Draper Polishing Machine for Large Optics: Restore or Switch?

When we talk to shops about a draper polishing machine for large optics, the first thing most of them want to know is whether they should just buy a CNC machine and be done with it. For a lot of workshops the honest answer is still no. The overarm Draper looks old-fashioned, but it keeps producing excellent surfaces on mirrors a metre across or more. A well-maintained unit is still a serious piece of kit.

Over the years our team has visited a good number of these workshops, from small two-person operations to busy sheds running three machines at once. We have seen what keeps these machines working, what usually breaks, and where the real money goes. Below we explain why the design still earns its place, how we approach restoring an older unit, and the point at which CNC really does start to make sense.

Why shops still rely on the Draper design

A Draper machine is an overarm polisher. The glass blank sits on a bed underneath, and a long arm hangs above it. The arm rotates while it sweeps the polishing tool back and forth across the surface. The idea is simple, yet it works very well on large mirrors. You get stable, repeatable results without a complicated control system or a programmer on the payroll.

Cost is the main reason people hang on to them. A CNC machine with similar capacity can cost several times as much, and that figure leaves out programming, fixtures and the time needed to train someone to run it. A used Draper with a sound frame usually only needs restoration before it goes back on the floor, which turns the budget conversation into something far more manageable.

Rigidity matters too. A heavy frame and a solid spindle keep vibration down even with a heavy blank. We once visited a shop whose owner had been running the same machine for more than twenty years. The frame had barely shifted, and the only serious work had been replacing the drive bearings twice. Maintenance stays simple because most failures trace back to a handful of drive components, and a competent local machinist can usually handle the repair without sending anything away.

There are limits, though. This type of machine suits spherical and near-spherical surfaces best. Once the geometry becomes complicated, the Draper stops being the right answer, and pretending otherwise just wastes time and glass.

How the overarm stroke shapes the surface

The arm combines rotation with a sweeping stroke, so pressure spreads fairly evenly across the blank. When the stroke covers the surface thoroughly, each zone of the mirror loses roughly the same amount of material. Get the stroke right and the figure develops steadily. Get it wrong and you can spend a week chasing an error you introduced on day one.

The polishing tool is the part that actually meets the glass. Speed and pressure decide how fast material comes off. More pressure cuts faster, but it also builds heat and can leave scratches. Experienced polishers rarely keep one setting for the whole job. They adjust speed and pressure stage by stage and keep watching the surface as it changes.

Tool size has to suit the blank. A tool that is too small for the diameter leaves the edge and the centre removing unevenly, and that shows up in the figure within a few passes. We keep the tool large enough to cover a good portion of the mirror diameter on each pass, so no zone gets missed.

Main components and where they tend to fail

A draper machine has a frame, a rotating spindle, the overarm and a blank support bed. Each part carries a different load and wears at its own pace. Knowing which part is likely to fail first saves a lot of guesswork when something starts to feel wrong.

Bearings and the drive assembly deserve the most frequent checks. When a bearing loosens, the machine gets noisier and the rotation centre begins to wander. Contact surfaces between the arm and the spindle degrade slowly, often so gradually that operators stop noticing until the figure drifts.

Balance springs and bushings are easy to overlook. A weakened spring means the arm can’t hold pressure evenly across the blank, and the error appears as a faint lopsidedness. Worn sliding surfaces bring vibration, and vibration leaves tiny pits in the glass that are miserable to remove later.

Restoring an old Draper: our usual sequence

Start by stripping the machine down and cleaning every component properly. Resist the urge to loosen everything at once. Measure wear on the drive components first, write the figures down, and only then decide what to replace. Shops that skip this often replace perfectly good parts and leave the real problem untouched.

Next, align the spindle and check that the overarm runs parallel to the bed. This step matters more than many people expect. A spindle off by a fraction of a millimetre is enough to lose the curve you are trying to make, and careful polishing will never fully fix a geometry error built into the machine.

Replace the bearings, retighten the whole structure and run the machine empty for a while. Listen for abnormal noise and measure vibration at the spindle. If the vibration persists, go back and recheck each step rather than hoping it settles.

