[Laser] Notes on obtaining optimal efficiency (Was: Fresnel LED TX - test over 94km)
Vlacilik, Radovan
Radovan.Vlacilik at siemens.com
Sun Sep 19 17:36:23 EDT 2010
Hallo Clint,
we would like to thatnk you for all the informations about optical setup
of the LED TX system. It is very helpful nowdays, when we are starting
to design our new LED TX. We will try to build new TX during the winter.
Kind regards
Rado om2zz
-----Original Message-----
From: laser-bounces at mailman.qth.net
[mailto:laser-bounces at mailman.qth.net] On Behalf Of C. Turner
Sent: Thursday, September 16, 2010 9:05 PM
To: laser at mailman.qth.net
Cc: laser-request at mailman.qth.net
Subject: Re: [Laser] Notes on obtaining optimal efficiency (Was: Fresnel
LED TX - test over 94km)
Greetings to the group.
Echoing Chris's notes, what follows is an email that I'd originally
composed (over the past several days) to Rado and his group, but it will
probably be of interest to all.
***
***
While I'm putting together something more comprehensive (e.g. web pages,
article, etc.) here are few things that we have determined that are
useful in using Fresnel lenses with LEDs.
For many of you, some of this is very basic stuff, but then for those
who *don't* know it, it will be new!
***
First, in any optical system (and assuming "perfect" optics, no
atmosphere, and that the wavelength of light was zero - none of which
are true) the divergence of a given system is related to the proportion
of the aperture of the lens and the size of the emitter. (Note: Laser
pointers have relatively low divergence by virtue of the fact that the
area from which the light is emitted is *extremely* small - so this
ratio is maintained even with physically-small lenses.)
What this means is that using such an impossible system (perfect optics,
zero wavelength light, no atmosphere, etc.) the collimated beam would
have zero divergence *IF* its source size were infinitely small and the
lens were of a finite size. Clearly, this cannot be true so let's
back-track to a real-world system using Fresnel lenses and LEDs.
In a system using Fresnel lenses we can ignore the wavelength of light
as our optics will be too imperfect to even get close to the diffraction
limit, and if we use monochromatic light we can also ignore chromatic
aberration. What we are (mostly) left with, then, is the "quality" of
the lens itself and the ratio of the source size versus size of the
lens.
For the size of the emitter (the LED) we are kind of stuck. With LEDs
we can only achieve a certain amount of "current density" for the LED in
terms of amps per square millimeter and this factor correlates directly
with the number of lumens emitted by that surface per square
millimeter. With the present technology this limit is roughly 1-2.5
amps per square millimeter for LEDs constructed in a manner similar to
the SST-90. This limit is imposed, in large part, by the fact that heat
dissipation becomes problematic at such current densities and further
(dramatic) increases are likely to require highly exotic materials and
techniques and would therefore not likely find their way into widespread
production anytime soon. The other factor has to do with the quantum
efficiency of the LED itself as well as the very ability to construct an
LED such that the photons emitted within the material can escape and be
useful to the designer.
So, with our number of photons per square millimeter being somewhat
"fixed" by the current technology, we must make do with what we have.
Now, one oft-repeated fallacy is that a "bigger, more-powerful emitter"
will increase the higher far-field flux (e.g. the "brightness" as seen
by the distant observer.) With a simple lens system, this is just not
true: A larger emitter will simply produce a larger, wider "spot" of
light (which would be a bit easier to aim!) but at a long distance there
is NO difference between the perceived brightness of large emitter and a
small emitter, assuming that each one was producing "X" number of
photons per second per square millimeter of the emitter. The key to
improve far-field flux density, then, would be to somehow increase the
*effective* number of photons being emitted per square millimeter of
emitter surface.
