I am looking at the specs for a system that monitors the three-phase AC voltage onboard an aircraft. It says,
I have never seen a tolerance stated like that. What does 90 +15/−0 volts mean? Is it exactly the same as 90 V to 105 V (97.5 ±7.5 volts)? Or does it mean the “probability curve” is more skewed toward 90 V, and tapers off at 105 V?
That’s weird.
Since the point is to determine where the shutdown threshold is, it would seem to be clearer to say “shutdown below 105 volts.”
Having the tolerance stated like that doesn’t make any sense to me.
Usually, yes, that’s what an asymmetric tolerance range means. It implies that 90 is the nominal value, and the actual value is allowed to be higher by up to 15, but is not allowed to be any less.
In this context, I believe it means the shutdown threshold is guaranteed to be above 90V - i.e. if the voltage drops to 89.9V, this device is guaranteed to respond. The actual threshold may be higher, but not by more than 15V - i.e. you need to allow for the possibility that 104V would trip it, but it’s guaranteed not to respond to 105V.
p.s. The reason it’s stated as 90 +15/-0, and not 97.5+/-7.5, is because 90V is the threshold this device is designed to respond to. Since it’s a safety device, it’s unacceptable for the actual threshold to be lower than the rated number. It’s supposed to shut down when the voltage gets down to 90V. If it shuts down a little earlier - e.g. when the voltage drops, but it’s still at 104V - that’s considered acceptable tolerance for this device. It still does its job.
Asymmetric tolerances are not uncommon. A hole diameter may have a +1mm/-0mm tolerance if it’s OK for the hole to be a bit bigger than the nominal value, but if it’s any smaller, whatever fits through the hole won’t fit. A car speedometer may have a tolerance of +10%/-0%, because it’s allowed to read higher, but never lower. But if you don’t actually specify for it to read 5% high and allow for +/- 5% tolerance. You design it to read as accurately as possible, but to err on the positive side as needed to make sure it never reads low.
In mechanical design, asymmetric tolerances are common owing to the physcial reality of manufacture (a drilled or milled hole will never be smaller than the bit size, but can be larger) or to assure positive clearance in a multiple part tolerance stackup. This is not statistical; it assumes that all parts are within the stared limits and anything outside would be rejected by screening proecesses, and that changes in dimension will not occur during the product lifetime such that it will fall out of tolerance.
There are analysis methods for performing root sum of square (RSS) which assumes values are distributed in either a single tail or symmetric normal (Gaussian) distribution, and second order tolerance analysis which produces a unique distribution and statistics based upon sample metrics, but those go beyond the simple dimensional limit tolerances you see on drawings.
For an functional tolerance such as voltage, I would assume those are ansolute limits, and 90 +15/0 volts essentiallmy means, “Maintain the voltage as close to 90 V as possible without going below, and less than 105 V, or else the generator will shut down in 2 to 4 seconds.”
Stranger
I believe this not a tolerance issue but a lost in translation issue. I have often had this kind of issue when a process engineer specifies SOP (Standard Operating Parameters), Trips and Ser points but the Controls Engineer programs them differently.
Let’s say the Electrical Engineer (the power guy) looks at generator protection or wiring protection. The Nominal voltage is 115V and the system is tolerant up to 105V. A voltage below 105V indicates a fault which overheats the wires or the stator conductors, if left running for extended times. So the Electrical Engineer makes a chart or a table for seconds to trip for voltages from 105V to 90V. Say it looks something like this :
90 V …2s
92 V …4s
94 V … 8s
…
…
105V…60s
The Control Engineer decides, this is too complicated to program AND gets no benefit to the pilot or airplane. So she picks the lowest allowable time (2s) and trips for all voltages below 105 V.
Someone else writes the operating manual and doesn’t know the history and writes it the way you describe.
I strongly suspect this is what happened.
Coming from the calibration metrology world, as I’d guess you’re pretty familiar with, I’d raise an eyebrow at that spec also. To concur with what has been mentioned earlier, my guess would be it is referring to a nominal 90 - 105 volts as a limit of 0 would make no sense as a measured voltage.
Thanks for the education on asymmetric tolerances. I had heard of it before on dimensional tings (as explained by **SOAT **and others), but had never seen it for electrical stuff.
I am guessing the probability distribution looks something like this, with 90 V being at the left end of the X axis.
