How not to write a mechanical drawing.

[T3sl4co1l]

I give this drawing a solid C-.
http://www.memoryprotectiondevices.com/datasheets/BC501SM-datasheet.pdf
It's absolutely, unfulfillingly, cromulent.

The shape is there, and you'll probably read it about as intended.

The dimensions don't mesh.  Absolutely no tolerances are noted (the terminal material can be negative thickness!).  The "PCB LAYOUT" is wrong.  Possibly.  I think.

They provide a 3D model for this part, amazingly!  It's an ugly extruded profile, with some extra bobs tacked on.  It doesn't quite match either drawing.

Tim

[Alexei.Polkhanov]

Tolerances are there - see "Tolerances unless noted" section in the stamp.

[void_error]

This reminds me of an SD card connector I used a few years ago, had to make a pattern for it and spent quite a lot of time figuring out pad locations.

[T3sl4co1l]

This reminds me of an SD card connector I used a few years ago, had to make a pattern for it and spent quite a lot of time figuring out pad locations.

Those are usually quite a mess, with the pins not in a neat row, nor a regular pitch, but all over the place, with dimensions for each.  But they're usually drawn okay.  I mean, so much of this is automated these days, you just tag the points you want drawn and there you go, including tolerances...

The most offensive tolerance in the OP example is the alignment pegs: with +/-0.25mm tolerance on diameter, there's no guarantee that they'll even fit into the recommended holes (even giving them the benefit of the doubt that the PCB layout is REF, not toleranced), let alone unambiguously (i.e., can assemble either direction).  With a +/-0.25mm positional error on top of that, the holes need to be 0.375mm oversize (plus the drill's own dia and position tolerances!), rendering them completely superfluous.

Always think clearly, and carefully, about what dimensions and tolerances you write down.

Tim

[Len]

Drawn By B.S.

Well, there you go.

[T3sl4co1l]

Tolerances are there - see "Tolerances unless noted" section in the stamp.

I said "no tolerances noted", and if I'm not mistaken, they never used that particular option...

Tim

[dmills]

See you that and raise you just about any connector datasheet ever, they are almost all just made of fail.

I mean come on guys, dimensioning is automatic these days, I should NOT have to be sitting there with a calculator going "That plus half the other minus the difference between...", also whoever thought that TWO different datums on a drawing was a good idea (TE Connectivity SFP Cages looking at you) needs a spanking.

I have a number of junk boards down to some mechanical weenie using a projection in a datasheet that apparently implies 'Bottom View' without actually saying it.

And yea, datasheet suggested layouts that fail to match the step models, just annoying.

Regards, Dan.

[suicidaleggroll]

I mean come on guys, dimensioning is automatic these days, I should NOT have to be sitting there with a calculator going "That plus half the other minus the difference between..."

I've never understood that with dimensioning in datasheets.  Why do all of them REFUSE to over-dimension a part?  So what if it provides dimensions between A-B, B-C, and [heaven forbid] A-C?  Will the world come to an end?  Why do they all provide just the bare minimum and make it a requirement to pull out a calculator to fill in the gaps?  With some complicated devices I've actually had to get out a piece of paper and draw the damn thing by hand myself, filling in all of the missing dimensions so I can draw in up in CAD properly without having to write out a 2-line algebraic equation to solve for a missing critical dimension they refused to stick on the datasheet.

[Mechatrommer]

Tolerances are there - see "Tolerances unless noted" section in the stamp.

I said "no tolerances noted", and if I'm not mistaken, they never used that particular option...
Tim

tolerance is ±0.25mm, except as noted. since there is no noted, by default use the ±0.25mm on all dimensions. if you dont want to hurt your brain, dont noted it, just use the provided dimensions as is, i dont see much problem with that.

I give this drawing a solid C-.
They provide a 3D model for this part, amazingly! It's an ugly extruded profile, with some extra bobs tacked on.  It doesn't quite match either drawing.
Tim

lucky for you. how about making your own? like so many parts that i bought off ebay.

