Basic PCB Design - Footprints - Match exact mech specs or round

[pigtwo]

Hello everyone!

I've just started working with OrCAD(and PCB design in general) and I've been creating some footprints for a basic board.  Often I've been finding that the mechanical specs will be in mm but I've been told by my boss that their designs were usually done in mils.  In the conversion from mm to mils, rarely does it come out to a clean number.  So my first thought was to just use as exact specifications as possible and let the manufacturer do the rounding.  But then I remember reading something Dave wrote a while ago to always have things on a grid and so I did it as though it had to be on a 5 mil grid rounding up or down accordingly. 

Which is the correct thing to do?  Are there any tips on how to do this with the least number of problems. 

Thank you!

[Landrew2390]

DISCLAIMER:  I'm a hobbyist and failures of my designs don't cost anyone money except myself.

Now that I've got that out of the way, I don't have a solid answer.  I use Eagle and draw the parts using whatever dimensions are listed on the datasheet.  I prefer to use mils and I design in mils, but like you, I've discovered converting millimeters to be troublesome.  The only problems I see with rounding the numbers is the potential for placing pads too close together or the pads coming out slightly undersized.  This can be checked quite easily, but it's another step that must be done for quality assurance purposes.

I don't know what software you're using, but I would check and see if it will allow you to draw the parts using metric and then install them on a design using imperial measurements.  Some software will do this and some won't.  As always, your mileage may vary and nothing will work 100% as advertised.

[T3sl4co1l]

Tell him mils are defined in terms of meters and lay it out how you see fit.

Rounding to 0.05mm (~2mil) is in the IPC spec, IIRC.  Pad size is almost always better larger than smaller (especially if wave or hand soldering is a desirable feature, the "toe" dimension should be made much larger).  Again, keep it within IPC spec and you'll probably be fine.  BTW, some manufacturers give footprints different from IPC guidelines; whether these are good or not is debatable.

Specifically, you should have IPC-2221 (general, PCB, and THT) and 7351B (SMT) handy.  They're worth getting, and I think can be found online.

Personally, I do the footprint in controlling units (in/mm as stated) and almost always do the PCB in mils.  In Altium, this is merely personal preference; you can switch units on the fly, so that in the end, it doesn't matter what is actually used.

Tim

[sleemanj]

Surely it's just a matter of changing the units when the boss is looking.

Boss: What's this 2mm! We don't take kindly to using metric around these parts.
You:  [ Edit > Units > Mils {or whatever Orcad has to change the units, I don't know Orcad} ] There, fixed, see a nice 78.74 mil.

If the measurements are provided in metric, design your footprint in metric.  The fact that it's not going to be a "round" number in mil is preferable to having a footprint that doesn't match the real measurements (even if it is by a small amount).

If you're talking about hole sizes mind you, there is a consideration there, if your fab is going to use drills (or router bits) sized in mil (and you'd have to ask the fab that, it might as well be that they actually use drills sized in metric) you might want to specify the finished hole size in mil (in a size you know the fab has) instead of metric so you know what you will get.

[pigtwo]

Great!  Thank you for the replies.  That's the answer I was hoping for.  Knowing about those IPC documents is really helpful.  I was going to a little too far with my rounding.  I was trying to fit it to my grid which in hind sight was kind of dumb.  Especially when my grid was 25 mil. 

I think I'm going to stick with mils for now, my boss doesn't really care I don't think.  He just said the previous engineer did it in mils. 

If I did use a lot of decimals, the manufacturer would just round them and not complain right?

Another thing as a quick question.  Is there a rule of thumb or a table that describes how large of a hole is needed for different sized leads?  To be safe I initially made them about 20 mils or so larger(in diameter or diagonal), but  now I'm little worried that I might be making them too big and making it hard to solder.   

[T3sl4co1l]

Fab data is all in decimal, and handled automatically by machines that work in whatever internal units they need; as far as I know, what you operate in, and export as, makes utterly no difference.  Might as well make use of those decimals; rounding them won't even save file size once it's all zipped up!

