How it fits together¶
There is one decision underneath everything else in XPressRT, and it is easier to see in a week at a bench than in a diagram.
Two people work on a test place for a battery management board. Anna tests. She does not program. Bernd programs. He is busy with something else and has no time.
Monday — Anna writes down what she needs¶
Anna has to establish that the supply voltage sits inside its window. She creates a data point:
Pruefling / Versorgung / U_batt — read from the unit, in volts, plausibly between 0 and 60
That is vocabulary, not equipment. There is no instrument yet, no connection, no Bernd.
Then she builds the step — in the editor, by choosing: read data points, that point, between 11.5 and 12.5, with the message Read the supply voltage. And a test case that says: Check the supply voltage is within its window.
Tuesday — the test runs. By hand.¶
Anna presses Run. No test stand serves that point, so the application asks her:
Read the supply voltage — Pruefling/Versorgung/U_batt [ ____ ] V
She measures with a handheld meter, types 12.1, and the test passes. The
report holds the value, the unit, the criterion it was judged against, the
verdict, the time — and that the reading came from the operator.
This is the part people expect to be a stopgap and it is not. It is a real test case with a real record. Only slower.
Wednesday — the gap is a list, not a rumour¶
Because tests name quantities rather than instruments, the application knows which quantities are being asked for and how often somebody is typing them in by hand. That is a work list, ordered by what it would actually save:
U_batt — used in 14 test cases, entered by hand 31 times in two weeks.
Bernd gets a case, not a request. And Anna was not blocked for a single minute waiting for him.
Thursday — Bernd connects the bench, once¶
Bernd connects the bus that is already there, and the test place announces what it can serve — a few hundred channels, with their names and units.
Nothing he does mentions Anna's data points. He does not know which quantities she tests, and he does not need to: he connects equipment, not test cases.
Thursday afternoon — Anna binds it, without Bernd¶
Anna opens her data point, and it now says the bench can serve it. She searches for batt, gets three of the hundreds of channels, and picks the right one. Unit against unit, direction against direction, range against range — the application checks that the two are talking about the same quantity before it lets her bind them.
The catalogue did not create the data point. It filled it. Anna went from what she needs to what the bench offers, never the other way around — otherwise she would end up with hundreds of data points that are a picture of the test stand rather than a vocabulary for testing.
Friday — the same test, automatically¶
Anna presses Run. Nothing is asked. The report holds the same line as Tuesday, except that the reading came from the bench instead of from her, with how old the value was when it was taken.
The test case did not change. The step did not change. The data point did not change. The only thing that changed is who answers.
Three weeks later¶
Anna needs the charge current. She creates the data point, writes her tests, binds the channel from the same catalogue. No ticket. No Bernd. It was not new equipment, only another quantity on a bus that is already connected.
That is where the dependency moves: an engineer is needed once per piece of equipment, not once per measured quantity. Bernd is needed again when a temperature meter arrives over the network — and not before.
And when it goes wrong, whose is it¶
The report does not just say failed. It says which of three things happened:
- 10.2 V, outside 11.5 … 12.5 — the unit failed. Anna's case.
- No value: the channel has not been received for seconds — the bench is disturbed. That is not a failed test point; it is not run. Bernd's case, or a loose cable.
- The data point no longer exists — a configuration mistake, caught before the run starts, with the list of test cases it affects.
Being able to tell those three apart is the whole reason a test names a quantity instead of an instrument.
Where this stands today
Monday to Wednesday is what XPressRT does now: data points, steps built by choosing, the operator reading the value, the record. Binding a data point to a bench — Thursday and Friday — is designed and not yet built. It is written here because it is why the parts you use today have the shape they have: everything Anna did on Monday and Tuesday stays exactly as it is when the bench arrives.