SparkFun Forums

[mooreaa]

So after looking at a bunch of Darpa Grand Challenge footage and lots of sites about the vehicles, I am let wondering how LRFs work...

I found two sites: http://www.pages.drexel.edu/~twd25/webc ... anger.html

http://www.repairfaq.org/sam/lr/

One seems to visually look at the beam and the other measures the time of flight. What techniques are being used on the spinning 3d laser scanners on the darpa cars?

The trick seems to be getting a good strong laser pulse that is detectable back at the source even in out door conditions. Does this mean increasing the intensity? If so how do they keep it eye safe? IR?

I guess where I am really lost is... what methods do they use to "detect" the laser beam. The webcam method seems silly as the laser beam may be very hard to see at distances especailly with a low res camera...

How are they able to see the laser bounced back?

Any insight would really help me sleep at night

Aaron

[Krogoth]

Don't lose too much sleep - there's a reason you can't get a decent laser rangefinder for less than $3000 (the Hokuyo 270 degree rangefinder).

They use a rotating prism or mirror to move the laser beam, and something called an avalanche photodiode with a highly focussed lens to look at the light bouncing off the object. Avalanche photodiodes are extremely sensitive and extremely expensive - they're the semiconductor equivalent of photomultipliers (the things they use in proper night vision systems), and require more than a hundred volts to operate. This does measure the TOF of the laser beam (and remember, light is ridiculously fast).

So, you need a reliable couple-hundred volt power supply, a super-expensive laser detector, an eyesafe laser, a very precise lens, and a very precise and reliable method of rotating the mirror. Oh, and nanosecond-accurate circuitry to determine the range. Don't even think of building one

[mooreaa]

Hmm but they sell hand held laser range finders for about 200-400 bucks. What kind of electronics are used in those? I see they have two lenses, I am assuming one if for the laser the other for the receiver.

Are there cheaper alternatives to this "avalanche" photodiode?

Like if one wanted to setup a primitive system for sensing like 100m or something is there a way besides vision to do this?

Still losing sleep :p

-Aaron

[mooreaa]

Basically I've been playing with vision, and day light effects can make it really difficult to see the laser. Perhaps you have some suggestions on being able to sense sub km ranges using laser?

[Krogoth]

Yes. Use the laser as a laser pointer, and ultrasonics to measure distance

I'd imagine that those cheap laser rangefinders they use for hunting, golfing etc contain the same laser diode/avalanche photodiode setup, but probably a very non-optimised one, hence the cheaper price. I think they're accurate to a metre or so.

Taking the speed of light to be 3x10^8m/s, it takes 3.3 nanoseconds to travel a metre. Therefore, you'd need to be sampling your photodiode at more than 300MSps (300 million samples per second) to get one-meter resolution, so your timer needs to be clocked at over 300MHz. I'd imagine those cheap rangefinders use some nifty tricks to do this cheaply, but I can't think of any.

Unfortunately, there are no cheaper alternatives to the avalanche photodiode. The cheapest I can find one is $70ish on digikey, but the supporting circuitry would probably be at least as much as this again.

What, exactly, do you want a sub-100m laser rangefinder for? And why can't you just buy a commercial one and hack it?

Will.

[winston]

Hmmm, I had vague hopes that there may be a reasonable cost laser rangefinder that could be used as a celiometer for measuring the cloud bases (it would only need to be accurate within 100 feet). My ponderings made me think that just the signal to noise ratio during daylight hours would be the biggest hurdle to overcome.

I guess it's cheaper just to fly the glider club's tow plane and just record the height at which you start scraping the bottom of the clouds

[silic0re]

If you carefully select the wavelength of you laser to be in a region that is naturally relatively dark, and add a selective filter infront of the photomultiplier/diode/detector that only allows the region of interest to pass, then you should be able to achieve a better signal-to-noise ratio.

The pseudocode for the algorithm may look like the following:

- Rotate to next position
- Take a few detector samples (baseline)
- Turn on the laser
- Continue to take detector samples until (new_sample - baseline) > threshold
- Turn off the laser
- Calculate distance based on the time beween the laser turning on and the threshold being passed

As was mentioned above, your resolution is going to critically depend on the sample rate. I'm not sure how it's done in the commercial units, but an alternative to a really fast ADC and microcontroller or CPLD/FPGA may be a fast analog comparator. I'm not really up on a lot of analog electronics, so you may have to research the particulars on that. Ideally what it would do is compare the baseline to the current reading, and just output a logic 0/1 when the difference passed a threshold. This would ideally eliminate the (potentially expensive -- as it's high speed) ADC, and perhaps even allow you to replace the microcontroller with a low-cost, high-speed variant (dsPIC?) that just used a timer and an interrupt pin to give you an okay (~2m?) resolution.

I hope that helps,
silic0re

[Krogoth]

Another solution to the SNR is to simply modulate the laser at a very high frequency (say, 1MHz+) and put a high-pass or band-pass filter as an analog front end between your avalanche photodiode and your timing circuitry. A very high speed analog comparator or Schmitt trigger would be essential since the fastest ADCs are only 500MSPS, and an ADC is complete overkill unless you want to actually analyse the returned signal (to look for reflections, multipath, properties of the target etc).

An analog circuit running at 1MHz+ will be a little difficult to design - think RF issues like trace capacitances/impedences, resonant frequencies etc. Although, if you can demodulate the whole return signal you may be able to get much better resolution (you're basically decreasing the frequency of the entire signal, so what happens a million times a second might only happen a hundred). I'm guessing thats what the commercial units do.

