our request disappears before the app sees it

RaviCarter0998 · 3 Mar 2025, 13:38 UTC

Closed
RA
RaviCarter0998
UR5e inspection bench: PLC request clears on a timer, sometimes before the app reads it. Ordinary job interface only; safety functions are separate. Ownership undocumented.

13 replies

PR
PriyaCarter1003
Replying to RaviCarter0998

Get both programmers to write the intended state sequence together, including who sets and clears each signal. Making the pulse longer might hide the symptom while leaving the two sides with different ideas about what acknowledgement means.

9 points
RA
RaviCarter0998
Replying to PriyaCarter1003

PLC author means 'request offered'. App author means 'request accepted'. Same bit name, different contract.

8 points
PR
PriyaCarter1003
Replying to RaviCarter0998

There's a useful starting point. Define the meaning of each transition as well as its owner, including what remains asserted until the other side confirms it. Names alone won't resolve that disagreement.

10 points
JA
JasperCarter0986
Replying to RaviCarter0998

Does the handover documentation include what should happen when the PC restarts halfway through, because we inherited a cell where both programs seemed reasonable individually and neither expected to meet the other's old signals on startup?

17 points
PR
PriyaCarter1003
Replying to JasperCarter0986

Include that in the contract. Normal completion, timeout, reconnect and restart need defined outcomes, not an instruction for the operator to keep clearing bits until the screen looks ready again.

18 points
RA
RaviCarter0998
Replying to PriyaCarter1003

No restart definition exists. We're reviewing the job state diagram before changing either program.

8 points
PR
PriyaCarter1003
Replying to RaviCarter0998

Use the missed-request trace as one example for that review. It should explain the observed loss, and the revised design should specify how receipt is confirmed despite different polling schedules.

7 points
CA
CallumBell0621
Replying to RaviCarter0998

Our training bench had two writers clearing one status bit. Could you check the actual writers against the proposed ownership table, not only the comments?

18 points
RA
RaviCarter0998
Replying to CallumBell0621

Found an old cleanup routine clearing request too. Comments never mentioned it. Added to the review.

22 points
JA
JasperCarter0986
Replying to RaviCarter0998

That sounds familiar; who will own the final interface document when both programmers leave, so maintenance has one place to look instead of choosing whichever comment seems newest?

23 points
LU
LucyBrown0915
Replying to JasperCarter0986

Assign a joint acceptance test too. Each side passing its own test isn't enough if the tests assume different clearing rules. Exercise the complete job interface, including interrupted exchanges, in a controlled environment before deployment.

1 points
PR
PriyaCarter1003
Replying to LucyBrown0915

And make the operator recovery procedure match those tested states. Otherwise the software contract will be precise and the shift instructions will still say 'restart and try again', which reintroduces the uncertainty.

8 points
RA
RaviCarter0998
Replying to PriyaCarter1003

Controls lead owns the draft contract. Joint tests and recovery instructions pending; no live interface change yet.

7 points

Discussion closed

This discussion is closed to new replies after six months without activity. Last activity: .