# LM2678 PCB Layout question / review

After my initial question about switching regulators I had decided on trying to use an LM2678 for an adjustable/variable voltage form 5-24V and 5A max with an input of ~30VDC.

The application will be a stepper motor tester for varying motors, hence the adjustable voltage.

In the datasheet, they mention that the PCB layout is very important for this device:

They also give this layout pattern which I admit to not fully understand:

I made this schematic here with the following PCB layout.

• The heatsink might change, but the actual Schottky diode and the LM2678 have the tab connected to ground, so I would like to share the heatsink.

• J1/J2 pin headers are just for demo, in real design they will get replaced by beefier screw terminals of course.

• RV1 will actually be replaced by a front panel potentiometer and not be on the board; it's for reference here only.

And once again, with the zones filled (but harder to see silkscreen):

My questions would be:

1. The bold traces from the design are all very small and copper pours are used for ground and L1 to C3. Other track widths are 2mm. I tried moving the feedback wiring far from the inductor. Would this layout seem acceptable?

2. Is there some trace/part (I'm sure!) that could be improved?

3. They say prototype etc, but I only need to make two or so of those boards so I'd like to get this right and obviously on a breadboard this cannot be much tested layout-wise, so what other guidelines would there be to follow, or is this a bit of a tall order without tinkering on a real PCB?

• I'd add a small input capacitor of e.g. 100nF as close as possible to the LM2678. See datasheet: It may be necessary in some designs to add a small valued (0.1 μF to 0.47 μF) ceramic type capacitor in parallel with the input capacitor to prevent or minimize any ringing. – Huisman May 6 '20 at 20:17
• Next, I would split the 100uF into two parallel 68uF in order to reduce the total ESR. I would bring R1 closer to U1 and RV1 directly below it, in order to reduce total trace length. Connect D1 and L1 on the top layer (red layer). – Huisman May 6 '20 at 20:28
• neither the actual Schottky diode nor the LM2678 have the tab connected How? Schottky diodes in 2-terminal TO-220 (except TO-220FP which is completely insulated) packages have the tab connected to the cathode (and thus to the GND). And the regulator's tab too is connected to GND. If you are using insulators, don't. They reduce the heat transfer capability. – Rohat Kılıç May 7 '20 at 6:41
• @RohatKılıç Yes indeed you are right about the diode after reviewing it again. But where in the LM2678 datasheet do you see that tab is grounded?? – namezero May 7 '20 at 7:05
• The datasheet rev Apr.2013 shows that the tab is connected to GND but the latest datasheet doesn't. However, the eval board has DDPAK-pkg on it and the tab is grounded. Most likely TI use the same die in both DDPAK and TO-220, so they should have connected the tab to GND in both packages. You can do a continuity test with a multimeter, if you worry about. – Rohat Kılıç May 7 '20 at 8:45