That's not easy. Since we really can't increase the luminosity-per-unit
area of the LED, we are stuck with trying to decrease the *apparent*
size of the emitter. While it's easy to go the other way, reducing the
apparent size of the emitter WHILE still maintaining high efficiency
(e.g. not losing most of the photons in the process) is a very difficult
task: In 2009 during a visit to the optics department at the University
of Arizona, I posed this very question to a number of the professors and
after thinking about the problem for a while, they agreed that it was a
vexing issue and that they'd have to think more about it, but didn't
think that whatever solution there might be would be particularly easy
to do. My own searches on the "GoogleWeb" about this topic have also
proved fruitless. (The use of LED-based "fiber optic" launchers come to
mind, but these devices often have rather poor overall efficiency - that
is, they usually harvest relatively few of the photons emitted by an LED
of large source size.)
Now, several things may come to mind if the goal is simply to *decrease*
divergence:
- The use of a negative lens between the LED and the Fresnel. This will
certainly reduce divergence, but it also spreads the light being emitted
from the LED over a wider area and most of it will "miss" the Fresnel
entirely, lighting up the area beyond the edges of the lens. The
overall result is a much LOWER far-field flux than if you'd done nothing
at all!
- Mask the source. By placing a mask with a hole in front of the LED
(and as close to LED as possible) one can reduce the divergence of the
beam. Doing this will NOT reduce the amount of light reaching the
distant observer, provided that this pinhole is large enough to overcome
the optical inaccuracies of the lens itself: This statement may
surprise some, but it's true! What IS reduced, again, is the
beamwidth. Doing this has pros and cons: On the plus side, since
divergence is reduced, crosstalk may be reduced in a full-duplex system
as the beams are better-contained and less transmit energy is likely to
be intercepted by a parallel-aimed receive system due to light
"spillover" from Rayleigh and other types of scattering as the transmit
and "receive" beams diverge. On the downside, with reduced divergence
it is more difficult to aim the system. It should also be noted that
putting a mask in front of such an LED can affect it thermally as it
will prevent energy (in the form of light) from escaping the immediate
environs of the emitter and this additional heat source should be taken
into account.
Neither of the above will actually improve far-field flux - which is our
desired goal. What can we do, then, to improve that?
What we are left with is to make sure that as much light as possible is
directed toward the Fresnel lens. For this, a positive lens is
typically used to "confine" the cone of light being emitted by the LED
to the apparent angular dimension of the lens at the focal length.
Taking the (imperfect) analogy of "illuminating" a parabolic dish for
transmit purposes, the energy that "spills" around the outside edge of
the dish is simply wasted as it never hits the dish and has no
opportunity to be directed toward the distant station. On the other
hand, if one were to confine the beam excessively - that is, produce a
small diameter "spot" - one would, using the dish analogy, be using only
a small portion of its aperture and lose efficiency that way - and this
analogy applies to "illuminating" the back of the Fresnel lens.
What is the optimum, then? Based on in-field empirical measurement
where I had a system in which it was possible to measure the far-field
optical flux while being able to vary the distance of this "secondary"
lens to the LED (thus affecting how "wide" the circle of light from the
LED was) AND the distance between the secondary lens/LED assembly and
the Fresnel lens (to adjust overall focus) I determined that optimum
far-field flux occurred when, using a square lens, the edge of the
"circle" of light from the LED just hit the edges and the corners were
slightly darkened. This shouldn't be too surprising, as that's about as
good a "fit" as one can hope for when trying to impose a circle atop a
square!
One point that is worth mentioning is that as one reduces the size of
the "spot of light" from the LED/secondary lens assembly, two things
happen:
- As the spot of light is reduced by moving the positive secondary lens
farther away from the LED, the apparent size of the LED is magnified.
This will increase divergence.
- As the lens is moved farther away, less of the light being emitted by
the LED will actually hit the secondary lens.
In reality, it is this second point that contributes the most to
inefficiencies in the system. If one studies the "beamwidth" charts for
the LED that plot intensity with the off-axis angle, one can see that as
one moves off-axis, the intensity deceases. However, if you consider
that as you widen the angle of the light from the LED, the area being
illuminated increases exponentially and one sees how it can be that a
very large portion of the light being emitted by the LED is not being
directed forward! It is important, therefore, that the lens be situated
such that it "gathers" up as much light from the LED as possible.
