Pad Holes Under SMT for Heat Sinking and other questions

Hey there, so I got a couple of newbie questions. I’m using this part: Rohm Semi LL Mosfet, there’s no footprint for a HSMT-8 Package, so I started creating one. Until I sat down and started doing the drawing of this, I had not realized how small the part was 3.3x3.3mm! Anyways, I’m not going to be driving it anywhere near it’s maximum current, however since I had to design a footprint, I wanted to include some provision of sinking away some of the heat generated by this device. Here’s what I did:

I would greatly appreciate if someone could tell me whether or not this is permitted or would create more headaches for other devs.

I’m quite new to Kicad and e-cad in general, but I’ve been a mechanical cad user for 20 years, From what I can tell, it seems the footprint editor simply snaps the part to the “center” of the pad, and then sort of ham-handedly allows you to snap the pad to the grid, or input an x-y coord. Anyone have any suggestions or a write-up on how to input this information via DXF? I can crank out DXFs by the pound.

Thanks all!

So as an addendum… I did try implementing this footprint in the design, the DSN export feature did nothing but complain. It’s still something I would like to revisit at some point, but apparently now is not that moment.

That footprint looks like a PowerPak 1212-8 I got here:

If you do heatsink vias - or any overlapping pads/through holes that should belong to the same copper/potential really - you need to give them the same pin number that you use in the schematic symbol, otherwise KiCAD will jump up and down when it checks the layout. I’m pretty sure you don’t have H & H1-H6 in the symbol as pins, do you?

20 years of CAD experience? - you’ll be fluid with ECAD in 1 month, KiCAD is not that deep. Just library organization etc. will take longer to grasp and get right as it takes a while for you to work out what you want after you’ve been doing it (probably wrong) for a while… Welcome to the fun :wink:

As for placing pads to certain XY coords… there is the manual input (which you found already) and then there is wizards or scripts to create ready footprints to use inside and outside of KiCAD. I haven’t come across a DXF tool for something like that yet… but KiCAD eats DXF for outlines or any lines really pretty nicely, just not pads.

PS: if you are new to all of this and there is even a slight chance you need to touch that IC with a solder iron during troubleshooting/etc. I suggest to extend the pads a little bit further outside the housing outline, as otherwise you got no chance to get those pins heated.

1 Like

In my opinion, one of the VERY useful features in KiCAD is that overlapping pads WITH THE SAME PIN NUMBER are merged into a single piece of copper. This not only lets you put vias into a pad (for either heat sinking, or increased current capacity) but also lets you build up pads with complicated outline shapes, using rectangles, circles, ovals, etc, as basic elements. (Yeah, there are some rules about how much the pads need to overlap before they’re interpreted as electrically connected but a little experimentation will teach you.)

Here’s a footprint I recently did for the surface-mount DPAK package.

There are actually 38 pads identified as “4”: An SMT rectangle for the package’s tab connection, an SMT rectangle for the solder paste stencil opening, and 36 through-hole pads functioning as vias. The pads are a bit oversized, not only to permit access for manual soldering, but also to make the footprint tolerable to the IPAK package, and the SOT-223 package.

The large tab connection is also defined for direct connection to a copper pour zone - a feature you may want to change if you truly must hand-solder to this footprint.

I did a somewhat similar footprint for TO-220 packages.

TO-220_Horiz_ThermalPad_Alt.kicad_mod (8.8 KB)
ATPAK_DPAK_Thermal_Vias.kicad_mod (10.2 KB)



KiCad can import DXF, but in what I’d call a modest-minimal manner.
You can, for example, import PAD outlines, and visually confirm your final footprint is valid.

Currently suggested as enhancements, are Snap of Footprint to Entity origin (eg circle centre) or Entity centre (rect centre) or entity End (outlines), to help DXF footprint pathways.

You can query any line, and manually copy X then Y then R as a workaround now, which means create of a DXF with outlines, and centres, would be useful for non-grid based parts.
(Also suggested is copy X,Y,R to clipboard and paste from clipboard)
Having 2 sessions active during this clone process can help.

However, there is some work being done on smarter DXF to Footprint conversions, like this thread -

I think that targets RF, but it could be extended to general footprint create ?

