LM51772:LM51772RHAR - high coil temperature and incorrect voltage regulation

I noticed an interesting thing, if I set PCM to 30%, so in this case the output voltage depends on the input voltage in the range of 17-24V, L21 does not heat up and remains cool, but exceeding (down or up) the voltage by 10mV suddenly increases the current drawn from 60mA-70mA to over 300mA. This causes an immediate increase in the coil temperature. Please help or give me some suggestions on what to change or improve.

Below are the oscillograms

• VIN DC:19,9V/ VOUT DC:22,97V

SW1 / SW2

COMP (yellow) DC:0,8V / SS (blue) DC:3,7V

the assumptions were as follows:

Vin: 10-36V

Vout:24V

Imax: ~7A

The oscillograms show the situation without load (only 2mA per LED), with the software setting PCM = 5%

In addition, the system heats up to 100 degrees Celsius, as does the coil. This is not normal. I have 2 leyauts, one made on a 4-layer pcb, practically identically built as the devkit, and the other made completely differently. In both cases it is identical.


I don't know where the problem is, I replaced all the transistors and 5 LM chips. The coil always heated up. I'm just not sure anymore but I think the waveforms looked different for boost and buck. That is, there was a situation where for buck one of the SWs was at 24V and the other had some waveform, for boost it was the other way around. Now I see that only the filling changes. In that case the LM chip didn't heat up. Currently, there are no waveforms only when the Vin supply is the same as Vout

HI Edgar,

unfortunately there is no scaling shown on the plots - so I do not get any information out of them.

Please provide scope plots (showing 5-8 switching cycles  and   5-10ms) when the output voltage drop appears:

• VIN / VOUT
• both sides of the Inductor (SW1 / SW2)
• COMP / SS

Please ensure the scaling for x and y axis for each channel is visible/provided.

It looks like the device gets a continuous restart - this could be over current in the inductor or voltage drop on VCC2, BOOT1, BOOT2

Best regards,

 Stefan

Hi Stefan, Of course I will take pictures.
I also have new conclusions.
I have a second converter on the same board only for 12V. It seems to work properly, at least better than the one for 24V. The LM51772 chip does not heat up, and the coil itself heats up but to about 50 C without load. With a load of 1A the temperature rose to 65 C. In this converter different transistors and a different voltage divider ratio were used.

Returning to the next observations regarding the problematic 24V converter, I did a test, namely: I changed the resistors R394, R395, R396, R397 on the gates from 1R to 3R. And unfortunately after 2 seconds from switching on the transistor Q51 burned out. I soldered a new one and the same after a few seconds. So I soldered resistors with a value of 0R. The transistor is no longer damaged, but everything heats up again, including the U39 chip. Maybe this is a clue.

Hi Romek,

for me this looks like issues with the dead time of the MOSFETs - maybe you get shoot through.

For checking this out you would need measure the gate signals of the two Boost side and the two Buck side MOSFETs.

Best regards,

 Stefan

Gate Oscillogram on Q48 and Q51

Gate Oscillogram on Q49 and Q50

SW1 and SW2

 thermal:

I also noticed that if I just turn it off, meaning I bring the ENABLE signal to a low value close to GND, the U39 chipset (LM51772) stops heating up because the transistors are not working. The same happens when I set the PMC to a value of e.g. 10% and lower the supply voltage to e.g. 23V. Then the transistors stop switching, the Vin voltage = Vout and the chipset stops heating up. It follows that the problem is either with the high current required to switch the transistors, which causes the driver in the chipset to heat up a lot, up to a temperature of about 100 C. I would like to point out that the longest path to the transistor gate is no longer than 20mm.

Hi Romek,

can you further zoom in e.g. 20ns/div for both edges, to proper ensure there is no dead time or shut through.

Can you let me know what „BW limit close“ means. Note: Do not enable a BW limit for this checks.

Thanks, 

Stefan 

BANDWIDTH LIMIT OFF (close):  full bandwidth.

Do you need any additional measurements?

Hi ErKa,

Thanks for the results.

The engineer is out of office today. You can expect a response tomorrow. Thanks

Best Regards,

Hassan 

Hi Romek,

it is not easy to see as the scaling is still a little bit high but i think the dead time is OK.

For the first image above ("Gate Oscillogram on Q48 and Q51") it is still strange that the edges looks much slower then for the others.

It would be good to check the design with another type of MOSFETs, i think the one we have used in the LM5177EVM-HP should fit as well.

SQJ422EP-T1-GE3

Best regards,

 Stefan

Hi Romek,

if efficiency at low power is so important it would be better to have PSM mode enabled instead of FPWM.

Is there a dedicated reason why you have enabled FPWM mode.

To further investigate on losses within the MOSFET and to select and optimized MOSFET for this design, have you already checked the quickstart calculator:

  LM51772 Design Calculator

This (after entering all the design data and MOSFET parameters) does show the estimated losses in the different components.

Best regards,

 Stefan

Hi Stefan,

I noticed a certain relationship. The higher the VIN voltage, the more the circuit heats up, even up to 110°C. Unfortunately, I do not have the transistors you mentioned, the delivery takes more than 10 days. I decided to experiment and installed the transistors I had. I used four CSD18514Q5A. The temperature of the circuit dropped to 70°C. The coil is still hot, but the current consumption dropped from 250mA to 200mA. I decided to change the transistors to CSD18512Q5B. This time the temperature of the circuit increased to 130°C. The current consumption increased to 260mA. So I have a feeling that the problem is with the charge on the gate, which causes the circuit to heat up. I have ordered other transistors with the designation CSD18563Q5AT. Most likely, the temperature of the circuit will then be lower. Unfortunately, this does not change the fact that such current consumption is unacceptable without load. The device was designed for cyclic operation, but on battery power, so it is unlikely that the power loss will be so significant. The power drawn from the power supply is about 10 watts. I don't think the problem is the path layout, because, as I mentioned above, I built another PCB with components laid out similarly to the devkit. I wonder if there is an error in the schematic, perhaps some values or connections are incorrect.
Below I am sending oscillograms from several measurements, maybe it will indicate where the problem may be. 

CSD18512Q5B
Vin=22,4V Iin=250mA, 5,6W, 

Q49 i Q50



Q48 i Q51

Vin=24V Iin=50mA, 1,2W

coil:

coil:

Q48 i Q51

After disconnecting one probe, the waveform looks a bit different. Maybe it's because of the capacitance probe? I don't know

Q49 i Q50

In any case, whether the voltage is below or above 24Vout, the system and the coil heat up very much.

have the settings defined via I2C. Most are defaults, except for the d9 register where I set it to 5%. From the i2C defaults, it appears that PSM mode is enabled (register d1). I also used a calculator early on, but it doesn't explain the situation when there is such high component heating. Isn't this some kind of shoot-through problem? Why does the insertion of a resistor, e.g. 3.3R in the gate circuit of the transistors cause damage to one of them - most likely due to the fact that the lower Mosfet has not turned off and the upper one is already conducting. Have you encountered such a problem during your trials and tests? What recommendations do you have for further action?