LPC2378 temperature testing

[soren_schuler]

I am looking at using the LPC2300 series at boarder lining temperature range 90+ degC for a customer. I know the chip is rated to industrial temperature range but it there any possibility it can go higher? Is there any production data which can tell me its test range?

I have tested a few trails in a oven to see how far I can push it and I find that some work well over a 105degC and others fail around 95degC.

Can you help me understand why that is and what happens when it fails?

It is really important for me to know as I worry it is not the MCU that fails but a external part.

What do you think?

Your answer is really appreciated as this is dogging me. I would love to use a different chip but currently it is not an option.

Kind Regards,

Soren 

[Mizuz]

Hello
I am using SPC5746CBK1AVKU6 microcontroller in a project. We use PLL of microcontroller to achieve 160 MHZ frequency. For the final part under the design, several environmental tests have been defined according to the standard, in which the temperature changes from -40°C to +85°C in 140 cycles with a gradient of 10 degrees per minute, and the part must be fully functional. In another test, the temperature increases from -40 with different time steps to +85 degrees in humidity. In this sample of tests, after reaching the temperature of +85 degrees Celsius and the passage of about 30 to 45 minutes, the microcontroller loses its function and the I/Os related to fail safe are activated.
We are at a critical stage of the project, and since the datasheet of the microcontroller states that the operating temperature is up to +125 degrees, we did not expect to encounter this problem.
Also, I read the things mentioned in this forum and by the way there was a similar register in this microcontroller which was set to 0 by default but we changed it to 2 but the problem was not solved.
Please guide us on what the root of this problem can be, and can we fix this problem by configuring certain registers and pass the desired tests successfully?

[soren_schuler]

Hi Bernhard,

Okay got it.

Does this include the use of the XTAL inputs?

Do you have any suggestion to test the chip to identify the area of first failure?

This task is somewhat challenging as the internal firmware we are running limits my insight as I do not have a full knowledge of who creates what failure.

The situation I am requires me to find a way to predetermine a chip passing at 105degC.

For new designs do you have a recommendation to a good chip with equal performance and operation modes but tested at -55 to +125degC?

Kind regards,

Soren

[CarlosCasillas]

Hi Soren,

Other fact to consider is, although chips may operate at temperatures higher that the indicated on their specifications, their proper functionality is not guaranteed, by operating out of specifications.

Hope this will be useful for you.

Best regards!

/Carlos

[soren_schuler]

I completely agree. Using this controller our side its spec is risky and really a bad idea. However my testing has proven a larger variety of outcomes. I have perhaps 20% passing at higher temperature and the rest failing.

So I am expecting them to fail. But my job is to understand the failure mode to whatever degree is possible.

Given this I have been testing failing MCU's to see if I could find what the failure modes are.

Something I have noticed is that when some start to fail ALARM out pin show a pulse from 1.74V to 1.82V.

Afterwards it goes to 0V and is not active again. I have been able to power cycle the MCU and get it back on line but the ALARM out stays at 0V.

What would cause this to happen? It is obviously some to do with the RTC. The design I am looking at uses an external xtal on XTAL1 and 2 and the RTC pins are floating.

[bernhardfink]

Hi Soren,

the RTC is in its own power domain VBAT, but may be influenced by the power path coming from the standard VDD supply. The reason for this pulse could be a crash on the ARM bus system.

But do you say that this ALARM pin is then out of function forever?

The ALARM pin goes high when an RTC alarm is generated, so by default is should be on low.

More general info:  we have specified an ambient temperature of -40°C to 85°C. There is quite some margin to the maximum junction temperature of 125°C. So with a good power supply and without any influence of external components (stable clock from an external oscillator etc) the LPC2378 should work up to 105°C ... 110°C.

Starting up on such high temperature could be a problem, the analog circuits of the power-on reset block could have a spread. For higher ambient temperature you should try to increase the thermal coupling of the package to the outside, for example connect as many pins as possible to the PCB ground plane, apply a heat sink etc

And yes, there is quite some variation in the silicon process which mainly take effect when you operate them at higher temperature.

