tachometer accuracy

[wangp1283]

How accurate can an tachometer be? in %

In addition, I need to measure the linear velocity of a linear movement. The maximum velocity is 0.1 m/s.

What is the best way to get a accurate measurement of the linear velocity as a function of time? (and not too expensive).

Is it possible to get an accuracy within 2%?

Thanks.

 

[MikeHalloran]

1. How accurate do you need it to be? How much money have you got? It seems to me that one limit is display resolution. For an analog tach, that means the width of the needle, and the developed length of the scale over which it travels, and the acuity of your eyeball. For a digital tach, it means plus or minus one count of the least significant 'real' digit. By 'real' digit, I mean that a cheap digital tach could be made to read '00' in the last two digits all the time.

1a. You also need to decide what sampling interval, or resolution in time, is appropriate for you. E.g., you can count revolutions for a minute, and get an accurate number, but it won't tell you the instantaneous angular velocity of the crank between firing events.

2. You can measure displacement and time with extreme accuracy these days. So if you could measure displacement every second, or the time it takes for your subject to travel 0.1m, you could have excellent accuracy.

2b. Same as 1a. If you want instantaneous velocity, it costs more.





Mike Halloran
NOT speaking for
DeAngelo Marine Exhaust Inc.
Ft. Lauderdale, FL, USA

 

[Warpspeed]

Agree with Mike, the linear pickups used for digital calipers and CNC machinery can read position to much better than 0.001 inch over considerable length. With suitable electronics, position, time, and velocity could be measured to incredibly fine limits. Five digit resolution should be fairly easy.

 

[wangp1283]

Well, let me ask in another way.

There is no problem with "width of the needle" and the "acuity of my eye" because the result it not read by a human but rather a computer. A computer uses this information to control a system.

I realize there are issues with sampling intervals, cost. ect...

But I just want to know if I could get a tachometer so that the measured rpm at any instant is no more than 2% off than the real actual rpm, reasonably cheap, so that it could be used on a car.

 

[Warpspeed]

That should be easy. I recently built a digital tachometer for a dynamometer project that measures time interval between the teeth of a rotating optical chopper disc. The microprocessor measures the time interval to 500nS resolution. The quartz frequency reference for the microprocessor is probably accurate to at least three digits without any adjustment or calibration, and so should the Rpm readout. Even 0.1% accuracy should be quite easy to achieve.

If you try a bit harder 1Rpm resolution and accuracy in say 9,000Rpm would be feasible. Digital electronics and microprocessors are wonderful things for counting, and if you add or divide two numbers, or count events, the accuracy will be absolute. Digital readouts are very good too for exactly the same reasons.

The only possible source of error or uncertainty is the time measurement. Even the cheapest quartz clocks and watches are accurate within a few seconds per day these days. Percentage time error in that case being quite low.

You can make it as accurate as you want. Revs per minute is easy. How about revs per hour or revs per month ?

 

[magnograil]

Wangp1283 - You also have to consider error from acceleration. If you measure pulses over a time period, you only know quantity and not the rate of change. If you measure time between pulses you only know the average time. If your computer is reading several inputs then the actual RPM can be very different by the time the last sensor is read compared to when the RPM average was read.

 

[Warpspeed]

In this case there is a 36 tooth slotted disc, measured speed is fully updated every ten degrees of shaft rotation. Even under some pretty extreme acceleration, the speed indication will follow on fairly closely, and it reads to one tenth of an Rpm resolution every ten shaft degrees.

The trick is to use a microcontroller with the input capture feature. It logs the arrival time of each incoming pulse edge in hardware. The processor can then read this sixteen bit counter at any time afterwards, (provided it is read before the next pulse arrives).

From there you can log the instantaneous speed, average, calculate acceleration, or do whatever else you want.

 

[MikeHalloran]

You can get even better resolution by sensing the ~180 teeth on the flywheel's starter gear. With that kind of angular resolution, you can measure the angular acceleration associated with each firing event. And that means that the instantaneous 'real' rpm is not single- valued.

When you decide how to decide what the 'real' rpm is, yes, you can measure it very accurately for not a lot of money. You just have to decide what you want.



Mike Halloran
NOT speaking for
DeAngelo Marine Exhaust Inc.
Ft. Lauderdale, FL, USA

 

[Rob45]

Mike:
In fact, I use the (113) teeth of large Diesel truck engines to measure not only RPM, but also flywheel angular acceleration as a measure of the impulse that produces driveline torsional vibrations.

And of course, position, speed, and acceleration can be measured measured by radar.

 

[GregLocock]

And while we're off on a tangent, modern ECUs look at the jitter in the rpm signal to detect misfires.

You should also bear in mind that your 'accuracy' is limited by the laws of signal processing - specifically the shorter your sample time, the less accurate your resolution.

max Resolution = 1/sample time, so if you sample for 1 second the best you can manage is to within 60 rpm, anything less has no meaning.





Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Warpspeed]

True, but in practice the problem is not electronic. Nanosecond time resolution is possible and feasible in the digital electronics.

The problem is always in picking up a clean signal from the rotating part to begin with. Optical and magnetic pickups can have a lot of jitter for all sorts of reasons, and that is what will limit you in the end.

Interesting about the misfires Greg. What does the ECU do about a detected misfire ?

 

[GregLocock]

Dunno, initially it just logs it I think. If it was persistent you'd get a check engine light - a dud spark plug is a quick way to kill the cat, as the excess oxygen from that cylinder will be continually detected, so the mixture will be richened. The California emissions rules have forced a lot of cross checking between sensors, that's a good example.







Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.