LM3485: LM3485: frequency switching problem

user4598500

Part Number: LM3485

Hi,
I have a problem with LM3485 switching frequency. I made simulations in WEBENCH, in LTspice and everthing was ok - frequency was about 409 kHz. So I've designed custom board with LM3485 buck DC/DC converter 24 V - 5 V (max 2 A) I took parameters of components from my simulations (schematic form WEBENCH in attachments).

What I've got?:
Output voltage is 5 V, but switching frequency is about 4 MHz when I have Cff on PCB, and 1.3 MHz after I removed Cff! So temperature of transistor was very high (very poor efficeincy) when I reduced impedance of load and I must replaced PFET twice during tests.

I am attaching screens from my scope (without Cff and load was 430 Ohms) and schematic from WEBENCH (is the same on my PCB).

over 5 years ago

0 Samuel Jaffe over 5 years ago

TI__Mastermind 25590 points

User,

Please label the scope shots. I'd like to be sure I know what I'm looking at.

Please also share the layout as well as the part numbers for the FET and diode.

Thanks,

-Sam

0 Samuel Jaffe over 5 years ago in reply to Samuel Jaffe

TI__Mastermind 25590 points

Oh I see the scope shots are just different time scales. Please send scope shots for a 2A load with and without CFF. Please also put VIN and VOUT on the other channels.

0 user4598500 over 5 years ago in reply to Samuel Jaffe

Prodigy 140 points

Dear Samuel,

Sorry for unclear description. Yes, there are 3 different time scales on screens, in order: 500 ns, 5 us and 500 us per div and this is PGATE output.

I've tried with three different transistors: STN3P6F6 (default), FDT458P (on screens), and with bigger one IRF9530PBF (I had it on my desk) - result is the same in all cases, frequency is too high. Diode is B540C-13-F. Output capacitors: 6SVPC120MV (ESR = 21 mOhms) and I also tried with EEU-FM1V471 (ESR = 19 mOhms)

I can't send screens for 2 A load, because transistor would break down before I will make screen shots. I have simulations in WEBENCH and in LTspice, both are similar - it should be 409 kHz frequency on PGATE. It is also equation (equation 4, page 11) in datasheet and calculated frequency is about 409 kHz.

Regards,
Pawel

0 Samuel Jaffe over 5 years ago in reply to user4598500

TI__Mastermind 25590 points

Pawel,

In that case I will need to see your layout. Please share and I'll take a look.

Thanks,
-Sam

0 user4598500 over 5 years ago in reply to Samuel Jaffe

Prodigy 140 points

Sam,
This is part of my PCB, only parts from attached schema are assembled.

Regards,

Pawel

0 Samuel Jaffe over 5 years ago in reply to user4598500

TI__Mastermind 25590 points

Pawel,

Thanks for sharing the schematic and layout.

I believe this is an issue with excessive FB ripple caused by the layout. The FB components (C15, R21, R22) connect to VOUT on a large plane far from the output capacitor. I believe this noise is triggering the 10mV hysteretic comparator at every transient causing an immediate toggle of the SW node. To test this I recommend scraping off a bit of solder mask near the connection point of the FB trace to FB plane and adding a 1uF ceramic cap (see below).

0 user4598500 over 5 years ago in reply to Samuel Jaffe

Prodigy 140 points

Sam,

Unfortunately it didn't help. I've added this capacitor and I've tried to solder capacitors in different places on my PCB but switching frequency is still the same.

I think that the next step is to move this circuit to universal board.

0 Samuel Jaffe over 5 years ago in reply to user4598500

TI__Mastermind 25590 points

Pawel,

I think you are right. Here's the link to the EVM.

-Sam

0 user4598500 over 5 years ago in reply to Samuel Jaffe

Prodigy 140 points

Sam,

I have some additional thought - on TI eval board for LM3485 it is also long trace from output to FB, without any additional capacitor. Are you sure that it works properly?

Regards,

Pawel

0 Samuel Jaffe over 5 years ago in reply to user4598500

TI__Mastermind 25590 points

Pawel,

The FB trace on the demo board connects to the VOUT plane very close to the output capacitor C2. C2 will be a quieter place to connect a FB trace.

If you're having doubts, test the demo board before you start swapping components to your BOM. That's a good test to do anyway in any situation. Let me know how the test goes.

-Sam

Power management

Power management forum

LM3485: LM3485: frequency switching problem