NavStrobe History

Results as of March 4th, 2003:

Executive Overview:

First let me say that it is possible to build LED based Nav Strobes, BUT...

The current technology in LEDs is still about a generation away from feasibly meeting the requirements of Aircraft NavStrobes, in either cost or performance.  You can meet the performance (and form factor requirements) of a common set of aircraft strobes (Whelen or Aeroflash) with a set of 60 Luxeon 5-Watt Portable White Emitters, but it would cost you $2,760 in LEDs alone; hardly a cost effective solution.

You can meet the cost requirements to some degree with a field of 318 White www.superbrightleds.com LEDs, but it would still cost you around $636 for the LEDs alone, and to accommodate a minimum of 600fc (derived as a minimum output from the testing of some Aeroflash Strobes) the form-factor grows to 6.5"L x 3.5"W x 2.5"H with a lot of LED driver circuitry to make it all work in a manageable fashion.  This design would take so much assembly time and would in my opinion be fragile (lots of LED drive circuitry to break), that it really is not feasible in any sort of volume.

LED Position Lights are Possible, but at near $100 per assembly, I don't see them as a cost-effective replacements for incandescents.

Problems:

Initially, I had hoped to 'overdrive' the Luxeon's to increase their light output (see the NavStrobe History for increasing the visual brightness of LEDs).  Since the Luxeons would only need to be 'ON' for about 50-milliseconds out of every second for the strobe (a 5% duty cycle) they would only produce 5% of the heat that a 100% duty cycle would produce.  So, the hope was that I could overdrive the LEDs by 500%, but for a very short period.  Alas, this is not possible with the Luxeons as they are very near their Thermal Run-away Limit operationally, and pulsing high currents, even for short duty-cycles at less than 100-microseconds will damage the device.  The Thermal Run-away Limit is the internal current limit where the LED stops producing light and only produces more heat, leading to a thermal run-away in the die that will eventually damage the LED.

The reason that I started investigating the Luxeons more thoroughly is that I was not happy with the 'Chia-pet' style LED Array as I saw it as too labor intensive to build and too fragile (too many parts to break!).  That and the form-factor issues made me decide that it just was not commercially viable to build.

So where are we?  One of three things need to happen to make LED Strobes commercially feasible:

Solution 1:

Luxeon Emitters fall in price to under $5 (they are currently $23.00 in quantities of 100) and their static output increases from the current 21fc per LED to 60fc.
(Hint: Since Luxeon's have just been released at the 5-watt (21fc) level this will probably take awhile.)

Solution 2:

Luxeon Emitters' 'Thermal Run-away Limit' is raised so they can sustain 5X over-current.
(Again, Since Luxeon's have just been released at the 5-watt (21fc) level this will probably take awhile.)

Solution 3:

www.superbrightleds.com White LEDs have a 3x increase in performance (and 'Thermal Run-away Limit!) while they maintain less than $1.30 per LED.
(I'll be watching their products, but it will probably be a generation or more before this happens.)

Conclusion:

We are not there yet (duh), but as soon as I see one of the three possible solutions above happen, then I will revisit the issue.

December 2002:

I have been thinking about LED strobes recently and I went through an evaluation exercise to see if something like this could be built....I think it can.

I have been discussing this concept with the Cozy-List, and I've think I've accounted for most of the major design issues.

The design below is based on 141 Super-Bright White LED's and 6 Super-Bright Red or Super-Bright Green LED's from www.superbrightled.com.

Each White LED puts out 6000mcd (the minimum guaranteed is 5000mcd), which I hope to visually boost by a factor of 2 to 3, by pulsing them with 62ma of current at a 30% duty cycle.  (For more on this technique, see Jim Weir's article here.)  This should produce a strobe of greater than 438cd output, worst case.

Note: For now this is an experiment.  Any number of unforeseen problems could cause it's demise.  If you have strobes, keep them!  If you are about to buy strobes and are 6-months away from flying; buy the strobes!  

But, if you have some time to decide on whether this could work, then come back to this site, and I will try to keep you updated on my progress.

The Current State of Affairs

I have received my initial order of the Super-Bright LEDs and I am organizing my testing of these parts.  I need to test whether these parts will perform as I am anticipating (I don't want to build the PCB below until I can confirm whether the LEDs will work as intended.)

