The CMM Meets The Boob Tube

As most of us still remember, before the advent of today's plasma, LED, and LCD flat panel televsion sets and even rear projection systems, TV sets contained glass picture tubes. A glass picture tube is a large vacuum tube that uses an electrically accelerated stream of electrons to paint the picture onto the inside surface of the glass screen. A coating of special phospors on that surface converts the energy carried by the electrons to colored light. In most color picture tube designs, a thin precisely curved piece of sheet metal (called a shadow mask) perforated with hundreds of thousands of tiny holes just behind the glass screen keeps the colors from getting mixed together improperly.

The glass tube itself is assembled from two pieces. The rear piece is shaped like a large funnel. The other piece is the face panel that forms the front of the picture tube, the part we watch. During manufacture, the two pieces are cemented together to form the finished tube.

In 1971 RCA was having difficulties with the shapes of the interiors of their picture tube face panels. The company that was molding the face panels for them argued the face panels were of the correct shape but RCA had a different opinion. So they asked us to produce a servoed CMM system capable of reading some 800 points on a face panel.

We had not done anything in the way of servoing a CMM in response to a soft probe's deflection up to that point. The wish was for the CMM to automatically follow in Z direction (the vertical axis) the shape of the part as sensed by a soft probe while X and Y were driven through the desired targets on the part's surface.

The soft probe that we used was an ElectroJet LVDT with a 0.200 inch total range tipped with a tungsten carbide ball 0.157 inches in diameter. Its electronic package produced an analog voltage proportional to the probe's deflection. It was adjusted so that the midrange point resulted in an output of 0 volts. Deflecting the ball tip upward from that center position produced positive voltages proportional to the amount of deflection. Allowing the ball tip to extend downward from the center position produced negative voltages in a similar way.

Sheffield manufactured the ElectroJet products in our own plant. Although most such probes would be equipped with an internal spring to extend the probe's tip (in this case downward), this ElectroJet was assembled with no spring. Instead, the ElectroJet was equipped with an air line. Air supplied via a regulator set for a very low pressure caused the probe tip to be urged downward with an extremely light force. This resulted in a minimum of part deflection caused by the gaging force as well as a minimum of drag on the part surface as the probe was driven across the countoured part. This became important because RCA wanted the system to not only measure the glass panels but also the thin curved shadow mask sheet metal parts.

When in operation, the part program prompted the operator to perform various manual steps to establish the part's coordinate system. Then it engaged the servos and drove to a starting point just above the part's surface. There it turned on the tracking facility which caused the Z axis to stop accepting its velocity directives from the computer program and, instead, to receive its continuous analog velocity command directly from the probe's analog output. Since the probe was not touching the part at this point, the probe tip was fully extended, the analog output was a large negative voltage, and so the Z axis drove downward.

As it reached the part surface, the probe began to be deflected upward reducing the negative analog voltage. This caused the Z axis servo to begin slowing down because the analog velocity command was becoming less and less negative. As the probe reached its midpoint and its analog output dropped to zero volts, the Z axis servo interpreted this to mean to drive with zero velocity. So it came to a stop with the probe being very near its midpoint.

Meanwhile, the program was using an analog-to-digital converter to take many readings per second of the analog voltage coming from the ElectroJet's electronics. When the readings showed that the ElectroJet's deflection was very close to its midpoint position, the X and Y axis servos were directed to drive the probe at a fairly low speed to a new position such that the tragectory would take the probe through a series of target positions on the part's surface. The program also monitored the X and Y positions as it did this and recorded the CMM's X, Y, and Z axis positions as well as the deflection of the ElectroJet probe at the last sampled position just prior to reaching each programmed target and the first such set of readings again just after passing the programmed target.

When the X and Y axes reached the end point of the intended move, an interpolation problem was solved for each programmed target to compute what the Z position and the probe deflection was just as it passed the programmed target. The deflection was scaled into a dimension and added to the Z axis position. The computed Z measurement value was stored in a list of results and the calculation repeated for each additional programmed target across the part. When the Z values for all the target points of the entire scan were done, an inspection report for that scan was generated comparing the result positions with the nominal Z values for the target points. Tolerances were applied and out-of-tolerance points were flagged to make them stand out. The inspection report was printed and then the process repeated for the next scan across the part.

