LM3481: Boost converter suddenly starts beeping and output voltage drops at higher currents

Hi Tobias,

Thanks for using the e2e forum.
The first two pictures you attached are not loading on my side (I think these should be the schematics), could you re-attach these pictures?
The waveforms are displayed correctly, so I am able to see these.

We have a quickstart calculation tool for LM3481. Would it be possible to fill in this calculator and attach it as well?
https://www.ti.com/tool/download/LM3481-BOOST-CALC 
Based on this it will be easier for me to review the design.

Thanks and best regards,
Niklas

Hi Niklas,

thank you for checking!

I try reattachijng the schematics:

WEBENCH

my implementation

Sure, I filled in the excel sheet (everything I could), here are the screenshots from it:

I hope all images load correctly now

Best regards

Tobias

Hi Tobias,

Thanks for the attachments. They load correctly now.

The first check I recommend is the stability.
If the system gets unstable at higher loads, it would explain the Vout drop and abnormal switching behavior.
The bode plot of the calculators show a rather aggressive compensation.
It recommends maximum crossover frequency of 2kHz to avoid RHPZ instability. This calculations shows 4kHz crossover frequency.
We also recommend 60degree or more for phase margin for a very stable system.

Could you check if the ESR value of C8 and C9 really is 5mOhm? This value has a strong effect on the compensation.
If it is correct, I would recommend to try replacing R4 with a 1kOhm resistor and check if the slower compensation improves the behavior?

Please also check if this saturation current of the inductor can withstand the expected peak currents of 20A.

Sense resistance of 3mOhm is rather small (current limit at >50A). I would recommend to use 5 or 6mOhm for R9 to reduce the margin.

There is no gate resistor at the FET. We normally recommend measuring the SW signal and add gate resistance or snubbers in case there are strong overshoots or oscillation. I am not sure if this is also the case for this design. Either way, you can consider adding footprints for these in the next revision.

70W in an asynchronous design is generally rather large and the losses on the forward diode will decrease the overall efficiency. If the design works with the cooling in place, this application is fine. For similar designs, I it worth considering a synchronous Boost design instead to reduce diode losses. A suitable part for this would be LM5122 or LM5123.

Best regards,
Niklas

Hello, thank you for going into this and for the recommendations Niklas!

Stability is something I was unaware of. Now I see there is much more complexity.

C8 and C9 have ESR 12mOhm, my bad. But I guess, thats not too distant. I tried now attaching few more capacitors in parallel, but without noticable change.

I replaced the R4 with 1K, but also no noticable difference.

The inductor can take 20A (with 20C rise) according to datasheet.

Thanks for the Rsense and gate resistor recommendation, I will definetlly include it in the next revision.

Good to know about the other options as well, I wanted to avoid complexity, thats why the asynchronous as a choice, efficiency is not critical.

What seems to have an impact, is the C7. With YAEGO 0805 (56pF) I can only reach 5,4A, with WE 0603 (1pF to 68pF tried) 5,8A to 6A. Could this be a useful lead?

Best regards

Tobias

Hi Tobias,

Thanks for the feedback and additional info.

20A inductor current limit might be rather close to the edge. Due to the low efficiency, peak inductor current will increase. Once the inductance drops, it can lead to a snowball effect which decreases inductance further and increases current even more.

My strongest suspicion is that the inductor may not be suitable for these specs.
We normally recommend an inductor which has its saturation current higher than the current limit set via the sense resistor.
Please also make sure the power supply can support these current ratings. Otherwise Vin can drop, which makes the behavior worse.

With modifying the sensing filter (C7 & R8), you can affect the triggering of overcurrent protection. However, this does not solve the root cause of triggering overcurrent protection in the first place. The goal is to achieve normal operation without overcurrent during all load conditions, so replacing the inductor should be more effective.

