Sensing An Extremely Long Distance With Precision

[shearstrength]

Anyone could recommend an economical way to sense a steel plate at a distance of 16 feet with precision to 2 decimal places?

I've found Ultrasonic Sensors that can sense the distance, not the precision but costs a fortune.

Any response will be greatly appreciated!

 

[IRstuff]

Laser rangefinders can easily achieve better than 1 mm at that range. The cheapest with that level of performance run about $4000

TTFN

 

[ScottyUK]

2 decimal places - but what is the basic unit? Metres? Feet? Microns?

I agree with IRstuff that laser instrument is technically the best solution, but you pay for the accuracy and contactless measurement. The better microwave sensors are just about able to resolve down to 0.01m, but again these are not cheap.

How big is the target?

What is your budget?



------------------------------

If we learn from our mistakes,
I'm getting a great education!

 

[nbucska]

what is the min/max. distance? Can you use other type of
sensor?

<nbucska@pcperipherals DOT com> subj: eng-tips

 

[shearstrength]

Answers to the questions above;

(1) The units are to be ±.10" (approx. 1/8")
(2) The target will be as big as the sensor will need it.
(3) Budget will depend on the customer ($4000 too expensive!)
(4) Distances? Min.=4 FT, Max.=15-16 FT.

 

[nbucska]

Evidently you are not interested in the steel plate -- it
is only mounted on the part you want to locate.

Could you describe the actual problem in some details?

<nbucska@pcperipherals DOT com> subj: eng-tips

 

[logbook]

One approach would be to reflect light off the object and measure the delay. We are talking about an array of high intensity LEDs as a source. Pulse them at a duty cycle of say 0.01 to get higher brightness without burning them out. Use a repetition frequency of say 10kHz. The round trip distance is between 2.5m and 10m giving delays up to 33ns. You will definitely need to have a ‘zero’ adjustment done at the minimum distance to take out errors in the system. You should be able to measure a delay of between 8ns and 33ns with enough resolution. You could use a emitter coupled pair of bipolar transistors to steer current into a sample and hold capacitor for example. You haven’t told us enough to interface this signal to your system. Maybe another sample and hold into a moving coil meter or an ADC into a computer system.

 

[IRstuff]

The problem is the the time range, but the required accuracy. 1/8" round trip requires 21 ps resolution, and roughly 5 ps accuracy for your primary timing standard.

Even with good instrumentation, this is a non-trivial time interval to measure.

TTFN

 

[logbook]

Good point, IRStuff. I was thinking in terms of the earlier statement of "two decimal places" , which I interpreted as a per-unit figure. In other words 1% resolution. I am not particularly bothered by the 21ps resolution (I just calculated 16ps resolution), it is this figure in the overall 33ns, requiring one part in 2000 resolution. I doubt that a simple circuit would have that kind of linearity, especially when done by a mechanical engineer!

We are talking about a 12 bit ADC or better. Perhaps a 16 bit ADC and careful calibration to linearise the response. In any case this all seems more than would be desirable.

 

[IRstuff]

I used the 1/8".

AND, that's assuming that you've ensured that the current source has been adequately corrected for linearity and temperature effects.

I've been involved in the development of this type of ranngefinder, and there are additional issues when your pulsewidth becomes a significant portion of the resolution, e.g., slope and amplitude of the return will also affect the error. You'd also need to use a constant fraction discriminator to ensure a consistently threshold point for the return processor.

Very likely, the resolution is low enough that there will be noise issues as well.

TTFN

 

[logbook]

Looks pretty hard then.

My next thought was a "static" scheme with a laser, or at least a well collimated powerful modulated light source. If the laser and the receiver were 1m apart, and the laser were mounted on a stepper motor controlled positioning system, the angle of the laser could be adjusted to bounce the beam off the target and back to the sensor. The angular measurement would need to be very tight and there would be some trigonometry involved to compute the distance. The problem then is that if the target did not stay parallel to the measurement plane as the target was moved, the angle required would be all wrong.

Ultrasonics may be easier in terms of the velocity being lower by a factor of 1,000,000. The problem then becomes one of resolution. There would be insufficient resolution in a time domain pulse modulation system. Maybe a time domain pulse could establish the rough position and then a continuous sinusoidal beam could be used, the phase shift determining the position more accurately.

