SparkFun Forums

[GTBecker]

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

Yeah, that was a cool little LRF project. Toward the bottom is an oscillograph of the laser output pulse. The pulse is ~28nS wide.



That is due, in part, to the simplicity of the flash-pumped laser, but a higher-quality external resonator produced a 7nS pulse, Sam said. Talk of <100pS range resolution here is probably unrealistic with this technology.

BTW, the image of the physical device has been degraded, it appears. Unless my mind has also degraded, the original photos of the prototype were sufficiently clear to reproduce the construction. Perhaps this is a sign of the times.

[NleahciM]

It's an idea. You can get position and frequency out of these. If you're willing to drop almost $50, you can get a continuously sensitive length of 37mm with a 1uS rise tme. This should be fast enough for a 1D scanning laser rangefinder for short range. Long range may still be an issue, yes, but that may just depend on your choice of optics.

The self-resonant would probably fit into mooreaa's idea of a distance degrading sensor. If all you did was frequency count, it would have an exponential taper off towards long distances in trading accuracy, although that near field resonant zone may be a bit too much for the electronics you'd normally choose.

(Addendum)

Time of flight is in Pain the butt region. This is what I've gathered from much of my own research and from Sam's (Have they updated anything there in the last year or two since I last looked?)

The most effective method appears to be peak detection of the pulse. Problem is that the lasing isn't always nice and accurate. Most of the older TOF rangefinders used a solid state pumped laser. Good power and short pulse, but the peak could be anywhere inside of the pulse, so the shorter the pulse the more accurate your ranging. This was obviously coupled with your analog accuracy or digital timing jitter to determine your best accuracy. A photodiode started the timing circuit and the PIN/APD stopped it. I don't know if this is still as critical with high energy diode lasers anymore.

I know Sam has an old paper on a handheld TOF solid state rangefinder on his site... (I'll add it here if I find it)

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

8 years old, still interesting

I had always thought it'd make the most sense to drive the laser with a really sharp pulse and then have your timer circuit time both the rising edge and falling edge of the pulse. With some calibration I would think you could take the average of the two pulse lengths and get decent data.

[NleahciM]

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

Yeah, that was a cool little LRF project. Toward the bottom is an oscillograph of the laser output pulse. The pulse is ~28nS wide.

[img]http://www.repairfaq.org/sam/lr/Image13.gif[img]

That is due, in part, to the simplicity of the flash-pumped laser, but a higher-quality external resonator produced a 7nS pulse, Sam said. Talk of <100pS range resolution here is probably unrealistic with this technology.

BTW, the image of the physical device has been degraded, it appears. Unless my mind has also degraded, the original photos of the prototype were sufficiently clear to reproduce the construction. Perhaps this is a sign of the times.

Surely it's possible to get faster speeds than that? I mean, some optical communication systems modulate lasers into the GHz regions. That suggests to me that it should be possible to get pulses with lengths in picoseconds. Am I missing something?

[GTBecker]

... it should be possible to get pulses with lengths in picoseconds.

Sure; indeed, Ti:sapphire femtosecond pulse lasers are commercially available, but probably not in your budget, and they certainly aren't handheld. Here's one: http://www.newport.com/Tsunami-Ultrafas ... talog.aspx

I think you'd find that producing a single clean 1nS optical pulse is more difficult than modulating a laser with a 1GHz continuous carrier, too.

BTW, OT, but interesting; well, I think so, anyway: a 100fS optical pulse is, start-to-stop, ~0.003mm long in air. At 650nm, that's six cycles of red light.

[NleahciM]

... it should be possible to get pulses with lengths in picoseconds.

Sure; indeed, Ti:sapphire femtosecond pulse lasers are commercially available, but probably not in your budget, and they certainly aren't handheld. Here's one: http://www.newport.com/Tsunami-Ultrafas ... talog.aspx

I think you'd find that producing a single clean 1nS optical pulse is more difficult than modulating a laser with a 1GHz continuous carrier, too.

BTW, OT, but interesting; well, I think so, anyway: a 100fS optical pulse is, start-to-stop, ~0.003mm long in air. At 650nm, that's six cycles of red light.

Sure, it's easier to modulate at that speed than produce single pulses, but that (I would think) would be more to do with your drive electronics than your laser diode. All of the laser diodes meant for optical communication that I've seen have not been terribly expensive, either. Unfortunately I think most all of them were IR, which is annoying.

[GTBecker]

... that (I would think) would be more to do with your drive electronics than your laser diode.

Absolutely right - if you are driving a diode; the diode capacitance and the driver limits the risetime.

Sam's LRF project doesn't use drive electronics, per se; it fires the laser with a Kodak disposable camera flashlamp that pumps a ~150µS 2-joule flash into a Nd:YAG rod, storing that energy until a passive optical switch allows the rod and resonator to lase, which dumps the energy into a briefer output pulse.