Best regards,
Niklas

Hi Niklas,

Im using the coil suggested by the WEBENCH simulator, so it should be suitable I believe (https://www.digikey.de/en/products/detail/coilcraft/SER2915L-472KL/21386994?s=N4IgTCBcDaIDoBcAEBlAogJTATgIwFYAZAWgBYB2MAaUJAF0BfIA). It has multiple current limits specified. Maybe I posted here less relevant one.

Power supply is big enough (up to 35A) and I havent experienced noticable voltage drops on the input. But I was not looking for spikes, if that matters.

ok, then I will look no further to these compoments and maybe check more the compensation. Also Im thinking about checking the MOSFET. The one I selected has a bit more gate capacitance then the recommended one.

Regards

Tobias

Hi Tobias,

Then I have to take back my previous comment. Seeing the inductor specs, the ratings are higher than I initially expected and I agree it should fit for the application.

Regarding the compensation, as the design expects a duty cycle of >50%, there is the risk of subharmonic oscillation, which can make the system unstable.
The calculator recommends a Rsl of 0 Ohm, so it does not calculator that there will be oscillation problems, but this is also a footprint that can be added for the next revision. Adding e.g. a 100 Ohm resistor might improve system stability.

Regarding the MOSFET and diode, do you see a strong heat-up of these components?

Best regards,
Niklas

Hi Niklas,

They do heat up, but its hard for me to say if too much. I use bulky heatsinks (V8818N- https://www.digikey.de/en/products/detail/assmann-wsw-components/V8818N/3511433) and the temperature rises at 4,5A output to 60deg C (schottky) and 55deg C (MOSFET)

Best regards

Tobias

Hi Tobias,

55 or 60deg C should still be okay for a board design. However, if the peak load is higher than 4.5A (the initial post mentioned 6A peak), I am afraid temperature will increase further.

Defining what is too high depends on your application goals and the temperature ratings of each component. LM3481 has a junction temperature of 150deg C, so there is no risk from the controller side. The bottleneck will be the diode, as most losses occur here.

Was changing the MOSFET part helpful in improving the stability behavior at higher loads?

Best regards,
Niklas

Hi Niklas,

I just tried different MOSFETs that I have ordered. It is interesting to see how much difference do they make. All of them were worse than what I currently use, with some of them I could only reach 1.5A before the voltage dropped significantly.I tried IRLZ34NPBF (max. 4.3A), IPA60R1K0CEXKSA1 (max. 0.1A), STP16NF06 (max. 1.8A), IRLZ14PBF (max. 1.5A), IPP011N04NF2SAKMA1 (max. 1.7.A), GSFH0970 (max. 2A).

Looking at trhe results, it seems to me, the gate charge and Rds on  should be as small as possible, while input capacitance is probably not so critical. I will try a few more, but it is hard to find something similar to the Workbench recommendation in THT package.

Hi Tobias,

Thanks for the update.
Unfortunately, I am not a full MOSFET expert, so I cannot give any specific component recommendations.
However, the most crucial parameters to keep in check are Vds max voltage and Rds_on resistance, just as you already discovered based on your tests.
MOSFET either break by overvoltage (therefore check Vds and Vgs ratings), or by overheat. This means Rds_on can be more important than the maximum current value they state.

You can search for a few reference design using LM3481 or similar boost controllers (LM5155, or TPS40210) and see what MOSFETs are used in these designs. 
https://www.ti.com/reference-designs/index.html?keyword=MSP430FR6989#search?keyword=lm5155

Best regards,

Niklas

Thank you for the reference designs Niklas!

They all have SMD MOSFETs, so maybe in the next version I try.

Now, I just tried adding a MOSFET gate resistor (10R) and it just got worse (only max. 4,3A). I see, I will have to look more closely on the whole circuit, if I want to make this run as I dont see an easy fix.

Thank you for all the support, I will give an update in this thread, if I have anything new.

Best regards

Tobias

Hi Tobias,

Thanks for keeping us updated.
Please let me know if any new questions come up.

Best regards,
Niklas