IR wavelength in photoelectric smoke detectors?

[klugesmith]

Just energized the LED from a photoelectric smoke detector, to get an idea of its color. 

For the on-axis beam to hit the detector on-axis, it needs to be deflected by about 45 degrees (forward scattered).
We know that fog along roads scatters yellow light less than blue light. 
Near-IR wavelengths are better than visible light for seeing clearly through haze.
So short wavelengths should be best for detecting smoke at low concentrations.  Not counting other factors, like...
Efficiency of LED and detector
Ability to pulse LED at high current without having to boost the battery voltage
Discrimination between smoke and other kinds of aerosols found in living and cooking areas.

Forward voltage at 1 mA was only about 1.2 volts, and the plastic molding for detector is dark.  Both suggest infrared.
Confirmed by digital camera photos with 10 mA in the LED.


How do you tell 780 from 850 or 940 nm, without a spectrometer?  They are progressivly farther along the tail of human visual response.
I was doing shortwave fiber-optic datacomm in the late 1990's, when 850 nm VCSEL's displaced 780 nm CD reader lasers.  You could distinguish those colors by the visual brightness of exposed ports (at standard power levels deemed to be eye-safe).
Isn't 940 supposed to be completely invisible?
At 10 mA this LED has a faint red glow when you look into it.  Enough to tell the difference between on and off, in a dimly lit room.
I would have expected a more obvious red glow at 780 or even 850 nm, but have no reference LED's for subjective comparison.

Anybody made or seen an LED spectrometer that goes into the infrared, using diffraction grating and digital camera?
Are there standard-product "colored" optical filters that can separate invisible LED wavelengths?

[davekni]

I'd thought that near IR LEDs were visible because of the tail in emission spectrum of the LED extending below 700nm rather than ability of the eye to see above 700nm.  Is there information to the contrary?

[shrad]

I would use optical filters used in astronomy (I think omega bob on ebay or something but I didn't look for a long time) or forward voltage.

Another solution would be fluorescence and comparison but you have to find the proper dyes and have reference leds for color reference of the fluorescence.

[Uspring]

I'd also try a diffraction grating. That can give you a roughly quantitative wavelength value. There are some post by Physikfan on this forum, where he looked at the invisible UV side of e.g. Hg spectra. On the IR side digital cameras often use IR filters to avoid false colors, since the red dyes used on the sensor are often wide open to IR. On my camera I have no difficulty seeing the transmitter of my TV remote, though.

[johnf]

One used to be able to get frequency doublers that you could then use the naked eye and a wavelength colour chart to get the IR wavelength

and a quick check they are still available
https://www.alibaba.com/product-detail/800-20000nm-0-8-2um-infrared_60670803956.html

[klugesmith]

Thank you, John.  Have you used those cards, and do they really frequency-double?  So the "fluorescent" output color depends on the wavelength of IR excitation?

I'm not seeing that in the cited web page from alibaba.

Or, for example, Newport Optics, which might be selling a very different physical mechanism:
https://www.newport.com/p/F-IRC1-S

Looks like the IRC1 outut peaks at 625 nm, which would come from halving 1250 nm; but 1250 is out in the tail of wavelengths to which IRC1 is sensitive.  How do these cards work, if not by frequency doubling?  Are they precharged, into some metastable state where a low energy photon can trigger a higher energy electron-state transition?


Now back to Dave's question. 
Does the marginal visibility of NIR wavelengths come from short-wave tail of LED emission spectrum, or long-wave tail of human visual response? 
I bet some of each, but can attest that 850 nm _lasers_ are marginally visible.  Their spectral width is vastly narrower than LED's. 
With a bit of hunting, I finally found a chart with logarithmic vertical scale, since linear scales are useless for this discussion. From Finisar, about a VCSEL diode product.
https://www.finisar.com/sites/default/files/downloads/application_note_vcsel_optical_modes.pdf
 

Here's a pic juxtaposing LED and laser spectra, but without the log scale.


[edit] Attacking from the other side, I just looked up the standard visibility curve and put it on a semilogarithmic chart (first time for me).


"Do not look into laser with remaining eye."
In terms of absolute radiant power or intensity, laser safety standards are tougher on near-IR than any other wavelengths.  For near-IR the eye is clear, and the lens focuses pretty well, but the protective reflex to close eyes or turn away from dazzling light does not kick in.


