Tracking Old CNCs Without a PLC: Logs vs Current Sensors vs Cameras (₹)

You can measure an old CNC fleet with no PLC, no MES and no network port in three ways: operator logs, a current sensor clamped to the spindle drive, or a camera watching the machine. They differ sharply on cost per machine, how much the machine must be opened up, and how much you can trust the data. The one rule worth remembering: current sensors need a cut into the electrical panel; a camera is the only method that needs no contact with the machine at all.

By the Mama Editorial Team. Most Indian machine shops run a mixed fleet — a few newer VMCs beside Fanuc 0, old Mitsubishi and Siemens controls, and manual lathes that never had a controller to begin with. None of them will hand you a clean "I am cutting" signal. Here is how to get one anyway, compared in rupees.

Key takeaways

• Three realistic options without a PLC: manual logsheets, spindle current/power sensors, and cameras.
• Manual logs are near-zero capex but self-reported — they capture breakdowns and miss the micro-stops that cost the most.
• Current sensors are objective and cheap per machine, but need a licensed electrical cut-in to the drive panel and machine downtime to install.
• Cameras are the only method that touches nothing on the machine — essential when the panel can't be opened, controls are too old or varied to tap, or one view covers several machines.
• Cost (indicative): logs ≈ ₹0 capex / high labour; current sensors ≈ ₹10,000–30,000/machine installed; cameras ≈ near-zero incremental if CCTV exists, else ~₹3,000–8,000/camera shared across machines in view.

Why old CNCs are hard

A modern machine on a network exposes its state over a protocol like MTConnect or OPC UA, and an MES reads cutting, idle and alarm directly. Pre-network machines don't. The controller is a closed box, often decades old, and retrofitting a data interface is expensive and sometimes impossible. So you stop asking the machine and start measuring a proxy for cutting: a logbook entry, the spindle motor's current draw, or what a camera can see. Each proxy is a different trade between money, intrusion and trust.

Option 1 — Manual logs (and the hour-meter)

The operator writes down starts, stops and reasons; a cheap hour-meter on the panel backs it up. It is the default in most shops because it costs nothing to start.

What it gets you: a rough availability picture and a reason code for big stops. What it misses is exactly what matters — the two-minute jam repeated forty times a shift, the slow-running spindle, the changeover that quietly ran 50 minutes instead of 30. These are self-reported by the person whose output they reflect and recorded after the fact. Logs are a fine first week to prove the problem exists, but a poor system of record, for the same reason hand-figured OEE over-reports: the un-logged minor stops are the biggest bucket, and they wreck the maintenance and downtime picture too.

Option 2 — Current / power sensors

Clamp a current transformer (CT) around the spindle-drive or main supply conductor, feed it to a small logger or IoT gateway, and you get an objective trace: current spikes when the spindle loads up to cut, drops to an idle baseline when it doesn't. It is the classic "no-PLC" retrofit and it works on almost any electric machine.

Strengths: objective, continuous, nobody has to remember to write anything, and the signal even hints at load (cutting metal draws more than cutting air). Costs are modest per machine — a CT plus a monitoring module runs roughly ₹10,000–30,000 per machine installed (indicative; simple energy-meter builds are cheaper, industrial machine-monitoring gateways dearer).

The catch is in the install, not the price. The CT goes inside the electrical panel, around a live conductor, which means a licensed electrician, the machine powered down for the fit, and a change to equipment that may still be under an OEM or AMC condition. On a 40-machine shop that is 40 panel entries and 40 short downtime windows to schedule. It also tells you the spindle is loaded — not why it stopped, and not whether the operator is waiting on material or a crane. You still can't see the floor.

Option 3 — Cameras

A camera pointed at the machine reads cutting the way a supervisor does — spindle turning, chips flowing, coolant on, a cycle lamp changing colour — and timestamps every transition. Crucially, it touches nothing on the machine: no panel, no power-down, no AMC question. If you already run CCTV, the incremental hardware cost can be near zero; where you add a camera it is roughly ₹3,000–8,000 per camera plus a share of storage, and one well-placed camera can watch several machines at once, so the cost per machine falls as the aisle fills. You can often run the analytics on the cameras you already own.

Video does more than on/off. Because it sees the floor, not just the drive, it catches the reasons a current sensor can't — the machine stopped because the pallet of blanks never arrived, the operator is at the tool crib, the crane is busy. That is the same signal that turns "the line felt slow" into a ranked, costed record in production-line monitoring.

The honest limits: a camera needs an unobstructed sightline to a cue — spindle motion, chip flow or a status lamp. A machine idling but powered can look like it's running if nothing visibly moves; a pillar or an open enclosure door can block the view. So the hard part of cameras is not detection, which is now commodity — it is placement: which camera, what angle, watching which machine, with the sightline to its running cue clear.

The comparison, in rupees

Indicative October 2026 figures — treat as directional, not quotes.

Where video is the only option that needs no cut-in

Three situations where cameras aren't just convenient but the only practical method:

1. You can't open the panel. OEM warranty or AMC clauses, or a safety/compliance policy, forbid third-party cut-ins. A CT needs the panel; a camera doesn't.
2. The fleet is too old or too varied to tap cleanly. Manual lathes with no drive to clamp, or ten different controller generations, make per-machine sensor work slow and fragile. A camera is indifferent to the controller — it watches the spindle, not the signal.
3. One sightline covers several machines. A single camera down an aisle can track four or five machines at once; the per-machine economics beat wiring each one, and you also get the floor context a sensor never will.

Current sensors remain the better pick when the panels are accessible and you specifically want per-machine load data and nothing else. Often the right answer on a big mixed shop is both — sensors on the newer accessible machines, cameras down the aisles of old iron — and for the ₹ payback logic across either path, see does camera-AI pay for itself.

How to start this month

• Week 1: manual logs on your busiest machine, just to size the problem.
• Week 2: pick the method per machine — panel accessible and you want load? sensor. Old, closed, or shared view? camera.
• Then: measure cutting hours, not "switched on," and feed the number straight into your machine hour rate and spindle utilisation.

This is the problem Mama was built for: old fleets with nothing to plug into. Send a short phone walk of your shop and Mama reads the feeds — your existing cameras over RTSP/ONVIF with no replacement, or cameras we add — then every morning the owner gets a plain WhatsApp note: which machines sat idle yesterday and what the lost hours were worth in rupees. It watches machines, not people; the output is where the hours went, a camera plan for the blind spots, and a proposal — never a ranking of operators.

FAQ

How do you monitor a CNC machine with no PLC or network? Measure a proxy for cutting instead of the controller: a manual log, a current sensor on the spindle drive, or a camera watching the machine. Sensors and cameras give an objective, continuous signal; logs are a quick way to prove the problem but too coarse to rely on.

Do current sensors damage the machine or void the warranty? A correctly fitted current transformer is non-invasive electrically, but it goes inside the panel and needs a licensed electrician and a power-down, so check OEM/AMC terms first. A camera avoids the question entirely because it never contacts the machine.

Can a camera really tell if a CNC is cutting? Yes, when it has a clear sightline to a cue — spindle motion, chip or coolant flow, or a cycle/status lamp. It cannot reliably distinguish an idle-but-powered machine from a running one if nothing visibly moves, which is why placement and an unobstructed view matter more than the detection itself.

Which is cheapest per machine? Manual logs have almost no capital cost but a high accuracy/labour cost. Current sensors run roughly ₹10,000–30,000 per machine installed (indicative). Cameras are near-zero incremental if you already have CCTV, and because one camera can watch several machines, the per-machine cost falls as coverage grows.