Fail-safe shaft rotation sensor

[hevans1944]

I have a shaft-mounted cast iron pulley with six spokes that is motor-driven with a v-belt. A Cherry GS100701 gear rotation sensor with an internal permanent magnet is mounted in a position to sense the passage of each iron spoke past the sensor head when the pulley is turning.

I need a circuit that will use the presence of the sensor pulses to energize a fail-safe relay when the shaft is rotating at normal speed. If the shaft stops or slows down from its nominal speed, the fail-safe relay must de-energize within one shaft rotation (six pulses). If normal shaft rotation resumes, the fail-safe relay must re-energize. All circuitry should operate from a single DC power supply in the range of 5V to 15V. The shaft rotates at a nominal 550 RPM.

My first thought is to use a 555 timer but I don't have one handy to "play" with. The negative going edge of each pulse would reset the timer while the following positive going edge would trigger it. If the pulses occur fast enough, the 555 would not "time out" before the next pulse reset the timer. Unfortunately, if the timer is reset, its output goes from high to low, which is the same thing that occurs if it times out.

If the pulses slowed down or stopped, the 555 would time out and de-energize the fail-safe relay by clocking, on the negative going edge of the 555 output, a D-type flip-flop whose D-input is connected to the Q-not output the D-type flip-flop. The low transition of each sensor pulse resets the D-type flip flop with each pulse. The reset state of the D-type flip-flop energizes the fail-safe relay.

Hopefully, the reset condition overrides the clock on the D-type flip-flop allowing the timer to be re-started on the positive going edge of the sensor pulse without setting the D-type flip-flop when the 555 output goes low because of the reset.

Note that the sensor pulses must be AC coupled because the shaft can stop with with a spoke under the sensor head, which would hold the D-type flip-flop in the reset state and keep the fail-safe relay energized.

I realize there is a possible race condition if the 555 times out just before another pulse occurs while the shaft is slowing down. This would cause the D-type flip-flop to be set by the 555 and then immediately to be reset by the following pulse. Worse, the condition could repeat until the shaft speed slowed sufficiently because this condition would re-start the timer. The only solution I can think of for this race problem is a second timer, triggered by the time-out of the first timer, that would inhibit the reset input of the D-type flip-flop.

An alternative, and possibly simpler, circuit would just differentiate and integrate with an RC network the positive-going edges of the sensor pulses. After a sufficient number of pulses (whose pulse width and repetition rate vary with shaft speed) are accumulated the integrated output would energize the fail-safe relay. The integrator would have to be "lossy" so it would decay in output between pulses. Only if the pulses occurred fast enough would the integrator output rise enough to energize the fail-safe relay.

And finally, a third engineering overkill solution would just apply the sensor pulses to a microcontroller and let the software figure everything out.

Does anyone have any ideas on how to solve my fail-safe shaft speed rotation sensor problem? I need to bread-board a solution on Monday, September 29, 2014.

 

[Harald Kapp]

Uisng a 555 is basically not a bad idea, although a microcontroller might be more precise.

Assuming tha you're only concerned with the rotational speed being too low (not too high), you'd operate the 555 as retriggerable monostable multivibrator (Google). Each passing spoke triggers the 555, therefore as long as the spokes pass in sufficiently rapid succession, the output of the monoflop stays active. If the wheel turns too slow, the monoflop will reset itself between triggers (the difference to your idea here is that you do not use pulses fromn the sensor to reset the monoflop, you just let it time out). On reset the monoflop's output becomes inactive. You can use this transistion from active to inactive for example to set a flipflop. The flipflop will stay set until you reset it once the wheel turn at nominal speed.

Your considerations lack a mechanism for disabling the supervisor circuit during the time the wheel starts up. During the time the wheel has not yet reached its final speed, the supervisor will see too low a speed and trigger the failsafe mechanism. How are you going to handle this?

A microcontroller solution could fulfill both purposes in one device by e.g. applying a dead band (waiting time) during start-up of the wheel during which it will ignore too low a speed.

 

[hevans1944]

Thank you, Harold Kapp, for pointing out the obvious. Retriggering, NOT resetting, the monostable is the obvious solution. No need for an additional flip-flop. The 555 output can directly drive the fail-safe relay coil, with the same race caveat that I noted in my original post. I will try to see if I can live with that.

Start-up considerations are unimportant. The fail-safe relay must de-energize for at least a few milliseconds to allow an interlock system to trip, but the interlock system is independent of this circuit. After the shaft reaches normal rotation speed the fail-safe relay will remain energized and the operator can then manually reset the interlock system.

