Ben Heck’s Mechanical Television Part 2 Episode -- Episode 233

I vote you revisit this project.The lessons learned are invaluable because understanding the dynamics of the issues you encountered will help designers of servo-mechanisms appreciate the problems they will be faced with. It is also a good demonstration of several failure modes in modern machines. My response provides purely mechanical solutions as that seems to be the intent of building a "mechanical TV". As I see it there are three issues with the design. 1) servo control of the two discs, 2) rigidity of the base and 3) propagation delay from the "camera" to the "projector".

You are using a toothed belt which is a good idea because it eliminates slip but there is a lot of flex in the belt and with the holes in the record being so small that flexure will add a lot of speed distortion - enough that the strobed light in the projector may not be turned on at the right time. I work for a company that makes vibration spectrum analyzers and loose belts add a lot of vibration to machines. Think of a guitar that has the strings attached to fixed locations on the guitar. Loosening and tightening a string is equivalent to the belt flexing. Note the difference in pitch of the string as it is loosened and tightened. If the pitch frequency is halved by loosening then the string is moving twice the distance. Even with a toothed belt that flexing considerably changes the speed of the load. When the belt is at maximum tension (arced towards the center) the load is slowed down and when the belt is at minimum tension (arced away from the center) the speed is increased. In the old audio tape days this was called wow and flutter. The conventional solution in tape and record systems was to add a capstan which keeps constant tension on the tape using springs as the servo-mechanism.

The capstan may not be good enough because recording and playback occur at different times on a tape player but in this experiment you are capturing and transmitting light pulses synchronously in real time. Another approach would be to have a larger gear attached to the motor which is mounted between the two discs. This gear has to have the same gear pitch as the gears attached to the records. It will slow the turning speed slightly but the motor gear is acting as a differential and both sides are affected equally and the gear reduction should also add a little more stability as slight changes in motor speed will be reduced. By using the differential gear, the tensioned portion of the belt is the same distance to both records and there will be less distortion because the tensioning sections of the belt (behind the direction the belt is traveling at the motor's gear) always have fairly proportional tension. The driving end of the belt (being pushed by the motor) just flexes and no longer affects the system due to the differential. With this approach you may have to add a slight friction braking to both records and adjust one of the brakes until you have good synchronization thereby maintaining constant velocity on both records. Imbalance and warping in the records will also cause distortion as imbalance cause velocity changes in the discs. The differential and brake methods should resolve most of the problems caused by imbalance and warping as long as the base is reliable. No matter what you do, the two disks have to be held rigidly in place and the tension on the belt must be constant. In the video it looked like foam board was being used and that alone could have caused the system to fail.

The last thing to consider is delay through the detection and light driving sources. I would measure the delay and then add a little lag in the playback record by either changing the angle of the mounting holes or elongating the holes on the playback side. I would opt for the slotted holes because the imperfections in the driving system are probably more a factor in distortion than the elliptical dashes caused in the playback. The holes on the outer diameter of the record are traveling at much higher velocity than the inner holes and consequently are going to be more affected by delays. You might have to do both, add a slight phase shift by adjusting the mounting holes on the discs and elongating the holes.

As a side note, another approach to solving the MOSFETs burning up would be to use PWM to drive the DC motors. The motors windings, being inductive, will have linear currents but the power dissipated by the MOSFETS would be more dependant on the transition time. I would put good Schottky diodes across the MOSFETS and not rely on the internal diodes to shunt the flyback voltage when the transistor turns off. The power supply will need low inductance paths and good bypassing to keep the flyback current from driving the supply regulator out of regulation by reverse biasing it. Even with PWM you should look at understand the power dissipated by the MOSFET. Otherwise it is best to use a proven driver circuit because a MOSFET failure can be catastrophic to your system by introducing high voltage, high energy pulses into your low voltage circuits! I've seen boards destroyed beyond repair with nearly every semiconductor destroyed and heavy traces (0.125 inches wide) burned to carbon. Your decision to go to the transistors used in the drill was a good one. Since you are using them in a linear application I would suggest putting 10K 1/4 W resistors in series with the gates and put protection diodes on the output of the opamp driving the gate. The 10K resistor limits the current to safe levels the delay caused by gate capacitance won't be significant with the frequencies the DC motor uses. If you use PWM then the resistor should be reduced to 100 Ohms or the switching time of the MOSFET can cause excessive heating,especially in an H Bridge implementation. The 100 Ohm resistor minimizes the effect of capacitive loading due to the gate capacitance of the MOSFET and will act like a fuse during catastrophic failure of the MOSFET but the damage to the rest of the system should be minimal.

In the Name of Purism

The approaches I've presented are mechanical solutions that would have been available in the 1920s and don't add electronics to the project as presented. A slightly more purist approach would be to use incandescent lights instead of LED which would have been around in the 20s. You would probably have to bias the lights all the time to keep them warm so their response is regular and therefore predictable. By using incandescent solutions the lag induced issues become quite apparent. A purely mechanical approach to this project would use a hand crank instead of the electric motors and and would use candles instead of lights altogether. Then the challenge is how do you modulate the light? Perhaps a slotted disc turning at slightly lower speed than the record to introduce a slight delay in the light which causes the horizontal scans. But there are enough challenges in the project just as it is.

