Skip to main content
A
AScha.3
Super User
February 11, 2024
Solved

H743VI - pin dies away -- reliability problem of chip ?

  • February 11, 2024
  • 1 reply
  • 2330 views

Hi,

Just yesterday on my H743VIT6 (rev. V ) on the running TFT : picture faded away...white screen.

I just happened to look at the display , thinking ...ok, maybe one of the thin wires just broke now...

TFT (small, 1,8" , ST7735 on 3,3V) connected /soldered about May 2022 , on SPI2 , at 48 Mbit . :)

(Not touched/changed the wiring since then - so no static problem etc. possible.)

 

Today i tried to fix the problem - but everything still connected !

So took the scope to check the signals, beginning with reset ...all look ok, except SPI data !

Here super slow rise time...ok, no useful signal for the TFT at 48Mb. (PC3_3 is sick ?)

First check short to next pins - no, all free. Then impedance of pin unpowered : high, as cmos should be.

Hmmm....try other pin: in Cube could choose PC1 as SPI2 data tx , so i changed to this and soldered the wire also here. (Soldering iron is antistatic.)

Hey, TFT display ok, as before.

But was the pin the problem - after 1 1/2 years running ?? So i set it as gpio out , hi speed, and put simple gpio toggle to it, this its doing now (looks like heavy capactive loaded - but just the 20pF of scope probe there):

AScha3_0-1707651489653.png

Seems this pin lost its drive power all by itself !

Now the crucial question : first sign of dying chip ? or why can this happen ?

(I never seen something like this on any cpu , by STM or others.)

And cpu just at 200MHz core clk, (for longer life...), about 30,3° C case (IR therm.).

 

 

    This topic has been closed for replies.
    Best answer by AScha.3

    Hi @STOne-32 ,

    maybe i killed it -- is it very sensible to "overcurrent" (>1mA ) ?

    Because i just made the write data to display a little faster on F303 with other TFT ( ILI9341 ) that day,

    at first using HAL_SPI_Transmit() , needs 305 ms for a 8x10 lines grid (like a DSO sceen);

    then tried direct register write : 6 ms ! 

    Then i thought : try same on the H743 + ST7735 TFT ; after some time ( 2 hours !!!!) , i got it working (who had the idea to make this SPI so over-complicated ??? ), then it draw one screen -- and faded away forever...

    So i suspect this SPI write at more speed - was too much for the tiny switch.

    Simu at this speed (with 60 ohms switch) shows > 20mA peaks , and about 7mW loss :

    AScha3_0-1707665876847.png

     

     

    And for now its just on my audio-player, one cpu . Next week i will check at work, whether there is any of this over sensible pins used as IO output .

    Thx for your offer of a failure Analysis , but i think we know now what happened.

    Just - it should be some warning in the ds /rm , to avoid any load > 1mA on this pins, otherwise destroy the pin output - not only small notice in errata:

     AScha3_1-1707666348939.png

    + But is pin still ok for analog use -> ADC3_IN1 ?

    ed:  I tested reading in ADC3_IN1 , shows +/- 2 digits error at 16 bit resolution (1Msps , 8,5 sampling time);

    this is as ds data, so input for adc still ok.

    -> I will use it as adc input and this still working fine.

    1 reply

    S
    STOne-32
    Technical Moderator
    February 11, 2024

    Dear @AScha.3 ,

     

    We released last year a documentation  and silicon errata on _C pin usage when configured as digital and not direct analog input for ADC .

    https://www.st.com/resource/en/errata_sheet/es0445-stm32h745xig-stm32h755xi-stm32h747xig-stm32h757xi-device-errata-stmicroelectronics.pdf

    IMG_6432.jpeg


    I would suggest to have a failure analysis to see if the same cause or another one .
    Cheers,

    STOne-32 

    A
    AScha.3Author
    Super User
    February 11, 2024

    Dear @STOne-32  ,

    My errata sheet from 6/2022 ... ok.

    But still super slow switching now - after 1 1/2 years working without obvious problem as spi tx.

    AScha3_0-1707656792606.png

    ...and switch is on. (I never changed/touched sysconfig.)

    So the "analog switch " died away ?

    To get such a signal with 20pF load, i tried simple simu :

    AScha3_1-1707657238676.png

    Looks similar - if "switch" has about 15k ohm now ! ?

    What else could i check ?  Or is it just -- switch dead now, because "overload" with some 48Mbit data ?

    But pin still ok for analog use -> ADC3_IN1 ?

    S
    STOne-32
    Technical Moderator
    February 11, 2024

    Hi @AScha.3 ,

    indeed it seems the signature of the switch , may be a kind of electromigration -  How many devices are impacted please?
    If possible to send it for a failure Analysis to confirm .

    STOne-32 

    Terms & ConditionsAccessibility statement