EEVblog Electronics Community Forum
General => General Technical Chat => Topic started by: axemaster on November 20, 2019, 06:29:48 pm
-
Hi all, this is a rage post about Bottom View in datasheets.
Over the past 2 years I have probably lost around $7000-10000 and weeks of development time due to datasheets using Bottom View for their pinouts. They do this, I am distracted and fail to notice, it gets through design review unnoticed, and the result is a PCB with mirrored footprints and a total disaster.
Yesterday I assembled and started testing a 30W power supply I designed for use with a radar system computer. And guess what? Some of the MOSFETs were mirrored! With the result that the power supply shorted out!
I have complained about this before to companies, and they have actually updated and fixed their datasheets in response. So people understand that Bottom View is a mistake.
I know a lot of industry people are here on EEVBlog. I hope you will have a conversation with whoever reviews datasheets at your company, and a rule is put in place that Bottom View is never tolerated.
It would also be great if we could start posting links to datasheets here, whenever we encounter Bottom View. To call some attention to this very real and frustrating problem!
Parts with bottom view problems!
https://www.diodes.com/assets/Datasheets/DMP2006UFG.pdf (https://www.diodes.com/assets/Datasheets/DMP2006UFG.pdf)
https://www.analog.com/media/en/technical-documentation/data-sheets/1460fc.pdf (https://www.analog.com/media/en/technical-documentation/data-sheets/1460fc.pdf)
-
I've had some JST connectors nearly get me. The footprint was bottom view, but wasn't labeled as such. Luckily the 3d model tipped me off when the alignment post didn't match.
I guess the JST wire to board connectors get used a lot in cheap consumer single sided PCBs enough that "copper side" or whatever it was labeled made sense with the right context.
-
Yes, the vagueness of some data sheets in that regard is very annoying. Because of that, we've taken to reviewing them carefully and determining which view is what, and then noting directly on the data sheet next to the view(s) whether they're top or bottom, and also clearly labeling pin 1 of the device (and a few others if necessary to be clear on the counting direction/order).
Footprints are also double-checked against the datasheet by at least one person in addition to whoever generated it. Between all of this, errors are normally caught before things go out to fab. The only thing I can think of that has burned us recently was an SD card socket where we had an attack of dopeassity and placed it 180* out so it faced the wrong way on the board and the card couldn't be inserted.
-Pat
-
Both the parts you link to have either "top view" or "bottom view" clearly printed next to the relevant diagrams.
That you fail to notice this and, to quote you, "They do this, I am distracted and fail to notice, it gets through design review unnoticed," strongly suggests that the problem isn't with the data sheets but with your self confessed inattention compounded with a design review process that clearly fails to do its job.
How manufacturers present their data sheets isn't the problem here and I don't quite see why you fail to see that.
If I was your customer/employer and it was my money you were spending and I saw you post "Over the past 2 years I have probably lost around $7000-10000 and weeks of development time" on here and attribute it as "due to datasheets using Bottom View for their pinouts" and then admit you'd been inattentive and your design review process failed, then you'd be in line for a very 'difficult' conversation.
-
I agree with your main point, having to mentally flip things in your head is asking for trouble, always design from the top down or front to back when you can...
But I also agree with Cerebus here, your examples are just not convincing enough. The diodes image is a very clear 3D model, its not an ambiguous 2D view.
The Analog datasheet is worse, its an ambiguous image, but at least they do clearly write "bottom view". I'm thinking that traditionally TO-92 pinout was shown as bottom view (google image search seems to suggest this), and LT being old school keeps with that standard.
-
I am distracted and fail to notice
Sounds like you need to pay more attention rather than complain... :box:
-
:-DD Can you let me know how I can apply for your job, because I reckon I could save your company $7000-10000 and weeks of development time for every design I would be involved with!
Everyone's talking about the DMP2006 which is blindingly obvious but the LT1460 datasheet contains two entries for the word "bottom". One is for the TO-92 package, which surely you aren't using as it's beyond obsolete. The second, for completeness, is a note saying:
5. LEAD COPLANARITY (BOTTOM OF LEADS AFTER FORMING) SHALL BE 0.102mm (.004") MAX
-
While it is irritating when datasheets are not clear, I'm going to have to agree that the problem here is largely your lack of attention to detail. If something is not clear, get a second set of eyes on it, contact the manufacture, or order a sample of the part to examine. It's the old "measure twice, cut once" saying in carpentry.