Practically speaking, DCX (Double Convex) secondary lenses are
undesirable for two reasons. The first of these is that in order to be
"strong" enough to be useful, they will probably be somewhat "fat" and
this will prevent one from placing the LED very close to the plane of
the lens to achieve the desired result. The other problem with that
type of lens, light that is far off-axis (which means that it's also
near the edge of the lens itself) will more likely reflect off the
curved surface lens - if it hits the lens at all. (That is, eventually
the angle gets to be too steep!)
At the very least, a PCX (Plano-Convex) lens is the best of the most
common types, with the flat surface toward the LED. A positive meniscus
(PMN) lens works very nicely, but these are much more difficult to find.
When adjusting such a system, optimal far-field flux will occur when the
following are true:
- The "circle" of light from the LED just spills beyond the edges of the
square lens.
- The lens is as close to the LED as possible.
For the systems that I have built, the LED and secondary lens either
touch, or are within a millimeter or two of touching. What this means
is that the "strength" of the secondary lens system should be chosen
appropriately: Too "strong" of a lens and you "under-illuminate" the
area of the Fresnel when the LED and lens touch each other (that is,
the circle of light is too small) while too "weak" a lens means that
it's placed so far away from the LED that much of its light doesn't even
hit the secondary lens. This also means that, compared to the size of
the LED, the secondary lens will be quite large (e.g. 48mm diameter in
the case of a Luxeon III LED.)
Now it is true that the use of any positive secondary lens will increase
divergence, but the fact is that you will end up with better far-field
flux with a well-matched system as described above with higher overall
divergence than a mis-matched lower-divergence system.
***
This brings to mind another consideration: The F-Ratio of the Fresnel
being used.
If the F-Ratio (the ratio between the len's diameter and focal length)
is too low (less than 0.5 or so) we can see that it might be difficult
to spread the light over a wide-enough angle to effectively illuminate
the lens while if the F-Ratio is larger (say 2 or more) then the problem
of being able to narrow the LED's beam to "just" illuminate the Fresnel
arises. If one runs the numbers it becomes clear that it is simply not
possible to efficiently illuminate a hypothetical Fresnel with a very
high F-ratio with a *single* lens as that would imply a lens with
impossible-to-make attributes.
In my opinion, the optimal F-Ratio is somewhere in the range of 0.6 to
1.5 - with the preferred range being somewhere around 1.0. In this
range the incident angles aren't unreasonable and it is practical to use
a single lens for the secondary. I have constructed Fresnel-based
systems at each extreme of this range and I have found a slight
advantage (10-15% overall far field intensity) at the short end of that
range - but this difference represents only a few dB of recovered
signal. One of the difficulties of a system with a very short F-Ratio
is that uniform illumination of the Fresnel itself becomes problematic
as the distance from the light source to the edges of such "short"
Fresnel lenses is significantly greater than the distance from the
center of the lens to the light source and naturally, falloff in light
intensity occurs toward the edges.
***
For aligning an LED-based Fresnel optical transmitter I set up a "test
range" over a distance of about 175 meters in which I place, at the far
end, an LDR (CdS cell) attached to a 555 oscillator feeding the audio
input of a low-power FM transmitter (a Handie-Talkie on a 40dB
attenuator, actually) over which an audible tone would be transmitted:
The higher the frequency of the tone, the more light being detected. By
using a laptop running the Spectran program I can make repeatable
measurements of relative intensity by measuring the frequency of the
tone, and by varying the current of the LED being tested by known
amounts, direct ratios of tone pitch and light intensities could be
established, allowing indirect comparisons between different
arrangements and LED transmitters.
At a distance of 175 meters it seems that this is sufficient to be
deemed "infinite" for purposes of focusing a Fresnel/LED system: This
isn't quite true, but the actual difference seems to be only a fraction
of a dB and the extra difficulty in testing over a much longer range
wasn't considered to be worth it. (30 meters, on the other hand, is too
short for proper focusing! A much longer test range would make it more
awkward to set up (having to traverse to opposite ends of the range)
while very far distances (e.g. several kilometers) may make the path
more subject to the vagaries of atmospheric attenuation and
scintillation, thus complicating measurements.