Holes should be 5-10 mil oversize (finished plated i.d.).  More and you need a lot of solder to fill, it fits too loose during insertion, and is probably weaker (easier to pull out, especially for soft leaded solder).  There's also danger of shrinkage and cracking or delamination (especially for the pad 'barrel' away from the PCB laminate it's held in, and from connections to internal layers embedded therein) when there's a big blob of solder.

Holes can be as tight as +2 mils, but insertion gets very difficult, and solder doesn't wick well.  Most fabs don't do better than 3 or 5 mil tolerance, so you want to be above that, hence 5-10.

I do tend to round hole sizes together, just for neatness really, so as not to clutter the drill drawing.  Actually, the fab should be smart enough to do this, and would do a better job rounding to the desired size, based on their available tooling.

For vias, I always use 12 and 20 mil i.d. as standard, with 8 mil as the smallest, reserved for fine pitch or dense circuitry.  You can get smaller vias, but they have to be microdrilled or laser drilled, at added expense.

Annular ring is usually 5 mil minimum, 10 preferred.  18/8 is the smallest via I use; larger go up roughly in proportion.  30/12 is enough for small pogo-pins to hit (good for assy/functional test), and 40/20 is good for power connections or soldering test wires into.  Pad ODs up to 60 mil or more are good for regular pogo-pins, and can be made as SMT or THT (with any size hole, up to, oh, 40 mil maybe) pads.

My preferred '0.1" pitch header' padstack is 65 mil OD, 40 mil ID, all round, except for pin 1, which is a rounded rectangle pad shape, 10 mil corner radii.

Tim

[pigtwo]

Thank you, that's super helpful. That should be enough to get me far enough into this project.  I'll have to start reading those IPC documents. 

[charlespax]

Personally, I do the footprint in controlling units (in/mm as stated) and almost always do the PCB in mils.  In Altium, this is merely personal preference; you can switch units on the fly, so that in the end, it doesn't matter what is actually used.

I'm on board with this. When making a footprint I do exactly what the manufacturer specified: metric or imperial. Then I always layout my board on an imperial grid. Just make sure the placement crosshairs of your footprint are in the correct spot.

[piranha32]

Hello everyone!

I've just started working with OrCAD(and PCB design in general) and I've been creating some footprints for a basic board.  Often I've been finding that the mechanical specs will be in mm but I've been told by my boss that their designs were usually done in mils.  In the conversion from mm to mils, rarely does it come out to a clean number.  So my first thought was to just use as exact specifications as possible and let the manufacturer do the rounding.  But then I remember reading something Dave wrote a while ago to always have things on a grid and so I did it as though it had to be on a 5 mil grid rounding up or down accordingly. 

Which is the correct thing to do?  Are there any tips on how to do this with the least number of problems. 

Thank you!

Metric v.s. mils is more of a holly war for me. The units are equivalent, it's easy to convert between them, and if the board specs say that you should use imperial raster, do it. However, for designs where the specs say nothing about raster, I would rather standardize on metric.

In the past imperial was *the* standard, however it's very quickly becoming obsolete. Just few years ago most of the datasheets had dimensions either in inches, or dual. Today it's becoming more and more difficult to find a datasheet where manufacturer bothers to convert dimensions to inches. With magnetic pads it's not a big deal, but I find it easier to use metric raster with metric parts.

You may, or may not not listen to my opinion, however Tom Hausherr from Mentor Graphics claims that switching to metric raster increased their productivity, and the fact that most of parts are defined in metric raster is one of important factors: http://blogs.mentor.com/tom-hausherr/blog/2011/03/23/do-you-use-imperial-or-metric-units-for-pcb-design/

[Landrew2390]

http://blogs.mentor.com/tom-hausherr/blog/2011/03/23/do-you-use-imperial-or-metric-units-for-pcb-design/

piranha32, that link really does sum up the entire Metric/Imperial debate.  The industry has immense inertia against change, but the metric system really is superior when it comes to any kind of engineering task.  I think the biggest thing that is preventing the metric system from being fully implemented in the United States is the lack of knowledge among the general public.  As commerce continues to march forward with the Metric system, I fully expect that to change, but it will be many decades before the full conversion is achieved.