I vote you revisit this project.The lessons learned are invaluable because understanding the dynamics of the issues you encountered will help designers of servo-mechanisms appreciate the problems they will be faced with. It is also a good demonstration of several failure modes in modern machines. My response provides purely mechanical solutions as that seems to be the intent of building a "mechanical TV". As I see it there are three issues with the design. 1) servo control of the two discs, 2) rigidity of the base and 3) propagation delay from the "camera" to the "projector".

You are using a toothed belt which is a good idea because it eliminates slip but there is a lot of flex in the belt and with the holes in the record being so small that flexure will add a lot of speed distortion - enough that the strobed light in the projector may not be turned on at the right time. I work for a company that makes vibration spectrum analyzers and loose belts add a lot of vibration to machines. Think of a guitar that has the strings attached to fixed locations on the guitar. Loosening and tightening a string is equivalent to the belt flexing. Note the difference in pitch of the string as it is loosened and tightened. If the pitch frequency is halved by loosening then the string is moving twice the distance. Even with a toothed belt that flexing considerably changes the speed of the load. When the belt is at maximum tension (arced towards the center) the load is slowed down and when the belt is at minimum tension (arced away from the center) the speed is increased. In the old audio tape days this was called wow and flutter. The conventional solution in tape and record systems was to add a capstan which keeps constant tension on the tape using springs as the servo-mechanism.

The capstan may not be good enough because recording and playback occur at different times on a tape player but in this experiment you are capturing and transmitting light pulses synchronously in real time. Another approach would be to have a larger gear attached to the motor which is mounted between the two discs. This gear has to have the same gear pitch as the gears attached to the records. It will slow the turning speed slightly but the motor gear is acting as a differential and both sides are affected equally and the gear reduction should also add a little more stability as slight changes in motor speed will be reduced. By using the differential gear, the tensioned portion of the belt is the same distance to both records and there will be less distortion because the tensioning sections of the belt (behind the direction the belt is traveling at the motor's gear) always have fairly proportional tension. The driving end of the belt (being pushed by the motor) just flexes and no longer affects the system due to the differential. With this approach you may have to add a slight friction braking to both records and adjust one of the brakes until you have good synchronization thereby maintaining constant velocity on both records. Imbalance and warping in the records will also cause distortion as imbalance cause velocity changes in the discs. The differential and brake methods should resolve most of the problems caused by imbalance and warping as long as the base is reliable. No matter what you do, the two disks have to be held rigidly in place and the tension on the belt must be constant. In the video it looked like foam board was being used and that alone could have caused the system to fail.

The last thing to consider is delay through the detection and light driving sources. I would measure the delay and then add a little lag in the playback record by either changing the angle of the mounting holes or elongating the holes on the playback side. I would opt for the slotted holes because the imperfections in the driving system are probably more a factor in distortion than the elliptical dashes caused in the playback. The holes on the outer diameter of the record are traveling at much higher velocity than the inner holes and consequently are going to be more affected by delays. You might have to do both, add a slight phase shift by adjusting the mounting holes on the discs and elongating the holes.

As a side note, another approach to solving the MOSFETs burning up would be to use PWM to drive the DC motors. The motors windings, being inductive, will have linear currents but the power dissipated by the MOSFETS would be more dependant on the transition time. I would put good Schottky diodes across the MOSFETS and not rely on the internal diodes to shunt the flyback voltage when the transistor turns off. The power supply will need low inductance paths and good bypassing to keep the flyback current from driving the supply regulator out of regulation by reverse biasing it. Even with PWM you should look at understand the power dissipated by the MOSFET. Otherwise it is best to use a proven driver circuit because a MOSFET failure can be catastrophic to your system by introducing high voltage, high energy pulses into your low voltage circuits! I've seen boards destroyed beyond repair with nearly every semiconductor destroyed and heavy traces (0.125 inches wide) burned to carbon. Your decision to go to the transistors used in the drill was a good one. Since you are using them in a linear application I would suggest putting 10K 1/4 W resistors in series with the gates and put protection diodes on the output of the opamp driving the gate. The 10K resistor limits the current to safe levels the delay caused by gate capacitance won't be significant with the frequencies the DC motor uses. If you use PWM then the resistor should be reduced to 100 Ohms or the switching time of the MOSFET can cause excessive heating,especially in an H Bridge implementation. The 100 Ohm resistor minimizes the effect of capacitive loading due to the gate capacitance of the MOSFET and will act like a fuse during catastrophic failure of the MOSFET but the damage to the rest of the system should be minimal.

In the Name of Purism

The approaches I've presented are mechanical solutions that would have been available in the 1920s and don't add electronics to the project as presented. A slightly more purist approach would be to use incandescent lights instead of LED which would have been around in the 20s. You would probably have to bias the lights all the time to keep them warm so their response is regular and therefore predictable. By using incandescent solutions the lag induced issues become quite apparent. A purely mechanical approach to this project would use a hand crank instead of the electric motors and and would use candles instead of lights altogether. Then the challenge is how do you modulate the light? Perhaps a slotted disc turning at slightly lower speed than the record to introduce a slight delay in the light which causes the horizontal scans. But there are enough challenges in the project just as it is.