I intend to test the steady-state and pulsed outputs of these LEDs and graph their light output performance while driving them in the same manner as a strobe unit would be driven.  Once I have this data, I will then proceed with PCB Design and Manufacture, if warranted.

December 10th 2002 Update:

I have ordered a light meter that can record Light Intensity down to 100-microseconds.  That should be good enough to measure the output of these LEDs!

While I have been waiting for the light-meter to arrive, I prototyped a drive circuit to get an idea of how these LEDs might respond to pulsed current control.  The results are encouraging, but I will need the light meter to really be able to quantify the results.

So in the mean-time, (while waiting for the light meter) here are some JPEG's and Movies of the prototype in action.  The pictures and the movies don't really give you a good rendering of the intensity (the strobes are MUCH brighter than they appear in the pictures and movies) and the color of the Green is a much deeper Green than appears in the photo at left.  Also, in the movies, the flash SEEMS to change in intensity from flash to flash.  This does not happen in real life;  I believe it to be an aliasing effect between the frame rate of the web-cam at 30Hz and the pulsing of the LEDs at 2400Hz, with about a 133-microseconds 'ON' pulse width.

Now, when I receive the Light-Meter, I should be able to make some more objective measurements!

Movies

The Single LED Strobing
A Row of 10-LEDs Strobing
A Green Nav LED and a Row of 10 LED's Strobing

December 13th 2002 Update:

I had a minor 'mental heart-attack' when I realized that I had been using Candelas and Foot-candles interchangeably (hint: they AREN'T interchangeable!)  Candelas measures the intensity of Light AT IT'S SOURCE, while Foot-Candles measure the amount of light that falls within a steradian angle on the inside of a spherical surface 1 FOOT AWAY.  SOOO...1 Foot-Candle equals about 12.57 Candelas (there are approximately 12.57 steradian angles within a sphere); follow that?  So when I realized this, I immediately thought my solution was about 12.57 times too dim.  But...hold on, my light meter measurements put the output of the white LEDs right where I was previously expecting them to be (at around 5 foot-candles)...what is going on here???

I then realized an important fact.  AN LED's mcd (milli-candela) rating is measured by taking the total light output and dispersing it over the inside of a 1²-Meter surface area Sphere; this is then reported as that LED's (milli)-Candela rating.  But in reality, the LED actually takes its light and focuses it in a narrow 30-degree beam, so all the light in the milli-candela measurement that is spread over the inside surface of a sphere is now focused into a fairly narrow 30 degree beam.  How close the candela rating is to the actual foot-candle number depends on the focusing of the LED lens.  (Hint: 30 degrees seems to be close to the steradian angle, but I haven't done the math yet.)

So in the end, it meant that my spreadsheet was off, but probably not by too much.  Now I need to get the actual test results with the Light-Meter to see how much the above misconception on my part will hurt me; if at all.

Now on to the Light-Meter Measurements

As I stated previously, I started with one LED in a darkened room and took some measurements.  My first test was with a single LED at 20mA of steady state current.  This gave me a result of 5 foot-candles...so far so good.  The minimum candela rating guaranteed by the manufacturer was 5cd and as it turns out, the 30-degree focus angle of the LED is close enough to the steradian angle that in this case the candelas are roughly equal to the final foot-candle number (for 30 degree LEDs anyway...I may have to redo my numbers for the Green LED which has a 45 degree focus angle).

Something's wrong capt'n! 

I then started my pulse testing of the LED's with a 62mA pulse at a 30% duty cycle, and the Light-Meter is reporting 5fc.  Hmmm....Something is wrong here.  I am now running over three times the current and I can see that the LED is visibly brighter, but the Light-Meter is reporting only 5 foot-candles (the same as the 20mA steady state current test).  I then decide to experiment with the Light Meter a little.  I try a 50% duty cycle and I get 9fc.  Okay... I haven't changed the intensity of the LED (visually it looks the same as the 30% duty cycle and the current through the LED IS the same at 62mA.)  So, maybe the light meter is averaging its samples?  To test whether this is the case, I decided to go back to the 20mA steady state current (i.e. 100% duty cycle) test and measure again (yep, 5fc).  Then I give it a 50% duty cycle at 20mA (visually, the LED still appears to be the same brightness as the 100% duty cycle; again see Jim Weir's article), and the meter now reports 2.7fc.  Okay, so it looks like the meter is averaging.  This is going to make the  testing a little more computationally intensive as I have to think in terms of duty-cycle and flash time (for example, this light meter will report a 100fc 0.5-second, 50% duty-cycle flash as 25fc.)  

