They now use intrusive soldering that solders surface mount and through-hole components at the same time:
The idea behind intrusive soldering, also known as pin-in-paste soldering is simple: solder paste is printed onto or around a through-hole pad, and the through-hole component is passed into reflow with your SMD components. The molten solder paste then fills in the through-hole and attaches the component pin.
I used to work on an SMT line, and pin in paste was the bane of my fucking existence. The parts (mainly connectors) were rarely within tolerance, and a leg or two would consistently miss their holes, if not outright rejected by the inserter.
The main problem was that it interrupted the line. We’d have to stop and inspect each product, then reposition or replace the connectors, before the reflow oven. It also ran the risk of damaging the connector, the PCB, or even the inserter head if the insertion force was too high. We had a higher rework and scrap rate compared to similar SMD-only products, but using pin in paste meant that wave soldering could be skipped altogether, and I guess someone above my pay grade determined that it was better in terms of finances.
This is just my own experience. I don’t know Rpi manufacturing practices.
It is very common in manufacturing for removing a process step to save a large pile of money even if more human labor is needed.
As others have already implied there is always the possibility for technology to get better and thus the inspections not be needed, or perhaps automating the inspections.
Accounting is often weird. It makes no intuitive sense why just-in-time should save so much money, but it does.
I feel like just-in-time is one of the things that makes most intuitive sence. You used to tie 1M in Inventory and now you tie 100k. Stick the other 900k in a bank account (worst case scenario) and collect the interest.
I’ve done some sm work but as repairs and upgrades … it definitely was /easier/ to remove and replace: that was for sure. I’m unclear on if it ultimately had a higher real world failure rate though.
Personally I’m hopeful that their reasoning for this is increasing the quality of what does hit shelves even if there is a higher on line failure rate. They can’t always be the cheapest (and recently haven’t been) but if they can double down on “It just works” for a slightly higher price… I’m here for it and I imagine other makers likely will be as well.
They now use intrusive soldering that solders surface mount and through-hole components at the same time:
I used to work on an SMT line, and pin in paste was the bane of my fucking existence. The parts (mainly connectors) were rarely within tolerance, and a leg or two would consistently miss their holes, if not outright rejected by the inserter.
So if I’m reading that right - higher failure rate on the line but those that passed I’d imagine have a higher rate of success?
The main problem was that it interrupted the line. We’d have to stop and inspect each product, then reposition or replace the connectors, before the reflow oven. It also ran the risk of damaging the connector, the PCB, or even the inserter head if the insertion force was too high. We had a higher rework and scrap rate compared to similar SMD-only products, but using pin in paste meant that wave soldering could be skipped altogether, and I guess someone above my pay grade determined that it was better in terms of finances.
This is just my own experience. I don’t know Rpi manufacturing practices.
It is very common in manufacturing for removing a process step to save a large pile of money even if more human labor is needed.
As others have already implied there is always the possibility for technology to get better and thus the inspections not be needed, or perhaps automating the inspections.
Accounting is often weird. It makes no intuitive sense why just-in-time should save so much money, but it does.
I feel like just-in-time is one of the things that makes most intuitive sence. You used to tie 1M in Inventory and now you tie 100k. Stick the other 900k in a bank account (worst case scenario) and collect the interest.
That’s some interesting insight - thanks 👍
I’ve done some sm work but as repairs and upgrades … it definitely was /easier/ to remove and replace: that was for sure. I’m unclear on if it ultimately had a higher real world failure rate though.
Personally I’m hopeful that their reasoning for this is increasing the quality of what does hit shelves even if there is a higher on line failure rate. They can’t always be the cheapest (and recently haven’t been) but if they can double down on “It just works” for a slightly higher price… I’m here for it and I imagine other makers likely will be as well.
So, would your suspicion be that it’s causing them more failed boards in production?
I guess if it’s reducing returns, that might be something they’re accepting as a tradeoff?
How long ago was this? Could the process have become more acurate in the time you’ve been away from it?