Meeting Notes

Sections

Craig did quick PID tuning at startup.
Flows were relatively easy to tune due to the fast-responding nature of the processes.
However, most flow rates operated at the low end of valve openings, which introduced hysteresis — forcing him to make the tuning intentionally sluggish.

There is only one digital output (DO) tag controlling all the pumps (16 pumps split across 4 banks).
Somewhere in the hardware, a single DO is converted into multiple outputs, possibly via relays.
Craig wasn’t clear on the exact setup — he believed it uses a single low-voltage (e.g., 1124V?) signal to trigger multiple relays.

🟨 Note: “1124 Volt” might be a transcription error. Possibly meant “24V” or “120V”. Needs confirmation.

Discussion of reversed digital logic, where tags use -1 and 0 instead of 0 and 1.
Craig confirmed alignment with Opto support’s explanation.

MC described a scenario: a chemical pump failure (e.g., belt break) should trigger an alarm.
Craig explained:
- There's no direct feedback from the pump — only a pressure sensor is available.
- Since they’re diaphragm pumps, the pressure signal is very noisy.
- Initial attempts at simple averaging and first-order filters didn’t work.
- Eventually moved to a second-order filter for a better signal.
- Interlock logic: if the pump is commanded ON, but pressure doesn't rise above a setpoint, it’s assumed to be not pumping, and an interlock is triggered.

Pressure is filtered; the filtered average is compared to a setpoint.
If average pressure < setpoint, it implies a pump issue (e.g., broken, tank empty, etc.).

🟨 Section with “Kempon power Bo is turned on” has transcription artifacts. “Kempon” likely means “chem pump” or “chemical pump”.

Sudhir suggested detecting noise levels:
- If pressure signal is noisy, pump is likely on.
- If signal is flat, pump is off.
- Proposed using absolute variation (e.g., standard deviation) to define alarm logic.
Craig insisted that his implementation still relies on filtered average pressure, not signal noise.
Melissa and Sudhir observed that some code may differ across channels, possibly due to testing variants or outdated uploads.

Craig confirmed:
- The control block calculates average pressure.
- Interlock logic lives in a separate calculation chart.
- He may have tried different approaches per channel.

All agreed: regardless of the method (noise-based, average-based), the chosen strategy must produce repeatable and reliable results.

Chris Nietch raised concerns:
- Tubing length and routing changes (due to experiments) may alter baseline pressure.
- Could also affect noise or surge behavior in readings.
- Craig agreed — longer tubes might reduce noise but alter dynamics.