My head just tends to be full of open hardware and open manufacturing stuff, and one idea I have had for years and always wanted to try, but never could for want of access to an electronics lab, is a method for measuring distances which would be super useful for cnc machines, and enable a lot of things in the grassroots open manufacturing world.

I'll explain my idea, then why it would fit in well/enable a lot of stuff.

It is not time of flight, or laser interferometry. We are trying to set up an oscillator which oscillates at a frequency directly dependent upon the optical path from emitter to receiver, and arrange for this path to be close to a single straight line (with the rays retracing their path, or something very close to it, the way there and then back again).

Ok, so you take a photodiode or more suitable detector, and a laser diode. Of a type used in fiber optic communication, so good for high modulation frequencies, etc.

You either place them closely side by side, facing in the same direction, or perhaps use a semi silvered mirror or other partially reflective beamsplitter, so that the laser diode emits along the same axis as the photodiode receives.

Conceivably, the electrical characteristics of the laser diode could even be used to receive the return signal, but it would be a weak signal as the light diverges since being emitted from the laser, and again after striking the ball bearing, so idk if that could be made to work.

The system emits and receives within a cone that is relatively wide, ideally.

Now, the reflector is actually a ball bearing , not a corner cube reflector as used with laser interferometry. This is because they are much easier to align, and cheap. And no matter the orientation, the distance to the center of the reflector is always the same as the distance to the surface doing the reflecting back to the reciever. They can be very accurately and readily mounted on things, too.

Ok, now instead of measuring time of flight, we measure the frequency. We can do this to a very high degree of accuracy using phase locked loops and counters and I'm sure many other techniques within ee that I have no clue about. And off the shelf reference oscillators, which for a couple bucks we can get parts per billion accuracy. accuracy, not precision. nanometers per meter. I understand there will be other sources of error, but still. my point there is that as reference sources of accuracy - for time, distance, force, whatever - frequency is super cheap and readily available.

ok. so it's a bit like time of flight distended measurement, but with a huge number of pulses, and you merely measure the frequency of the oscillator, rather than trying to detect and measures the length of a single pulse.

A single pulse is inherently harder to measure accurately than to just take the average of a hundred thousand pulses.

So now, I think there might be issues with receive sensitivity and ambient light noise sources. filters, shading to reduce the receive angle cone, etc. might be used. Even if the apparatus has to be draped in opaque fabric, the alternatives are so expensive and unavailable, that's ok.

Ok, now part of why this is valuable is that you could attach three of them to a plate. Now the measurement of the distance from the center of the ball bearing to three points in space, with a high degree of relative accuracy (relative to previous points the bearing has been located at, rather than to an external reference point) is still extremely useful. The bearing can touch some known points in the machine, record the reflector position, and from that point onwards know the position of the reflector relative to the other components of the machine, or the part being operated upon. So now we have absolute accuracy.

That's a massively, fundamentally useful thing, to calibrate the machine. Secondly, suppose you have a milling machine, and you have the ball bearing mounted on the top of the machine I.e. the top of the spindle motor. So it is visible from above the machine. Then mount the sensors above the machine. Now the sensors can work no matter where the spindle is, or what is in the tool holder. It becomes a closed loop machine, like using glass scales, but there is no concern about the orthogonality of the mechanical axis, the straightness, the alignment of the scales with the axis of motion, thermal expansion of the axis during operation, or anything.

It could be used with a touch probe in a coordinate measurement machine. Indeed, with 2 measurement systems tracking the location of 2 ball bearings, some distance apart from each other, and mounted on a wand with a touch probe, you could conceivably even have a totally freehand CMM. That is, with none of the 3 linear axis structure. As long as the views of the bearings do not become occluded. The system should still be able to achieve sufficiently high time resolution I think.

This could imbue cheap reprap or repstrap (which are built with stuff like metal rods from the hardware store) like machines with extreme accuracy for cheap. Accuracy of motion, of course. I understand that the actual forces involved with printing or cutting etc will introduce plenty of errors, however it would still be extremely useful in many cases to have this foundational accuracy to draw upon.

