Need advice from math gurus: ballscrew>sprocket>chain force



  • I posted this image earlier. It is a way to have a swing arm with sprockets on it that that allow constant tension on the belt. The idlers are on a swivel plate that is held in the center position by tensioner springs. If there is excess tension in either direction a switch gets depressed. This worked out well on the previous high torque worm gear design. It is more quickly tripped versus current sensing on the motor.
    954 x 544 - 189K
  • Here is the concept for the back side idlers acting as stress detection. Springs hold a swivel please in place. The LOAD side will only see a known load unless there is obstruction so the springs can be set to allow the load to not trigger the switch, but any stress on the idler above the load will press the switch. On the slack side, there should never be a load on that side other than belt tension.

    I have looked for a while but cannot find a formula for a force acting on the idler pulley to get some estimates for the load on the springs normally and under stress.
  • Debbie Downer chiming in again:

    I can't help but imagine all kinds of nuisance trips with this mechanical sensing method. It might be just as cost effective to employ a shaft load sensor, commonly used for web tension control (?).
  • T ChapT Chap Posts: 3,563
    edited May 4 Vote Up0Vote Down
    Thanks for the suggestions, I will look at the shaft load sensors but at quick glance they are crazy expensive, more than I would spend on this. As for the swivel plate with two idlers, I have the chain version working for years with no problems using switches only and several springs to set the force. A smaller spring sets the normal running load tension for the swivel and two heavier springs kick in on a overstress motion which add more restriction at father travels. On the new design, a heavy spring will sit between a fixed plate and the load cell. If the spring compresses beyond normal running loads, the snap switches will get pressed as a fail safe in case the loads ever fail. The switches are only a back up in case of load cell failure and a real jam. With the load cells, as the window goes up, there will be a known average load(once it is in motion) so I can mask the initial acceleration phase. If during lift, the load goes above a user defined upper limit OR drops below a user defined lower limit, the system can shut down or retract. If during the lowering of the window the load exceeds a max and min load, the system will retract as it may first assume the window has hit something going down. I will mechanically calibrate the load on the load cells to fit nicely well under the max rating. The idlers will have several options for mounting to set the deflection needed to use a proper idler tension force based on static belt tension, working load on the belt, span etc. For example 1 inch deflection can be a given idler tension force while lifting the window. In this new load cell > spring > switch scheme, the swivel plate will hardly be able to move in any direction more than +/- .25". In practice, only one load cell should ever see a load. I only used the load cell on the non working side of the belt for redundancy with the switch on that side. This gives two methods of sensing for the working side of the belt and two for the slack side.

    I will get in the sparkfun load cells and HX711 boards today and build a special enclosure to house those with and LCD display to see and set all the parameters of the 2 load cells. The LCD will show upper and lower load limits for OPEN, upper and lower limits for CLOSE, options for defining the average load sample size and also method to use mean load values and discard transients.

    BTW with the original spring/switch concept, you never just read the first switch event, you so something like:
    If Switch1 == 1
      waitcnt(1_000_000 + cnt)
      If Switch1 == 1
         waitcnt(1_000_000 + cnt)
         If Switch1 == 1
            waitcnt(1_000_000 + cnt)

    This way you can toss out transients after analyzing what a transient might look like, then set a sampling method to mask it.
  • The suggestions for the load cell and HX711 combo worked out incredibly well. I bought the 50Kg load cell and HX711 board from spark fun. There was an object posted in obex. I wish I had of experimented with these load cells years ago. Thanks to whoever put that object in there, I used the read raw and with minor changes to show it on the LCD I am very pleased. Too bad nobody carries this HX711 but I intend to try to locate some and build a board for it.
    723 x 551 - 98K
    453 x 467 - 65K
  • Nice job! Sounds like you're having fun.
Sign In or Register to comment.