Resolving PAC3000 Thermocouple Noise on Injection Molders

The symptom is a Type K barrel zone set at 280 °C whose reading jumps between 180 and 360 °C and trips the high and low operating limits. A heated steel barrel has too much thermal mass to move 80 °C in a scan, so the real temperature is not changing. The signal is being corrupted somewhere between the junction and the A/D converter.

Two field patterns show where to look:

• In one base, seven thermocouple modules share an 11-slot base, and only some of them are erratic.
• In another installation, a module reads stable with one thermocouple connected, gets worse as more are added, and is very erratic with all 8 connected.

That second pattern points at interaction between channels, usually ground-referenced (common-mode) voltage. It does not point at a failed converter.

Keep the Machine Running, But Do Not Call It Fixed

Get it running, then fix it properly. The usual night-shift fixes each have a cost:

Quick fix	What it does	Why it fails
Time delay on the high/low limits (for example, ignore out-of-range for less than 10 s)	Stops nuisance stops	Also delays a real overtemperature trip. The PID loop still acts on the bad value, so heater output chatters.
Asking for hysteresis in the card	Would add a deadband to the reading	Hysteresis belongs on the alarm comparison, not on the measurement. It does nothing for the control loop input.
Heavy averaging on the raw value	Smooths the trend	Averaging a 180/360 °C swing gives a believable number that can still be wrong. It hides the fault instead of removing it.

Leave the time limiter in place as a temporary bypass while you work, and log that it is active.

Check: Trend the raw channel value for the bad zones at the fastest scan you can. Record the minimum, the maximum, and how often spikes occur. This baseline is what every later step gets compared against.

Confirm the Junction Type on Every Probe

"Grounded" in thermocouple terms means the measuring junction is welded to the sheath inside the probe. "Ungrounded" means the junction is insulated from the sheath. Clamping a sheathed probe or its shield to the barrel at the thermowell does not make it a grounded thermocouple. It only grounds the sheath.

The distinction matters here:

• A grounded junction puts the thermocouple signal conductor directly onto barrel steel.
• With several grounded junctions landed on one multi-channel module, each channel is referenced to a different point on the machine.
• Any voltage difference between those points appears at the module inputs as common-mode voltage. Heater leakage current flowing through the barrel and frame is a common source of that difference.
• If the channels do not have enough channel-to-channel isolation, that voltage couples between channels. This is why the reading degrades as more probes are connected.

Read the module's isolation and common-mode ratings from its datasheet before deciding on a fix.

1. Lock out the zone and let the probe cool enough to handle.
2. Lift both thermocouple leads at the module terminal block.
3. Measure resistance from each lead to the metal sheath with an ohmmeter.
4. Record open circuit (ungrounded) or continuity (grounded) for each probe.
5. Compare the results against the purchase specification. Mixed lots are a known cause: a supplier ships grounded junctions where ungrounded were ordered, and the wrong probes produce exactly this jumping reading.

Check: Every probe on a given module reads the same junction class, and that class matches the spec. Replace any probe that does not match before moving on.

Kill the Heaters and Watch the Trend

Heater bands are the most likely source of the injected voltage. Two mechanisms are common:

• Leakage through degraded band insulation into the barrel.
• Switching noise from the SSR or contactor coupling into thermocouple wiring that runs nearby.

A hot barrel holds its temperature for minutes, so you can drop the heaters and still read a meaningful value.

1. With the barrel at temperature, open the circuit breakers feeding the heater bands for the zones on the affected module.
2. Watch the raw trend for several minutes.
3. With a meter, measure AC volts and then DC volts from a thermocouple negative terminal to panel ground. Do this once with heaters on and once with heaters off.
4. Re-energize one zone at a time and note which band brings the noise back.
5. Insulation-test that band to ground per your site procedure. Moisture-soaked bands leak more when cold and after downtime.

Observation	Likely cause	Next step
Stable with heaters off, erratic with heaters on	Heater leakage or switching noise coupling into the thermocouple circuit	Find the offending band; fix grounding reference and wire routing
Fine with one channel wired, worse as channels are added	Common-mode interaction between grounded junctions at different potentials	Junction check, then ungrounded probes or better-isolated inputs
Fault follows the module to a new slot	Module hardware	Official support / replacement
Fault follows the terminal block or field wiring	Termination, wiring, or probe	Re-terminate, inspect the head, replace the probe
Fault stays with the slot	Base or backplane position	Official support
Reading all over the map, also on vibration or touch	Excess bare wire shorting at the probe head, or a loose termination	Re-terminate at the head and the module
Only some modules affected; ambient panel temperature normal (the reporting site measured below 105 °F)	Noise, not heat	Focus on field wiring and grounding

Check: Readings go flat with the heaters off, and the voltage from thermocouple negative to ground drops when the bands are de-energized. If the reading stays erratic with the heaters off, skip to the hardware swap section.

