4
\$\begingroup\$

I am working on a power supply that regulates the voltage coming out of a 3S LiPo battery. The buck portion of the circuit works fine and outputs 5+ A with acceptable ripple. The boost portion is giving me some problems.

I require it to work at 20 V, 5 A but it only works up to ~3 A, after that it starts making audible noise (even though the switching frequency is 300kHz) and one of the MOSFETs (HDVR1) instantly overheats.

From the datasheet of the device (LM5175) I know that the HDVR1 switch should be constantly conducting during boost operation and the mosfet that I used is rated at well over 3 A, so it should not heat up as it does.

The diagram of the circuit is pretty much a copy-paste of this schematic from a reference design by Texas Instruments.

My schematic

I am aware that it is likely to be a routing problem, but I don't know where to start. I did some analysis on the board, but it's not easy because the MOSFET fries after 4-5 seconds.

The FB signal does not change from buck to boost operation, and the waves on the SW1 and SW2 nodes seem to be OK. The signals on the gates of every MOSFET are as predicted and the rest of the board doesn't heat up.

During operation under 3 A measured efficiency is around 97%. All of that makes me think that there is not a clear and horrible error, but that something more subtle must be going on.

I received much help from reading answers to other questions in various situations in the past, but this time I have no idea of how to proceed.

EDIT 1:

On the gate of HDVR1 I get this behaviour during high load condition:

HDVR1 high load

@peufeu said that it should be around VIN+7V and at the beginning it's 18 V, which is right for a battery partially discharged.

The messy section on the left looks like this:

enter image description here

It lasts for around 6 ms, which doesn't mean anything to me. Also, it oscillates at ~27 kHz, which has little to do with the 300 kHz working frequency.

While on the right there's more or less this:

HDVR1 high load - right part

These are the BOOT1 and BOOT2 signals on no load operation:

BOOT1 - no load

BOOT2 - no load

While they are the same under load:

BOOT1 - high load

BOOT2 - high load

I'm adding here a couple of screenshots from the routing. I'm sorry if they are messy; this was planned to be a prototype.

Only traces

Top traces and planes

Bottom traces and planes

\$\endgroup\$
24
  • 1
    \$\begingroup\$ If it's continuously conducting (no AC on SW1) you're not driving the bootstrap capacitor. If SW2 is always operating you may be able to connect a cap from SW2 to BOOT1 instead. If that doesn't exceed Vgs for HDVR1. Also, 20V 5A implies about 10A in HDVR1, is it rated (and heatsinked) for that current? \$\endgroup\$
    – user16324
    Sep 3, 2017 at 15:59
  • 1
    \$\begingroup\$ Yes, putting a scope probe on HVDR1 should help here, to have Q1 stay on continuously, HVDR1 must stay at about VIN+7V... Put the probe in, then run the board for 1-2 seconds, don't let it burn, kill the power, and post the HVDR1 trace... If it goes down as the bootstrap cap discharges then you have your answer, but if this is the case, then how can this chip even boost? Weird... \$\endgroup\$
    – bobflux
    Sep 3, 2017 at 17:02
  • 2
    \$\begingroup\$ I used the LM5175 in a previous design of mine (old job), had also some problems with stability under higher load currents. However, i heard from TI-support, there's a new LM5176, with an improved stability compared to the LM5175. Maybe ask support, or give the new chip a try. \$\endgroup\$ Sep 4, 2017 at 11:10
  • 1
    \$\begingroup\$ "While trying to get the BOOT1 and BOOT2 on high load waves it suddenly stopped making noise and the current went up to value." Interesting. Did your probe push a cold solder joint and made it contact? Or layout problems?... \$\endgroup\$
    – bobflux
    Sep 4, 2017 at 13:11
  • 3
    \$\begingroup\$ I don't understand how you can mention in the question as well as the comments that it is probably a routing/layout problem, then omit everything that would let someone help you if it indeed is that problem. Show us the PCB. Or at least the critical switching node (the mosfets, input and output caps, and inductor in one screen shot). If you're trying to protect the design... it doesn't work. So anyone coming across this would only be able to rip off a bad layout and you'd still have a better, improved layout to keep to yourself (assuming someone can help you). \$\endgroup\$
    – metacollin
    Sep 5, 2017 at 7:47

1 Answer 1

1
\$\begingroup\$

Set up a TI WEBENCH of your design and find out where your schematic or PCB layout differs from the recommended WEBENCH schematic and layout. It only takes a few minutes and can shed some light on the issue very quickly.

\$\endgroup\$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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