[ludzinc]

I mean come on guys, dimensioning is automatic these days, I should NOT have to be sitting there with a calculator going "That plus half the other minus the difference between..."

I've never understood that with dimensioning in datasheets.  Why do all of them REFUSE to over-dimension a part?  So what if it provides dimensions between A-B, B-C, and [heaven forbid] A-C?  Will the world come to an end?  Why do they all provide just the bare minimum and make it a requirement to pull out a calculator to fill in the gaps?  With some complicated devices I've actually had to get out a piece of paper and draw the damn thing by hand myself, filling in all of the missing dimensions so I can draw in up in CAD properly without having to write out a 2-line algebraic equation to solve for a missing critical dimension they refused to stick on the datasheet.

It's done because of compounding errors.

Define each dimension as a distance from an origin, and the error remains constant.

For example take a 0.100" header with +/- 0.010" error, with a single line of 10 pins.

If you show each pin as being separated by 0.100", each pin could then have an error of 0.010" from the other pin, to the (ridiculous) point that the 10th pin's error could be as great as an extra 0.100" away from where you expected it to be.

BUT

If you show each pin as being increments of 0.100" from the datum, each pin is constrained to only 0.010" error.

Get my drift?

[Mechatrommer]

I've never understood that with dimensioning in datasheets.  Why do all of them REFUSE to over-dimension a part?  So what if it provides dimensions between A-B, B-C, and [heaven forbid] A-C?  Will the world come to an end?  Why do they all provide just the bare minimum and make it a requirement to pull out a calculator to fill in the gaps?  With some complicated devices I've actually had to get out a piece of paper and draw the damn thing by hand myself, filling in all of the missing dimensions so I can draw in up in CAD properly without having to write out a 2-line algebraic equation to solve for a missing critical dimension they refused to stick on the datasheet.

It's done because of compounding errors.
Get my drift?

that not what he's saying. compounding  error should not be an issue these days, unless the machineries are really fucked up. tolerance usually has something to do with "length" of a mass, not "distance" between 2 masses (or voids/holes). for example, you cant make a perfect 1mm hole, you cant make a perfect 1mm cube, etc, surely you cant make a perfect 1mm distance between 2 objects but you can make a safe bet that the error will not be compounded due to how machinery and gearing works. sometime you will find dimension so small that if you take tolerance into account, it will have -ve length, but with some sensical mind, that should not be happening, unless the machineries are really fucked up. you can make a safe bet, you only take +ve tolerance into account. drawing are usually dimensioned to their nominal (designed) value, tolerances are due to machinery or material imperfection. a wiser thing to do is redraw based on nominal values and only take tolerances into account on critical part, such as you dont want two separate components to be sitting next to each other at a distance of 0mm, put extra tolerance to that based on tolerance info in the drawing, or you'll want pad area bigger than anything probable, but can use the nominal distance given etc etc. this can be lengthy, long story short, if you usually play with this thing, you'll know which part to take care the most.

now what suicide asked is why not redundant dimensioning, few reasons, such as to avoid cluttering usually on more complex part that you'll not know which is object line which is dimension line, other reason is probably they want to show more critical dimension compared to the non-dimensioned part, so you can work out tolerances based on the more important dimension, fwiw, imho.

[dmills]

I am actually fairly sure the real reason is that the old ISO/BS standards for mechanical drawings heavily discouraged over dimensioning due to consistency concerns back when drafting was done on drawing boards.

The reality of universal use of CAD has not quite propagated to the mechanical drawing standards yet.

Still doesn't explain tosh like that SFP cage however.

While we are on the subject, how about dimensioning footprints as between centres plus pad details rather then size of pad plus size of gap? Almost everything ECAD want to know centre to centre distance when doing step and repeat.

Regards, Dan.

[Kalvin]

No tolerance needed as the parts are manufactured to match the stated dimensions

[T3sl4co1l]

I mean come on guys, dimensioning is automatic these days, I should NOT have to be sitting there with a calculator going "That plus half the other minus the difference between...", also whoever thought that TWO different datums on a drawing was a good idea (TE Connectivity SFP Cages looking at you) needs a spanking.

