STM32H745 power up issue

[ABora.4]

Hello,

We are using the MCU STM32H745 in our current design.

We have faced the issue of working MCU properly and below are the problem,

Case 1 : At power ON, some time the MCU work properly and sometime does not work properly, when it is not work properly, after pressing the reset switch multiple time, it will start working properly but not every time. In the case of working properly, it takes current of around 300mA but sometime the current will reduce to 100mA, and the interfaces are going to stop.

Case 2 : After then, we remove the SDRAM and disable the other interfaces like I2C, SPI, ADC and keep only UART interface ON and check the log on dock light over UART. In this case, the situation is the same, at power ON, some time it works and some time it is not working and using the reset switch, some time it works. 

So now we have only MCU STM32H745 is mounted on board and we expect that it should be work properly at least at power ON.

Note: This board is working properly with full software and hardware for 30 Days and suddenly, it stops working and behave as mentioned above.

As of now, we are planning to check the voltage level, power on sequence with respect to reset signals.

Kindly provide the steps to debug the issue.

Regards,

Alpesh

[ABora.4]

Hello,

I have removed the MCU and solder the new MCU.

After replacing new MCU, we observe that, it takes around 1A peak current and stable at 400mA at this time we observe that the MCU is heating.

so, we remove the MCU again and give power to board (without MCU) at this time it takes 70mA current which is normal. Also check the resistance between 3.3V and GND which is around 8K. Also check the resistance on BGA pin of VCC and GND which is OPEN. Now, we solder the new MCU and check the resistance of 3.3V to GND, it is around 8K. Before given the power, we remove the hardware which is operating on 3.3V and then give the power but the same thing is happening, it takes current around 500mA and MCU is heat.

Please provide the suggestion.

Regards,

Alpesh

[AScha.3]

Hi,

so new cpu draws 400mA - maybe possible at max. speed, but not on a not programmed, new cpu.

If you keep nrst at hi , cpu in reset - still so much current ?

So even after check: board has no short, but next cpu again too high current ---> I have no good idea to help.

At work, if this happens on a board (out of 100 boards), i would say: throw it in the box with "unknown problems" and forget it. (or to the electronics trash...).

[TDK]

If a brand new chip is heating up, taking 400+ mA of current, there is a hardware problem. Possibly a short between pins, or an overvoltage condition, or something else. Possibly a poorly soldered BGA.

Going back to your earlier chip, it is not normal for an erased chip to take 70 mA. It should take something much closer to 5 mA. If other components are making up the difference, then that's okay, but if it's only the chip, that indicates an issue.

Can you provide a schematic?

[AScha.3]

Right.

Thats why i didnt use BGA , you cannot see the pins after soldering ; you have to relay on your 100% perfect Reflow process. On a 144pin lqfp i still can check pins and solder , if needed.

[BarryWhit]

> After replacing new MCU, we observe that, it takes around 1A peak current  

> we remove the MCU again and give power to board (without MCU) at this time it takes 70mA 

> we check the resistance on BGA pin of VCC and GND which is OPEN.

Every sign suggests that whomever is doing the soldering is doing a bad job.

[ABora.4]

Hello,

I have made further detail testing and found that when I mount the FB1, FB12 and R1, the current goes up to 500mA and heat the MCU. These are the components for the internal DCDC to make turn on. Find the image for your reference.

So, what it should be the issue?

Regards,

Alpesh

[TDK]

Edit:

• FB1 should definitely not be there.
• VFBSMPS should not be tied to VDD.
• Ground planes should be tied directly, but it seems like R1 does that.
• I don't see why FB12 is present, although I don't think it is hurting anything (provided R1 is placed)

See reference figure in datasheet:

@ABora.4 I updated this reply. My original reply had some issues.

[AScha.3]

@TDK  , 

i think he using direct smps connection, VCAP connected to vfbsmps :

But i wouldnt do this - seems to dangerous to me. If anything wrong with smps hardware or setting - kill cpu.

[AScha.3]

The separate gnd , connected just by R1  , is bad idea. There should be one full ground plane.

+

Or dont use the smps, but the LDO , if so much fear to get EMI from smps :

[ABora.4]

Hello,

Thanks for the suggestion.

As this is a BGA package so I cannot make difficult changes but I can do as mentioned below,

1) Remove the FB1 to stop the internal SMPS.

2) Remove the L1 to isolate the SMPS output to LDO/VCORE input.

3) Give 1.2V externally to the VDD_CORE which is going to the LDO/VCORE.

Below is the reference image.

Kindly check and provide your Input.

Regards,

Alpesh