[SOI_Sentinel]

Neither are comm laser systems particularly powerful. Most comm lasers have a transmit and a recieve, and with fiber coupling they get a significant percentage of their power from point A to point B. Here we're looking at 1/r^4 plus reflection losses. If you look at that article, you'll see the instantaneous power is in the watt range.

Plus figure with circuit capacitance, inductance, and other variables, are you sure you're going to get your pulse out at this exact nanosecond? Yes, diodes are a lot closer to reality on this, but you still have to build up the diode's own resonance chamber and this may bit a bit different than you expect. The uLRF, as GTBecekr has pointed out, uses a flashlamp pumped solid state ND:YAG laser with Q switching. You actually don't KNOW when it actually fires, so it needs an optical switch diode to start the timer. Might need something similar for a high power diode given that we don't know what the lasing threshold is exactly, but a little less critical.

(Addendum)

Speaking of other variables, I was hoping to find an MCU with multiple gated asynchronous timers in it, but no go yet. The FPGA/CPLD route Busonerd is thinking about is a normal way, but I was hoping a bit more off the shelf.

[GTBecker]

... for a high power diode [] we don't know what the lasing threshold is exactly...

True if the laser diode driver just slams the diode with a square pulse from zero current. There is a period below the lasing threshold where the diode acts like an LED and yields little light; each time it is turned on and off this way it must pass through that dark delay. The threshold moves with temperature, too.

If the diode is biased at the threshold, the laser never stops oscillating and can be switched between high and virtually-zero output very quickly. This is true even with a cheepie laser pointer slice.

For some perspective on what is and isn't fast for pumped solid-state lasers, Newport's DPSSs produce "Extremely short pulses, typically <15 nsec..." http://www.newport.com/HIPPO-Diode-Pump ... talog.aspx

[GTBecker]

... multiple gated asynchronous timers...

Note that Sam's LRF uses a 150MHz range counter. That, and faster, should be easy to do today.

[SOI_Sentinel]

Very true about the lasing threshold. That takes a bit more experimentation than simply slapping on a cheap red pointer to the project. I'm attacking this project from the perspective of someone who isn't aware of optically stabilized laser driver circuits nor their necessity for diode power control and protection. Cheaper we can make this work, the better.

The same goes for the async timers. a 150MHz counter is not readily possible in discrete logic. Neither does it appear to exist in flash MCUs of any size. I was thinking of analog clock skewing into gated external timer inputs, but the lack of async is killing that. For instance, a high end dsPIC33 has 9 16 bit timers. Only Timer 1 can run async. If they all could, I'd take 8 of them and skew and gate them from a 20 MHz oscillator. Of course, by this time, it might just be worth it to be looking at an FPGA. Speaking of which, theoretically a modern CPLD would be able to do 150MHz counting with some headroom. I was avoiding the option on the thought that there might be a bit more wide spread appeal if this could be done with a common MCU and some discrete logic.

[GTBecker]

> ... a modern CPLD would be able to do 150MHz counting...

A Lattice GAL22V10 clock can run at 250MHz. I'll try a 10-bit counter for grins. Two might provide a 20-bit 250MHz counter for <$10.00. Add two bits of gated RF prescaling and you've got 22-bit 1GHz counter. Maybe.

[busonerd]

5 gsa/sec sampling is available in discrete logic.

See: http://www.onsemi.com/pub/Collateral/MC10EP445-D.PDF

Use that to sample the receive stream at 5 ghz. That then gives you 625mhz 8 bit data coming out of the ser->par. A fast FPGA can sample that, or have another ECL level to div that again. Bonus points for using multiple phase-offset clocks to run 4 in parallel, thereby increasing the sample rate.

Feed that into an FPGA and use that to do the analysis. Should be quite precise, esp if the input binary signal is run through a high speed comp so it has a really fast rise time. Pricing looks not too bad.

Cheers,

--David Carne

[GTBecker]

... at 5 ghz...

Let me see you breadboard 200pS logic.

[busonerd]

hah, who's talking about breadboarding?

You'd need at least a 4 layer board to do proper signal integrity techniques, and quite possibly need something exotic like Rogers for the top layer /w the diff pairs on it. Also, QFN32's don't exactly lend themselves to breadboarding.

Its not quite as bad as 5ghz though - that guy samples on both edges, not on one edge, so the fastest signal is a 2.5ghz clock .

Still, its a design challenge, but thats the fastest solution I've seen so far [props go to icee on #electronics for finding that guy].

Cheers,

--David Carne

[SOI_Sentinel]

Now, you're talking theoretical frequency yield, I'm talking effective. I'm not going to pin numbers that high on something I've not seen simulated I'd look at their higher end yeilds, as I see they have the logic content for a fully pipelined counter of nearly any bit depth we'd need that could work decently well. Same with the Xilinx hardware. I'm thinking 3.3V anyway.

Another neat trick with some added hardware:

http://www.k2.t.u-tokyo.ac.jp/fusion/La ... eTracking/

(and I need to stop dropping in 10 minutes after you)