LM2676 switching frequency

Hello Martin,

we would like to take a look at the schematic for a quick review.

Could you please share it?

Regards,

Giuseppe

Hello Giuseppe,

I basically used the schematics TIs webbench produced from my inputs. I wanted 2.5V 0.8A from 24V input. I add my schematic as a pdf. J1 is not set, so on/off pin is floating. As a load I use three 10 Ohms resistors in parallel, target load is also purely resistiv. D9 is a SMB Schottky diode, currently Farnell 2101192, elcos are standard types.

Meanwhile I watched my board a little and found that it will always turn to 130 kHz after some time. I even saw a situation where it switched between 130 and 260 kHz back and forth.

I really can't remember such a behaviour from my earlier LM2676 boards.

I hope someone can help

Martin

Hello Martin,

the schematic looks fine.

Could you please provide me the following information:

1. scope shots of the input voltage and output voltage

2. scope shots of the switching point voltage, boost pin voltage and feedback voltage

3. PCB layout

Regards,

Giuseppe

Hello Giuseppe,

that sounds like you have an idea...

Can I send you the enquired information outside the forum please. Posting layouts to a forum is not so welcome in my company.

Just send me an email with your email adress and I will send you some files.

regards,

Martin

Hello Martin,

I took a look at your layout and I have seen something that could cause the circuit to be very noisy.

A very important rule for a proper PCB layout for switching power supplies is to keep the AC current loops as small as possible. In a buck regulator the primary switching loop consists of the input capacitor, high side FET and diode. Minimizing the area of this loop reduces the stray inductances and minimizes noise and possible erratic operation. On your layout, D9 is not tightly connected to the IC (high side FET) and the input cap (C20) and that will cause the AC to be huge (the switching current goes from the bottom layer to the top layer through vias and that's a very bad practice).

Another good practice is  to have a ground plane on the printed circuit board as a means to connect the quiet end (input voltage ground side) of the input filter capacitor to the output filter capacitors and the GND pin of the regulator. Place several vias from the ground side of the input capacitor to ground place, that will minimize the path for AC current. A general rule of thumb is to maximize the low impedance area (VIN,VOUT, GND).

The ESR of the output cap introduces an additional zero that will provide a phase boost hence it will help the stability.

Introducing an additional 100nF output cap (ceramic therefore low ESR) will vanish that effect since the equivalent ESR will be the parallel of the electrolytic cap ESR and the ceramic cap ESR. Please try to remove the 100nF cap (C21).

Try to remove C21 and please note that in the webench design there is only an electrolytic output cap.

Regards,

Giuseppe

Hello,

I would like to give a final result to the community, maybe someone else can benefit from it.

First I tried to enlarge the Cout capacitance with a standard capacitor of 470µF. This had two effects, 1. Switching frequency turned to 260kHz and 2. Ripple was reduced compared with a 100µF standard el. Cap. You could have expected less ripple because of the higher capacitance and twice the frequency but ESR of both capacitors is unknown.

Then I ordered a number of electrolytic caps with different capacitance and ESR. One type with 470 µF promises a Z of 120 mOhms at 100kHz. This one gave me a ripple of 20mV which was just above the wide bandwidth noise. Then I had a type with 150µF and 35 mOhm ESR which is the ESR value that is given in the webench schematic. With this one I don’t see any ripple, output voltage is flat down to the noise level.

So my conclusion is:

Webbench suggests an optimum capacitor. If I don’t need a perfectly flat output I don’t need such a high quality capacitor. But going too far from the given ESR value is dangerous. Using standard electrolytic capacitors should only be done in combination with at least two times the capacitance better 4 times.

Martin

Hello,

I came across the same strange behavior, this time with a LM2670-ADJ: it has a switching frequency of only 130 kHz. I assume the reason is some “Pulse skipping”: every second switching period has zero duty cycle. Not sure what the real reason is my question at this stage is regarding previous advice:

“Try to remove C21 and please note that in the webench design there is only an electrolytic output cap.”

Yes, but also noted that in Application note AN-1135 “LM267X 3A, 5A Evaluation Boards” (SNVA013D) there is also output ceramic capacitor. So which is right?

Regards,

Darko

Hello Darko,

I would say that you need lower ESR either by using an electrolytic capacitor with a lower tan d or by using one with higher capacitance (2x to 4x the suggested value of web bench). That is what helped me.

Good luck 

Martin