 

[shrad]

VCSEL have a narrower spectra due to their nature, this is a bit like the OPSL and all the vertical, jet and thin disk cavities

if you use an off the shelf laser diode, especially consumer grade like cd, dvd or bluray diodes, you have a wider spectra especially when you have abused them, and not all that spectrum are coherent

this is due to partial optical damage occurring with overcurrent, overtemperature or back reflection which will change the properties of the emitting facet and spread the emission (a burnt laser diode can behave like a led after abuse)

[klugesmith]

Some fiddling with a DIY spectrometer, using part of a CD as a reflective grating, shows promise.
Turning the grating is easier than turning the view angles, but makes the arithmetic slightly less trivial.  Can do it either way with this setup. 
First pass had Violet extreme at 414, 448, 454, and 543 (2nd order).
Green at 553, 547, 573, and 590 (2nd order).
Red end at 944, 678, 726, and 681 (2nd order).  Might be a blunder in there, but I report what the book says.


It won't be trivial to tell 850 from 940 nm LEDs, when digital camera is the only way to hunt for the sweet spot to see the spectrum.

This should wait until a general cleanup reveals my real diffraction gratings.
Repeating the subjective "visibility of good 850 nm VCSEL", from 20 years ago,
should wait for my real shortwave optical-fiber transceivers to turn up.  Or for someone else to try it and write in.

[klugesmith]

Bit o' progress after learning how to remove the metal layer from a CD, to make a transmission grating.
Trouble is, it's not very efficient.  Compare the brightness of the diffracted spectrum with the direct view of source here:


Here's what it looks like, in quick handheld 1-second snapshot of the smoke detector LED with same camera:


With a bit of mechanical fixturing and a dark background, might be able to get an image of diffracted rays.
Better yet: get a "real" grating and use the detector diode that came with the smoke detector, with no IR-cut filter in the way.  Could get fancy & put a focusing lens in front of it.

p.s. the LED in picture still does have a visible red glow, in this run at 25 mA.  I bet it's nominally 850 nm; 940 would be invisible and 780 or less would be much easier to see.
In the trail-cam / security cam business, you get to pick low-glow or no-glow (850 or 940 nm illumination).
  https://shop.naturespy.org/camera-trap-support/whats-difference-low-glow-leds-no-glow-leds/

As mentioned in OP, I believe the choice of wavelength (and scattering angle) matter in smoke detectors with good sensitivity and selectivity.

[shrad]

I would say the led has a low pass coating, or at least something approaching...

[Uspring]

As mentioned in OP, I believe the choice of wavelength (and scattering angle) matter in smoke detectors with good sensitivity and selectivity.

Perhaps the idea is to avoid Rayleigh scattering from small particles, which strongly favors shorter wavelength and to select simple reflections from larger particles. I really don't know, this is just speculation.

[klugesmith]

Just learned something about tilted diffraction gratings, and drew some pictures to show the point.
A better forum to search & write about spectrometry is probably laserpointerforums, but I haven't been there in years, and have old friends here.

Before buying a new diffraction grating, it seemed proper to run the numbers & know what pitch (d) to shop for.
It's nominally 1600 nm for CD's.   1000 nm (25400 lines per inch) is common in inexpensive plastic film transmission gratings sold for teaching. 
Less easy to find is 2000 nm (500 lines/mm).

Discovered that 1000 nm is a pretty fine pitch for what we're talking about in this thread.
For normal incidence, you get nice wide dispersion of visible light, but NIR at 950 is approaching the cutoff wavelength (equal to d) for first order diffraction.
The cutoff wavelength for second order diffraction is d/2, 500 nm, same as what deviates by 30° in 1st order diffraction (sin a = 0.5).  We could see the violet end of 2nd order spectrum, maybe blue, but yellow and red angles would need sin a values greater than 1.

We can extend the range substantially, and get much better linearity, with oblique placement of the grating. 
That favors spectra deviating toward "normal to grating", at the expense of those deviating the other way.
I drew it for 30° tilt, which seems like a good value for the NIR study if a 1000 line/mm grating must be used.
Now the first order spectrum reaches 950 nm comfortably, in fact I bet it would have little trouble at 1300 nm.
Second order gets the whole visible spectrum, including rays deviating by more than 90°.  I'll believe that when I see it.
But on the other side of transmitted ray, even the first order is almost lost except for UV and violet.


p.s. just picked up some reference LEDs whose nominal wavelengths were marked on the bins in store.
There's one with part number and datasheet, saying peak w = 950 nm +/- 20, whose molded package has a faint blue-violet color.
Couldn't find any IR LEDs below 940 nm.
An article on a trail cam website mentioned having one person at the office who could marginally see the flash from a "no glow" camera.