The manufacture of accurately shaped objects is critical. It is a fundamental question, where this accuracy is to be obtained. Conventionally, you must buy highly accurately made parts. Then they must be accurately aligned. It is very expensive and labor intensive to us regular people. This system would really enable a lot. The accuracy flows from cheap electronics and the speed of light in air being well known and constant.

Does anyone see why it would not work? It seems to me the main issues are getting a strong enough return signal, and/or designing sufficiently sensitive receive electronics, keeping the noise low enough, suppressing harmonics.

Ok, I found a desktop to at least draw a diagram. I can't log into my google account on this computer, though.

Yes, it is in a way time of flight, but the thing is that we can take a millisecond to determine the frequency. We can take the average frequency over a millisecond, or hopefully 50 microseconds, maybe.

So that way jitter doesn't matter as much. As you say, in TOF, when you are trying to determine the timing of a single pulse, yeah, jitter is a major problem. To accurately determine distances is not that practical. Most time of flight measurement systems, as you say, don't get this sort of accuracy.

This uses the averaging of a great number of pulses, instead. If the path length is 1 meter, we get 300 million pulses per second or so.

The frequency would be pretty high, I don't know how practical it is to measure a frequency like that directly, with say a counter, but I suppose with the techniques developed for programmable radios etc. it should be doable?

To measure a frequency directly and accurately between say 3 Ghz and 300 Mhz. Directly or indirectly.

Figures 1 and 2

Regarding the received signal strength, yes I agree that looks like the biggest barrier (I would reply to your answer, but it says I need 50 reputation to do that, so speaking here is the only way I can say anything). I mean, yeah, that's where the design talent comes in, to accomplish that at a reasonable price.

A small amount of light might land back on the laser, but yeah, it's the same light we are talking about, that tiny bit that comes back from the ball bearing, so I wouldn't expect it to interfere with laser operation. Well, plus any stray reflections from any windows etc. which would be much stronger.

In the optical department, wavelength and polarization filters could be used.

Regarding the challenge of keeping the system pointed at the ball bearing, yes.
Ideally, I was thinking the laser output is not a collimated beam, but say a 30 degree cone. So anywhere within the cone, the distance is still measured. Alignment of the receiver etc. such that the optical paths are coincident should be done and wouldn't be demanding. This way there are no moving parts.

If a lens was used, to collimate the output, and greatly increase the amount of light hitting the detector, then yeah, it's not as simple. I would hate to have to physically point it at the ball bearing, but I guess you could have a servo mechanism that did that, with a ball joint kind of thing. Essentially the emitter/detector system would just have a much smaller cone, say 0.5 degrees instead of 30, and the system would have to track the 3d position within that cone, then when it got to the edge of the cone, command the electromechanical mechanism to move the system so that the cone again shines in such a direction that the reflector is near the center of the cone. You see what I mean. If you moved the ball bearing too fast one day, and the mechanism couldn't keep up, then poof, it would loose track of the reflector. But whatever, that could be dealt with or tolerated. The question would be if it is necessary to incur that expense, complication and additional mechanical tolerance loss that the mechanism imposes, in order to achieve the required return signal strength.

Radios routinely handle received signal strengths a picowatt or so. If the diode is say 10 milliwatts, I think we should be ok. It's a lot better emitter/reflected signal strength than radar. But a lot of those techniques work cuz you know the frequency of the modulated signal. In this case it's light.

