Cambridge Underground 1994 pp 20-23

In Search of the Ultimate Surveyor

By Mark McLean


Two summers spent grovelling around in muddy Austrian holes with a misted-up compass and a broken tape measure had convinced me that there must be a better way of cave surveying. I am an electronic engineering student, so for my final year project I proposed a cave surveying instrument. I had already done some thinking about the design during the previous summer, and investigated what might be feasible. I proposed to build a combined compass, clinometer and rangefinder that would store the readings taken for later download to a PC.

The instrument was to be in two parts, a main unit containing the compass, clinometer, microprocessor and half the rangefinder, and a transponder unit containing the other half of the rangefinder. It was to be used by holding one corner of the main unit at one survey station, and pointing it at the transponder which was held by the other survey station. At the press of a button it would then measure and log the length, bearing and inclination of the survey leg. In addition there was to be provision for specifying which corner of the main unit was to be held next to the survey station, and for an echo-type ultrasonic rangefinder for measuring passageway dimensions.

Sound too good to be true? Unfortunately it was!

The Compass

This was a flux-gate compass based on a design in Wireless World, October 1982. The heart of a flux-gate compass is a toroidal core with a primary winding on it. A pair of pickup windings are wound across the full width of the toroid at right-angles to each other. The magnitude of the voltage in each winding is proportional to the magnetic field in that direction, so the arctangent of the ratio of the two voltages gives the direction of magnetic North. In my design it had to be able to work regardless of its inclination to the horizontal, so it was designed with two toroidal cores set at right angles, such that it could measure the direction of the magnetic field vector in three dimensions.

Unfortunately I never succeeded in getting it to work to an accuracy any better than 45 degrees, and that with the compass horizontal. The underlying problem was that I failed to understand how the properties of the core affected the operation of the compass, and used a core of the wrong type which did not saturate easily enough. As a result I had to increase the current flowing in the primary windings so much that they became significantly self-heating. This caused the zero offset to drift significantly as the coils warmed up, making calibration almost impossible,

The only advice I can offer to anyone else who wants to build an electronic compass is to make sure you use the correct core. Supposedly an accuracy of better than one degree is possible.

The Clinometer

In principle it should be possible to calculate the clinometer reading from the three dimensional magnetic field vector direction and the local angle of dip. The local angle of dip could easily be measured by the instrument if it was known to be held level. Therefore I decided not to have a separate Clinometer, but to use the compass. Therefore the clinometer didn't work either. Not only did it not work because the compass didn't work, it also didn't work because there were only two perpendicular compass toroids, not three.

A much more sensible approach would be to use a separate clinometer and use it to help make sense of the outputs from a pair of perpendicular toroids. Possible ways of building a clinometer are:

  1. Optically encoded shaft with damped pendulum attached.
  2. A pair of accelerometers at right angles (expensive but nice).
  3. A rotating semicircular metal plate held near a similar fixed semicircular plate. The capacitance between the two plates will change depending on the amount of overlap.

The Rangefinder

This worked well and was fairly simple, but was fundamentally not cave-proof. It was a hybrid infrared/ultrasound system. The main unit sent out a pulse of IR light. When the transponder unit detected this it replied with a pulse of ultrasound. A timer in the main unit counted at 245.76kHz for the time it took from emitting the pulse of IR to receiving the ultrasound response. This gave the range of the transponder unit to a resolution of about 1mm. The accuracy was to within 50mm over a range of 10m and then tailed off to 300mm at a range of 20m. The main source of inaccuracy was the slow rising edge of the pulse of ultrasound, which is inherent in the fact that to produce the ultrasound a physical diaphragm must be made to vibrate at 40kHz. The wavelength of the ultrasound was in any case 8mm.

In addition to the main rangefinder there was also an echo-type rangefinder built into the main unit. I didn't have time to test this but it is probably good enough to take quick measurements of passageway dimensions.

The most significant problem with this rangefinder for caving use is that the ultrasonic transducer element cannot be sealed; from the environment if the sound is to escape freely. It is therefore very vulnerable to being damaged by water or mud. I considered putting a thin membrane over the transducers, but a physicist friend calculated for me that at 40kHz a membrane of mass 3g per square metre would absorb half the sound energy passing through it. It would also hardly be cave proof itself.

