LM2736Y (but buck in general). Can I use a reverse polarity protection diode in series with Vin?

Hi again, another question relating to the LM2736Y. The data sheet layout guidline says: There should be a continuous ground plane on the bottom layer of a two-layer board except under the switching
node island.

But the application note it suggests for further reading (SLVA139) says: . Some
people suggest that a copper island, exactly the same size/shape as the switching-node island be created
on the opposite side of the PCB, connected through several vias. This is supposed to prevent 'capacitive
cross-talk' to other traces and to enhance thermal dissipation. But this obviously also leads to the
breaking-up/partitioning of the Ground plane. This defeats the very purpose of Ground plane as it can
cause strange effects arising due to the odd current flow patterns in the now divided Ground plane. In
general, the Ground plane should be kept continuous/unbroken as far as possible, or it could
behave like a slot antenna. For the switching node therefore, the best option is to keep the amount
of copper around it to the actual minimum requirement.

So which is it? The data sheet and application note are contradictory.

Cheers. Matt.

The way I read that, we are recommending that you do not follow "what some people suggests". Keep the GND plane whole. That is what the evaluation board does (except for the VFB trace on the outside edge of the bottom side.

Hi Matthew,

A reverse polarity protection diode can definitely be used in series with VIN.  Since it forward biased during normal operation and hopefully does not need to protect against reverse bias very often, the type of diode is usually not critical.  A Schottky diode will have slightly less power loss and less heat, but other than that, both a conventional and Schottky would be fine.  A PFET is usually not needed.

It should not affect the switching waveform very much so long as the input capacitance is between the cathode of the diode and the input of the LM2736. 

Cheers John,

that's what AN1229 says, but the data sheet for the LM2736 specifically says the opposite "There should be a continuous ground plane on the bottom layer of a two-layer board except under the switching node island."

Thanks for your advice.

I would like to use a  schottky diode between the + supply and Vin pin. I'm a little concerned about the reverse leakage of the schottky though and whether the leakage current would allow a reverse voltage in excess of the lm2736 maximum rating of -0.5V.  The schottky can leak as much as 0.2mA at 25'C, which would certainly create a voltage if the Vin pin is high impedance in reverse voltage.

Penny for your thoughts?

For your amusement, have a look at the schematic for the Arduino MEGA2560 ADK which employs the LM2736 www.arduino.cc/.../ADK_MEGA_2560-Rev2-sch.pdf. They use a diode across Vin and ground for reverse polarity protection, but I can't see how that would work. For one, there is no fuse in series with the diode, so if the supply is capable of greater current than the diode is rated for the diode will burn out. Second, even if the diode survives, there will be more than 0.6V across it, which is greater than the absolute maximum rating for Vin of -0.5V.

Oversight, or am I missing something?

The typical way to provide reverse voltage protection is to have a series connected diode from VIN+ to the input of the regulator. Fig.1 in this app note shows it: www.ti.com/.../slva139.pdf

I agree with your assessment of the schematic you sent.  I am not sure what the reason is for the placement of the diode in the schematic you sent - it would only protect if the input power supply did not cause it to burn out.  A fuse would not help because then the reverse voltage would still be applied to the terminals after the fuse burns out.  If powered from a USB port which typically would detect overcurrent and shutdown, this might be ok.  This would not be true if powered by a battery. 

Perry

Matthew,

I agree with John, and second his opinion that you should follow the app note and the EVM design. I have seen that advice also about keeping the ground plane away from the switching node before and it is theoretically appealing because it minimizes switch node capacitance. However, the reality is that it usually makes common-mode conducted and radiated EMI worse because the E-field from the switch node is not as well coupled to the GND plane. Coupling the E-field from the switch node into the GND plane minimizes EMI that radiates and makes it easier to filter on the board with an input filter.

Overall, I think you are better off with a complete GND plane rather than a cutout.

Perry