[klugesmith]

Hey, this is probably going to be wrapped up more easily than I'd expected.

Episode N:  Walked into a science store in an industrial part of the city.  Bought a really cool little hand-held spectroscope.  This way of shopping is better than getting stuff via Amazon or Ebay.

The instrument has, above the mid-plane, a slit on the right and black space on the left.
Below the midplane is a back-lit scale on the left and black space on the right.
Scale reads ..|....7....|....6....|....5....|....4....

Episode N+1.  Rain was expected at home late that night, causing a scramble to put away things that had been stored outdoors for sorting and getting rid of.  Found my greatest grating (from electronics flea market a year or two ago) and my old reference LED's (including some 850's).  Here's the smoke detector LED reflected in hand-held grating, then after turning the grating enough to bring diffracted LED image into view.
 
Next step is to get quantitative & use spectral reference sources.  Anybody notice a fluxmeter sense coil in these pictures?

[klugesmith]

Sorry to be dribbling stuff in. 
Turns out my old wavelength-reference LED's, just like the new set, didn't include any 850's.  Just a bag of 940's with a part number on the Digikey label.
Wired one up & saw its diffracted image in the big reflective grating, but first attempt shining it through the black-box spectroscope was a failure.

Then got over a much lower hurdle. 
Pulled the laser diode assembly from a CD player head, an edge emitter almost certainly at 780 nm.  Biologically visible red light even at 10 mA.  Soldered on 3 wires, for feeding LD current and reading the monitor diode current.  Cranked it up to a little above threshold, after taping it onto the entrance slit of spectroscope.  The slit is long enough for an incandescent flashlight to peek through just above the TO-can laser package, or maybe it's reflecting from visible metal surface of the can.

The intensity of bright spot off to the left varies with the laser current.  Don't know whether the LD has already been damaged, like shrad warned about.  Can a chart of optical power vs. drive current be used as an indicator of health?

Now I want to measure the actual angles associated with spectroscope scale, compute the grating pitch, and figure out where the scale marks for 850 and 940 nm belong (if anywhere).   We saw that nonlinearity kicks in rapidly as the absolute deviation angles become large, and here they are pretty small.  Hey -- I got some "UV" LEDs the other day from a bin labeled 410-415.  They should have a 2nd order diffraction angle same as 1st order for 820-830 nm.

Anybody got a recommendation for a non-semiconductor NIR wavelength reference source, like some kind of gas discharge lamp?  Or optical filter, like the multilayer units from omegabob on ebay?

[davekni]

Looks like you are way past the stage of looking for crude estimates, but I'll share this idea anyway.  If you can find an 850nm LED, use it as a photodiode.  A 940nm light source will produce almost no response.  An 850nm light source will generate a reasonable response.  780nm light will likely produce a yet-higher response, unless the LED package (for the photo-diode one) has filtering.

I haven't personally tried the above test with IR LEDs, but have with both visible and UV.  So far, all the LED's I've used as photodiodes behave as expected, responding to wavelengths shorter than or equal to their emission wavelength, but not to longer wavelength light.

[klugesmith]

I'll have to try the LED-as-photodiode idea with the matrix of 6 or 8 LED wavelengths on hand.  Thanks for the suggestion.

This evening I started with a "UV" LED, and saw a surprisingly extended spectrum.  Apparently from broadband fluorescence in the clear molded package.

The spectrograph showed 2nd order diffraction with that source and the plain old fluorescent room light, far away from the reference scale.  I added reference marks by drilling some 1 mm holes in the box, mostly through side wall at a very shallow angle.
Now those marks are out of focus, and like the main scale they are pretty sensitive to parallax when position of eye or camera changes. But vastly better than nothing.  Indicated by upward-pointing arrows here:

The source here is the UV LED in middle of slit, in front of a flat mirror that presents view of a T12 fluorescent lamp overhead.
Downward facing arrows point to 2nd order lines for
yellow-orange 578 nm (would be 1st order for 1156 nm)
green 546 (1092)
blue-green 494 (988)
blue 436 (872)
"UV" LED 410-415 nm, as labeled at store.  (Would be 2nd order for 820-830 nm.)

Plan for next round: bring back the IR LED's and CD laser.  Leaving the fluorescent lamp on in the background is probably a better scale reference than the little spots of drillium (a word once popular in the lightweight bicycle world).

