2
\$\begingroup\$

I've got a doozy of a situation:

Transistor pads

On this circuit I had made, I routed an IRL530N D2Pak transistor with Source and Gate switched....the large heat pad should be the gate input, but instead I routed it to ground. Furthermore, I added 2 heatsink pads on either of the transistor, with traces from those pads going to the Gate pad.

To compound the issue, I added 11 small copper-plated through-holes on the Gate pad that runs through the board to the ground plane on the bottom side. Unfortunately (for this situation), I added 2 inner layers (power and ground); those through-holes are part of the inner ground plane as well.

Here's a couple options I was going to take, all of which could possibly damage the board:

1) Heating issues may arise

  • a) flip the transistor 180 degrees
  • b) cut the Drain pin off
  • c) solder the Source pin to the square pad
  • d) cut the traces going to the heatsink pads
  • e) cut trace from the small right pad
  • f) bridge connection between heatsink pads and the 2 small pads
  • g) solder underneath of transistor (Gate) to 2 small pads (the left small pad will be the Gate trace
  • g) wire the non-SMD Drain pin that I cut off to the resistor to bridge that connection

2) May cause a short somewhere or I may accidentally end up removing the entire square pad (this would be my first time running this kind of procedure)

  • a) use a thin knife to cut around each hole (top and bottom of board) to break electrical connection
  • b) remove the parts of the copper off the top of the board
  • c) cut trace from small right pad, then wire it to ground
  • d) wire the square pad (now the Gate) to the pin that used to be connected to the right small pad
  • e) do nothing to the left small pad or the trace (would still be the correct route for the Drain)

3) Inner power and ground planes may short, or eventual re-soldering may flow through and come in contact with the inner planes

  • a) drill through each of the 11 holes
  • b) cut trace to right small pad, reroute to ground
  • c) route square Gate pad to the pin originally routed to right small pad

So far that's what I'm thinking of doing...I don't have enough expertise (or even a minimal understanding) of the cause-and-effect risks involved, so any opinion or alternative solution would be greatly appreciated.

If the board is a goner, let me know too...I only purchased 3 for a relatively cheap price, so I'll take the hit if that's the case.

  • Anthony
\$\endgroup\$
  • \$\begingroup\$ Ahen the big pad is the drain - not the source as your question implies. \$\endgroup\$ – Andy aka May 25 '18 at 18:36
  • \$\begingroup\$ Set the transistor on the PCB upside down. Use wires to connect the pins to the correct pads. If heating is a problem, solder a piece of copper-clad PCB material to the giant pad of the transistor as a heatsink. You may need a heat gun to do that. Don't cook the transistor. \$\endgroup\$ – mkeith May 25 '18 at 18:55
2
\$\begingroup\$

People make mistakes all the time, and learn with more experience not to do those things. After a few mistakes like this you check all of the pad layouts with the datasheet (sometimes this doesn't help because the datasheet can be wrong), regardless you need to do some rework.

You can cut traces, solder components by tacking them on to an existing pin (like an SMT resistor), and wire things together. The best thing is 30 gauge wire (usually comes in blue) it's small enough that for low current signals (like a 7mil trace or 10 mil trace) it makes a great substitute if you need some rework. You can also use high gauge wire for larger currents.

Sometimes it's advantageous to buy the same part in a different package for prototyping. Like you said, don't short stuff out. You can even cut traces and solder bridge them back together, rework is a part of circuit design.

enter image description here
Source: Nexlogic.com

The time you can get in to trouble with prototyping is parasitics and differences between wires and traces. If your signal needs a particular resistance, inductance and capacitance then these differences will show when you change a trace for a wire and do some rework. These differences will be manifest in the miliohm, uH-nH range, and pf range which also means that high speed +40Mhz signals could have a problem. Large currents will also see a difference.

The last thing is if the design operates in the uV to nV range there will be thermal noise introduced from the solder junctions of the wire (vs no thermal junctions with a plain copper trace on a PCB)

The engineers at linear usually do all of their prototyping without PCB's at all and solder all of their components together. Something like this:
enter image description here
Source: http://www.computerhistory.org/atchm/an-analog-life-remembering-jim-williams/

\$\endgroup\$
1
\$\begingroup\$

The middle pin and tab are the drain.

If you have that connected wrong and with a bunch of thermal vias to ground rather than wherever the drain should go, your best bet is probably to mount the transistor somewhere else and connect it through wires. If it's a switching supply, rather short wires. For example, you could solder the MOSFET to a thin (eg. 1mm) piece of copper clad PCB and glue that to your PCB with short wires for the 3 connections.

\$\endgroup\$
  • \$\begingroup\$ Ah, right, thanks for the correction. That idea with copper clad PCB glued to the board is brilliant, I will be doing that instead. Should I put electrical tape over the original pads to prevent any accidental shorts? \$\endgroup\$ – Anthony_b May 25 '18 at 18:18
  • \$\begingroup\$ Yeah. If I read this answer before my comment I wouldn't have commented. Kapton (polyimide) tape is preferred over electrical tape. Among other things, it is much more heat resistant. \$\endgroup\$ – mkeith May 25 '18 at 18:57
  • 1
    \$\begingroup\$ If the back is not copper clad, then obviously no need, otherwise follow mkeith's comment and use a layer of polyimide tape or a silicone heat pad. \$\endgroup\$ – Spehro Pefhany May 25 '18 at 19:10
  • \$\begingroup\$ Just to clarify, you can solder the tab to the cladding and slightly bend the feet up.. put a layer of polyimide tape under the feet- then you can attach wires to the feet. \$\endgroup\$ – Spehro Pefhany May 26 '18 at 2:26
0
\$\begingroup\$

I would like to expand on @Spehro idea and suggest that the additional PCB does not have to be glued, nor the original pads covered by tape.

If you make very thin adapter PCB exactly the same as old pads on both top and bottom and then cross-connect them with vias you can solder this PCB to the board in place of the component and solder FET to it at the same time. This way you will preserve much of the heat dissipation capability.

Here is quick drawing of possible arrangement. Note that it needs good isolation coating on that middle trace. You can come up with better layout.

enter image description here

Update: here is a bit better layout which avoids crossing the trace on main board.

enter image description here

Update: Hmm... this would add a lot of parasitic capacitance. I think you can cut down on copper pouring in the bottom layer (or both) and still have enough heat dissipation capability left.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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