Machining Monday

How a Big Machine Learns to Measure Itself: The Glass Scale

You cut a part at 8 in the morning and it measures dead on. You run the same program all day, same tool, same offsets, same everything. And somewhere in the afternoon the parts start creeping out of tolerance. Nothing changed on your sheet. You did not touch a thing. But the machine is making a different part than it made this morning, and you are the one who has to explain the scrap.

Most machinists chalk that up to the tool, or the material, or the shop warming up. Sometimes it is. But here is the part most machinists never think about: on a lot of machines, the control does not actually know where the table is. It is guessing. And in the afternoon, its guess is worse than it was at eight.

The Machine That Does Math Instead of Measuring

Here is how a standard machine finds a position. There is a servo motor turning a ball screw, and on the back of that motor is a rotary encoder counting how far the motor turned. The control knows the pitch of the screw, so it does the arithmetic: this many turns times this much travel per turn equals the table is right here. That is called semi-closed loop. The loop closes at the motor, not at the table. The control is watching the screw spin and calculating where the table ought to be.

That works fine as long as the screw behaves exactly the way the math assumes. The problem is that a ball screw is a long piece of steel doing hard work, and steel does not hold still. It wears, it has tiny pitch errors built in from the factory, it winds up under load, and most of all, it grows when it gets hot. The control never sees any of that. It is still doing the same clean arithmetic on a screw that is no longer the length it thinks it is.

Where the Ball Screw Lies

Run a screw hard and it heats up from the friction in the nut and the bearings. As it heats, it expands, and a warmer screw is a longer screw. Now every turn of the motor moves the table a hair farther than it did cold, and the control has no idea, because it is still counting turns and multiplying by the cold pitch. This is exactly why the morning part and the afternoon part come out different.

This is not a small effect. Renishaw put numbers on it: on a screw-driven axis with no direct feedback, thermal growth can drive positioning error to around 100 microns in about 20 minutes of running, and the length change on a ball screw can reach 150 microns per meter. A hundred microns is four thousandths of an inch. On a lot of work that is the whole tolerance band, gone, before your coffee is cold, and the control still thinks it is nailing the position.

A semi-closed machine is not measuring where the table is. It is calculating where the table should be, based on a screw that grew since the last time anyone checked. The glass scale is what turns that calculation back into a measurement.

What a Glass Scale Actually Does

A glass scale stops the guessing by putting a real ruler on the axis. It is a linear encoder: a strip with an extremely fine grating ruled onto glass, mounted along the travel, with an optical read head that rides the axis and scans the grating as it passes. The graduations are tiny, spaced down in the micron and sub-micron range, so the head can report position to a fraction of a thousandth. Heidenhain has been building these since 1961, etching chromium lines onto glass with a process fine enough to make scales accurate from 10 microns per meter down to half a micron per meter.

The important part is where the scale reads. It does not watch the screw. It reads the actual position of the slide, right there on the axis, and sends that back to the control. Now the loop closes at the table instead of at the motor, which is why it is called full closed loop. The control is no longer trusting arithmetic. It is looking at a measurement of where the table truly is, and it corrects to that. Backlash in the screw, pitch error, wind-up under load, thermal growth, all of it gets caught and zeroed out, because the scale sees the real position no matter what the screw is doing.

The Trick Is the Glass Itself

Here is the clever part, and it is easy to miss. If the machine grows with heat, why does the ruler you bolt to it not grow right along with it and tell the same lie? Because it is glass, and glass barely moves. The scale is made from glass with a thermal expansion coefficient down near zero, under 0.1 millionths per degree, which is a tiny fraction of what steel does. The iron castings expand, the screw expands, the whole machine breathes with the temperature of the shop. The glass scale sits through all of it and stays very nearly the same length it was cold.

That is the whole point. You want the one thing in the system that defines position to be the one thing that does not change with heat. The screw is a moving part made of steel and it will always drift. The scale is a stable ruler made of glass and it holds still. When the machine reads position off the glass instead of off the screw, it is reading off the honest part.

Why It Matters More on a Big Machine

Every machine drifts, but a big one drifts more, because error stacks up over distance. A short screw on a small VMC does not grow much, and the semi-closed math stays close enough for plenty of work. Stretch that axis out to two or three meters and the story changes fast. A longer screw has more length to expand, more pitch error accumulated end to end, and more room for a small per-turn error to add up into a big number by the far end of the travel. That 150 microns per meter is per meter, so the longer the axis, the more the machine can lie to itself.

That is why glass scales show up first and matter most on large machines, boring mills, big turning centers, gantry machines, anything holding a real tolerance across a long axis. The scale reads the true position at the far end of the travel, where the screw math has drifted the furthest. It is also why the good builders put thermal stability and direct scale feedback at the center of their big machines instead of treating them as options. A machine trusting its ball screw is doing math. A machine with a glass scale is measuring. On a big machine, across a long axis, in the afternoon, that is the whole difference between a part and a paperweight.

Resell CNC Take

When we appraise a used machine, whether it has glass scales is one of the first things we check, because it tells us what the machine was built to hold and how it will behave on a long run. A large machine with direct scale feedback is a different tool than the same size machine running semi-closed off the screw, and the price should reflect it. If you are buying for tight work across a big envelope, tell us the tolerances you hold and we will confirm the machine reads its axes off glass, that the scales and read heads are healthy, and that the feedback the machine promises is the feedback it actually has.

Frequently Asked Questions

What is a glass scale on a CNC machine?

A glass scale, also called a linear scale or linear encoder, is a precision ruler mounted along a machine axis that reports the true position of the slide to the control. It is a fine grating etched onto glass and read by an optical head. Instead of letting the control calculate position from the ball screw, it measures the real position directly, which lets the machine correct for backlash, screw pitch error, and thermal growth.

What is the difference between semi-closed loop and full closed loop?

Semi-closed loop measures position with a rotary encoder on the servo motor or ball screw and calculates where the table is from how far the screw turned. Full closed loop adds a linear scale on the axis that reads the actual slide position and closes the loop at the table. Full closed loop catches errors semi-closed cannot see, including backlash and thermal growth in the screw.

Why does a ball screw lose accuracy as it heats up?

Friction in the nut and bearings heats the screw, and a hot screw expands and gets longer. A semi-closed control still calculates travel using the cold pitch, so every turn now moves the table slightly farther than the control assumes. Renishaw data shows this thermal error can reach roughly 100 microns in about 20 minutes, with screw length change up to 150 microns per meter.

Does a glass scale get affected by heat?

Very little, and that is the point. Glass scales use glass with a thermal expansion coefficient down near zero, under 0.1 millionths per degree, far less than the steel of the screw and castings. As the machine heats and grows, the scale stays very nearly its original length, so it remains an honest ruler while the rest of the machine drifts.

Do all CNC machines have glass scales?

No. Many machines, especially smaller ones, run semi-closed loop off the screw and hold accuracy fine for their work. Glass scales are more common on machines built for tight tolerances and on large machines with long axes, where screw error accumulates over distance. On any machine it can be a factory option, so two of the same model may differ in whether they have direct scale feedback.

Are glass scales worth checking on a used machine?

Yes. Whether a machine has glass scales tells you what tolerance it was built to hold and how it behaves across a long run, and it affects value. On a used machine it is also worth confirming the scales and read heads are healthy and reading correctly, since a damaged or contaminated scale undermines the very accuracy you are paying for.