Strange behavior of the LM25011

Hi Gerhard,

At low load the part enters DCM which looks like burst mode or pulse skip mode. That is normal. If you remove the ripple circuit the circuit will match the evaluation board circuit. The regulation should improve.

Thanks and regards,

Vijay

Hi Vijay,

sorry, get your post today, were some Troubles with EMail notification ...

I removed the ripple circuit, helps Little bit. But still get strange behavior. I also bought a eval board, here the waveforms looking good.

I originally copied the schematic of the eval board, just using different inductor (other brand). This inductor differs in the DC resistance, it is Little higher than the original. I also used a 100mOhm current shunt, instead of 75mOhms. Coukd this make the difference?

I also carefully checked the schematic and I build two boards to get rid of simple Errors.

With best regards

Gerhard

Hello Gerhard,

Vijay is out of the office this week and next so I will pick up the conversation.  

The larger current sense resistor could make a difference in the line regulation because the ripple on teh sense resistor is fed to the FB comparator as a ramp for pulsewidth modulation.  See the block labeled FCIC CONTROL in the datasheet block diagram.  

The LM25011 regulates the valley (minimums) of the ripple waveform presented to the FB comparator.   When the valley ripple falls below the internal 2.5V reference, the ON TIMER turns on the high side FET and the ripple reverses from negative going to positive going.  The regulated valley voltage will differ from the average output voltage.  More peak-peak ripple amplitude means more error and more line votlage dependence at VOUT.  The external repple network adds to this error along with the larger CS pin current sense resistor  

Could you reduce the CS pin resistor, removed the external ripple network and retest the line regulation (Vout vs Vin)?

Hi David,

I already removed the external ripple network and that helps. Now the dependency between Vi and Vout is histrory. Unfortunately, the chip switches on for very short time, switch off and switch on again, but now for the correct time (correct means that the output voltage is correct).

Problem is, that this short puls introduce very high frequencies to the circuit.

I also get new sense resistors, but I have to assemble a new board, or Change that resistor. Hope can do this during the week.

With best regards

Gerha

Hi Davisd

I changed the sense resistor to 75mOhm and keep the external ripple network removed (all three parts unsoldered!!) I just see the spikes where the chip turns on, for what reason ever ...

I made shots, the filename describes the voltages and currents.

My first question: How can I remove this spikes?

My second question: Is it usual to have such big amount of HF at the output?

Thanks a lot

With best regards

Gerhard

Hello Gerhard,

Apparently the reduced sense resistor pushed the design past a critical ripple requirement that is described in the datasheet.    A larger value sense resistor or smaller inductor value will boost the ripple and stop the "double-pulsing" behavior that you are seeing.    Page 12 of the LM25011 datasheet explains that 25mV ripple is needed at the CS pin for stable, fixed frequency operation.  This ripple is the peak-peak change in the CS voltage during the off-time of the high side buck switch.  See Figure 19. 

The waveforms you provided in the Zip file look normal except for the double-pulsing.  I haven't seen HF noise in your recent attachments

Hi Gerhard,

I talked with David about the issue you are having. In addtion to increasing current sense resistor, we think increasing output capacitance is another way to help achieve stable operation. You now have 20uF ceramic cap in your design, however the equivalent output capacitance could be only 10uF when considering the 5V DC bias, which is not enough to assure the stability based on my calculation. In order to figure out the root cause, I would suggest you to increase the output capacitance aggressively, say increase the output capacitance from 20uF to 100uF or evern higher. After figure out the root cause, you can optimize the output capacitance according to your requirements.

Thanks.

Haifeng Fan

Wide Vin DC-DC Systems

Hi David,

thats the reason I used the external ripple network.

Now I have something to do, see posting from your colleque. Hope I get it today ...

Smaller inductor I have to order, Needs some days.

With best regards

Gerhard

Hi,

ok, will do in the next moments, hope I can get out the problems. Interesting is, that I just copied the EVAL boards schematic ....

With best regards

Gerhard

Ok,

here the measurements with 220µF at the output. Generally, there is no difference with or without the capacitor. I removed the capacitor, no change in the results. Still 'extra systols', only change is a liitle bit fewer 'jitter' between the cycles without load, but thats quite normal.

I also checked the CS pin and I see 50mV ripple, so this should be more than enough, but only at full cycles. Interesting is, that the ramp starts little bit higher than the ramp of the full cycle, maybe that is something ...

Facit: Lowering the sense resistor dont remove the extra systolics and agressively increasing Cout didn't. As I tried to copy the EVAL design, the only part which could make the difference is the inductor. I can try to made a board with an inductor with smaller DC resistance, this is closer to the inductor used on the eval board, other values I have to order ..

With best regards

Gerhard 

Hi Gerhard,

The EVM board works well, but your design does not. Since you mentioned that the only difference between your design and the EVM is the inductor, it will be interesting to see how it goes with the EVM if you swap the inductor on the EVM for the inductor you are currently using in your design(change the inductor on the EVM to the inductor you are using in your design, and test the modified EVM to see if you have similar problem) . This might tell us if there is anything else which is overlooked in your design.

Thanks

Haifeng Fan

hmm, will check if I can do that ...