[klugesmith]

A new horse appears in the final furlong: 850 nm gigabit optical transceiver. I found some in 1x9 form factor, which predates GBICs and SFP's. 1x9's are electronically the same and easier to connect if you don't have a socket.  The associated datacomm standards are for runs of up to 300 or 500 m in multimode quartz fibers.  My day job 20 years ago included designing-in parts like these and helping to write the standards. 
 

Here's one datasheet: https://datasheet.octopart.com/HFBR-0535-Avago-datasheet-10309772.pdf
It conservatively gives min & max wavelengths of 830 and 860 nm, so these VCSEL's are not appropriate as wavelength references for serious work.  Same datasheet says spectral width, maximum, is 0.85 nm RMS.  A representative optical power is 0.35 mW (-4.6 dBm) into the fiber, which is deemed eye safe & won't be exceeded after any single-point failure.

 In service, with a DC-balanced data stream, this might be switching between 0.05 and 0.65 mW, for an extinction ratio of 11 dB.  A 1010 data pattern at 1 Gb/s would generate sidebands 500 MHz above and below the 352 THz carrier, for wavelength deviations of +/- 0.00121 nm.  Inconsequential here, but not in 1550-nm DWDM signaling.  There, standard channel frequencies (light colors) are integer multiples of 100 GHz or 50 GHz in the low-loss window around 193 THz.

The business end is bigger than the spectroscope entrance slit.  Here's a kluge to admit some spectral reference light, made from a sheet metal EMI gasket that came with one of the 1x9's.  Part of first session with improved "optical bench" fixturing.  Camera and spectroscope and source under test can easily be returned to fixed positions, each with axis horizontal at a standard height above the work table.

I think there's a notable difference between the overhead T12 fluorescent lights at two stations in my garage.  They seem to differ in spectral brightness of the continuum, in between the Hg lines. Might be an artifact of poorly controlled camera exposure. The place with bench power supply, for running CD lasers and transceiver modules, has the less favorable light.
 
Here's first and only picture with all parts together.  Speckles!

I figure the speckly spot is from 1st order diffraction of laser at about 855 nm.
A little to the right of blue line from 2nd order diffraction of the standard 436 nm mercury line.
This has been fun and educational for me.

Oh, and the "850 nm" light is visible in the dark from 1 meter away when eye is within a couple inches of axis. Not as a star-like point, but it might help to have a white light feature to focus on.
Visible from 1/2 meter with bench light on.  Next exercise is to observe whether the beam is more extended vertically or horizontally, or is round because it's from a VCSEL, or fills the metal sleeve of fiber optic connector.

[klugesmith]

Got a shot with original smoke detector LED, battery powered, under the good reference light.


Surprising answer: 895 nm, if my interpretation of the mercury reference lines is correct, +/- about 5 nm.
Never heard of LED's at that wavelength or even close.
I'd rather send the part out to some volunteer with a better instrument, than make a more serious spectrometer at home.

I'm sure smoke detectors only turn on the LED for brief pulses, maybe with 100 mA or more (good use for a little supercap).
Guess I could try this specimen at higher currents and instantaneous temperatures.

The picture juxtaposes a spot of 895 nm IR, by first order diffraction,
with spots of 436 nm blue light by second order diffraction. 
Camera's optics will handle them differently, perhaps contributing to IR spot being more out of focus.
But those spots are close to lens axis, center of the field of view, so it's hard to imagine they are displaced differently because of their wavelengths.

[klugesmith]

Followed up with an unknown clear LED from a remote control,
and a blue-tinted 1080-1071-ND "Emitter IR 940 NM 100 mA".

At first I thought there was a detectable difference in spectrograph view, but it's hard to show that objectively & quantitatively.
Linear interpolation between the mercury lines puts both of these LED's at 935 +/- 2 nm, more or less.

[shrad]

You'll be glad to know your setup has nothing to be ashamed of, most monochromators and simple spectrometers rely on the exact same principle

Simpler ones have a rotating grating with charts against wavelength ranges, and a simple photodiode to manually tune to a peak and read the wavelength on the chart, and more complex ones have a set of parabolic mirrors focusing on a linear CCD to get a spectrum reading.

Another easy option is FFT on a simple michelson interferometer, somewhat easy if you don't have a grating but some mechanical tools and optics

That's something I always wanted to make as a project, but life always have other projects