New Teensy MicroMods defective

only about 5% are faulty at the most.
I would regard that still, as a disaster.

Yeah, I have to admit I do agree with this statement.

The form factor of MM is so much more appealing for repeated projects/products. However, the amount of required testing and inevitable issues kind of balances out the scales. If the form factor of MM and reliability of PJRC could be achieved – I believe sales and satisfaction would be exponential.

I do wonder – are other MM products are experiencing the same fail rate and issues? or is it solely the Teensy MM?

Either way, I'm thankful for the current improvements, but even the improved fail rate of T_MM is wildly high compared to any other product in this category and has lead to a lot of discouragement and frustration. I'm hoping and crossing my fingers for a resolution soon.
 
I do wonder – are other MM products are experiencing the same fail rate and issues? or is it solely the Teensy MM?

Or is it even still an issue for Teensy MicroMod?

Certainly Sparkfun has heard from several people and been very responsive. Hopefully they they are able to fully resolve these issues. It's been long enough that they may already have done so.
 
Or is it even still an issue for Teensy MicroMod?

Unfortunately, it's still an issue. Again – Sparkfun has made huge strides and demonstrated phenomenal customer service through it all, but it still remains an issue.

I just received a fresh batch of 20 this week, and 3 of the 20 failed a pin test right out of the box.

I can put up with receiving dead on arrival units, because Sparkfun has phenomenal customer service and always makes it right.

However, the real trouble with this issue is that 'good' units may passed SEVERAL pin tests, then pass additional testing after packaged/programmed in a unit, but then fail 'in the field' with no warning.

By fail in the field, I mean some pins MIGHT randomly experience intermittent connection. In my case, a project that utilizes an i2c display and micro SD card – the screen loses connection resulting in my programs to freeze. Or in other instances – communication with the SD card is lost. Or in other instances the unit becomes unprogramable or loses serial connection.

Sometimes pins lose connection for a few milliseconds, sometimes for a few seconds, sometimes for a few minutes, sometimes all together.

In my ongoing experience, 5-10% have defective pins upon arrival, and an additional 1-5% fail at a later time. Sometimes days later, sometimes months later.
 
It surprises me anyone would be using a product in a production/commercial environment that has a 3 to 5% failure rate on arrival - that failure rate indicates production flaws that may cause the parts to fail in the field over time.

Its time that PJRC stepped-in and developed a Teensy 4.1 derivative that is intended for high density situations and surface mounting - either castellated pins or an M.2 style connector. I've used hundreds of Teensy LC, 3.2, 4.0 and 4.1 and never had one fail. But the hacker community style through hole solder pins on the Teensy range of products are a bit too 19th century for this day and age for commercial users.
 
- either castellated pins ...
I would be very happy to buy Micromod-sized T4s with 1.27mm pitch castellations that I could solder onto PCBs myself before selling them on to customers. The convenience and quality control aspect of having the iMXRT IC plus memory and bootloader soldered for you and then an easily soldered daughter board, is very attractive. There are so many people using Teensys on finished products, that a form-factor like this is ideal.
 
I think a big plus on a revised TeensyMM would be to have the MCU and hold-down screw symmetrical about the centerline. Having them offset must place asymmetrical stresses on the thin PCB and solder ball connections under the MPU. Most of the M2 memory cards I've seen have the screw slot on the board centerline and major chips located symmetrically about the board axis.
 
I would be very happy to buy Micromod-sized T4s with 1.27mm pitch castellations that I could solder onto PCBs myself before selling them on to customers. The convenience and quality control aspect of having the iMXRT IC plus memory and bootloader soldered for you and then an easily soldered daughter board, is very attractive. There are so many people using Teensys on finished products, that a form-factor like this is ideal.

Agreed. Castellation would be the best option. The Micromod M.2 connector is not that easy to route around as it is a high-density focal point for a lot of tracks - really needs a 4 layer PCB to be efficient, but with castellation on all 4 sides a 2 layer PCB could be used.

