Part 2 - The CMM Gets Automated

Following our creation of the first computerized CMM system in 1967, the next major advances came from computer software.

We developed one program that permitted generalized manual scanning of any of the CMM's axes as a function of any other axis using a solid ball-shaped probe tip. For example, it could be used to record a probe's Y axis as the probe crossed any of a set of preprogrammed Z axis target points. That satisfied some of the customers who needed a scanning capability.

There were others that needed a scanning capability involving polar coordinates so we wrote another scanning program that performed the trigonometry necessary to record polar radius as a function of polar angle target points. This was especially useful for measuring the shapes of parts such as cams.

One program was developed for the purpose of measuring the contours of projectiles such as artillery shells. This program was the first of which we knew that performed a best fit to a circle. Later, this same computational algorithm would find its way into many other programs for our CMMs.

All of these programs, and others of the period, were designed for use on manually operated CMMs.

In 1970, the touch probe was still five years in the future. The very first servoed CMM measured the locations of holes in a steel plate using solid tapered probes. Here is the story of how it came to be.

In 1969, Sheffield had built a very specialized super accurate measuring system for the Atomic Energy Commission that was servoed. In order to be so accurate it was also very, very slow. In 1970 we decided to try to put servos on a typical shop CMM.

The overall concept was pretty simple. Initially, there would be a typical manual operation to establish the location and misalignment of the part on the CMM's table. Automatic operation could then commence.

The servos would be engaged and the Z axis would be driven upward high enough for the probe to easily clear the part. Then the X and Y axis servos would be driven horizontally to bring the tapered probe directly over the nominal center of a hole to be inspected. Then, with the X and Y axes remaining motionless, the Z axis would be driven downward far enough to bring the lower tip of the tapered probe into the top of the hole. Then all of the servos would be completely disengaged causing the CMM to behave as a manual CMM for the moment. By having the Z axis counter-balance set for a net slow downward drift, the tapered probe would descend by gravity deeper into the hole until the taper of the probe was fully seated in the hole. In the process, the X and Y axes would be urged to the hole's actual center coordinates by the action of the probe's taper. After a suitable delay to allow the machine to settle, the positions of the X and Y axes would be read by the computer and a suitable evaluation of the hole's position could be reported. Then the process would repeat for each additional hole to be inspected.

The mechanism for engaging and disengaging the X and Y axis servos was straightforward and gave little trouble. The Z axis servo system was a little more involved. Let us begin with a description of the Z axis mechanism for the CMM in its manual configuration.

The Z axis consisted of a steel shaft which traveled vertically in a set of bearings (not shown) mounted in the CMM's carriage. An optical grating (also not shown) was affixed to the side of the Z shaft. A reading head in the carriage optically monitored changes in the position of this grating, and therefore the position of the Z shaft, as it traveled up and down.

The Z axis shaft was perhaps 1.25 inches thick, 1.5 inches wide, and 30 inches long. Although it had a hole running its full length to allow for a probe knockout feature, it still weighed something like 20 or 25 pounds. The opening at the bottom of the Z shaft was machined with a slight taper to allow for probes to be inserted and removed.

It is not practical to expect a CMM operator to lift a 25 pound Z axis shaft every time he/she wants to move the probe from one hole to the next. Therefore, a mechanism is provided to counter-balance the weight of the Z axis shaft and probe. It consists of a long spriral wound spring similar to a clock's main spring located inside a cylindrical drum. One end of a thin steel band is attached to the drum's outer surface and wound around the drum. The other end of the band is attached to the side of the Z shaft near its lower end. In normal use, the spring is carefully wound up to the point where the Z shaft neither rises nor falls if the operator's hand is taken away while not in contact with a part. As mentioned earlier, for this application the counter-balance was adjusted to yield a slow downward drift of the Z axis shaft when released.

Finding no practical way to drive the Z axis shaft directly with a servo, we did the next best thing. We drove the drum and relied on gravity to keep the steel band tight, yielding the equivalent result. To do this, we attached a large spur gear to the side of the drum. It was engaged by a smaller gear driven by a DC torque motor. A magnetic clutch located in the drive train between between the motor's shaft and the smaller gear could be engaged or disengaged to allow the Z axis to be driven by the motor or to allow it to float freely as in a manual CMM.

The motor itself also included a device called a tachometer. A tachometer, which is basically a small DC generator, produces a signal that allows a servo driving circuit to control the current required to keep the motor turning at a desired speed. The desired speed was to be computed from one moment to the next by the computer program which compared the actual location of the Z axis with a mathematical model of where the computer thought it should be during the duration of a move.

Although such a DC torque motor can be very powerful when it needs to be, this motor was not going to be required to lift the entire weight of the Z axis shaft since almost all of that weight was being offset by the counter-balance mechanism.

We managed to program a DEC PDP-8/L minicomputer to read a part program block from paper tape, run the servos to measure the part for the previously read part program block, and print the results for the part program block that came before that, all at the same time. In dramatic contrast with today's computer systems, this small machine had only 4,096 12-bit words of memory. In modern terms, that would be 0.000006 gigabytes.

