Frightening rare failure... Any wisdom?

This is a long shot...
Squeeze RT1062 between thumb (top of pcb) and index finger (bottom of pcb).
Keep squeezing while trying Blink restore. You'll probably need help from your other hand as time passes on.

Maybe try the "squeeze" on MKL02 chip as a second test.

Could look for activity with oscilloscope on signals between RT1062 and Bootloader chip MKL02 (U2). Access to signals on via's on bottom of pcb? No way to interpret this activity, but if there's nothing happening, that can't be good.

Could look at 24MHz crystal with scope. It doesn't matter if it goes off frequency (as long as it starts up). Only interested to see if it's oscillating at some frequency near 24MHz. As pointed out earlier, scope probe capacitance makes it hard to measure crystal performance. But, lets see if it even oscillates in the first place.
 
Last edited:
I tried to squeeze technique on both failing boards, on the RT1062 and on the MKL02. I used plastic spacers and a vice grips with gentle and moderate pressure. No change in the results.

As for the crystal, I think you're talking about the one that's on the RT1062. I use an oscilloscope every two years whether I need to or not!😉
I'm looking for a 24 MHz signal on one of the 4 pads, I expect (with respect to GND? or with respect to another pad on the crystal?)
1755219600257.png


Now this is above my engineering pay grade.

Might I also be contacting SparkFun and/or PJRC directly?
I expect they have vastly superior knowledge and tools to assess this now-detached failing Teensy.
Is anyone aware of a precedent for actual manufacturing defects in Teensys?
(I certainly could have damaged these two Teensys, but they have received the same gentle treatment in my hands as a hundred other Teensys, and I'd like to know what I need to do differently)
 
The 24MHz crystal is the larger one with 4 pads next to the RT1062.

I just measured one and get about 800mV with respect to ground. The frequency was still pretty accurate at 23.9995MHz even with the scope probe on it which surprised me a bit. Most crystals I have tried to measure were not as well behaved. Be sure to set the probe and scope for x10 to minimize the loading on the crystal.

The active pads are the one nearest the Program button and the one on the opposite corner. The one nearest the Program button had a fairly nice sine wave look to it. The other pin had a little more of a triangular/distorted waveform look to it. I would post some scope captures but too hard to probe the little crystal and do much else. Measure a good one first to see what to expect.

Regarding possible Teensy build defects, I think that is fairly unlikely though SparkFun does not yet have the long stellar track record of building these that PJRC had. PJRC quality was truly remarkable and I have processed thousands of those boards without any issues.

Because all board assembly houses do make mistakes on occasion, a lot of that quality traces back to the test fixtures that Paul designed to catch build defects before they could make it to the field. Paul has confirmed that those test fixtures went to SparkFun and are being used, so that testing part should be consistent, though they are using a different board assembly house.

In any case almost all build issues result in permanent defects, whereas your failures are occurring over time and also sometime after your own assembly process is complete as well, so it does make it a bit of a head scratcher as to what may be going on.

The only thing somewhat unique about your build from the pictures is that your are permanently soldering everything together in a fairly compressed format. If you were somehow damaging the boards in the process, the failures would probably show up right away. Fractured solder balls under the BGA or other fractures solder joints being one possible exception which is what BillFM was probably referring to with his suggestion.
 
Thanks, @KenHahn

I’ll check out the crystal as you specify. If the crystal is not oscillating, what should I assume from that?

My design is for space efficiency (it all fits in the palm of your hand). I’ve prototyped a thicker device due to this issue, allowing stacking of boards with headers… at least to allow swapping out a failed component without trashing the whole assembly.
This comes at some cost, of course, and I was hoping to find a root cause that didn’t require a design change.

Of note, my assembly process involves unpacking and seating the Teensy-with-pins into 48 holes on the daughter board and soldering in place.
No stress to the Teensy.

If it’s a Teensy issue in the end, I expect it will occur again (I have dozens more from the same batch). I’ll run each new Teensy overnight in a test fixture (male headers on an otherwise functional daughter board). The only stress will be inserting it into the headers.

I’ll report back.
 
Crystals are pretty simple devices and usually very reliable. The load capacitor values are important and if incorrect values were installed, that can sometimes prevent the crystal from starting up or doing so reliably. Since yours worked and then didn't, I wouldn't suspect issues with the value of the load capacitors.

For that reason I would recommend turning the Teensy on and then checking for the signal. Having the probe on the crystal during startup might be enough change in capacitance to induce a startup issue. Something I didn't mention earlier is that I tested the crystals on both PJRC and SparkFun built boards and they looked and behaved the same on the O'scope.

I am not sure what the 2.2M ohm resistor to ground is for. Maybe helps dampen harmonics or something, but probably not required for basic operation. It is the small black component between the 2 crystals.

Cold solder joints on the crystal or caps could be possible, but the solder quality of both the PJRC and SparkFun boards looks pretty much identical under a microscope, so not very likely. If you have access to high magnification, it might be worth looking around in the area of the crystal and MKL02 chip for anything that looks odd.