[amyk]

Converting between units and leaving "ugly" decimals is far better than rounding off too much and finding out, much later, the part won't fit right. Been there, done that.

[piranha32]

http://blogs.mentor.com/tom-hausherr/blog/2011/03/23/do-you-use-imperial-or-metric-units-for-pcb-design/

piranha32, that link really does sum up the entire Metric/Imperial debate.  The industry has immense inertia against change, but the metric system really is superior when it comes to any kind of engineering task.  I think the biggest thing that is preventing the metric system from being fully implemented in the United States is the lack of knowledge among the general public.  As commerce continues to march forward with the Metric system, I fully expect that to change, but it will be many decades before the full conversion is achieved.

Things are changing. Few years ago only some manufacturers were providing metric dimensions in datasheets. When I was preparing slides for SMD soldering class few months ago, I had to check 3 or 4 datasheets before I found imperial dimensions for basic passive elements. Today most SMD parts are based on metric raster. As the industry is shifting away from TH components, and the manufacturing and design is moving to metric parts of the world, the only reasons I can see for imperial raster to still being used in a few years are old standards, old habits, and general inertia of the industry as a whole.

[Landrew2390]

It's a good change.  Something else that is driving the transition is the miniaturization of parts.  With increasingly smaller parts, it makes more sense to work in a more precise unit.

Now we just need to get NASA on the metric system and away from this mixed system they're insisting on using.  A good deal of their current equipment was designed 30 years ago and they insist on continuing to use the Imperial System for compatibility. The problem is all their instrumentation and trajectory data is in Metric.  So we have to read the telemetry, crunch the numbers, convert to Imperial and then tell the system what to do.  It's nuts to do that when you can change over and avoid that mess altogether.  But that's government agencies for you.

[McBryce]

It's a good change.  Something else that is driving the transition is the miniaturization of parts.  With increasingly smaller parts, it makes more sense to work in a more precise unit.

Now we just need to get NASA on the metric system and away from this mixed system they're insisting on using.  A good deal of their current equipment was designed 30 years ago and they insist on continuing to use the Imperial System for compatibility. The problem is all their instrumentation and trajectory data is in Metric.  So we have to read the telemetry, crunch the numbers, convert to Imperial and then tell the system what to do.  It's nuts to do that when you can change over and avoid that mess altogether.  But that's government agencies for you.

Wasn't that the reason why the Mars Climate Orbiter crashed and burnt? You'd think that would have been enough to kick them into the 21st Century.

As far as PCBs are concerned. If I'm doing stuff with thru-hole parts (yes I still design retro computer hardware as a hobby) then I use mils because it fits nicer with the DIL pin spacing. If I am doing an SMD layout I use mm.

McBryce.

[Landrew2390]

The Mars Climate Orbiter was one of the many projects that has suffered that problem.  Another one was a shuttle mission to repair a satellite in LEO.  They started their spacewalk and discovered their wrenches were Imperial and the bolts were Metric.  It's kind of hard to run to the hardware store when you're +150km above the Earth's surface.

[void_error]

As far as PCBs are concerned. If I'm doing stuff with thru-hole parts (yes I still design retro computer hardware as a hobby) then I use mils because it fits nicer with the DIL pin spacing. If I am doing an SMD layout I use mm.

McBryce.

I often find myself mixing Imperial and Metric on the same board, switching between Imperial and Metric grids. I do the board outline with mounting holes in Metric and use Imperial if I have SMD passives around SOIC or DIP parts, if I have ICs on the Metric system then that's what the grid size for associated components will be. In the end it turns out quite nicely.