If you are skilled at DXF create, maybe you could help there ? Done any Python ? :wink:

Some suggestions for DXF to Footprint smarts could include

  • Block name to Pin Name, outer block to footprint name
    (no names, use sequence order, defaults)
  • Layer name aware
  • Circle sense, for DXF.Circle -> Pad.Circle
  • Concentric circle sense, for DXF.Circles -> Pad,Drill or Pad.Smd
  • Rectangle sense for DXF.Rect -> Pad.Rect
  • Oval sense for DXF.Oval -> Pad.Oval
  • Rounded Rect sense for DXF.RoundRect -> Pad.RoundRect
  • R/O/RR sense of enclosed circle/oval for (offset) hole/slot generate
1 Like

What pgms do you use for mechanical CAD & what do you suggest for reading & writing DXFs ?
It seems there is almost enough smarts in a DXF.TEXT record, to map onto a Footprint Pad, which could make a DXF transport for Pad handling a little easier.
( eg TxtStr -> PinName, X->X, Y->Y, Th -> DiaY, ScX*DiaY -> DiaX, Just -> PadShape, leaves Drill XY ?)

For DXF, I prefer Q-CAD, I actually like the older version (~2.2) when they forked off LibreCad off Qcad there were a number of things that changed, and made x,y entry much more difficult and pushed most of the effort of absolute measurement off to the scripting console. If you take the time to learn the relatively rich scripting system qcad (or librecad) there’s practically nothing you can’t do. At this point, because of the complexity of much of my work I do much of the single part creation in solidworks, export it to dxf, and then if I have to do a laser-cutting layout I create a block in q/librecad do my layout and then explode blocks at the end (This is what breaks apart the elements into single lines, it makes modification afterwards much more difficult so it needs to be your last step).

The major difference between solidworks for dxf and q/librecad is the ability to quickly modify elements, and to set relations between the lines. So if you’re designing something where there’s a radius that is at the intersection of two lines with an acute angle between them (like if you made a knife blade round on the end) Solidworks allows you to just set the radius tangent to the other two points. Q/Libre kinda requires you to trim the lines, and requires a lot more manual effort in order to get the radius and line placement right. This mostly comes from Q/L treating every element as a mathematical construct, rather than a distinct element whose relation to other elements is more important than the pure math aspect.

Generally, I would suggest starting with Librecad, it’s arguably the better of the two, and it’s free. At this point my goal is within 2 years to completely eliminate windows from my environment, which is a major reason why I’m supporting Kicad at least for the moment with my efforts at creating footprints, and dxf board outlines. Here’s the page for that:

So unfortunately, I don’t do python, probably much to my detriment as I’ve largely actively avoided it for some years now. I could probably pick it up relatively quickly, as I’ve been writing PERL and C since my teens. (pushing 40 now, has it been that long?)

At this point, I would probably be better off working on the MCAD/ECAD stuff, and I still need to learn how KICAD handles this. If the KICAD project is looking for pointers on how the MCAD world does this kind of thing, let me know, I would be happy to help.

I’m trying to remember how long I’ve been using KICAD at this point… few months maybe? I understand electronics (dad’s an IEEE) but it was never my primary calling, and ended up doing mechanical. However, I’ve been working on a lot of automation systems for work (I build automated ammunition loading and processing machinery hence the name) and while I was previously relying on my dad for help, he got pretty sick earlier this year (about the time I started in with KICAD) so it became clear I was going to have to rely on myself to do what I wanted. It’s been a fun experience, a lot of my friends use Eagle, but it’s workflow is frustrating, and it’s not open source. I would much rather invest the time in developing work flow in a product that has seen development effort in the last decade (Eagle’s interface is horrible), and I simply can’t afford altium or some of the other commercial products out there.

I’ve been looking at different options as far as soldering some of the SMD/T parts, a large group of my friends have a makerspace, and usually use hotplates and solder paste for this process, given the surface area of the HSMT-8, I figured that would be the way to go, and the heat-sinking component would allow faster heating and flow into soldering that particular part, and I could simply solder the rest by hand. 0604’s are easy, but are really the practical limit for what I can do by hand. Here’s what I was designing:

It’s essentially a level converter (not sure on the terminology) to allow an arduino to turn on higher amperage 12V devices. I needed 3 of these to run a block of pneumatic solenoids. I ended up getting aggravated last night, and went from the part I was originally planning to use and went with an FDN537N, which they call a “SuperSOT-3” but it seems to be functionally identical to the SOT-23 package, just taller. I think this dropped my total amperage capacity by 2A, and will probably impact my duty cycle, but it shrunk the board footprint, and I really only need these for my proof of concept, later I need to redesign it with an I2C GPIO expander and put all of the parts on a single board.