Perhaps what I really need is a Flash Meter, but those tend to be tuned to photographic film's spectrum response, not the Human Eye's (which is why I went to the Light-Meter approach; its spectrum is tuned to the Human Eye).  This light meter will respond to a flash as short as 100 micro-seconds, BUT it then averages that over a one second period (so a 100uSec 10000fc flash is recorded as 1fc!!!)

Please bear this in mind when you look at the table below.

The Light Meter Test Results

As you can see from the above table, I was getting near my expected visual peaks (after I accounted for the Duty-Cycle averaging of the meter).  Test 3 looks to be the best approach for the final circuitry, as it produces the visual intensity I need (which is a minimum of 16fc per LED peak) and it lives within the maximum current ratings of the LED (which is 70mA peak).  I will probably reduce the duty-cycle to under 40% to keep the keep the average current under 30mA, which is recommended in the LED spec-sheets.

*Note: Test 4 and Test 5 are not recommended as they violate the maximum current spec and therefore might result in reduced lifetime ratings.  (Although they were nice and Bright!)  I also noticed with Test 5 that the light output turned a definite blue color.  I suspect that either the quantity of light was too high for the phosphor in the LED (and hence the phosphor was saturated, resulting in the leaking of shorter wavelengths) or perhaps the additional heat was affecting the chemical properties of the phosphor in some manner.  In any case, when I returned to more acceptable current levels, the light output returned to its normal white color.

So, it seems the LEDs are bright enough, but I may want to double check these numbers with a flash-meter, just to be sure.  (Anybody got a high end digital light meter that can measure down to 100 micro-seconds?)

After some double checking, I'll get on to the task of building a full test board (with all the required LEDs) to compare with a set of Whelen Strobes on a friend's Velocity.

December 22nd 2002 Update:

I decided that I should test some of the light distribution patterns of the LEDs; just to make sure that I'll meet the requirements of the FARs, and after some intensive testing which included testing one white LED to destruction (sometimes you just gotta know where that corner is!), I am much more confident that this thing can be built.  The light distribution pattern is much tighter than I initially assumed, but its intensity doesn't fall off as rapidly as I assumed either; so the net result is these two factors cancel each other out.

I've also decided to redo the Concept PCB as a result of a conversation that I had with my PCB vendor.  It turns out that they can print PCBs down to 8-mils thick (for comparison, take two pieces of 4-mil plastic sandwiched together...that is the thickness of this PCB!)  As a benefit, these super thin PCBs can be bent along a curve...this should make the LED mountings much easier.  All I need to do is make a form to bend the PCB to the required angles and then solder on the LEDs.  I'll come up with a new CAD model after Christmas and update the web-page then.

As for physical mounting, I've purchased some wingtip covers from Wicks that may do the job.  I am designing the layouts of the PCBs to be able to fit in the space within the wingtip cover.  Again, after Christmas, I'll update the web-page with my progress.

December 27th 2002 Update:

I have completed the new PCB CAD File and I have included the images below.  I have left a picture of Version-1 at the top for comparison.  The reasons for the change are manufacturability and robustness.  I felt that the "Flying LED" design of the Version-1 Concept was too fragile.  I also needed to accommodate some additional LEDs to solve some light distribution problems.  This meant that the concept grew in height (from 1.6 to 2.45 inches) and length (from 4.5 to 6.43 inches) but stayed nearly the same width (2.4 versus 2.5 inches.)  

The downside to all of this, is that it will not fit into the Wingtip covers that I bough from Wicks (the covers narrow too rapidly to accommodate the LEDs at the top of the "hump".  So a custom cover will have to be made.

I have also updated the Spreadsheet to include the new LED performance numbers I gathered during my December 22nd Testing (see above.)