Perhaps the signal to the laser could be modulated at a fixed 50 ghz or something, whatever is close to the max it can handle. Then the system can filter the output from the detector to a very high degree, 100 db or something, to eliminate noise from stray light etc. Then it can be highly amplified, smoothed, and the resulting output is then used to modulate the laser diode. The point of all that being to make the detector+associated electronics highly sensitive to the signal from the laser diode, and nothing else.


closed as unclear what you're asking by brhans, Dave Tweed Oct 2 '18 at 5:15

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • \$\begingroup\$ Welcome to EE.SE. Your post is a difficult read due to strange sentence structure, missing capitalisation (title and post) and punctuation. You could probably remove much of the text to outline your proposal and ask the question. A sketch would be a big help. What does "it's a bit lounge time of flight" mean? \$\endgroup\$ – Transistor Oct 1 '18 at 18:11
  • \$\begingroup\$ Sorry, I wrote it all on my phone. Yes, that is a passage that got unreasonably garbled. It was supposed to say "it is a bit like time of flight". \$\endgroup\$ – user7311634 Oct 1 '18 at 18:59
  • \$\begingroup\$ Hit the edit link when you get back to a computer and see if you can refine it a bit. \$\endgroup\$ – Transistor Oct 1 '18 at 19:02
  • \$\begingroup\$ A diagram of the mechanical setup and a schematic of the circuit would be very helpful for us to understand the proposal. \$\endgroup\$ – The Photon Oct 1 '18 at 19:04
  • 1
    \$\begingroup\$ @user7: There is still a load of typos in your post. It needs a major edit, careful proof-reading and explanation of what this oscillator is and how it's supposed to work. I reads as though you think you can count lightwaves. If so then calculate the frequency of, say, 600 nm red. Explain your "for a couple bucks we can get parts per billion accuracy" statement too. \$\endgroup\$ – Transistor Oct 1 '18 at 19:27

What you are describing IS time-of-flight, if I'm understanding what you're saying.

You're just using the time-of-flight to control the period of an oscillator (by emiiting a new pulse each time you receive one), and then measuring the resulting oscillator frequency in order to determine distance.

This has all kinds of problems, not the least of which is making sure that the electrical delay from receiver to transmitter (completing the feedback loop for oscillation) is stable on the order of a few fs (10-15 seconds). For example, drift or jitter of 6.67 fs would represent a path length error of 0.1 mm, which is the sort of precision you seem to be hoping for.

  • \$\begingroup\$ To achieve 6.67 femtosecond jitter, the Vnoise/SlewRate must support that. With 1volt/nanosecond SlewRate, the Vnoise must be Tj * SR = 6.67e-15sec * 1volt/1e-9 sec = 6.67 e-6 voltd or 6.6 microVolts RMS total integrated noise. \$\endgroup\$ – analogsystemsrf Oct 2 '18 at 3:10

You either place them closely side by side, facing in the same direction, or perhaps use a semi soldered mirror or other partially reflective beamsplitter, so that the laser diode emits along the same axis as the photodiode recieves.

If I understand what you're saying, you have a problem here. If the return light can also reach back to the laser's output facet, it's likely to cause problems. Many laser types are sensitive to back reflection. It can cause mode hops (optical frequency changes), power fluctuations, etc.

Now, the reflector is actually a ball bearing , not a corner cube reflector as used with laser interferometry.

Here, I'd be worried that very little of the light reflecting off the round reflector will actually reach the receiver. The reflection off this reflector will be a strongly diverging beam. If the beam path is more than a couple millimeters, you'll probably have multiple 10's of dB of path loss before you reach the receiver.

You'll have to very carefully validate your optical path loss and receiver sensitivity to know if this can work or not.

It becomes a closed loop machine, like using glass scales, but there is no concern annoy the orthogonality of the mechanical axis or anything.

I'm not clear how you'll keep the laser aimed directly at the center of the ball bearing if the ball bearing is moving around.

  • \$\begingroup\$ If enough unmodulated laser light could be reflected back to the laser itself, then mode spacing becomes a measurement of interest: a measurement of distance-to-reflector. Could frequency span from mode-to-mode be measured? A ball bearing reflector does seem like a poor choice, especially on a vibrating machine. \$\endgroup\$ – glen_geek Oct 2 '18 at 2:21
  • \$\begingroup\$ @glen_geek, conceivably yes. Practically, the mode hopping behavior is probably too chaotic to be useful. \$\endgroup\$ – The Photon Oct 2 '18 at 2:54

Not the answer you're looking for? Browse other questions tagged or ask your own question.