LM2731: Different results produced by same design for LM2731XMF/NOPB

Hi Jonathan,

I'll loop the WEBENCH team to answer your question.

Hi Zack, 

Please note that there are no changes in the efficiency equations. There has been update in mosfet parameters after this post on discrepancy in efficiency curves 

Thanks and Regards,

Atul

are you sure the mosfet parameter errors are this severe? we're talking about a 270% increase in power dissipation for the ic and 300% in other webench designs we have done

What is the source of this difference?
Is it a webench error or an error in the part’s datasheet?

Hello Tom,

Please let us know the condition or share the design where you are observing 270%  mismatch in the efficiency.

We try to match the Webench efficiency with board data with +/-3 percent mismatch with exceptions at light load conditions. 

Thanks and Regards,

Atul

Hi Atul. this is frustrating. i have already shared the design reports and hilited the differences

below is another screenshot showing the same design generated by webench on 5/1/2020 and then on 8/5/2020. ic power dissipation jumps from 18.412 to 49.922 mw. thats a 270% INCREASE!!!

again, i ask, are you sure the mosfet parameter errors are this severe? 

What is the source of this difference?

Is it a webench error or an error in the part’s datasheet?

since this device does not meet our expected efficiency 89% can you please provide ASAP an alternate design with a different ic that will.

errors are this severe? What is the source of this difference?Is it a webench error or an error in the part’s datasheet?

Hi Atul

can you please provide me with an update to my last posting?

thx

Tom

Hi Tom,

Apologise for the delayed reply.

We checked the design and found that we made changes in the mosfet parameters (rise and fall time) some time in MAY due to which there is change in the efficiency( looks like the new efficiency is matching with the board.)

We usually fine tune mosfet parameters like rise time, fall time which are difficult to calculate to have the WEBENCH efficiency close to the bench results at different operating conditions. This in turn affects the IC_PD as this is converter device.

Due to change in rise and fall time , the switching loss has increased which contains the term (rise time+fall time)*fsw and the efficiency is decreased. Also the the values are calculated with number of iterations.

Since the design is for light load, even a slight change in parameters, the change in power will be high percentage wise. 18mW to 45mW in this case. 

We observed that TPS61040 provides better efficiency for this condition.We will get back to you on the part that can be used after consulting the product line. 

Thanks and Regards,

Atul.

Hi Tom,

You can select TPS61085.

hi Zack, thanks so much for the information. we looked at the TPS61085 and compared it to the TPS61087 since the TPS61087 has a smaller package size and we have tight space constraints. the only difference i saw was a larger current rating of 3.2A. please let me know if you see any issues going with this part.

we are also looking at a 5.5V design version, same specs (3.3V-3.7V input, 50ma out, >90% eff, shielded L, ceramic C small fp). would this controller be suitable for this lower voltage version as well? webench seemed to recommend LM2698MM-ADJ.

on a separate note, we are having a similar issue with another webench design using the TPS61170QDRVRQ1 for a 24V design. instead of getting 87% efficiency, we're getting 75%. ive included the report for your reference. what i would like to know is if there are other controller ics including this one that might also be modeled with the wrong parameters. we are also looking at a 22V version in hopes to get a better efficiency result.WBDesign22 24V 50mA 85C new design.pdf

Hi Atul,

thank you for the reply on the mosfet paramaters. however, can ypu please clarify your comment below? did the change in rise and fall times occur in the datasheet specs or the modeling parameters entered into the webench tool?

Due to change in rise and fall time , the switching loss has increased which contains the term (rise time+fall time)*fsw and the efficiency is decreased. Also the the values are calculated with number of iterations.

thanks,

Tom

Hi Tom,

There is no change in datasheet specifications.

The rise_time and fall_time specifications are not provided in the datasheet. These are the parameters used within WEBENCH model to calculate the internal FET switching losses, using below equation -

FET Switching loss = 0.5*Vin*Iout*(rise_time + fall_time)*Fsw.

We reviewed the Power dissipation model for this product and realized that there was significant error w.r.t bench data – due to incorrect internal FET parameters. We corrected the model in May – this lead to update of the Webench model.

Thanks and Regards,

Atul.

Hi Tom,

1. TPS61087 has same Iq, Rdson with TPS61085. You can selects it.

2. What's the output voltage of the 5.5V design version? TPS61023 is a good choice but its maximum Vout is 5.5V. We would suggest you set Vout to 5.45V to leave some margin.

3. For TPS61170 case, just a reminder, to measure the efficiency accurately, the input and output current should be measured by DMM. Normally the power supply monitor number is not accurate. Also measure the input and output voltage close to the IC input and output capacitor because the cables usually have voltage drop.