Very cool, this is exactly what I wanted to do. I’ll give it a shot later tonight (going out to dinner) and try my hand at updating the HSMT-8 Package again. I’ll try creating a dxf outline in SW first and see if I can get that to integrate as cleanly as it’s supposed to.

I got a pizza oven for that and some lasercut aluminium parts to get the stencil aligned for the paste and some stainless parts as trays for a fish tank air pump supported pick&place and later reflowing.
Really need to write that down some time I guess as the whole setup runs nicely and is good for prototyping of maybe 10-20 boards.
But I really just followed what I did find about that on the net.

Hmm, thanks.
I already have LibreCAD, and looked at QCAD.
One of my peeves with LibreCAD is likely this made x,y entry much more difficult - seems strange to remove xy entry ?
QCAD seems crippled and has nag-screens to try to $ell upgrades…
I was hoping I had missed something, that was better than LibreCAD…

Yea, I wish I had a better suggestion… this is why I still use Qcad 2.x versions, I actually bought the “pro” version god knows how many years ago (maybe 2000-2001?) and I still use it. For a 2d cad system it is quite good. LibreCad still has x,y input, but it does it through the console, it’s not at all intuitive, Qcad has a popup box in the upper left that allows you to enter points this way, very similar to autocad.

I have not tried any of the freecad, freescad, recently, I tried them some years ago when i was considering purchasing solidworks, and the interface was not intuitive enough, and the toolchain wasn’t powerful enough to make it a viable option. There is Onshape, which is a 3d system very similar to solidworks, it’s very very good for what it is. Moving from SW to onshape is a very intuitive process.

It looks like Qcad is still ~$30, which if it still has the same features it used to is a very good and very reasonable price, especially if you do any lasercutting. The company I contract all my work out to takes a dxf that has all of the cuts you want it to make, just totals up the time it takes to run it, (typically 100in-per min +1s for each pierce) so it makes understanding your costs and controlling them very easy. I may hit up the Q-cad people and see if my professional license is still good and try the current one. I know last time I tried the free version there were some things that should have been there that were missing.

I hope you’re OK with deviating your thread a bit further, but did you ever try parametric modelling instead of the x/y coordinate input?
The tool I use (Autodesk Inventor) works pretty much flawless that way and I know that FreeCAD has similar features to get from ‘sketch’ to something-useful.

This video cuts right to the chase (narrator is not loud, but not needed to understand the concept):

Any CAD tool that can’t do that is essentially binding at least one arm up behind your back.

I have had moderate success with an electric skillet purchased at a Thrift Store for US$6.00 . See:
or just seed a search engine with “skillet reflow solder”.

The main point to remember is that, for prototypes or feasibility demonstrations, nothing is nearly as critical as you think. Some hints based on experience:

  • The enclosed skillet, rather than an open hot plate, seems to help keep the board and components at a uniform temperature (top to bottom as well as across the horizontal surface).

  • A transparent (glass) cover on the skillet would be VERY helpful, but almost impossible to find on inexpensive skillets.

  • Place about 3/8" - 1/2" (10mm - 12mm) of common sand in the bottom of the skillet. This evens out any hot spots due to the heating element’s placement, and provides a uniform contact surface for conducting heat into the board.

  • I use brand-name eutectic tin/lead solder paste. (Sorry, but I forget the brand at the moment.) It came in a plastic syringe with 3 or 4 extra applicator tips. For my applications, shelf life in my desk drawer is at least twice as long (and climbing) as what the mfgr claims for storage under ideal conditions. I use a medium sewing needle to plug the end of the solder paste’s applicator tips between work sessions.

  • You’re gonna need some magnification. A medium- to high-magnification desk magnifier is the bare minimum. Get a decent one that maintains focus across the whole field of view, and doesn’t put rainbow fringes on the edges of objects. That doesn’t necessarily mean the most expensive unit you can find, but expect a retail price around US$100 (+/- 1 dB).

At work I have a binocular toolmaker’s microscope that’s almost worth its weight in gold. Around US$200 from Amscope - possibly this one:

  • Get a box of plain wooden toothpicks, and a pack of fine-gauge sewing needles. Use these to apply solder paste, nudge components around, and clean solder balls from between fine-pitch pads. Tweezers from the hobby store are also useful - it takes practice to become adept with the reverse-action style of tweezers but I think it’s worth the effort.