Fascinating.

It looks like fully 100% of all SFP related products are, one can only assume -- willfully, intentionally underspecified.

I'm looking at a 20 pin card edge connector right now, which happens to be an SFP type.  These are not even sold as card edge connectors, but as "pluggable"! 

I noticed a pattern: FCI, Amphenol, Molex, TE, all of them: all of them dimension and omit exactly the same aspects.

Huh.

So I wonder if there's an underlying standard there.

Which, "SFP" being what it is, sounds like there should be.

So, don't blame the manufacturers in this case, I suppose -- the source material itself is just that shitty.  Horse mouth:
http://www.snia.org/sff/specifications

Have fun (or not).

Naturally, the manufacturers can't sell their products as "to spec" if they include extra specs.  Same problem as, for example, the broad tolerances on plain old TO-220 (but not TO-220AB, which is okay), or the SOD-214 series.

(Specific problem I'm having: no one provides a pin length dimension, so I cannot calculate the solder heel joint for ANY of these parts.  Disturbingly, the standard itself provides PCB layout, and I guess manufacturers are entrusted to design their parts so that a satisfactory joint is formed.  This is the reverse of normal order, and I find it somewhat disturbing.)

Tim

[dmills]

Naturally, the manufacturers can't sell their products as "to spec" if they include extra specs.  Same problem as, for example, the broad tolerances on plain old TO-220 (but not TO-220AB, which is okay), or the SOD-214 series.

I had not spotted snia.org before, so thank you for that, looks somewhat scarily like it has useful things.

The other nightmare package for this sort of thing is the humble D2Pak, every one of the damn things is different at least in the amount of metal on the bottom.

SFPs, also things like SD card sockets (Another complete pain in the tail), and any circular connector that is PCB mount....

Regards, Dan.

[T3sl4co1l]

The other nightmare package for this sort of thing is the humble D2Pak, every one of the damn things is different at least in the amount of metal on the bottom.

Ah yep, only thing you can hope for is to get most of it!

Related subject: PDSO-8 style packages.  I don't think there's a JEDEC standard for these (or, there are many...).  Naturally, this leads to a great diversity in packages you'll actually see.  The biggest variation is in the metal sticking out the side of the 'tab' part: sometimes it's indicated on the drawing, often it's not dimensioned, and never is it indicated on the PCB layout (as a region for soldering pad area, or for keep-out to avoid shorts!).

And probably, manufacturers don't want to specify anything additional, as their packaging source may change as well...

SFPs, also things like SD card sockets (Another complete pain in the tail), and any circular connector that is PCB mount....

Ahh, speaking of -- there are a number of connectors where it's impossible to produce a "safe" footprint, per IPC-2221A or per the board fabricators' design rules.  1mm pitch THT headers are one case.  The holes can't simultaneously be:
- Large enough i.d. to fit the pin tail (including pin and drill tolerances), and
- Enough annular ring to meet IPC/fab rules (usually 5 mils), and
- Enough clearance (pad to pad edge) to meet fab rules (usually 6 mils clearance).

THT headers aren't usually very dense (e.g., 2 rows), so routing them at least isn't horrible.  I've seen circular connectors where the above three are already challenging enough, but there are pins inside a surrounding ring of pins, which are essentially impossible to route!  Four layer boards, with custom pad stacks, are seemingly required for these beasts!

Tim

[george.b]

Drawn By B.S.

Well, there you go.

Approved By T.B.

Approved by The Boss?

[ChunkyPastaSauce]

Why do all of them REFUSE to over-dimension a part?  So what if it provides dimensions between A-B, B-C, and [heaven forbid] A-C?  Will the world come to an end?

Yes 

Take your fully defined example:

Then tolerance:  (AB = +-0.1) AND  (BC = +-0.1), result (AC = +-0.2)


Now over-define the drawing by including the derived AC dimension........but closer look:

Consider one top dimension and the derived dimension

Tolerance: (AB = +- 0.1) AND (AC = +-0.2), result (BC = +-0.3) ... uh-oh
Notice BC doesn't match original drawing! Now BC tolerance is +-0.3 rather than +-0.1, the meaning is very different

The drawing lost its consistency by including the derived AC dim... because one lost the information on the dependency order of the tolerance (always exists) stack [when you over define, the solution structure becomes circular, you throw out the starting point of a stackup...which matters].

