Reset failure during reliability test. STOne-32..

[hwimh]

Posted on July 11, 2009 at 19:19

Reset failure during reliability test. STOne-32..

[hwimh]

Posted on May 17, 2011 at 13:06

Hi STOne-32!

During reliability test of 10-device, 3 devices got reset failure. (STM32F103C8) Test was done by a mechanical robot, switching the power on and off 3,000 times a day. At second day, 3 devices were not working anymore. I found the reset pin is dropped as 0.3V. Reset pin has 10k pull up and 0.1uF Cap to GND as normal. If reset pin is connected to 3.3V directly, reset is OK and the device working good. PCB is not concerned, because I raised reset pin from PCB. And I omitted the capacitor too. So I think the reset pin was corrupted! Do you have any reliability test report about the reset pin? And what is your suggestion for the reset pin? Sorry for my English. ^^; Thanks for reading.

[16-32micros]

Posted on May 17, 2011 at 13:06

Dear hwimh,

To get some details about our Micros reliability and Qualification here are the reports : http://www.st.com/mcu/familiesdocs-110.html#Quality%20-%20Product%20Documentation Now, Back to your case, I highly suggest you to get in touch with your distributor or direct ST technical contact for a deep and further Failure analysis since it seems you have damaged the reset input signal For Investigation. By my experience, These kind of failure are generally due to Hardware and Board issues either decoupling or having an excessive stress (signal shape and noise) that is above the operating conditions and may be above the Absolute Maximum Rating of our micros that may lead to this behavior. Therefore a deeper analysis by a specialist is required. In a General way, The Forums {here} are not suitable for this kind technical support and we can just give you Hypothesis and theoretical interpretation. By the way, Why you are performing these tests with a special Robot ? I'm interested to know as it is not in my specialty :( Cheers, STOne-32. [ This message was edited by: STOne-32 on 11-03-2009 20:29 ]

[design5]

Posted on May 17, 2011 at 13:06

Hi Fellows:

The ''reset circuit'' design, as shown in the data books, does not provide a series 1K0 resistor from the RESET pin to the RC junction. Consequently, there is unlimited instantaneous current from flow from the Capacitor to the RESET pin. This may cause silicon failures over time, especially if the current flows through the substrate catch diodes. Most likely this is the cause of the failures you have observed. Furthermore, the ''reset circuit'', without the 1K0 series protection resistor, will not work reliably with the watchdog timeout circuit, and may also cause damage to the internal open drain sink transistor (unlimited instantaneous current). Given a large enough external capacitor, the 20 us watchdog pulse is severely damped by the external capacitor (without the series 1K0 resistor) to the point that the watchdog may not be reliable. MODERATOR: I have previously recommended to ST that the ''reset circuit'' diagram be corrected in the PDF documents to include the 1K0 series protection resistor, however, to date, that change has not been implemented. HWIMH: I strongly recommend that you install the 1K0 series resistor then repeat your tests and report back on this forum with the results. I expect you will have no failures with the protection resistor installed. Thanks, Garry. [ This message was edited by: design6 on 12-03-2009 00:23 ]

[hwimh]

Posted on May 17, 2011 at 13:06

STOne-32: I say it's a 'Reliability Test'.

It's my responsibility to offer reliable product to my customer. It may be severe problem if reset failure happens after some time. And I'm already contacting with local ST technician too. Thank you very much! design6: Thanks for your smart advice! I will try. And how about only using 10K pull up, with no Cap? Will it cause any problems? [ This message was edited by: hwimh on 12-03-2009 02:26 ]

[design5]

Posted on May 17, 2011 at 13:06

Hwimh:

No problem using a 10K pullup without capacitor. In that case, you don't need the 1K0 series protection resistor. The 1K is only required when you have an external capacitor. Larger capacitors cause more problems. The CPU has an internal pull up resistor on RESET, so, you can actually operate with the RESET pin not connected to anything (except the JTAG port). In my designs, my RESET pin has no components, except JTAG. So, for your next automated testing, could you prepare 5 boards without the capacitor, and 5 boards with capacitor plus 1K series resistor? I would be really interested in the test results. Thanks, best wishes, Garry. [ This message was edited by: design6 on 12-03-2009 02:32 ]

[lil-vince]

Posted on May 17, 2011 at 13:06

Hi all,

Hwimh have you made some other test ? I´m designing a new board with the STM32 and I´m interested by your test results. Thanks, Vince

[lil-vince]

Posted on May 17, 2011 at 13:06

Hi Hwimh,

Thanks for your answer, I think I will let it floating (multiplexed with JTAG (floating) and RTS from the UART (MAX232)) like design6 says. But where have you seen that the Reset pin need a 10k0 pull-up with the 0,1uF? I just see in AN2586 the 0,1uF Cap for the push-button... (and it´s not needed for the JTAG) Thanks Vince

[hwimh]

Posted on May 17, 2011 at 13:06

Hi Vince,

The test was not done by me, but my customers.(two different companies) As a result, with 0.1uF Electrolytic Cap: 3 samples got failure (10 samples, 10,000 times) with 0.1uF Ceramic Cap: Passed (5 samples, 10,000 times) with no Cap: Passed (5 samples, 10,000 times) But, remember the result may vary with your board or test conditions. I'm sent the error sample to ST for decap. Thanks.[ This message was edited by: hwimh on 30-04-2009 06:25 ]

[design5]

Posted on May 17, 2011 at 13:06

Hi Folks:

Inside the STM32, the RESET pin has a fixed 40K resistor that is always active. Only if the hardware is located in an externally noisy environment with the possibility of picking up noise spikes on the RESET line, would it be necessary to lower the impedance by an external resistor, such as 10K, 4K7 or even 2K2. Otherwise, the internal 40K works fine. If, however, somebody needs to implement an external Reset Pushbutton, then simple contact debounce can be achieved by using a capacitor from RESET pin to ground. The external 10K pullup resistor would give shorter debounce times for large capacitors. However, in all cases, whenever there is any external capacitor, a series resistor MUST be used to limit the discharge current from the capacitor into the RESET pin. Furthermore, the internal 20 microsecond watchdog timeout pulse will be dampened, and may become ineffective, by using an external capacitor without a series resistor. Simply put, the STM32 data sheet is WRONG. Someday, hopefully, it will be corrected. Garry.