Teensy 4.1 vs 3.6 -- Thermal Performance Data

methods

Member
We had a 4.1 sputter out in some thermal testing. During validation, I collected the following data:


Teensy 3.6
Ambient Temperature ~80C
Soak Time ~2hrs
Airflow minimal to none
Mounting - standard header set, about 1cm off the PCB (floating in free air)

Pass, no indication of abnormal behavior ;)

IMG_20210215_170618_compress91.jpg

Shown above is a report-back
* Degrees F
* Degrees F trunk (for CAN)
* Degrees C, calculated from Degrees F

+/- a few degrees

IMG_20210215_170623_compress84.jpgIMG_20210215_170623_compress84.jpg

4 measurements on the chamber including
* Fluke 87V
* Thermal Controller
* Teensy 3.6
* another random meter

I will post an update when we reach the limit, but as far as I am concerned the Teensy 3.6 is ready for any reasonable environment.

...

When I run the 4.1 thru a similar test

* Around 52C it performs
* Above that things get dicey

It fails the above test, and I dont have time to neck down on it, but... generally speaking... the Teensy 4.1 (as shipped) wont perform as well thermally as the Teensy 3.6. Not a huge surprise, as ... lol ... it is like 600Mhz in an Arduino form factor

But
Data for what it is. If anyone wants to piggy-back a test (running a specific loop or test) let me know and I will help where I can.

...

Overall
Teensy is great, we are very happy with it, I really appreciate the efforts.


To end on a positive note::
With the Teensy 4.1 I can output a 20Mhz square wave by reading my input, toggling it, and writing it back out. That is... unbelievable. Teensy 4.1 for Arb of the year.

-Schindler
 
What is the purpose of testing the board at 80C ambient? HALT, HASS, or ESS ?

Extended environmental verification? Your product's V&V ? If so, dunno if data has any meaning other than the board itself; that is, environmental tests are done for system-level verification and validation. For components and modules, should just use the manufacturer's ratings.

I have done some HALT and HASS series on systems using the T3.2 and T3.5, and most problems with these boards will occur below -30C. Above 60C, with any amount of vibration, the stuff that fails first is not the Teensy.
 
Any chance I could talk you into trying this test with the clock speed configured to 528 and 396 MHz?

Use Tools > CPU Speed before uploading to set the clock speed.

NXP's extended temperature parts are rated only up to 528 MHz. That's also the threshold where the CPU's voltage jumps from 1.15V to 1.25V, so even though it's about 15% change in speed, it could have a major effect on usable temperature range.

My guess is the commercial & industrial rated parts exactly the same silicon, though it's not clear if NXP tests them differently or "bins" the parts depending on test results.
 
Absolutely

Temperature Testing is coming up again in the next week or two. I would like to provide supporting information to the degree possible. I really appreciate the depth of hard work it takes to bring something like this to the community. Whatever we can contribute we will.

If there is a Standard Stress Test I can run it, otherwise... I basically spin the processor. I will try to normalize the test code so I can share it. Usually consists of

* Present Menu
* Take Selection from User or timeout to baseline
* Execute
* Repeat

Open Source Example
Botletics 4G Module

I was always a fan of the menu-type interface. Botletics has a shared example (he is a good guy, helps a lot)

https://github.com/botletics/SIM7000-LTE-Shield/wiki
https://www.botletics.com/products/sim7000-shield

I built on some of the code in that repository. Still to this day write every program about like that.

...
Anyhow, will report back once I have some findings.

-methods
 
Back
Top