WEBENCH® Tools/LM5165: Efficiency drop: Switching Frequency is not as following RT resistor setting with more than 70% duty cycle

Hi Francois,

We are looking into this issue. We will get back to you on this soon.

Regards,
Atul

Hello Atul,

Thank you for your support.

Further to the Switching Frequency programming and measured issue, I would like to add the following comments.

1. With a similar LM5165 DC-DC circuitry but with conversion 3V3 to 2V09 (instead of 2V44), it appears that the efficiency at 100 mA load is close to Webench simulation of 94%.
2. However the switching frequency remains much lower with 70KHz than the RT programmed value of 310 KHz,
3. For the 3V3 to 2V44 LM5165 circuitry, the drop in efficiency is assumed to be probably due to oscillations in the switching waveform, but this has to be confirmed by adding a resistor in series to output capacitor and stabilising the switching.

The switching frequency is always much lower than programmed - unfortunately  this is the main issue, as programming optimum value found by simulation drives this choice.

I look forward to hearing from you soon.

Best reagrds,

Hi Francois,

Could you please let us know whether the above observations are on the bench or the tool as we are unable to reproduce in the tool.
Could you share the design with us for both before and after the changes using shared public link. Also if possible ,share the snapshots of efficiency dip observed if you cannot share the design.

Regards,
Atul

Hello Atul,

The frequency of 70 KHz is within a LM5165 implementation on a PCB, the 310 KHz is within Webench.

I have carried out more tests yesterday and the switching frequency is capped to 70 KHz - no matter how low the RT resistor is set at 100-135 mA load (20K tested).

The datasheet indicates a formula for duty cycles above minimum:  R(RT) [Kohms] = Vout [V] x 10K / (FSW [KHz] x 1.6) - which so far works with Webench simulation but not within a hardware implementation.

This is with a typical feedback with a cap in parallel with the upper divider feedback.

Unfortunately, I cannot share the design on a public link, Atul.

As I said, one design has proper efficiency (3V3 to 2V09) and the other not (3V3 to 2V44) with same circuitry except divider values with RT resistor very similar.

I look forward to hearing from you soon.

Best regards,

Francois

Hi Francois,

Could you please share the Webench design to the following mail id : atul_k@ti.com so that we can also look into it.

Below are the steps to share the design.

1) Click on Share design as shown in the below image

2) Enter the mail id and click on "share this design"

Thanks and Regards,

Atul

Hello Francois,

Please can you provide me with the Vin min/max specification and can you confirm Vout is 3.3V @ 100mA?

Also please can you provide me with your schematic used to test the board?

Thanks.

David.

Hello David,

For the Webench simulations the Vin range is 3.2V to 3.4V with 3.3V nominal.

I used the same range for checking the switching frequency and the discrepancy is significant at light load and heavy load.

For Vin = 3.3V nominal and Vout = 2.44 V:

I load = 1 mA => Webench FSW = 62.9 KHz, Implementation FSW = 2.9 KHz,

I Load = 30 mA  => Webench FSW = 313 KHz, Implementation FSW = 36.4 KHz,

I Load = 100 mA  => Webench FSW = 313 KHz, Implementation FSW = 70.0 KHz,

FSW[KHz] = Vout [V] x 10K / ( R(RT) [Kohms] x 1.6 ) = 313 KHz

I will check if I can share the design with TI support.

Can you tell me if we have a NDA already in place?

Best regards,

Francois Arnaud