  • It took quite a while to learn how to effectively apply solder paste with the syringe, toothpicks and needles. Practice, practice, practice. One of my biggest problems was applying too much solder paste. On a well-tinned board, it takes only a moderate film - not a blob - to do the job.

  • I had problems with components getting jostled out of place as I worked at placing other components on the board. My exceptionally inelegant solution was to touch just one pad of each component with a fine-tip soldering iron. In just a second or so the solder paste will begin to flow, anchoring the part. Don’t worry about the thoroughness or quality of the joint - it will correct itself when the joint reflows along with the rest of the pads.

  • Using a thermocouple probe I found that the thermostat control on my skillet was reasonably accurate. My reflow procedure is roughly:

    • Preheat the skillet (with sand in the bottom) to 250F. Let the controller go through at least one complete on/off cycle.
    • Carefully place the board, with solder paste and components in place, in the skillet. Press lightly to make uniform contact with the sand.
    • Close the skillet and set the thermostat to 400F. (425F if the board has connections to many large copper-pour areas, or has several parts with large thermal mass - such as ferrite core inductors.)
    • Wait for the thermostat to reach the set temperature and cycle off - about 5 minutes.
    • Start timing. After 2-1/2 minutes, remove the skillet cover and quickly get a visual confirmation (using the desk magnifier) that all of the solder has reflowed smoothly. If any of it is still lumpy or granular, replace the cover and wait another 30 seconds.
    • If there are only a few troublesome areas - such as the tab connection of a DPAK or other power package - touch the pad with a soldering iron tip, possibly even flowing a small amount of fine-gauge wire solder into the joint.
    • When all of the solder has reflowed smoothly, turn off the skillet and leave it open. Within 30 - 60 seconds you should see the solder freeze. At this time you can lift the board out with tongs and place it someplace to finish cooling.
  • Inspect the board under magnification, especially between pins of fine-pitch parts. Dislodge solder balls with a toothpick (preferred) or needle. Clear solder bridges with a soldering iron and solder wick.

  • Scrub the board with isopropyl alcohol and a stiff nylon-bristle brush (e.g., the nylon brushes in this: )

  • Put at least a few (non-tented) vias into the power-pad connections of packages that have power pads, even if you do NOT connect to the vias on the back side of the board. After reflow soldering, look at the vias on the BACK side of the board with a little magnification. If you see that solder has flowed into the top of the vias and filled most of them, it is reasonable to assume that the solder melted and flowed under the power tab, securely adhering the tab to its pad. If the vias are clear, or mostly clear, of solder than the power tab may not be adequately soldered to its pad.



No, after the electric skillet gets used just once to heat up leaded solder and rosin flux it NEVER sees the kitchen again! (But it might get used to accelerate some epoxy curing, or to expand a tight-fitting bearing before I drop it on a shaft.)


1 Like

So one of the projects I have no idea what to do with, is I built an arduino based temperature controller for a heat treating oven I have. It’s electrically powered, with two zones (coils) each with about 3000W, it then uses a thermocouple to measure the temp at the top, side and bottom, and there’s an attachment for a N2/H2 valve so you can keep a mostly inert atmosphere. (the N2 will purge most of the O2, and the H2 will rapidly turn it into water).

You could very easily take the same hardware and code, and make a much smaller reflow rig, instead of using the big heating coils, a few mcmaster-carr cartridge heaters, and a big chunk of aluminum would make a fine hot plate. At the 500F or less mark, CO2 can be used as a cheap and handy purge gas you could generate by just putting a little cube of dry ice in every time you opened the door.

At the moment, I’m patiently waiting for the boards I sent off to oshpark to get back from fab and delivered (oooh the anticipation!)

At the moment at least, I’m able to solder parts as small as 0604 by hand, I just tin the contacts put the part down with tweezers and then re-melt each pad. A lot of the parts, like the HSMT-8 would be pretty much impossible without a hotplate. There’s also this accelerometer I’m messing around with that’s same thing, all the pads are on the back.

Anyways, thanks for the very awesome reply. At the moment, I’m mostly using spools of solder my dad had sitting around, most of them are probably older than I am, and I found a quart bottle of rosin. This has been handy to flux the parts before I try to put them down, and makes the solder flow nicely.

I’ve run a variety of bullet casting machines for a number of years, those skills have translated well into assembling electronic parts.