[suicidaleggroll]

So stick "derived" dimensions in italics or similar, it's not an insurmountable, or even difficult problem to solve when making the drawing, and not doing so makes using the drawing a PITA...

[ChunkyPastaSauce]

As soon as you use a derived dim, then the tolerance stack is likely broken. In the above example, first and third drawings have different meanings, and the second is inconsistent.

Even if you are doing work with just nominal dimensions without a tolerance...consider just 6 feature A>B>C>D>E>F, then you would have to include derived dims AC AD AE AF BD BE BF CE CF DF, so a drawing with 6 dims turns into one with 15!
It explodes the more features on the drawing (simple part with 15 dimensions explodes to 105 dims)
 

[komet]

I make a point of always calculating the derived dimensions I need in my head. After doing a couple of connectors my mental arithmetic is top notch for several weeks.

I once finished a library part (a convoluted connector, as I recall), went out to the shops and immediately spotted a pricing inconsistency.

[Pjotr]

I make a point of always calculating the derived dimensions I need in my head. After doing a couple of connectors my mental arithmetic is top notch for several weeks.

I once finished a library part (a convoluted connector, as I recall), went out to the shops and immediately spotted a pricing inconsistency.

Same experience here. And when you put the pricing inconsistency on the agenda of the weekly group meeting, an opinion inconsistency pops up   

[T3sl4co1l]

For instance, this is why pin pitches are defined as "0.50 REF (x N)" and the total is toleranced.  No single pin can be off by more than the tolerance, because that would be awful, anyway!

This is part of my (legitimate) complaint in the OP, that they provide the pitch of two pins, yes, with a cumulative pair of wide tolerance measurements -- the worst case error is as great as the radius of one pin!

Tim

[suicidaleggroll]

As soon as you use a derived dim, then the tolerance stack is likely broken. In the above example, first and third drawings have different meanings, and the second is inconsistent.

You're right, the first and third do have different meanings, nobody is suggesting they change which dimensions are being specified, just add derived dimensions in addition to the rest.  The second is not inconsistent at all, it makes perfect sense.  A-B is 1.00 +/- 0.1 and A-C is 2.00 +/- 0.2, but they can't be +0.1 and -0.2 respectively at the same time or it would violate the third dimension of 1.00 +/- 0.1.  There is no inconsistency and no confusion.  It's the exact same result you'd come to with only A-B and B-C labeled, without the need to calculate A-C yourself, and did adding that third dimension really make the drawing so cluttered that it's unusable?  No, it looks just fine.

Even if you are doing work with just nominal dimensions without a tolerance...consider just 6 feature A>B>C>D>E>F, then you would have to include derived dims AC AD AE AF BD BE BF CE CF DF, so a drawing with 6 dims turns into one with 15!
It explodes the more features on the drawing (simple part with 15 dimensions explodes to 105 dims)

Talk about making mountains out of mole hills...Jesus
It's not an all-or-nothing thing.  All they have to do is use some common sense and add the handful of derived parameters that actually matter to someone trying to use the part (overall length/width of the part, distance between center of holes/pads vs edges, etc).  Most parts only need a few, and it's the same few that pretty much everyone who uses the part has to go out of their way to calculate themselves.  Even if they don't get all of the ones that matter to everyone, anything is better than nothing.

[suicidaleggroll]

While we are on the subject, how about dimensioning footprints as between centres plus pad details rather then size of pad plus size of gap? Almost everything ECAD want to know centre to centre distance when doing step and repeat.

Because according to some posters here, even though these are the exact same values that everybody who uses the part calculates themselves and enters in their CAD package, for some reason printing them on the datasheet would break the tolerance stackup, make the part unusable, and the world would come to an end.