Use the PLC built-in PID for conveyor speed trim

What the symptoms point to

Each behavior on the line traces to a specific rung in the trim ladder:

• Speed jumps between two values from scan to scan. Two writers hit the analog output word: the first-scan rung (STR SP0, LD V2021, OUT V2010) and the trim rungs behind C100 and C101. One word with two writers gives the drive a command that changes every scan.
• The belt speeds up while the loop is calling for slower running, then hunts. The trim word is base minus loop output, added onto base again, so the command mirrors the loop command about the base. The worked example below shows the arithmetic.
• The trim word turns into a huge positive number whenever the loop output is above the base. The subtraction goes negative there, and BCD word math cannot hold a negative value, so the stored trim wraps.
• The increase branch works and the decrease branch never fires. The decrease rung is gated by C100, the same coil the increase rung resets, so that branch cannot run as written. The rung comments also name V2121 where the rung reads V2021, so anyone editing by the comments touches the wrong word.
• The loop output sits at one end and never comes back. The process never responds the way the loop expects, so the integral winds until the output reaches a limit.
• The sample timers do not help. T14 and T15 with K20 pace the compare rungs; pacing changes how often the wrong arithmetic runs, never what it adds.

Restore production first

Put a known speed on the drive before touching the configuration. Skip the PID and the trim rungs and write the 60 Hz equivalent straight to the analog channel: 4095 × 60 / 120 = 2047.5 counts, so put 2047 or 2048 on the word the drive is wired to. Leave the captured setpoint on record and tell the crew the belt holds constant speed with no level correction until the loop is in.

Configure one loop in the setup

Open PLC - Setup - PID in the programming software and create the loop with a table start address. With the table at V5000, the three words that matter are:

All three are binary. The rest of the table, gains, mode and output limits, is laid out in chapter 8 of the CPU manual; read the offsets there. If the table starts at another address, the same offsets travel with it: setpoint, process value and output sit at start +2, +3 and +5 words.

The ladder that runs the loop is five rungs:

The rungs do the format work: BIN converts each BCD word before the loop sees it and BCD converts the output back for the card. Set the loop to Auto and drive C10 from the conveyor run command. The loop calculates only in Auto; in Manual the output word holds whatever it last held no matter what the error does.

Capture the setpoint in one scan

The operator watches the powder height and presses a button when it is right. C0 comes on for that press and copies the live sensor word into the setpoint word: STR C0 / LD V2010 / OUT V2000. Both words then share one scale with no conversion.

Drive C0 from a one-shot contact. If the button is held, the rung rewrites the setpoint every scan, the setpoint tracks the process value, the error stays near zero and the loop stops correcting. If the operator types a height instead, put the entry in the same scale as the sensor word, one decimal place as a ×10 entry (8.5 mm entered as 85), and read the raw sensor word at 0 mm and at 20 mm first, because where 4 mA sits on the count scale is decided by the input module. Clamp typed entries to the 0-20 mm range before they reach V2000; the same check is laid out in validating high and low HMI setpoints before they reach the PLC. If the raw word wanders while the level is steady, run the erratic analog input checks before touching the gains.

Give every V-memory word one job

The rewrite moves jobs between words the trim ladder used, so cross-reference every address before you download. In the trim ladder V2010 was the analog output and V2020 was the HMI's base frequency; in the loop design those meanings swap. The original ladder read its loop output from V3503, a word that only held a loop output if it sat inside the configured table:

Confirm exactly one writer per word, and take the analog channel's address from the module's I/O configuration so the output rung writes the channel the drive is wired to. Remove the HMI's base-speed entry and its write to V2020: kept in place, it becomes the second writer on the output word and the speed jumps return.

Check loop direction and Auto mode

1. With the loop in Auto and C10 on, hold the powder level steady near the setpoint and note V5005.
2. Step the level by hand, or step the setpoint one increment.
3. Watch which way V5005 moves.
4. Compare with the process: the speed change must move the powder level toward the setpoint.

An ultrasonic sensor over the belt reports distance to the surface, so the raw word can fall as the height rises. That flips the sign of what the loop sees, so set the loop action from this step test and not from the sensor label. If the output moved the process away from setpoint, reverse the action setting in the loop's mode word and run the test again.

One safety sentence belongs here: before the first run in Auto, clear the area where a sudden speed change matters, because a loop enabled with a wound-up integral can step the drive hard the moment it takes control.

Worked example: trim on a base inverts the command

Take the 0-4095 = 0-120 Hz output scaling with a 60 Hz base. The base in counts is 4095 × 60 / 120 = 2047.5, so the base word holds 2048. Suppose the loop output sits at 1835 counts for a moment, which is 1835 × 120 / 4095 = 53.8 Hz. That number is the complete, correct command. The trim math computes base minus loop output, 2048 − 1835 = 213, then adds the trim back onto the base: 2048 + 213 = 2261 counts, and 2261 × 120 / 4095 = 66.3 Hz. The loop asked for 53.8 Hz and the ladder delivered 66.3 Hz, 12.5 Hz off and pointing the wrong way. The level error grows, the integral keeps winding and the output runs to a limit.

One configured loop has none of this: V5005 reaches the channel unchanged and the drive follows the command. To hold the loop inside a band around the working speed, use the output limit words in the table. A 50-70 Hz band on this scale is 4095 × 50 / 120 = 1706 counts at the low end and 4095 × 70 / 120 = 2389 counts at the high end. Leave the loop the full 0-4095 range unless the process needs that bound; if 60 Hz really is the balanced speed, the output settles near 2048 on its own.

The decision

Recommendation: Delete the trim ladder, including T14, T15, C100, C101, V2021 and V2022, and remove the HMI's write to V2020. Configure one loop in PLC - Setup - PID with the table at V5000, run the five-rung ladder with V2000 as the one-shot captured setpoint and V2010 as the sensor word, and let V5005 own the full 0-4095 command; add the 1706 to 2389 count limits only if the process needs a band around 60 Hz.

Questions