Pin-compatible according to ChatGPT
At DeltaProto every project starts with an analysis of the data we receive.
For the Bill of Materials (BOM), one of the first things we have automated is checking whether everything is described properly, and checking availability in real time.
Because certain components are harder and harder to get hold of. For R's and C's we are willing to suggest an alternative ourselves. For the other components we happily leave that to an engineer.
So when we flagged a supply problem on a Nexperia transistor in a project, his reaction was:
"Oh, I will just ask ChatGPT for an alternative."
The question: which pin-compatible alternatives do you know for the Nexperia BC846W,115?
And fair is fair. The answer looked pretty convincing.
A nice list.
Same package.
Same type of transistor.
Comparable specifications.
Problem solved?
Well... not quite.
Because in electronics, pin-compatible is a little more complicated than:
- Same number of legs
- Same housing
- Roughly the same specifications
- Done
Because what about the actual pinout?
The hFE selection?
VCE(sat)?
The use in the circuit?
And maybe the best one: is the alternative available at all?
Because a perfect alternative you cannot buy is of course still not an alternative.
For questions like these, AI is a fantastic tool as far as I am concerned. But an answer that looks technically convincing is not automatically validated engineering advice.
That is why, once we receive a BOM, we automatically check the description and the current availability, and flag possible risks before they turn into a problem later.
Do you use ChatGPT or other AI tools to look for alternatives too?
And more importantly: would you put that first answer straight into your design?
#HappyEngineering