My own peeve with datasheets is when the dimensions are not fully labeled and they have just the bare minimum from which I can extrapolate the rest of the measurements relative to the information I am given. It is super irritating to have to go through the extra mental gymnastics, greatly increasing the chance of error.
Then there is the worst mistake I've made, more than once etching and drilling a board before realizing that I've etched a mirror image of what I wanted |O Less of an issue these days as PCBs are so cheap I only etch my own if it's something simple that I want right now.
-
I've only ever made that mistake less than 5 times over my life. It's one of the first things to check right after "is it even orderable", power supplies, and operating temperature.
And the few times I made the mistake it was with parts like RF splitters with pin numbers reversed anyways (thanks, Mini-Circuits) or bizarre LEDs that put the "orientation marker" on pin 4 and the second source is precisely the opposite.
-
I've seen worse, an IC available in two different packages with the same number of pins but different pin numbers, so changing the package without changing the symbol would result in the wrong pin out
-
Those of us greybeards who "bit our teeth" on vacuum tubes had a problem with the early DIL packages, as tubes are always "bottom side".
We adapted, & can use both conventions-----if in doubt, we check.
My favourite hate was opto couplers from Siemens, which used the same 6 pin pack as everybody else, but with a different pinout.
These were not anything special, just "bog standard" optos like everybody else made.
-
My own peeve with datasheets is when the dimensions are not fully labeled and they have just the bare minimum from which I can extrapolate the rest of the measurements relative to the information I am given. It is super irritating to have to go through the extra mental gymnastics, greatly increasing the chance of error.
Yep, presumably MCAD people drawing for their own audience, expecting us to be using the same tools they are, which I totally get if you're just a low level worker.
- Mech often dimension to edges of pads, but ECAD we never use that
- Mech people are taught to never over-dimension a drawing, but for a footprint its useful
Luckily MCAD ideas are slowly starting to make their way into the PCB software. Latest Altium supports equations in input boxes, and dimensions are much better at snapping to features than they used to be.
-
I'm sympathetic to the OP. When designing a board requiring lots of new part library entries it's easy to screw things up and the "is it top or bottom view" thing has caught me before, too. Although tangentially related, take a look at the attached excerpt from a CUI DC power jack datasheet which shows one of the worst attempts I've seen in recent memory to draw a component package.
-
Sure but 7000$ buys you a lot of PCB review contract time.
-
aint it great how sometimes getting a single part down correctly on a PCB and having it work rev 1 is actually good engineering ?
attention to detail is nothing to scoff at. otherwise it would be like 'oh this board only has 200 parts its no skill'. :-DD
-
Never had a problem with this. I read datasheets intensely and if there is any ambiguity, I get it clarified.
That being said, clowns at Honeywell shipped airflow sensors with a changed pin-out but the same old datasheet contained in each box. The part numbers stayed the same. The result - about $1,300 of sensors destroyed. Element 14 (the supplier) and Honeywell (the manufacturer) both failed to take responsibility and credit the cost of the sensors. The fallout is I avoid anything made by Honeywell, and Element 14 is has lost a lot of business to Digikey over the last 9 years. Sure people and companies make mistakes, but it is how they fobbed me off that is the problem. Ironically the company I was working for a the time is now owned by Honeywell :palm:.
Been very happy with Digikey... preferred R& D provider. They do a great job and their website is clearly superior. Delivery speed to Oz is outstanding too.
-
Yesterday I assembled and started testing a 30W power supply I designed for use with a radar system computer. And guess what? Some of the MOSFETs were mirrored! With the result that the power supply shorted out!
For through hole components, you can simply mount them on the other side of the board :)
Apart from that, your design review process is (a) ineffective and (b) doesn't improve in the light of experience. That's the second thing you should fix.
-
I've seen worse, an IC available in two different packages with the same number of pins but different pin numbers, so changing the package without changing the symbol would result in the wrong pin out
Recently on the forum I have seen a SOT23 transistor datasheet with the same pinout as any other but the pins numbered backwards :-//
-
I'm sympathetic to the OP. When designing a board requiring lots of new part library entries it's easy to screw things up and the "is it top or bottom view" thing has caught me before, too. Although tangentially related, take a look at the attached excerpt from a CUI DC power jack datasheet which shows one of the worst attempts I've seen in recent memory to draw a component package.
