1756-L72 - Pulse train Question - Missing pulses

[EICS]

I would like a sanity check on an application i am working on.

I have a proximity switch counting pulses with a Mark/space ratio of 1:1 on a rotating shaft. 1 pulse per revolution.

the prox is fed into a 1734-IB8 point I/O with an RPI set for 20ms. the CPU is a 1756-L72S with Safety Partner

I am loosing pulses when I go above approx 290rpm (290 PPM) so that is about 5hz. the filter on/off time setting on the module is at default (1m/s from memory)

maximum speed of shaft I am expecting is about 340RPM ( or less than 6 Hz) so I am close to what I require.


I am using a counter to get the pulse rate every 15 seconds then dumping that value into a register then resetting counter. and multiplying the register value by 4 for a rate/minute. the application is to compare head roller speed of a conveyor to the tail roller speed to check for slippage hence the 15 second sampling, I need to alarm & shut-down to save mechanical issues on belt if slippage or a jam up occurs.


should I be looking at a periodic task specifically for this subroutine or any other things i should be aware of?

 

[drbitboy]

the combined pulse falling and rising edges should be coming in no faster than 680edge/minute, which is around 85-90ms per edge at 340RPM, so 20ms RPI seems like it should be good enough. If it is not, I would work to solve that problem first.

If it is not solvable, then what speed resolution is required? A hardware circuit to divide the number of pulses by 2 would double the time between edges, but also double the aliasing of the measurement from around 1%+ (85 pulses per count at 340RPM) to 2%+, and more at slower speeds.

 

[EICS]

Many Thanks i will speed up the RPI and see what effect it has at first opportunity

the combined pulse falling and rising edges should be coming in no faster than 680edge/minute, which is around 85-90ms per edge at 340RPM, so 20ms RPI seems like it should be good enough. If it is not, I would work to solve that problem first.

If it is not solvable, then what speed resolution is required? A hardware circuit to divide the number of pulses by 2 would double the time between edges, but also double the aliasing of the measurement from around 1%+ (85 pulses per count at 340RPM) to 2%+, and more at slower speeds.