The global chip shortage has not made the life of the electronic design engineer an easy one, as products have been designed around whatever parts are available rather than the first choices. This has manifested itself in some unexpected ways, including as [CNX software] investigates, products whose multiple-choice bill of materials has led to mistakes being made in manufacture.
On the face of it, designing a PCB with two sets of footprints to accommodate more than one part choice is a clever move. But as Radxa found out with their Rock 3A single board computer, this could lead to a production mishap as some boards left the production line with a mix-and-match BoM in their USB PD circuitry which left them unable to operate from voltages above 5 V. The board has footprints for both an Injoinic and a WCH part, and the faulty boards appear to have the support components fitted for the other chip to the one on the board.
We’d join [CNX] in congratulating Radxa for coming clean, and we like that one of the options to fix it is to be sent the chip to fit yourself. We’re left rather glad that it wasn’t us on whose watch such a mistake occurred, as from experience we know these things can happen all too easily.
Has the chip shortage led to any similar production mistakes in your life? Let us know in the comments.
So that explains why OKDO’s launch of this new “amazing” Pi competitor has been delayed.
Maybe for a power circuitry and similar easily separatable sections it’d be better to modularize it so that you just swap the inpedenpent power board instead of using PCB space for alternative sections. Just a thought. Not my invention as so many devices i’ve seen has this kind of system. I just would’ve not taken the risk.
It would’ve prevented this sort of mess by making it easier to know which parts to put on the power board, since they are different boards, and by being able to swap the power board separately.
Though it would’ve required then to have a new connector or soldering the power board to the main board, i bet this mess ate all the profits anyway for long time and it wasted PCB space.
In general connectors are expensive. If you have the board space, board space is cheap (free). So just adding extra footprints is cheap. The only next step that’s important is to get the actual “assemble this” list correctly populated. It is possible to do it right and upfront everybody will say they can mange it and won’t make a mistake. But mistakes happen.
In my case I modified the “export the PNP file” program to listen to the “assembly variants” settings. If you decide to “We’ll do that manually on the PNP machine” you’re bound to get errors. Usually those are found when testing the first few off the production line. Here you see that it’s not a “would fail testing” error, but just an obscure feature that didn’t work.
Hello rewolff I’m electronic engineer and one of mine responsibility is to design and prepare PCB for production.
We are talking rather about big capacities not small production.
I must disagree that PCB space is cheep, one sq inch cost additional $1 which might look like nothing but this is also biggest testing enclosure, biggest device enclosure, biggest parcel etc.
In addition if you place more complicated footprints this occur as highly more complicated test procedures and test enclosures with additional test probes and of course longer test time.
Resiliency projects are one of most complicated projects to design and people should keep in mind that production of 100 devices is totally different than production of 1k, 10k or 1 milion devices annually.
Even small cost multiplied by 1 milion devices gives big numbers.
Don’t forget also that many PCBs these days are multi layer, high density boards that are VERY expensive. BGA chips simply cannot be used on a simple double sided boards, there isn’t enough room on two copper layers for all the traces they require to make connections with the outside world. Special laminates like Rogers for microwave circuits are even more expensive… and there are flexible circuits also…. even circuits on glass ITO substrates for LCDs
In this case in particular though, the board size was already stipulated since it had to be Raspberry Pi compatible.
Yeah, we also use assembly variants and I’m surprised they didn’t. I wonder if this change happened at the board fab/assembly house, like they were partway through a long term run and couldn’t reorder a part.
But then it would no longer be a single board computer j/k
I was looking at my CM4 snug in its carrier board, hat seated jauntily to one side, when I noticed the PCIe slot and visions of graphics cards danced through my head…
Wait… guys I think I’m doing SBC wrong
I populated 5v can transcirver on a 3.2v transcirver footprint space on my board. Boards shipped. I realized the mistake and had to recall and repair by hand to save the batch. It’s been hell.
Not exactly the same result bit chip shortage made me scavenge dev boards for the main logic chips as they are not a obtainable easily or fakes are around. Dev boards can be tested before scavenging and sent back if fakes/defects are found. Not so easy with bare chips when that are avail.
I worked at Cricut during the first year of the pandemic. Between chip shortages and everyone being at home wanting to learn new crafts they were getting slammed with orders. There’s at least 3 or 4 different boms that had to have different processors or leds but perform the same on the user side. That led to various issues and pains.
Just a question but what would it take to add a touch of old school tech to new era computing. I had a friend that had a switch big thing had A. B. C. And another dile with numbers. He had 3 different computers set 3 different ways one was over clocked one was this and that. But the switch made it possible to go from example 2.5v to 3.5v or even 5.0v. can something like that be designed In a smaller scale and made as a addition
Back in the early SMT days our boards included layouts for PTH and SMT RAM and Processors.
We find it slowing development at the moment. You go down the line of designing, prototyping and then are unable to purchase a component in suitable quantities.
So you go back to the drawing board……
How does the chip shortage effect the antennas for RFID labels? I just took a job in the RFID insertion field and it seems like antennas and inserts are worth thier weight in gold right now. Curious to know if this is price gouging by suppliers due to the shortage or if we (my company) are being played for fools.
The RP2040 flying in a military jet trainer.
Spent two weeks for evaluating a chip, when it was approved, chip wasn’t anymore available, two weeks wasted.
Take huge risks, choose parts that seems to fit buy them in at least year of quantity and evaluate afterwards.
I agree. If they don’t end up working for you, someone will buy them on the secondary market. But now we have to watch out for economic downturn. You don’t want to be the one holding the bag of 5 million chips when demand shrivels up.