Teensy in general are pretty robust. It is unlikely you are physically damaging them during your assembly process as you describe. Similarly it is hard to overheat the pins enough during soldering to cause damage.

It will be interesting to hear if you see any similar failures on your test fixture. You might also run some just off USB power without your baseboard to see if somehow there is any correlation with your baseboard in regards to this failure, though I can't image what that would be.
 
The 2M2 resistor is probably for biasing and may be a requirement - check the datasheet.
Click to expand...
You are correct. I didn't see it referenced in the NXP datasheet, it just shows the 2 caps, but looking a little further it is called out in the hardware development guide as a required biasing resistor.
 
@jmarsh - Oh, it seems right on target!
Thanks @Dogbone06 !

Interestingly, as I brought out my old oscilloscope finding that it is no longer supported on Mac. I do have a Windows box that I can try it.
That's not the interesting part… as I was plugging in the failing teensy, it was working!

I quickly loaded the Arduino environment and put on a modified Blinky program that printed to the console. It was able to boot and run for about five minutes. Now after making no changes, I get the nine blink series.

So, I think tracking down this alley could be fruitful for diagnostics. Perhaps a failing resistor is causing the crystal to fail to oscillate. I'm using a synthesizer chip on my daughterboard and it two has a crystal and a one mega resistor. I've had electrical engineers say that they don't know why that's necessary. In talking with the folks that make the synth chip, they said it is for consistent start up of oscillation.

I'll dust off my windows box and see if I can catch this crystal misbehaving.

Perhaps better yet, does anyone have a recommendation for an inexpensive but useful a oscilloscope, and if it's Mac based, it works with the latest Macs?
 
PicosScope has a range of USB scope that works on the latest Mac. The 2000 series is the low cost range.
 
I'm guessing this is a bad solder joint on the large bga (Ball Grid Array) package. The crystal is connected to the worst possible location on the package bga stress wise.
see:
https://forum.pjrc.com/index.php?threads/teensy-3-6-program-loader-and-i-o-glitches.40638/
https://forum.pjrc.com/index.php?threads/low-temperature-operation.59356/

The resistor post above looks interesting but unlikely and hard to measure in circuit.
On a good part in circuit I get:
466k ohm @ 100 Hz with Shannon LCR tweezers.
291k ohm @ 10 kHz with Shannon LCR tweezers.
800k ohm with regular ohm meter.
 
Well, that would be frustrating. I'm ordering some 2.2 megaohm resistors in the 0402 size from Digi key. I think I'll be able to test out the resistor theory adequately, but it sounds like if this is a BGA problem, I'm stuck with trying to detect the failures with testing before shipping, and indeed, before assembly, hopefully.

I'll let you know what I find out
 
My Picoscope 2204A arrived, and I am thrilled!
Here are my findings:
Probing both old (2023) and new (2025) Teensy that are working, I get a 1MHz oscillation at the probe location noted in the pictures.
The New Teensy is the previous failing B model. It often runs after sitting overnight without being plugged in. It is not showing the Bootloader flashing LED now. It's running Blinky.

It typically fails within hours if it has been running at all, but not yet.
I've added my code in a test more than(typical app stresses) and am awaiting failure.

Failing unit A does show the flashing bootloader LED (4 flashes) and the crystal is not oscillating

So, it appears that the 4-flash LED is associated with a non-oscillating crystal (cause or effect not determined)

I'll wait failure on Failing Board B, but does this ring any bells as to the cause of failure?

I have not tested or removed the crystal's resistor yet. They are on order.

Teensy 4.1 - 2023.jpeg
Teensy 4. 1 - 2025.jpeg

1755706026518.png
1755706081039.png
 
Sadly (and gladly), I have duplicated the problem.

I have been "stress testing" Teensy 4.1 units in a female-header-mounted manner (on my customer daughterboard) before soldering them in without the headers. All have run without trouble for at least an hour before use.

All have run well after subsequent mounting directly on my daughterboard, until one of the last few.

I recall during mounting I squeezed the Teensy against the daughterboard while reheating the first solder joint to remove a gap between the two boards. I was squeezing in the middle of the Teensy, likely right over the main IMXRT chip. Not gently enough.

After that, on power-up, the deadly 4-blinking boot-loader LED.

I've been lulled into a false sense of security over the last decade by the indestructible Teensy.
While this is unfortunate and perhaps new, it is thankfully avoidable, I believe.

I'll be treating the Teensy with more gentle maneuvers.

Thanks all for your input on this matter. In the end, I believe it was a mechanical failure caused by downward pressure on the IMXRT.

I'll post back here if there's any change to that conclusion.

Since this may be a change in manufacturing process that makes this more likely, do you think there's anyone at spark fund that would want to evaluate this or the failed boards?
 
Last edited:
You must log in or register to reply here.
Back
Top