As it turns out, optimal far-field flux occurs coincident with the
"sharpest" image of the LED die being projected. For LEDs like the
SST-90 in which the emitter is on a flat plane, this is fairly easy to
do, but for other types of LEDs - particular the "TIP" (Truncated
Inverted Pyramid) such as the red to yellow members of Luxeon family -
there is a degree of ambiguity as there is not single plane from which
light is emitted: In the case of these LEDs the optimum far field flux
seemed to occur somewhere between achieving a sharp image of the "top"
of the LED and its reflector cup although the difference isn't all that
great (only a few percent.)
***
Having tackled that topic, a brief word on the lenses themselves.
I've had the opportunity to "test" a number of Fresnel lenses of
different manufacture, sizes and F-ratios to determine how "good" they
are. While it is well known that Fresnels would be a poor choice if the
intent was to use them for imaging, the question is "How good/bad are
they?"
To answer this question, I first did some analysis using the star Vega -
which, at the time, was conveniently very close the Zenith at a
reasonable hour. Using this "infinitely-distant" and, for all practical
purposes, "point source" of light, I determined that most Fresnel lenses
would produce a "blur circle" of under 1mm diameter.
Because stars are rather awkward to use (they "move", they are fairly
dim - which makes initial setup rather difficult, they can be at awkward
angles, and one as at the mercy of weather) I decided to make further
measurement using another system.
For this, I drilled a 0.25mm hole (yes, I have a 0.25mm drill bit) in a
piece of 0.16mm sheet brass and placed behind it a high-brightness red
LED. The LED unit - the current of which could be switched between
"high" (about 60mA) and "low" (about 1mA) was placed a reasonable
distance away (about 13 meters). The "high" setting allowed quick setup
as it was bright enough to project a red spot onto a piece of paper to
determine aiming and focusing of the Fresnel while the "low" setting
avoided saturation of the camera's imager. The use of an LED provided a
monochromatic light source to minimize chromatic aberration which, with
white light, can make precise focusing more difficult. While this light
source was neither a true "point source" nor was it infinitely distant,
it was suitable for analysis of lenses of the quality of the Fresnels -
and comparisons between results obtained using the star Vega and the
LED-based test range using the same lens tend to bear this out.
To determine the size of the blur circle, a CCTV camera was used with
its CS lens removed. The imager of this camera was of known size and
thus, the dimensions of the individual pixels of the digitized picture
could be calculated. Light from the test source was optimally focused
directly onto the CCTV camera's imager and digitized so that the result
could be analyzed later.
The result of this testing was that almost every Fresnel lens produced a
blur circle with a size that was less than 1/1000th of its focal length
- and some where much better than this. In short, it would appear that
the size of the blur circle of an optimally-focused lens or reasonable
size (e.g. at least 200mm on a side) and focal length was well under 1mm
diameter. Unlike with conventional lenses, Fresnels don't produce a
nice, fuzzy circle (much less, diffraction rings!) but rather a
less-defined blob with streaks emanating from it, so the *exact*
determination of the size of the blur circle and the area over which
most of the energy is focused is a bit of guesswork.
This has several implications for a Fresnel-based transmit/receive
system:
- If the size of the emitter is significantly greater than that len's
blur circle, it is likely that the divergence of that system will be
most strongly determined by the ratio of the emitter's size and the size
of the aperture, rather than the quality of the lens.
- One would probably NOT use a detector with an active area less than
the size of the blur circle. Practically speaking, to intercept the
most energy and make aiming a bit easier the diameter of the detector
should be at least 2-3 times that of the "blur circle." One may not
want to use too-large a detector, as its capacitance can submerge weak
signals into the noise and the field of view may be unnecessarily
enlarged.
In these tests, several things were noticed:
- Overhead projector lenses are NOT good candidates for this use.
Simply put, these aren't usually designed to focus at infinity, so they
produce VERY poor results. If you have a Fresnel from an overhead
projector, don't bother with it UNLESS you have thoroughly tested its
performance!
