably not matter but is against the rules. Such a mistake usually means procedure was not followed. Thou shalt not rubber-stamp an ECO :-)

Not allowed. The schematic must contain all part locations that are on the PCB, whether stuffed or not.

If the schematic is wrong that means the netlist has not been verified. In this day and age of the computation machine there is no excuse for that. It is a major violation of procedure. Following procedures is important, any respectable pilot will tell you that.

The shocker to me, when talking to some engineers (not my clients'), is that some companies don't seem to have procedures in engineering. Only their accountants do. That gives me goose bumps.

Heck, I even have a procedure right here that outlines how to prepare the tax packages for my CPA, in great detail.

We always know exactly where that schematic comes from, as I have explained. This is a requirement in the medical device industry and I've been involved in numerous audits. All passed with flying colors.

You don't seem to understand how they work, that's why you think that. The government folks who audited us all signed off on these procedures. Not just US authorities. They know a thing or two about this. So what could be lacking if they did not think so?

board.

Which I seriously doubt will pass muster with the auditors we had in house over the years. But it ain't my employer, it's yours :-)

Ok, let me tell you a trick how to avoid the proliferation of schematic mistakes: You generate a netlist from the suspicious schematic, or better yet, all schematics the original schematic has "fathered". Then you compare that to the netlist used for the layout. Is that really too difficult or onerous a task?

If this is standard procedure (which it should be) then it's a non-issue because it gets done every time an ECO is signed that contains a new schematic variant.

Not true. Which schematic does the final test guy use? Which one does the guy at the board repair station use? Something with "R92 is 14.3k here" scribbled on by hand? Finding that in the shop is bull's eye for an inspector.

Highly doubtful that an inspector would allow a BOM to be the only documentation for a stuffed circuit board. Legally it might even be ok but the guys at final test and board repair will cuss about it. They'll want current schematics. You cannot understand a complicated circuit from a BOM and in many industries they are only allowed to have controlled documents on their desks. So how are you going to accommodate them?

--
Regards, Joerg

http://www.analogconsultants.com/

1234567AA (the

the second 1234567AB,

Gerbers, and everything in

1234567AB will just be

#1234567A.

different one versus

BOM (a

documenting the board. If

or may not

bare

and the CAD

do) have a

requirement.

file

are the

that

the

original design.

those

needed.

P/N and

rev B

boards

It's a BOM

document for an

in

latest

of the

first

documentation

wants

and

So I'll have to ask again because nobody answered that yet:

I assume we all agree that board repair techs, final test, QC and possibly service personnel all need "understanding conveyed to them by means of the schematic". Now which schematic will you put into their hands? An uncontrolled version?

That still doesn't make it right :-)

"But, but, your honor, I was merely following a dozen cars who all did the same speed and we even had some passing us!"

The final test guy is not part of the build team? The board repair tech isn't either? QC who pulls every umpteenth unit to verify some stuff is not part of the production process? Those guys all make decisions based on the doucments they have at hand. Now it becomes spooky ...

Dudes, careful, there's cracks in the ice you are walking on :-)

Then someone did not do the netlist check that was (hopefully) part of the procedure. And the manager who signed off on the ECO didn't catch that. Neither did all the other guys who signed the ECO. Very unlikely, unless a strong anesthetic was piped into the HVAC duct work of the building at that time.

Ok, there was a guy at Caltrans who allegedly rubber-stamped his own measurements on the tower foundations for the new Bay Bridge. Then obviously his boss rubber-stamped as well. If true, and according to some reprt it looks like it might, that was illegal activity. If they had followed process it wouldn't have happened. Raw data was allegedly discarded, and that could have been prevented if proper procedures were in place for that. In medical devices we make sure they are.

That could be fixed :-)

In the end it depends on your bosses and the amount of risk they are willing to take. Or their ability to see the risks in the first place. You might never see auditors because your products aren't mission-critical. Let's just hope nothing ever happens, like a fire or whatever. Because then someone will come and audit your procedures whether you like it or not.

--
Regards, Joerg

http://www.analogconsultants.com/

1234567AA (the

the second 1234567AB,

Gerbers, and everything in

1234567AB will just be

#1234567A.

different one versus

BOM (a

documenting the board. If

may or may not

the bare

and the CAD

do) have a

requirement.

schematic file

are the

all

schematic that

the

original design.

those

needed.

separate P/N and

new rev B

boards

It's a BOM

document for an

in

latest

of the

first

documentation

wants

be

and

the

The only schematic that can ever be printed is the released one. It's released as part of the PCB layout/Gerber package, all at a single rev letter.