I guess the point is, one is having to develop a measurement system which is more accurate than commercially available equipment. In this case the development timescale and budget could be excessive.

This probably means the original specification is unreasonably tight and/or the customer’s expectation in terms of cost/performance cannot be met.

 

[IRstuff]

I thought that the laser rangefinder that I bought was eminently reasonable at $4000:

http://www.sickcommerce.com/master/compare.asp?FAM=Measure&ITEMLIST=26639,26641,26643&P=4841

Note that the resolution is 0.125 mm, which is 0.5 mils; my memory was off a bit

Even at $40/hr of time, that's only 100 hours. You can barely begin the design with only 100 hours.

TTFN

 

[ScottyUK]

Hi IRstuff,

0.125mm is 5 mils isn't it? America's 1 mil is 1 thou to the English, and 25.4um = 1 thou.

Interesting exchange of ideas from you and logbook, although I think the OP wants an off-the-shelf solution. The resolution of the Sick product is impressive for a contactless system. Good link.





------------------------------

If we learn from our mistakes,
I'm getting a great education!

 

[IRstuff]

ScottyUK,

You're right; that's what I get for taking the datasheet verbatim.

I've been itching for the instrument we bought to fall out of warranty so I can take a peek inside ;-)

TTFN

 

[IRstuff]

The only other plausible approach would be to gin up some sort of optical encoder strip that would have to span the entire distance.

Not even that unrealistic. There are digital tape measures that can do that. Hook up the free end to the object and move it and record the reading from the tape measure:

http://store.yahoo.com/tylertool/mit2118digta.html

http://www.professionalequipment.com/xq/ASP/ProductID.1385/id.15/subid.126/qx/default.htm

Another possibility is GPS RTK system, but that's probably in excess of $4000, although the accuracy is probably only on the order of millimeters.

TTFN

 

[MacGyverS2000]

How about this setup...

Two lasers, several meters apart, pointing towards the stock at some arbitrary angle. The lasers are rotated until they each hit a detector in the middle (obviously only one is one at a time). The angle of the laser turntables at the moment the detector is hit may help determine distance to the object, as well as allowing the removal of error due to the stock's angle from parallel to the measuring setup.

The stock's angle can only be removed if the change is parallel to the ground, and not swaying back and forth on a hanging mount, but at least it's something to start with.

To be honest, this may or may not be feasible, I just don't have the brainpower tonight to determine if the geometry involved will really help solve the problem. Try it out...

 

[IRstuff]

1/8" over a distance of say, 20 ft, results in a required resolution of 500 microradians, which requires a high-resolution encoder, resolver or theodolite, again, no trivial matter or cost.

Even at that distance, the beam spread is probably too large to get an accurate measurement, but now, you're stuck with two precision angle stages to calibrate and maintain, as well as designing the circuitry for two 4-quadrant detectors, or possibly PSD's, the multiplexing for the two sources, a robust centroid measurement algorithm, etc.

You can do this with only a theodolite, but a decent theodolite is more expensive than the laser measurement unit. Your theodolite can simply be placed in a known position with the two endpoints also known. Assuming that you can keep the retroreflector on the target on a line between the known endpoints, you can triangulate the position of the target with a single measurement.

TTFN

 

[IRstuff]

As I indicated before, I really don't think that you can even adequately begin to design anything for 100 hours of labor time. If you really need this kind of accuracy, you can use the digital tape measures, which run less than $50, or buy a laser rangefinder.

Anything homebrewed requires much more careful design to ensure adequate accuracy and performance over temperature and time and other environmental conditions. After you've completed such a design, built-it and tested it, you'll still need to perform a design validation and develop the means to maintain calibration. Since it's a custom design, you'll probably wind up doing the calibration in-house, requiring documentation, procedures and equipment to do the calibration as well as keeping someone qualified and trained to do the actual work.

TTFN

 

[logbook]

Of course shearstrength hasn’t indicated whether he wishes to make 1 off or 5000/week. This choice would tend to influence the decision.

 

[shearstrength]

Check this unit out, it's within the distance and inexpensive too!

http://www.omega.com/ppt/pptsc.asp?ref=HH-DM