The timescale depends heavily on condition. A unit with a sound frame can be ready in a few weeks. A badly worn machine needing several parts remade can take a few months, particularly if you’re waiting on a machinist to finish custom work.

Choosing compounds and tools for each stage

Polishing compounds are normally matched to each stage. Cerium oxide is the usual choice for the final stage on optical glass, while aluminium oxide tends to appear in the coarser steps. Picking the wrong grit size is one of the most common causes of scratches we see, and it is almost always avoidable with a bit of discipline at the bench.

Tools come in softer or harder material depending on the stage. A soft tool holds compound well but deforms easily. A hard tool keeps its shape longer but removes material slowly. On mirrors with different radii, many polishers adjust tool hardness to avoid distorting the curve, and they learn the feel of it only through practice.

Cleaning between stages is easy to skip when you’re tired. One coarse grain left behind can cause a scratch right at the fine stage, and then you’ve lost a day. We wash the blank thoroughly, dry it and inspect it under raking light before moving on. It is slow, but far quicker than redoing a stage.

Keeping the curve under control

The simplest check is the sphere test, with a radius gauge laid across the surface. If a gap shows at the edge or the centre, the mirror has drifted away from the ideal sphere. A dial indicator gives quicker readings during the work, though it is less precise than the gauge for final judgement.

Roll-off comes up often: the edge ends up ground deeper than the centre. The usual cause is the tool leaving the edge too early, or uneven pressure when the arm reverses. Catch it early and a small change to the stroke will correct it. Leave it too long and the fix costs far more effort than it should.

Once the error is small, the sphere test stops being enough. At that point we move to interferometric measurement. The principle is covered in our guide on interferometric null testing aspheric optics, which explains why specialised equipment becomes necessary at the final stage.

Comparing costs for your workshop

The comparison below is only a rough guide. Always get current quotes, because prices move with size, age and condition.

A restored Draper is cheap to start with and suits people who already know the process well. Material removal is slow, and results depend heavily on the operator. A CNC machine of the same capacity needs a large investment and a technician who can programme it, but once running it gives strong repeatability and much higher throughput.

If you make a handful of mirrors a year, a restored Draper is usually the economical choice, and the skills you build on it stay useful. If you produce batches that must match each other closely, CNC begins to pay off. Don’t use someone else’s numbers as your benchmark. Ask at least two or three suppliers for written quotes before committing.

Signs it’s time to move to CNC

A few signs suggest the time has come. The first is when the required shape is no longer spherical or near-spherical, but freeform or tightly toleranced. The second is when output grows to the point where operators can no longer repeat results from batch to batch, however careful they are.

If you want to understand shapes beyond the sphere, read Freeform Optics Manufacturing: Beyond Spherical and Aspheric Surfaces. It explains why complex geometry calls for different tooling altogether.

Many workshops we have visited combine both approaches. The Draper handles the coarse stages and removes most of the material quickly, while CNC takes over for finishing. This keeps costs reasonable and still reaches the surface quality the job demands. Before you settle on a process, it is also worth reviewing our notes on preparing optics for surface roughness measurement, since the measurement requirements often decide the finishing route.

Frequently asked questions

Can a Draper machine polish convex mirrors?

Yes, but the tool and stroke settings differ from those used on concave mirrors. Convex surfaces are usually harder to control at the edge, so we check the figure earlier and more often than we would on a concave piece.

How long does it take to restore an old Draper machine?

It depends on the condition of the unit. Replacing bearings and realigning can be finished in a few weeks. A machine needing several parts remade may take a few months, so plan your production schedule around that.

Do I need interferometric testing after every session?

Not necessarily. During the coarse stages the sphere test is usually enough. Once you’re close to final accuracy, interferometric measurement is worth doing to confirm the result before the mirror leaves the bench.

How much does a Draper machine cost compared with CNC?

The gap is usually large, though the exact figure depends on size and condition. Request real quotes before comparing, and include programming and training costs on the CNC side of the sum.

How do I know when the bearings need replacing?

Typical signs are a louder machine, vibration during empty runs, or a rotation centre that drifts after each pass. If you notice any of these, check the machine before the glass is damaged, because repairing a scratched mirror costs far more than replacing a bearing.