A named PCB assembly is defined by its BOM, which calls out the bare board and all the parts. The parts reference designators are on the schematic and screened on the board, so they are locked to one another. The parts presence, types, and values are controlled by the BOM for each assembly version, so the values on the schematic may not be the values actually installed on a board. Everybody knows this, and it's noted on the schematic.

More complex assemblies have their own assembly number and BOM, and can call out PCB assemblies as subassemblies.

I can't stop people from scribbling on paper schematics, but the scribbles control nothing. Nobody is authorized to assemble anything that is not in conformance to released documents. Well, unless I say so.

John

1234567AA (the

the second 1234567AB,

Gerbers, and everything in

1234567AB will just be

board #1234567A.

different one versus

BOM (a

documenting the board. If

may or may not

the bare

and the CAD

do) have a

a requirement.

schematic file

are the

all

schematic that

the

original design.

same

of those

needed.

separate P/N and

new rev B

boards

board. It's a BOM

document for an

working in

the latest

integrity of the

those

first

documentation

wants

be

use and

the

That becomes very cumbersome if, for example, you'd do that to a filter board that comes with umpteen passband frequencies.

How can you print component values for new releases onto the schematic if it is still the same released rev A schematic as the PCB is? You wrote right there above that the only schematic that can ever be printed is the released one. So how does this new information get on there?

And that's what's different in very regulated businesses such as medical devices. The managers in production and in the clean room pay very close attention that this does not happen.

If it ever does happen and an unannounced inspector waltzes in that's just like violating the open container law. If the cop finds an open bottle of booze in the car it's the end of the journey. No matter how often the driver assures the cop that he never took a sip.

You could write and release an ECO that says that a "Larkin edict" spoken before any trip to Zeitgeist has the authority to release anything. Who knows, then it might be legit :-)

--
Regards, Joerg

http://www.analogconsultants.com/

1234567AA (the

the second 1234567AB,

Gerbers, and everything in

1234567AB will just be

board #1234567A.

different one versus

different BOM (a

documenting the board. If

may or may not

the bare

it

and the CAD

typically do) have a

a requirement.

schematic file

are the

all

schematic that

and the

original design.

same

of those

needed.

separate P/N and

new rev B

boards

board. It's a BOM

document for an

working in

in

the latest

integrity of the

are

those

first

documentation

wants

because

not be

use and

Now

the

We do that. Each version of the board has a BOM.

Like this:

ftp://jjlarkin.lmi.net/Filter1.jpg

ftp://jjlarkin.lmi.net/Filter2.jpg

V451 is the model number, which is officially assembly drawing 22A451. It's rev A. the -51 version is a 50 Hz Bessel filter, as noted in the

22A451.51A BOM file. How else could you have multiple versions of a filter board, if you didn't have a BOM for each?

This was my original design, which The Brat entered into PADS as the formal drawings.

ftp://jjlarkin.lmi.net/FilterBoard.jpg

It's a Sallen-Key with chopper amps, which means its offset is microvolts and its DC gain is precisely 1, so the filters can be changed in the field without requiring main board recalibration.

I've explained, about 40 times now, that we don't issue a new schematic just because we change component values. The BOMs control values.

The only time we change a schematic is if we intend to rev the PCB layout, at which time we roll the rev letter of the schematic, the bare board, and the PCB assembly, all together, all the same letter. And we issue new BOMs at that rev letter.

The parts values that are on a schematic reflect the base (-1) version of that PCB rev at the time of release.

You

If you mean parts values, it doesn't. The BOM is the formal definition of which parts are loaded. A given assembly can have several, or many, dash numbers, each with its own BOM.

Suppose you had a board like this, with, say, 20 possible cutoff frequencies, and a choice of Bessel or Butterworth. That's 40 variations already. Would you release 40 schematics? What would you call them, given only 26 letters in the alphabet? And what would you do if the board was changed a bit, creating 80 versions? That sort of thing is easy for us.

You don't allow people to mark anything on a printed schematic?

The CEO of my company is the final arbiter of any quality type issues. Lots of companies have formal provision for some person to be the tie-breaker for things like this. My QC person comes to me now and then to see if something is sellable, and we talk it over. And I can authorize someone to, for example, change a resistor if a board needs it, or epoxy something maybe, or repair a lifted trace.

We had a case like that a couple days ago. A plastic card extractor can, at extreme travel, swing over a ceramic cap and shear it off. They wanted to trim the plastic and wanted permission. Turns out it's just a bypass cap, so I told them to remove the cap instead. That decision got formalized into an ECO, and it's noted in the product's NEXT file, so the layout will get fixed next rev. In some one-off situation, I could say "just do it."

John