Set One Grounding Reference

Pick the fix based on the junction type you recorded:

• Ungrounded junctions: the input floats, and induced voltage can pull it outside the module's common-mode range. On another controller platform with ungrounded probes in ceramic tubes, tying the thermocouple negatives and the 24 V DC negative to machine ground stabilized readings disturbed by heater elements. Confirm this against the module's wiring diagram before applying it.
• Grounded junctions: do not add a second ground at the thermocouple negative in the panel. The junction is already grounded at the barrel, and a second bond closes a loop through the machine frame. The preferred fix is to replace the probes with ungrounded-junction probes. If the module's datasheet shows enough channel-to-channel isolation for the measured common-mode voltage, grounded probes can stay.
• Shield: bond it at one end only. In this installation the shield lands on the clip at the thermowell. If it is also landed in the panel, lift one end. Which end to keep is set by the module's wiring guidance.

Stop here if you are about to bond the negatives on a mix of grounded and ungrounded probes. Correct the probe mix first.

Check: Repeat the heaters-on trend and compare the spike count and amplitude against the baseline.

Swap Hardware to Separate the Module From the Field

1. Power down the base.
2. Move the P3-RTB terminal block, with its field wiring still attached, from a bad module to a known-good module.
3. Power up and trend.
4. Power down, move the bad module itself to a different slot in the base, and trend again.

Read the result:

• If the noise followed the terminal block and wiring, the problem is in the field.
• If it followed the module, it is hardware.
• If it stayed with the slot, it is the base.

Check: You can state where the fault lives: field, module, or slot. Only a field result continues down this procedure. A module or slot result goes to official support.

Load the Channels One at a Time

This repeats the "stable with one, erratic with eight" test on purpose, to find which channel pulls the others off.

1. Leave only one thermocouple connected on the affected module and confirm it reads stable.
2. Add one channel at a time, with the heaters in production state.
3. Trend all connected channels after each addition.
4. When the readings break up, disconnect the channel you just added and confirm they recover.
5. Recheck that probe's junction type, its heater band, and its wire route.

Check: You have a specific channel, or combination of channels, that reproduces the fault, and removing it restores stable readings.

Clean Up the Terminations and Routing

• Probe head: trim excess bare conductor so it cannot touch the housing or the other leg. Too much stripped wire at the head produces readings all over the map.
• Extension wire: use Type K extension wire end to end, with no copper splices. Observe polarity; in the ANSI color code the red lead is negative.
• Routing: keep thermocouple conduit away from heater band power and SSR output wiring. Do not share a conduit or raceway with them.
• Module end: re-seat the terminal block and re-torque the screws.

Check: Wiggle-test each probe head and conduit fitting while trending. The reading must not move.

Add Software Filtering Last

Once the hardware is clean, a light filter handles residual noise. The P3000 software provides the AVG (Average) instruction; its details are in help topic P112.

• Where to filter: apply the average to the raw channel value before it feeds the PID loop and the limit comparisons.
• How much: a barrel zone responds over minutes, so a short averaging window costs nothing in control. Keep the window short enough that a real heater failure still shows up promptly.
• Module settings: open the module in the hardware configuration and read which filtering or conversion options it offers.
• Alarms: put any hysteresis on the high/low limit comparison, not on the measurement.

Check: The filtered value tracks the raw trend without lag you can see, and it contains no spikes.

Verify End to End

1. Run the machine with heaters cycling at production duty.
2. Trend raw and filtered values for all channels on the affected modules for at least one full shift.
3. Confirm the spike count on the raw trend is zero, or small enough that the filter fully absorbs it.
4. Cross-check each zone at the thermowell with a calibrated handheld Type K meter. It should agree with the displayed value within your process tolerance.
5. Remove the 10 s out-of-range time limiter, or cut it back to a short confirmation delay.
6. Force a test overtemperature on one zone and confirm the high limit trips without an excess delay.
7. Record probe junction types, grounding points, and shield termination on the drawings, so replacement probes get ordered correctly.

FAQ