[muntron]

Taking the speed of light to be 3x10^8m/s, it takes 3.3 nanoseconds to travel a metre. Therefore, you'd need to be sampling your photodiode at more than 300MSps (300 million samples per second) to get one-meter resolution, so your timer needs to be clocked at over 300MHz. I'd imagine those cheap rangefinders use some nifty tricks to do this cheaply, but I can't think of any.

Actually the nifty trick is quite simple. Instead of measuring TOF, you only have to measure frequency

If you take the returned signal and use it to modulate the laser, the frequency output will be proportional to the distance traveled. Frequency measurement is fairly simple.

[GTBecker]

Indeed, the simplest optical rangefinder might simply be a laser pointer and a fast lensed photodiode that feeds an inverter whose output switches the laser off when it detects the return light. That forms an oscillator whose frequency depends mostly on the optical path delay.

As muntron says, one can just measure the oscillation period and calculate the distance that the period implies. At ~1nS per foot, a 100 foot path, two ways, takes ~200nS; since the laser state inverts each roundtrip, one cycle takes ~400nS so the frequency at 100 feet between the laser/detector and the reflecting object will be ~2.5MHz. Longer paths result in lower frequencies/longer periods.

More sophisticated processing can use phase, too, to determine range much more accurately.


Tom

[mooreaa]

Hmm I'd really like to play with this if possible.

I am still not sure if this can be done with a lower cost photodiode. Ones such as the Hamamatsu S6986 or the Sharp IS471F sensor seem to have a rise time in the micro second range so it would clearly not work.

http://www.junun.org/MarkIII/datasheets/S6809.pdf
http://www.junun.org/markiii/datasheets/IS471F.pdf

Are there any lower cost optical sensors out there that are high speed yet affordable? The cheapest avalanche diode I saw was about 80 bucks.

Also wouldn't it be necessary to trigger the laser as quickly? Are there any concerns in firing the laser? IE delays that may make it difficult to turn the laser on/off quickly?

One more, does anyone have any info on where to find any kind of "affordable" bandpass optical filter or lenses? I've looked at a few places and they also seem to be in the 100-300 range for optics.

[mooreaa]

I am tempted to buy a 2-300 range finder and rip it apart. But unfortunately even my scope won't be able to clarify if the signals are running above the 500mhz mark

Ideas?

[GTBecker]

At these frequencies you do not need a particularly fast photodiode but the examples you linked to are too slow, probably because of the integrated amplifier, and will not suffice; they appear intended for IR remote control, indexing and slow communication.

I've used a PDB-C104L (apparently no longer available after a company acquisition, but similar diodes are here: http://dkc3.digikey.com/PDF/T073/P2225.pdf). These are inexpensive silicon photodiodes generally used in reverse-bias conductive mode (do not confuse with phototransitors (slow) or CDS cells (slow as mud). You just don't need the speed and expense of an avalanche photodiode for this application.

Cheap laser pointers are actually pretty fast devices because there is nothing there but a few batteries, a laser diode slice and, usually, a limiting resistor. You can connect one of them directly to a signal generator and get modulated output, and you can drive them well with a simple 2N2222 switch or with a driver like a 74HC04.

Laser pointers usually radiate 655nm red, sometimes 635nm (brighter to the eye but not to silicon), so you will need a photodiode that is reasonably sensitive to visible light; most are, but prefer near IR. Later, you can play with ~900nm IR lasers which match silicon detectors better and which will improve S/N. Daylight will be your enemy until you progress through filters, lenses and PLLs, but you can get results with none of those.

Go to a Dollar Store or other surplus/discount house and buy a _bag_ of pointers, for $1 each if you're lucky. You'll need plenty since you'll be blowing them up as you push limits. Later you can try $20 5mW proper laser modules, but they generally need modification (removing a capacitor from the power-limiting circuit) to be toggled quickly; you'll blow them up, too, so save that for later.

For targets, plastic truck bumper reflectors will get you going. They are molded multi-corner-reflector arrays. Reflective tape works well, too; the tape is coated with glass beads that behave like corner reflectors, returning the beam to the source.

I'd advise against hacking a commercial rangefinder. I have three (Bushnell, Leupold and Browning). Inside you'll find highly-integrated, inaccessible logic and optics, in dry nitrogen in the Leupold. Don't waste your money destroying an instrument.

Have fun, but avoid the beam and the reflections. Still, you'll get dazzled, I guarantee.


Tom

[GTBecker]

I dug out a similar project and fired it up.

This is the visible return from a plastic bumper reflector mounted on a skinny tree that's 99 meters from the laser, which is a ~3mW 655nm pointer, a typical cheepie, modulated at 1.5MHz, in full Florida sunlight. You are looking through a 4.5" reflector telescope. The vertical interference is caused by a diffraction grating in the optical path, used for other purposes than ranging.




Tom

[NleahciM]

If you don't need cm resolution you will probably be OK using just a normal PIN (Positive Intrinsic Negative) photodiode. An APD (Avalanche Photo Diode) is essentially a PIN diode on crack. Decent ones are generally $100-1000. You reverse bias them with a very large voltage. The higher the voltage, the more sensitive they become. Generally you bias them with 50-250VDC. APDs are fun, but for your application they sound like overkill. APDs are what you'll see in military grade laser range finders as well as instrument grade photon counters.

TOF (Time Of Flight) rangefinding is, IMHO, the most difficult of the standard ways of doing laser range finding. I think it is best reserved for when you need crazy high sample speed or when you need crazy high resolution (sub cm) at long range. The webcam method is fairly simple and you can get good results with that. I've always wanted to try using a linear CCD array (like those used in scanners) instead of a webcam.

Phase difference rangefinding would probably make a lot of sense for you. The idea is to modulate your laser at a set frequency, and then compare the phase difference between the transmitted and received signals. Once calibrated properly this can give you very accurate results.