Yes, that's a very sucky drawing. It's the kind that makes sense if you've physically got a sample part in hand and can look at it while you figure out the drawing, but is about as much use as chocolate teapot if you're using it to decide whether to get the part or not.
-
a CUI DC power jack datasheet which shows one of the worst attempts I've seen in recent memory to draw a component package
What's the problem with that specifically?
-
Apart from that, your design review process is (a) ineffective and (b) doesn't improve in the light of experience. That's the second thing you should fix.
Ah, I agree. Mirrored footprints can be pretty tough to spot though, unless it's specifically part of the routine checks that are done through your design reviews (which is more often than not, not!)
Specifically for PCB footprints (and generally speaking, CAD parts), I've found that setting a specific review for them (generally speaking, a design review on CAD libraries) was a good idea. The key process there should be that no layout can start on a given board if the parts that have been created for this specific project haven't been reviewed first. If you make it part of the general design review, it's a safe way of overlooking that IME. Most teams will focus on the schematics, and then next on the placement and next, routing. A specific showstopper phase for libraries works better.
-
Apart from that, your design review process is (a) ineffective and (b) doesn't improve in the light of experience. That's the second thing you should fix.
Ah, I agree. Mirrored footprints can be pretty tough to spot though, unless it's specifically part of the routine checks that are done through your design reviews (which is more often than not, not!)
Specifically for PCB footprints (and generally speaking, CAD parts), I've found that setting a specific review for them (generally speaking, a design review on CAD libraries) was a good idea. The key process there should be that no layout can start on a given board if the parts that have been created for this specific project haven't been reviewed first. If you make it part of the general design review, it's a safe way of overlooking that IME. Most teams will focus on the schematics, and then next on the placement and next, routing. A specific showstopper phase for libraries works better.
That's sane.
I know that you can put PTH devices on the other side of the board, because I saw someone have to do that 30 years ago. The draft pre-production documentation had the devices drawn from the PoV of the chip designers, not the board designers.
-
a CUI DC power jack datasheet which shows one of the worst attempts I've seen in recent memory to draw a component package
What's the problem with that specifically?
How about: what are the coordinates for the locating pins? The only data unambiguously given are the diameter of the holes and their relative displacement on the y axis.
-
How about: what are the coordinates for the locating pins?
Ah, yes, that's a little bit annoying :)
-
I know that you can put PTH devices on the other side of the board, because I saw someone have to do that 30 years ago. The draft pre-production documentation had the devices drawn from the PoV of the chip designers, not the board designers.
Ahah, that certainly happens.
I've experienced way worse; the layout technician used a mirrored footprint for a flip-chip (with, mind you, 200µm pitch!). Another, very talented technician we had managed to solder the bumped die by hand with tiny wires, so we could have a couple boards to at least test the design (PCB yield time being pretty long for the kind of board) before submitting the corrected layout...
Fun times. That's after this that we set up a CAD library review process.
-
That power jack isn't too bad. Similar to this which is one I disliked but fewer unknowns/calculations to get it right. Every reference just seems random, luckily they had a lot of measurements.
-
The problem for the mechanical people is that they can't join the final dot, so to speak, because of tolerances. So there will always be a gap, albeit with sufficient other dimensions to infer what it would be, and the issue is merely where that's going to be. Plus they have to reference everything to a datum so, as someone mentioned about, the dimensions may well be perfect for them but pretty obscure for us.
The SDCard thing is tricky because the pins aren't symmetrical (dunno why - it wouldn't hurt them, surely). That dimension between the centerline of the holder and pin 4 could easily be put in, but then they couldn't use pin 8 centerline as the datum for the rest of the pads.
-
That power jack isn't too bad. Similar to this which is one I disliked but fewer unknowns/calculations to get it right. Every reference just seems random, luckily they had a lot of measurements.
Oddly enough, I don't find the SD card socket that bad, though offsetting the center pin from the centerline ever so slightly and calling out the datum for the latter is evil, I'll grant you that.
-
That power jack isn't too bad. Similar to this which is one I disliked but fewer unknowns/calculations to get it right. Every reference just seems random, luckily they had a lot of measurements.
Oddly enough, I don't find the SD card socket that bad, though offsetting the center pin from the centerline ever so slightly and calling out the datum for the latter is evil, I'll grant you that.
No it wasn't too bad but it was annoying. I rechecked it and my numbers maybe 3-4 times before calling it good. We used the CD so those pads were needed beyond just holding it down. Just an example I still had datasheet for since we still use it.