There is a lot of flex in the 0.8mm thick PCB used for Micromod/M.2 connectors. That flex is possibly a big contributor to the reliability problems, both during manufacture, and while in use. A castellated board could be 1.6mm thick, and is also fully supported by the underlying board..
 
It surprises me anyone would be using a product in a production/commercial environment that has a 3 to 5% failure rate on arrival - that failure rate indicates production flaws that may cause the parts to fail in the field over time.

Its time that PJRC stepped-in and developed a Teensy 4.1 derivative that is intended for high density situations and surface mounting - either castellated pins or an M.2 style connector. I've used hundreds of Teensy LC, 3.2, 4.0 and 4.1 and never had one fail. But the hacker community style through hole solder pins on the Teensy range of products are a bit too 19th century for this day and age for commercial users.


I designed a test board that tests the Micromod pins. So every Micromod product I ship out has a 100% working rate.
 
Whatever is going wrong, I'm a bit skeptical to believe it's primarily explainable by the thin size of the PCB. Many millions of M.2 form factor SSDs are made with PCBs that thickness. Most of them are loaded with BGA chips and other advanced parts.
 
What's different about the red M.2 board? Screenshot 2023-02-15 at 8.23.21 AM.pngScreenshot 2023-02-15 at 8.27.10 AM.jpgScreenshot 2023-02-15 at 8.26.00 AM.jpgScreenshot 2023-02-15 at 8.22.24 AM.jpg
 
It could be as simple as the FR-4 material used in MicroMod Teensys having lower flexural modulus (basically Young's modulus), meaning smaller force is required to bend the board.

The PCB on my MicroMod Teensy definitely feels much less stiff than e.g. my Kioxia M.2 2230 NVMe drive and my M.2 NGFF modems (mPCIe, 30x51mm). It is too small a sample to be sure, just a single datapoint. I wonder if SparkFun has tested the flexural modulus of their unpopulated boards, and compared the measurements to the manufacturer datasheets and what is generally available on the market?

At the core (pardon the pun), I think whoever has manufactured the PCB material (substrate? what's the term?) used in MicroMod Teensys may have used less glass fiber and more or weaker epoxy in the PCB construction, leading to even small forces being capable of bending the PCB too much, so that the spring pressure from the M.2 connector when the MicroMod is screwed down to the board, is sufficient to deform the PCB to break connections on the BGA pads. I would definitely do an industry standard flexural modulus test, at small deflections (say, matching those seen in MicroMods in practice), so sensitive equipment is needed (and if the modulus varies as a function of deflection, concentrate on the small deflections).

Note that flexural strength, the amount of force needed to deform the board (so much that it no longer springs back when released), is not involved here. Just the stiffness of the PCB material under bending forces.
 
The way to sort this problem out is a forensic investigation of failed boards to see what's actually happening with the solder balls. There could be many potential causes I fear, such as not being able to handle temperature cycling or vibration during shipping, inadequate reflow profile (cool spot in the reflow oven?), problems with surface contamination, inadequate strength of PCB, problems with the PCB itself such as vias not being sealed against corrosion, etc etc.
 
The way to sort this problem out is a forensic investigation of failed boards to see what's actually happening with the solder balls. There could be many potential causes I fear, such as not being able to handle temperature cycling or vibration during shipping, inadequate reflow profile (cool spot in the reflow oven?), problems with surface contamination, inadequate strength of PCB, problems with the PCB itself such as vias not being sealed against corrosion, etc etc.

Sound like a great idea! Can I borrow your scanning electron microscope for a week or two? :rolleyes:
 
Sound like a great idea! Can I borrow your scanning electron microscope for a week or two? :rolleyes:

Some medium scale and most large scale productions houses will have PCB X-ray inspection capabilities on their shop floor. If not they outsource it. Its pretty common, not just for retrospectively analysing failures, but also for ensuring ongoing production quality (sample and X-ray inspect say 1 in every 10,000 PCBs). Very expensive, but necessary for manufacturing high density boards. If fabricating MRX1062 processors, I would consider it essential to periodically inspect production samples.

We can't afford these capabilities for our small P&P line, but have used an outside lab to provide a report in the past.
 
Back
Top