Initially, we tested the system with the CMM's metal covers removed. We ran into a few snags here and there. For example, get a sign reversed in the computer program or reverse the two wires of a motor or a tachometer and you instantly get an axis running away to a machine end stop. Little by little, we ironed these issues out and the machine began to perform as intended. The moment when this machine had to be packed up to be loaded on a truck for transport to a machine tool show in Chicago was drawing very close and we had to concentrate on saving every bit of time possible.

Since it appeared the system was working properly, we added the CMM's covers and proceeded to some final checkout.

Then it happened. I typed an X or Y axis nominal coordinate that was off by about half an inch into a part program block. When the system came to this part program block, the machine drove up, then over to the location it thought was the nominal location of the hole, and proceeded downward as usual. When the probe drove down, it missed the hole due to my error and stopped abruptly on the steel part's surface about a half inch from the intended hole.

From the computer's point of view, the Z axis just did not seem to be positioned quite all the way down so it continued to direct the servo control to drive the Z axis downward at a slow speed. The servo control continued to send current to the Z motor, carefully regulating the current to maintain the Z motor's speed at that directed by the computer program. Since the probe had run into the part, the Z axis shaft could go no lower. As the motor continued to turn, the drum continued to unwind more band. It was also winding up the coil spring inside the drum even tighter than normal. The band became slack. At the time, we could not see any of this because the CMM's covers hid what was happening inside.

The motor unwound all of the band from the drum and then began winding it back up again the other way around the drum. All the while, the coil spring inside the drum was being wound ever tighter.

As we stood there watching the probe sit on the part, the Z axis suddenly lifted off the part and began to drive slowly upward, being pulled upward by the band as it wound up the wrong way on the drum. Of course, I had no idea why this was happening. It seemed as though the computer had come up with some new servoed motion I never intended. This all took place in a matter of a few seconds so there was very little time to think what to do.

Naturally, I pressed the Emergency Stop button to stop the action. That button was wired to release all the servoes from driving the CMM so it released the Z axis magnetic clutch. No longer being driven by the motor, the now very tightly wound coil spring began to unwind with great speed. Even before the Z axis could begin to fall very much, the coil spring fully unwound the reverse-wound band and began winding it back up in the forward direction again. Without the support of the band, the Z axis shaft began to fall. The falling Z axis shaft pulled downward on its end of the band while the winding of the drum was pulling up on its end of the band. By the time it ran out of slack, the impact force on the band was so great that it snapped the band in two near its connection point with the Z shaft.

No longer supported by the band at all, the unsupported Z axis shaft dropped in free fall and slammed into the part. After the initial impact with the part, there was a terrible sound coming from inside the CMM as the mainspring coil continued to unwind completely with the loose end of the band flapping around inside the covers like some maniacal runaway window blind. Then the system fell quiet.

After we figured out what had happened, we needed a solution and we needed it fast. The machine was so late in being readied for the show that the general manager directed the Marketing department to abandon any thoughts of presenting this wonderous new machine and to revert to their previous sales pitch at the show.

We put on our thinking caps. A small addition to the computer program did the trick. It simply watched the Z axis counter while the Z axis was being driven downward. If the Z axis stopped moving completely when it should still have been moving, the servoed move was interrupted, the Z axis motor was directed to move upward slowly for a few seconds to be sure that the probe had lifted off the part, and then an error message was presented.

As soon as the machine was fitted with a new band and the counter-balance adjusted, we tested this feature and it worked like a charm. So we quickly called the Marketing folks down to the machine and showed them that it worked and explained what it did to defend itself against a missing or mislocated hole.

The machine was packed up and barely made it onto the truck in time for its journey to the show.

One annoying observation was that when X and Y were simultaneously being positioned, they both drove at full speed until they independently reached their respective targets. This worked acceptably but when a move involving both axes took place, it drove along a diagonal angle until one axis made it to its target. The other axis then continued alone until that axis arrived at its own target. The result was a trajectory that often had a bend in it. It worked but surely looked odd.

On the plane ride up to Chicago for the show, I began to think about how we could scale the axis speeds so that both X and Y would get to their targets at the same time. I wrote the code while on the plane and hand-assembled a patch to make it work. When the CMM was set up at the show, I keyed in the patch through the computer's front panel and gave it a try. It worked.

We engineers and a technician who had built this prototype system were kept nearby at the show in case the machine would need any adjustments or repairs but it ran well all week long. The scariest part was that the sales folks kept putting a quarter over one of the holes to demonstrate how the machine was able to survive a missing hole. Every time I watched them do that, I just prayed that the quick fix would work as intended. Luckily, it did work every single time.

Soon there were people standing four deep around that system to watch the world's first commercially available servoed CMM automatically measure that flat plate with holes in it. It continued all week long, always four deep. We were blown away by the response. Clearly, the future of the CMM could be seen in this machine.