Introduction
Mistakes happen. That's just part of life! So it is important that when the mistakes happen to catch them in time and fix them. This is why we have an established Quality Control (QC) procedure that is applied to all of our microcontrollers.
We recently had a large batch of Nano Flip 3V3 fresh off the production line. Typically when we get a production batch done we have several hundreds to 1,000+ products to examine. So the QC step can take a significant amount of time and effort, but it is important to do right.
In this blog we are discussing three different types of QC issues identified in the recent large batch and what is done to correct the issues, if possible. But before we begin, we must first explain our QC procedure as it relates to the Nano Flip 3V3.
QC Procedure for the Nano Flip 3V3
The following steps are part of our in-house QC procedure as it relates to the Nano Flip 3V3.
- Check for obvious physical defects. This step is done automatically as each Nano Flip 3V3 is individually picked up by hand.
What does this test?
-
- Missing or damaged THT headers
- Missing or damaged SMD parts (e.g. the USB-C port)
- Issues with depaneling
- Silkscreen misprints
-
- Burn the custom bootloader (firmware) onto the ATmega328P. We have developed in-house tools to easily and quickly burn the bootloader. This tool works offline and does not require an IDE or even a computer besides another microcontroller.
What does this test?
-
- The SPI bus
- The Reset (RST) circuitry
- The ATmega328P and its clock source
- Onboard LED on digital pin D13
-
- Upload our custom "Getting Started" sketch. For this we use our own PTSolns IDE to make it seamless. In fact, we will have two or three instances of PTSolns IDE open and programming multiple boards at a time.
What does this test?
-
- The USB to UART pathway (CH340)
- Serial (Tx/Rx) to ATmega328 pathway
- Bootloader from previous step
- Onboard power circuitry
-
- Check performance based on the tests included in the uploaded 'Getting Started' sketch included tests.
What does this test?
-
- USB upload pathway
- Serial Tx pathway
- Reset circuitry
- ATmega328 execution
- Onboard LED and D13
- Analog pin read (A0), GND and 3.3V rail
- I2C bus (via scanner)
- Digital pin PWM (via fade)
- Onboard power circuitry
-
When considering each of these steps above and the parts that are tested, once the last part is passed one can be assured the board is in good working condition. At this stage the boards are sent to packaging.
The most common failures
After going through several thousands of boards, checking the QC steps for each, some statistics start to emerge. The most common failures are:
- Physical defects. This can include a THT header not soldered straight, issues with the USB-C port, part of the PCB panel remaining on the board (trivial fix). The rate of these types of defects is about 2-3 per 1000. However, some of these defects are easily fixed such as simply snapping off the PCB remaining from the panel. This is in fact the most common physical defect. Therefore, a more accurate physical defect rate is closer to 1-2 per 1000.
- Issues with SPI bus. Something is causing the bootloader not to burn properly. It could be a timing issue which is critical for burning the bootloader. This can arise from the RST operation for example. Sometimes we are unable to diagnose every time bootloader fails. The rate of these type of defects is about 1-2 per 1000.
- Issues with uploading the test sketch. This indicates a likely USB to UART issue. Several places could be the root cause from the differential D-/D+ pair, to the CH340, to the serial path to the ATmega328, as well as the main power circuitry. The rate of these type of defects is about 1 per 1000.
There are some very rare defects that we've only encountered once in the lifecycle of a board. These could include a silkscreen misprint, a PCB routing issue, tombstoning (very rare for our boards). The sum total of these rare events is about 1-2 per 1000.
Therefore, the total defect rate ranges between 0.4% to 0.6%. This is not too bad at all!
Mistakes happen, what's next?
Although we have implemented a multi-step quality control procedure, mistakes still happen. As every step is done by humans and we process large batches at a time, we would be untruthful if we said that an occasional bad board never makes it through. Though this is extremely rare, it happens.
If one of our boards is not behaving as expected, we always encourage our users to contact us first. We pride ourselves in providing top-class support and address each inquiry with a personal response (none of those canned messages!). Over the years we have found that if an issue arises it is most often not related to the quality of the board, but rather something software related, or even a bad USB cable (we've written an entire blog about this!).
What about the boards in the picture?
In the most recent QC check of a fresh Nano Flip 3V3 batch, we had identified three boards with issues.
The left board in the picture, which has a white wire attached through one of the mounting holes, did not accept the bootloader. Sometimes we spend a bit of time trying to find anything obvious that could be wrong. In this case, we didn't have time. With any bootloader related issues, we always mark defective boards by attaching the wire. This makes it very easy to identify later on. This board is taken out of circulation.
The middle board in the picture has a tiny bit of solder stuck inside the USB port. This prevents the USB-C cable from being plugged in! Although this is an easy fix to simply remove the solder, we believe the board is then no longer considered "new". Therefore, this board is taken out of circulation and our team will be using this for in-house projects.
The right board in the picture has a piece of the PCB stuck to it, that remains from the depaneling process. It is easy to remove this piece and sand the rough edge. We do this manually and put the board back into circulation.