Regards,

Bernhard.

[soren_schuler]

Additional question.
Any idea why the ALARM would start out at 1.72 V?

Is there process not being followed?

Is there an ALARM bit for this output which can be read?

Which are the exact registers which can trigger the ALARM out? and can they be read through the UART?

Soren

[soren_schuler]

Sorry for the constant bombardment :smileyhappy:

I am wondering if there is any temperature test date from NXP showing behavior above 85degC?

Soren

[bernhardfink]

Hi Soren,

is the VBAT domain powered or did you also leave the VBAT pin open?

If yes, the you should take this from your watchlist, if it's not powered then the behavior is not really defined.

For sure we made stress tests above 85°C, but this happens on a chip tester in a low level format and cannot be compared to the application test you are running.

From the point where the stack check fails the core could run wild, so weird things may happen.

Regards,

Bernhard.

[soren_schuler]

I totally get what you are saying. When keeps me on the case is the fact that I have tested several systems and they all seem to fail similarly.

The Alarm pin is not used by the system but do have a tp. I am a little unsure how it is used by the MCU. I read that it relates to the RTC.

Is there a clear definition of which functions, Registers are the trigger functions for the ALARM?

[bernhardfink]

The ALARM pin fully belongs to the RTC domain, so to have a useful funktionality this domain must be

1)  powered  ==> VBAT is connected to power, PCRTC bit in PCONP register is set

and

2)  clocked ==> crystal or oscillator is connected, or from internal prescaler

If these conditions are not met, the behavior of the RTC block in general is not really defined.

Regards,

Bernhard.

[soren_schuler]

Hi Bernhard,

I am wondering if NXP has the ability to screen LP2378 at 100 deg C for customers?

If so what would the cost be?

Could you put me in contact with someone at your firm who I can discuss this option with?

I would love to hear other options and or suggestions.

Thanks you,

Soren :smileyhappy: 

[soren_schuler]

Hello Bernhard

Here is the sequency i have observed.

Chamber Temp set 95 degC
Adjacent board temp report 102 degC

I chose to use following pins for observation

RS232 data stream Tx (from MCU)
ALARM pin
RSTOUT pin

1) ALARM out pin
 -intial value was 1.74 V (not 0 V as expected)
 -Pulse from 1.74 V to 1.82 V for 148ms
 -ALARM went to 0 V after pulse and never went active again

2) Firmware reported STACK CHECK failed

3) TX data stream stopped.

4) send Reset to MCU pin 24

5) not action observed.

6) Power cycled

7) TX data stream started up again

8) After a while it stopped again

9) I repeated 7 multiple times

10) ALARM nor RSTOUT never went active.

[bernhardfink]

The LPC2378 should also work on higher temperature, but you might need to recude the operating frequency. At higher temperature you normally increase the chance to run into race conditions on the digital blocks. The edges of the signals are less steep due to increased parasitic effects, so the rise and fall times are longer. Especially the connections to the outside, for example to the SDRAM, are affected from this. You can fix this problem most of the time by reducing the operating frequency, increase number of waits states on flash, reduce SPI frequency etc.

On the analog blocks like A/D converter, PLLs etc you may see malfunctions as well, no way to fix this.

Our spec specifies 85°C as max operating temperature, this is not the ambient temperature but the silicon die temperature Tj. If you operate the die on higher temperature it will not be damaged, but the lifetime decreases as the silicon is under increased stress.

Regards,

Bernhard.

[soren_schuler]

Also,

When the chip starts to fail I was expecting both the ALARM pin and RSTOUT to go active.

Is this the correct assumption?

[soren_schuler]

Hello Bernhard,

This information is fantastic. Would you have any idea why some chip do well while others fail? Is the variation of the design that big?

Is there anything specific that stands out to you give the two senarios?

I know you have already given a lot of if but I hoping to narrow down which effect is the most influential one in my case and find a way to measure it that I can use for a screening process.

Kind regards,

Soren