Background information:

This circuit is for Welch roughing (foreline) pumps that are connected to Varian VHS-6 oil diffusion pumps that evacuate the columns of a small 1.7MV tandem particle accelerator (IonX Tandetron). The original installation used turbomolecular pumps, but when both failed sometime in the 1960s (before I arrived in 1996) Varian diffusion pumps replaced them because of cost (not engineering) considerations.

I have two model 1374 Welch vacuum pumps with v-belt drives. Each of these pumps is used in the fore line of two Varian VHS-6 oil diffusion pumps. Neither Welch pump has anti-suck back prevention. When the pump stops, the inlet port and the fore line to the diffusion pump quickly rises toward atmospheric pressure. The effect of this is to force hot diffusion pump oil vapor out of the diffusion pump inlet and into the vacuum system. Clean up and recovery to a high vacuum condition is difficult and expensive.

A proper solution would be to replace the Welch/Varian pumps with high-flow turbomolecular pumps backed by large direct-drive rotary pumps with anti-suck back valves. Cost still prevents this solution.

Last week I retro-fitted each pump with a Cherry GS100701 magnetic Hall-effect gear rotation sensor mounted behind the spokes on the motor-driven pump shaft. This sensor has a built-in magnet for the Hall-effect sensor. As each cast-iron spoke rotates past its sensor, a negative-going pulse occurs on the open-collector output of the sensor. The open-collector output is connected to a 15 VDC power supply through a current limiting 2200 ohm resistor. The amplitude of the output is +15 V when a spoke is not present under the sensor. The output transitions to the saturation voltage of the open-collector output (near common potential) when a spoke is present under the sensor.

When the pump is operating normally, the sensor output is a series of square, negative-going, pulses whose width and repetition frequency are a function of pump shaft speed. I need to use these pulses to indicate that the pump shaft is rotating and de-energize a fail-safe relay when it is not rotating. When rotation is resumed, the fail-safe relay must re-energize without operator intervention.

When the pump shaft ceases to rotate (power failure to the motor, v-belt broken, or whatever other reason) the absence of the pulses must cause an energized fail-safe relay to de-energize and thereby open a form A normally-open contacts on the relay. The open contacts on the now de-energized relay are wired in series between the two pumps and a fail-safe interlock circuit that causes power to be removed from all electro-pneumatically controlled high-vacuum valves in the vacuum system, thus quickly closing the valves above the diffusion pumps. Heater power to the diffusion pumps is also removed when the fail-safe interlock circuit is “tripped”. Operator intervention is required to resume normal operation after a fail-safe interlock trip.

 

[hevans1944]

Oops again. The 555 is not capable of re-triggering to extend the timing interval unless additional circuitry is added to discharge the timing capacitor without affecting the state of the internal flip-flop. I am looking into this and setting up a breadboard trial. However, a Texas Instruments SN74LS121N is re-triggerable and still available. I just don't happen to have one handy to try out yet.

 

[Harald Kapp]

Right, I didn't think of the re-trigger issue.

This article shows a modified 555 monoflop with re-trigger capability (figure 11). It requires a positive trigger impulse. You could invert your sensor sinal or you can just trigger on the rising edge of the sensor signal - the frequency is the same as for the falling edge.

 

[hevans1944]

Another "oops". The SN74LS121N is NOT re-triggerable. The dual monostables SN74LS122N and SN74LS123N are both active and re-triggerable, so I should be able to purchase either on-line from a distributor.

After lots of surfing on the Internet the past few day I discovered this is an old problem, solved years ago by simply discharging the timing capacitor with the trigger signal thus preventing a 555 configured as a monostable from timing out until the trigger pulses go away. A simplie circuit I found is attached. I'll breadboard this tomorrow.

 

[KrisBlueNZ]

You can simplify that circuit. All you need is a capacitor that's charged up through a resistor, with a threshold detector connected to it, and a transistor (or similar) that fully discharges the capacitor on every trigger pulse.

You can use a 555 as the threshold detector.

The important thing is to ensure that the capacitor is fully discharged on each trigger pulse. You can do this using a second monostable, and it can be a non-retriggerable type, made from another 555 (or half of a 556), which generates a fixed length pulse on every trigger event. This pulse drives the transistor that discharges the capacitor. It needs to be long enough to ensure that the capacitor is fully discharged, but it can be a lot shorter than the time between trigger events.

If this is not clear, I can draw up a schematic.