Barcode scan timestamps to track part cycle time per workstation

You run a five-station manual cell in single-piece flow, roughly 100 pieces a day, and the improvement project starting this quarter needs processing time per part at every station. The system that fits is also the simplest: a barcode scan at the entry of each workstation plus one at the exit of the last — six readers feeding one logging host, and timestamps. The arithmetic does the measuring, so no per-scanner PLC is needed. This guide gives the wiring options, the data rules that keep timestamps trustworthy, a worked example from six stamps to a daily report, and the parts to order.

The decision: six scan points, one host, one clock

With one piece in the cell at a time, a station's cycle time is the gap between two stamps:

The whole system therefore reduces to six timestamped records per part. Six scan points at 100 pieces a day produce 600 records (6 × 100); at about 60 bytes per row — station ID, part ID, timestamp, status — a 90-day study accumulates 54,000 rows and 3.2 MB (54,000 × 60 B = 3.24 MB). Any logging host holds that easily.

Clock precision is not a constraint at this scale. A 7.5 h shift of 27,000 s at 100 pieces sets a takt of 270 s per piece (27,000 / 100). Stamps at 1 s resolution add at most 2 s of error to an interval: 2 / 270 = 0.74%. Stamp on receipt at one host and there is exactly one clock to trust.

What a scan actually stamps

The stamp marks the moment the traveler crosses a boundary. With single-piece flow and no queue, the interval is the station's process time. If the cell takes on queuing later, the interval becomes wait plus process; add a scan at every station exit — twelve points in all — only when you need the two separated.

Three rules keep the table trustworthy. Give each physical piece a unique ID: if one invoice line covers several pieces, append a sequence number so every traveler pairs with its own neighbours. Ignore a repeat of the same ID at the same station within 3 s: a double trigger lands inside about 2 s, while a genuine return to the same station takes minutes in a cell moving one piece per takt. And validate order: every ID must appear exactly once per station, in rising station order — a missed scan merges two intervals in a way no later edit can split.

Three ways to wire six readers to one host

The log line must say which boundary sent it. Three patterns cover every layout:

For a 90-day study the default is Ethernet: the log arrives self-labelled and standard runs reach anywhere in a cell. USB wins when the host can sit among the stations. Reserve wireless for the one or two boundaries a cable cannot reach. If the stamps must later trigger machine actions, that becomes a controller project, and the scanner input methods guide compares the serial, USB and Ethernet paths into a PLC.

USB handhelds with a station prefix

Keyboard-wedge readers on one PC all type into the same cursor, so program each reader to prepend its station code. The station-prefix programming guide shows the reader side and the preamble and serial capture guide the host-side parsing; the HID keyboard mode versus serial port guide covers when HID mode is the wrong pick. A wired USB handheld is the least wiring available at a boundary:

Ethernet readers over TCP

Each reader carries a network address, so the log line names its own boundary and no prefix parsing is needed. The TCP client or listener guide covers the two socket roles. Latency is a non-issue even on the slowest reader link: a 20-character line at 9,600 baud with 8N1 framing takes 20 × 10 / 9,600 ≈ 21 ms to arrive, under a hundredth of a percent of a 270 s takt, so stamping on receipt is safe. Past a switch or across the aisle, a Fast Ethernet handheld reaches the same host over the plant network:

Wireless handhelds with a base station

Where a run would cross an aisle or a fixed machine, a wireless handheld pairs with a base that sits on the network; the cordless base-station guide walks that link end to end. Reserve wireless for the boundaries that need it — each base is one more powered device to maintain:

Mounting the scan points

Handheld readers live in cradles at the boundaries; the operator scans the traveler while taking the part. A clean Code 128 label decodes on trigger in a fraction of a second, so the scan adds nothing measurable to a 270 s takt.

The boundaries you cannot afford to lose deserve a fixed reader over a part nest: the part's presence fires the scan, so a forgotten scan becomes impossible. Keep the traveler label inside the reader's focus band and print it matte — laminated gloss throws glare, which the barcode preprocessing guide treats alongside rotation and blur. A compact USB 2.0 reader covers a nest at its 60-120 mm reading distance:

Cable the runs along the cell frame. Locking RJ45 data cable in a 15 m length reaches a host cabinet at the far end of the cell:

Worked example: from six stamps to a daily report

Part 2317-04-07 crosses the six points, and the logger stores:

Subtracting neighbours gives the report row:

Cross-check by summing the intervals: 455 + 392 + 466 + 425 + 386 = 2,124 s.

Against the 270 s takt, station 1 at 455 s runs 185 s over, or 1.69× takt (455 / 270). That station caps the cell at 27,000 / 455 = 59 pieces a day against the 100 target. Pull station 1 down to takt and the same arithmetic gives 27,000 / 270 = 100. That is the report doing its job on day one.

Flag any interval above twice the station's rolling median. A part left sitting at a boundary between shifts shows, say, 14 h — 50,400 s against a 400 s median. Report it as dwell, not process time, and keep it out of the averages.

Checks before you order

Run five checks before the order goes in:

• Print 20 traveler labels and scan them with the intended reader at the intended mounting distance; label size and glare failures show up in minutes.
• Count the runs from the layout drawing — one per reader to the host or switch — and pick cable lengths from the drawing, not from the box.
• Reserve one switch port and one power point per wired reader and per wireless base.
• Build the three rules into the logger before day one: unique ID per piece, 3 s debounce, consecutive-station order.
• Trial one shift against five hand-timed parts. The stamps should agree with the stopwatch within a couple of seconds; if they do not, the clock or the framing is wrong.

When the usual answer is wrong

The design assumes the cell stays single-piece flow. If operators start batching parts, intervals stop meaning process time: restore one piece in hand, or add entry and exit scans at every station — twelve points — so each station is measured directly.

It also assumes the traveler label survives the process. A smeared or torn label produces exactly the missed-scan gap, so reprint on synthetic stock in wet or abrasive work.

And it assumes reporting, not control. The moment a stamp must trigger a machine action, move to a controller with a real handshake: the PLC handshake queue guide shows the queue that loses no scans, the scan-to-flag latency guide shows how to shorten the response path, and the PLC link planning guide covers the link itself if the host is a controller from day one.

Parts list and next step

The hardware list is one reader per boundary: six handhelds for operator-scanned points, or five handhelds and one fixed reader over the most critical nest. Start from the handheld code readers and fixed code readers categories, add cable from the cables page, or filter by interface and mount in the code reader finder. The host can be any small industrial mini PC mounted inside the cell cabinet. More station-floor patterns sit on the code reading topic page, and if you want the layout checked against a parts list, contact us.