0
\$\begingroup\$

I am brand new to layout design. I am using the TPS62136RGXR buck converter in my design. However, according to the attached image, the pin's clearance boundaries (0.2mm according to JLCPCB's assembly capabilities) overlap with adjacent pins. For example, Pin 2, 'SW', is overlapped by both Pin 1 and Pin 3's clearance boundaries. The SW pin in my design outputs up to 3A of current (not continuous, but rather pulsed as it is supporting a gate driver). I would like to connect it to a copper pour, which then connects to my gate driver and its associated input decoupling capacitors.

When attempting to place a copper pour encompassing this SW pin, the pour naturally avoids the clearance boundaries. Thus, throughout the entire length of the SW pin, the copper pour would be approximately 0.25mm in width, while the footprint's pad itself is 0.35mm in width. This leaves 0.05mm on either side of the pour where the pad extends with no connection.

Should I create the copper pour as I have done within the attached image and extend the width of the copper pour as soon as I possibly can? Should I simply choose a different buck converter (non-optimal)? The output of my gate driver has a low duty cycle of 0.5%, so perhaps the former would work? Any other solution/advice?

Image

\$\endgroup\$

1 Answer 1

2
\$\begingroup\$

Datasheet: https://www.ti.com/lit/ds/symlink/tps62136.pdf?ts=1719417594079

The pad pitch is 0.5mm (p.42):

enter image description here

IPC-7351 describes solder joint geometry and terms. This is identified as a flat no-lead solder joint. LGA or CSP style pads are similar, depending on whether the side flanks are wettable (and thus show a toe fillet), and the pad arrangement.

Heel is irrelevant here, as the bars cross the package. There are two toe fillets, and two side fillets. Table 3-13 (IPC-7351 Feb. 2005) suggests -0.04mm side and 0.2-0.4mm toe fillet (depending on component density class*).

Given the lands are 0.2mm minimum width plus 0.1mm maximum tolerance, I would be perfectly comfortable using 0.2 + 0.1 - 0.04 = 0.26mm wide pads on the PCB. This should be easy to fab most anywhere. They can be shaved down to 0.23mm or so to reserve a little more space for soldermask, but less than 0.2 I might be concerned with strength and alignment.

*And, more informally, how well you want to be able to hand-touch-up the solder joints using a soldering iron. This isn't elaborated on by IPC here, but from my own experience, fillets of 0.3mm are difficult to reach, short of a pointy (SMT rework) tip. You may therefore prefer larger toe dimensions for iron-assisted rework. Of course being a no-lead part, there isn't a whole lot of value an iron can provide, so this is of low importance here.

On the following page, TI suggests a SMD (soldermask defined) footprint, which not only requires tighter copper tolerances (oversized pads = closer proximity), but tighter soldermask tolerances as well -- usually this is advertised as LDI (laser direct imaged) as opposed to traditional photoimaged (or even inkjet or silkscreened) mask. A mix of SMD and NSMD (as they seem to suggest) sounds like a poor idea to me; there will be a systematic error between both imaging steps, and there doesn't seem to be any reason not to follow the same practice on all pads (they're all the same pitch and width range, and there's ample space between bar sides and pin heels).

For prototype, I would personally recommend a NSMD footprint, with 0.26mm (Cu) pads, and soldermask expansion by rule (typically ~0.07mm; the exact value hardly matters as proto fabs generally adjust it to suit their process anyway). This leaves 0.24mm space between pads, which should be producible by most anyone; the remaining soldermask web between pads may be too small to produce (<0.1mm?), and should be removed automatically in that case (or you can remove it yourself). Or as mentioned, somewhat thinner pads are also acceptable.

Relying on library footprints without verification against industry practices (such as IPC) or datasheet recommendations, tends to be a good way to learn lessons -- libraries are poorly vetted, may not have been produced at all (let alone in good yield), symbols range from standard to ugly and in all range of scaling; in all, you end up with inconsistent designs on both the SCH and PCB. There is much to learn, of course, both in working with EDA software itself, and in developing ones' own style and preferences, as well as where the solder hits the road, so to speak.

The best advice I can give, is always be playing around: try different symbols, schematic styles, figure out what's readable and what's not, work to that; check library pinouts and namings against datasheets, learn to read package drawings (drafting in general, and GD&T (Geometric Dimensioning and Tolerancing)), read up on IPC standards and how they're used, check footprints against drawings and calculations, make sure everything should fit; and draw whatever silkscreen and mechanical layers you find suitable for your process.

A closing note: in a professional context, the footprint dimensions should be ultimately up to the assembly team putting the board together. IPC recommendations are best used as a starting point: the least-bad way to solder any part, in any given style, by any CM (contract manufacturer). Actual dimensions may vary with the part (depending on how they tolerance it, and what the tolerances of real parts are), plating/lead finish, and even the solder formulation (alloy and flux) and how it's applied (stencil and thickness).

\$\endgroup\$
5
  • \$\begingroup\$ Thank you very much for your detailed response! Lesson learned about trusting footprints found online. I have decreased the pad width down to 0.23mm for all pads so that the soldermask. According to JLCPCB's capabilities, they state that a minimum gap for Solder Mask between pads is 0.2mm. With a pad width of 0.23mm and a soldermask expansion of 0.07mm, the solder mask web in-between each pad is 0.13mm. Does this render the soldermask expansion useless in this case? Could I decrease the soldermask expansion to meet to 0.2mm minimum? Or is that not possible for proto fabs? \$\endgroup\$
    – NickRand
    Commented Jun 28 at 0:58
  • \$\begingroup\$ Are they saying (or are you reading correctly) the distance between (copper) pads, two mask expansion plus one web width? \$\endgroup\$ Commented Jun 28 at 1:57
  • \$\begingroup\$ I am so sorry, I was looking at incorrect information. I see now that the required minimum web is 0.1mm, not 0.2mm. I will go for a 0.23mm pad width and a soldermask expansion of 0.07mm. Given the distance between pads and two mask expansions, this should leave room (0.13mm) for a soldermask web that exceeds the minimum requirement. Thank you very much. I have one more small question if you do not mind; should I extend the copper pour connected to my pin such that it encompasses the entire pin + a little extra around the pin? Or fit it perfectly according to the dimensions of the pin? \$\endgroup\$
    – NickRand
    Commented Jun 28 at 2:43
  • \$\begingroup\$ Cheers, that makes more sense. Pours, won't really be much room around these, you'll have to make supply connections (i.e. bypass cap) one side, and inductor the other. Trace or poly doesn't much matter, but you can widen the path as soon as it clears the neighboring pads. And yes the trace width will be "too narrow" for ampacity, but that's fine for short lengths where heat is still spreading out. More copper sticking out the opposite side will just reduce clearance or increase distance to nearby components; keeping the layout compact is priority. \$\endgroup\$ Commented Jun 28 at 4:21
  • \$\begingroup\$ Awesome, this is all exactly what I needed to know. Thank you very much for taking the time to answer my questions extensively. You have been a tremendous help and I greatly appreciate it! \$\endgroup\$
    – NickRand
    Commented Jun 30 at 4:01

Not the answer you're looking for? Browse other questions tagged or ask your own question.