- Larger F-ratios and/or focal length imply larger detector sizes for
optimal performance. Such lenses - since the grooves need to be made
with greater precision in order to provide a sharp focus at the greater
distance - tend to produce wider blur circles than "shorter" lenses.
Since these lenses tend not only to be more expensive and require a
larger physical structure to support the lens and accompanying
electronics, they aren't recommended for such use, anyway.
- This lenses with fine groove pitch (e.g. more grooves per millimeter)
appear to have higher scattering losses than those with "coarser" groove
pitches. The resulting blur circle of the "finer" lenses were slightly
"fuzzier" and it is likely possible that scattering losses would be
higher. Examples of lenses with "fine" pitches are those intended more
for imaging purposes (such as video screen enlargement) and those
imprinted on flexible vinyl.
***
These "page magnifier" lenses are relatively small, with a typical
"active" lens area of around 200mm x 260mm (although the size varies)
and often (but not always) include an "un-grooved" border along the
edges. This size makes them fairly easy to use and they are, in fact,
large enough to provide reasonable performance. Unfortunately, they are
rectangular, which means that their overall efficiency is reduced
somewhat because - as determined from previous experimentation and
measurement - optimal far-field flux seems to occur when the "circle of
light" from the LED is adjusted to *just* hit the edges of the long
dimension of a rectangular lens - and this also means that a portion of
the light will be cut off in the "narrow" dimension and lost. (This
amount of degradation appears to be approximately proportional to the
difference in area to the rectangular lens and a square lens of the same
"long" dimension.) Considering how cheap these lenses are, however,
they are certainly worth considering!
One surprise was that the very cheap, rigid "page magnifier" lenses
available from stationary/office supply stores (and, in some discount
outlets) provide very, small blur circles on par with or exceeding those
obtainable from much more expensive lenses! As it turns out, these
lenses tend to have relatively coarse groove pitch - an advantage in our
case.
About these rigid "page magnifier" lenses, there are a few things to
consider, however: They seem to be available in two ranges of focal
length. Most have a focal length the 250-330mm area (although most seem
to be at the longer end of that range), and others are around 600mm.
Again, the "longer" lenses (with an F-ratio of 2 or greater) aren't
particularly useful for our purposes as that F-ratios is awkwardly large
(implying difficulties in obtaining optimal LED transmission efficiency)
and would require a fairly large assembly to support the lens and its
gear.
I am working on a web page that discusses comparative "quality" of
different molded Fresnel lenses - complete with pictures showing the
resulting blur circles. It will be another couple of days before I put
it online. However, the upshot of the tests is this:
- Using a good-quality Fresnel lens (including those rigid "Page
Magnifier" types with around 330mm focal length) you will be "safe" in
using a photodetector of 1mm diameter or greater. (The BPW34, for
comparison, is around 2.7mm on a side).
- This also implies that if the apparent source size of the emitter is
larger than 2-3 times the "blur circle" of the lens you use that the
primary contributor of the divergence angle of a given system will be
the ratio of the apparent source size and the aperture and NOT the
quality of the lens. For the lenses that I have tested, that would mean
that as long as the apparent source size was 1mm or larger, you were
"source-size" limited rather than "lens-quality" limited in the amount
of divergence.
73,
Clint
KA7OEI
> Hi Jim,
>
> we were a bit hurry, because I finished the optics and driver a day
> before the test. Optical alignment is not optimal. I need to do more
> iterations to find optimal position of Fresnel and condenser lens. But
I
> think we rather make it in new equipment.
>
> Beam width is about 100mRad (2m spot at 20m distance). We know that it
> is very bad. But there is way how to deal with it. We need to have
> better lenses and beeter mechanical enviroment. Clint KA7OEI has good
> results - his beawidth is less than 10mRad.
>
> About the way of the communication - Robo saw the red light visualy
and
> he heard it very weak - audio signal.
>
> Kind regards
>
> Rado om2zz
>
>
______________________________________________________________
Laser mailing list
Home: http://mailman.qth.net/mailman/listinfo/laser
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:Laser at mailman.qth.net
This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
More information about the Laser
mailing list