When RCA received the system they discovered they were right and the glass face panels they were receiving from their supplier had undesirable variations in their contours. When they confronted the glass supplier with their measurement data, the glass supplier asked who sold them a system to measure the panels. The next thing you know, we were building another identical machine for the glass supplier who also wanted it to measure the steel mold plungers that were used to form the countoured interior surface of the face panels.

Other RCA plants ordered more of the Picture Tube Scanner systems and so did their glass suppliers. It spread to other TV set brands and ultimately all over the industry. When the dust had cleared, we had delivered something like 50 of these scanning systems.

One funny little story belongs to the end of this. Today, CMM counters do their counting in binary. Then the system converts the binary numbers to a decimal form if they are to be displayed. Back in the early 1970s, the counters were still counting directly in decimal and the axis displays were driven directly from the counters. The computer was an option.

Eventually, Sheffield created a metric option for the CMM decimal counters because not everyone in the world uses inches. This scaled the count pulses to account for the imperial/metric ratio. It also positioned the decimal point at a different digit on the display. The first such machine ordered with this option was a picture tube scanning machine destined to go to Germany.

I added a suitable option to the scanning software to allow reading the counters with the decimal point in a different position from that when it was counting in inches and to scale the read values to appropriate units for use with the servos and for measurement processing. We thoroughly tested the scanning system with the CMM in inch mode but the hardware crew was having a lot of trouble getting the hardware metric option to work correctly. When they did finally get it to work, the deadline was upon us to ship the machine so it went out the door without having actually been run in metric mode with the scanning software
When the machine was installed in Germany, it was found to work properly when operated in inch mode but the customer wanted to run it in metric mode. When operated in metric mode, the servos oscillated wildly with ever greater cycles until one of the CMM axis limit switches was tripped. This caused the servo system to shut down and the service engineer to call back to Sheffield to find out what to do to fix it.

It was a Thursday. I was directed to immediately have my photo taken and to hand walk an application through for a passport so I could catch a plane to Germany on the following Monday morning. There were plenty of people in the world who would have enjoyed a trip to Germany but I was not one of them.

Friday afternoon, I recalled that while that first metric scanning system had been on its way to Germany, one more scanning system had been produced with the same metric hardware option and shipped to an RCA plant near Marion, Indiana. RCA had no immediate intention of running the system in metric mode and, in fact, did not even receive the software designed to work in conjunction with the CMM counters in metric mode.

I knew one of their CMM operators well. I called him and asked him if he could get me into the plant on Saturday so I could see if I could resolve the issue without taking the plane to Germany. He was glad to help.
My wife had a sister who lived near the Marion plant so we planned a little one-day trip to Marion, Indiana for Saturday.

I drove from Dayton over to Marion on Saturday morning and the operator led me to the machine. When we got there, even before loading the new software, I pressed the button to switch the machine's counters to metric. As soon as I did that, it was immediately clear as to the cause of the whole problem. I could tell by where the decimal point went that the machine was then counting in millimeters. However, my software metric option assumed that it would count in centimeters. This was my first real exposure to metric as used in manufacturing and I assumed (there is that ugly word that always gets me in trouble) the counters would count in centimeters rather than millimeters because centimeters are closer in size to inches than are millimeters.

We loaded the computer with a copy of the new software that was giving trouble in Germany and duplicated the problem. After keying in a patch to correct the problem, the scanning CMM system ran properly in metric.

Monday, I did not report to the airport. Everyone was surprised to see me back at the Sheffield plant. I recounted my adventures on Saturday, punched up a corrected paper tape, and we shipped it off to Germany. The problem was gone and I didn't have to leave the country.

My passport expired years later. I had never used it even once.

There is one more story about one of the picture tube scanner machines. This one was shipped to Corning Glass Works in Corning, New York. The plant was located right along the Chemung River. During a period of extraordinary rain, the river flooded in June of 1972. The water level reached well up into the floor of the building above the one where the picture tube scanner sat. The machine was completely submerged in the river's muddy water.

After the river retreated, the machine was shipped back to our plant. I remember walking out to our shop and seeing the computer's paper tape reader with mud and water still in the pockets where the paper tape unfolds and folds up. There, Sheffield workers stripped the whole thing down basically to the iron and totally rebuilt that machine with a new computer and all new electronics. It was my understanding that insurance paid for the entire rebuild.