The range and risetime of the ultrasonic part of the rangefinder could be improved by increasing the magnitude of the drive voltage. I used a bridge configuration running from a 15V power supply (two alkaline PP3 batteries) giving 30V peak-to-peak. The transducer however is specified to withstand pulses at up to 60V p-p. I also planned to write clever software to sample the rising edge of the received ultrasound, as well as trigger the counter from it. It would then be possible to extrapolate back to the true beginning of the pulse and thus avoid errors due to triggering at an indeterminate point on the edge.

A much more reliable rangefinder could be built using a fully optical system, and if done properly it would be more accurate as well. The IR part of my rangefinder has a range of better than 30m using ordinary cheap IR LEDs, photodiodes and plastic lenses from Maplin. These devices though, would not be suitable for a fully optical rangefinder as their risetimes are too slow.

Professional laser rangefinders are very expensive, but it may be possible to build something with an accuracy of a few centimetres fairly cheaply. It should not be necessary to use a laser, with its attendant risks to the surveyor's eyesight.


This was pretty standard stuff and generally worked well. A 68008 processor was used purely because I already had an assembler and C compiler for the 68000. The 68008 is simply a 68000 with an eight bit external data bus. This makes both the CPU and the memory physically smaller. 32k of RAM was provided, enough for over a thousand survey legs. An RS232 port allowed data to be downloaded to a PC. I talked to Olly Betts, author of 'Survex' who said it would be easy to modify his program to accept the data directly.

The main unit contained a keypad for control of the surveying functions and editing data, and a 16 character by two line LCD to display readings and other information. The LCD was backlit for use in conditions of low ambient light.

Power for the digital circuitry was provided by four NiCad C cells, which gave 5V directly. This worked well and gave about five hours surveying time. A separate lithium battery backed-up the RAM contents so the instrument could be turned off while caving between survey stations, thus giving plenty of surveying time. Two NiCad PP3 batteries provided the positive and negative analogue power supplies. This was simple but had the disadvantage that as soon as one of the three batteries went flat, no more surveying could be done. It would be better to use a switched mode power supply running from a single battery.

A significant reduction in weight, size and power consumption could be achieved by using a higher integration CMOS microprocessor with more on-chip functions. An IC such as the 64180 consumes 10mA and combines a Z80 CPU with two sixteen bit timers, a UART and a synchronous seriai port. This would replace the two biggest ICs which between them use 300mA.


Proper surveying software was never written as the compass and clinometer did not work well enough to warrant it. Instead I wrote some software that demonstrated what the compass and the rangefinder were capable of individually. This software allowed the user to calibrate the compass by rotating the instrument through a full circle, and then tried to compensate for the temperature drift by using a look-up table. It also improved the rangefinders performance by taking ten readings, deleting any that differed too wildly from the majority, and then averaging the remainder.

Mechanical Design

No real attempt was made to make the instrument cave-proof. It was housed in a diecast aluminium box but was in no sense waterproof. Nor was it particularly shock resistant. The main unit measured 188 by 119 by 82 mm and weighed 1.8kg. The transponder unit measured 150 by 80 by 50 mm and weighed 300g. Quite chunky, but not unmanageable.


Much more work remains to be done. My project is now complete but I hope to follow it up now that exams are over. I intend to concentrate on improving the sensors before having another attempt at putting it all into one box. A possible intermediate stage is an electronic compass/clinometer. This would avoid the problems of not being able to sight a compass along legs with more than fifteen degrees slope and would have a clear digital readout. It would not have the datalogging functions, so the digital hardware would be simpler and a single chip microcontroller could be used.

I remain convinced that the idea behind the project is a good one. Hopefully another summer of squinting through conventional compasses and clinometers will engender some more bright ideas as to how it may be realised. I am very happy to share the work I have done with anyone else who is also interested in building cave surveying instruments. For a full copy of the report please write.

Cost (photocopying plus postage) should be of the order of £9 with software, or £5.50 without. Sorry, but it runs to 50 pages of text, 35 pages of drawings and 65 pages of software! Alternatively the text and most of the drawings could be supplied on disc.

> Index to Cambridge Underground
> Table of Contents for Cambridge Underground 1994
> Back to CUCC top page