Marc Zeitlin's Superposition of LEDs Question

When I first started discussing this concept with the Cozy Mailing List, Marc Zeitlin posed a question about whether the superposition of the LEDs would appear as bright as a point source strobe...to quote:

Do we know/believe that superposition holds for the case of lights, especially LED lights? 
In other words, is it definitely the case that 2 LED's of x lumens each are equivalent to one 
LED of 2x lumens, and that 100 LED's of x lumens each is equivalent to one LED of 100x lumens?

Hmmm.  Good question.  So in the back of my brain I started looking at the strobes on our planes.  

The most useful observation I had actually occurred at dinner with my wife at Pilot Pete's (a great restaurant at Schaumburg Airport here in Illinois).  While we were eating dinner, I was watching various airplanes taxi, take-off and land, and since it was night, the strobes were on.  What I discovered is that the strobes we use are not point sources in the strictest technical sense.  The view I had (from about 100 feet away) was that the strobe tubes actually lit up the entire glass focusing cover so that it appeared that the light was coming from the entirety of the glass cover (this was probably an effect of the focusing lens of the cover which is put there to try to direct most of the light outward in a horizontal direction to meet the requirements of the FARs.)  So this 'point source' was actually about 1.25 inches in diameter (when viewed abeam the plane) to about 2 inches long (by 1.25 inches thick) when viewed from a Fore or Aft position.  (Yes, I WAS SEEING AFTER-IMAGES!)

So, if this is the case of the Xenon strobes, then if I keep to roughly the same viewing dimensions, then given the same light output, they should appear to be the same brightness...right?

Now how to test this?  The final test of course, will be with the first working prototype, but I wanted to get an idea of how LED superposition might look from a distance.  So I arranged my lab proto board with Ten flashing white LED's out of my den window and went and took some pictures from the back yard.  I kept the light output down to about 80fc (I wanted to test LED superposition, not brightness!  I have already verified what these LEDs can do and I have the after-images to prove it!)

The photo on the left is from 100 feet away and the one on the right was taken at 50 feet.  In both photos, the bluish light in the center of the photo is the 10-LED strobe taken through my den window.  The 'vertical' appearance of the strobe is due to the vertical poles on my deck collimating the light, not the LEDs themselves.  My own observation was that I could not discern a clear 'vertical' shape to the LED's until I was closer than 50 feet.  Now with more LEDs you may have to be a little further away before they 'integrate', but I don't imagine that you would have to be too much farther away...the inverse square law is actually helping you a little here.

The blue color is more an artifact of the camera.  To my eye, the light appeared mostly white with a tinge towards the blue end (similar to the light a Xenon strobe produces.)

These tests seem to show that superposition won't be a problem, but I still plan to have a head to head (er, strobe to strobe?) comparison with a Whelen setup.

February 8th 2002 Update:

I finally got around to testing the light output of some Aeroflash Strobes that I borrowed from a friend...WOW!  The tested (and confirmed) peak was 1,100fc!!!

This put the problem of LED strobes in a whole new light for me (pun half-heartedly intended  ;-)

Producing an LED strobe that would meet the FARs (400fc) was hard enough, but now to find that a common set of strobes produced almost three times the FAR requirement really put it in perspective.  I could meet 500fc to 600fc with the current LED setup, but the cost and size (it already looks like an LED Chia-pet!) really get out of hand for 800 to 1000fc; which is where I think this setup needs to be to be competitive with Xenon strobes.

So where to go from here?  All is not lost.  I returned to a high-power LED product that I previously rejected due to problems with availability.  The availability of this part has eased up some, but it is still a wait (current quotes are in the 9-week range for product delivery.)  In the mean time, I have ordered some in-stock samples that should have the same electrical/luminosity characteristics (except they are Green instead of White.)  Once these samples come in, I will do an evaluation to see if they can meet the new required light output of 800fc. I think I can make this work, but I don't want to get too far ahead of myself until I have tested a working sample.

The upside of this is that the final unit (if it works) will be substantially smaller than the "LED Chia-Pet" concepts of versions 1 and 2.

I have included a sample PCB concept that would measure 1.5" Wide X 4.5" Long and 1.25" High.  I am calling this Version 3 for obvious, historical reasons.

After my testing, I will update this web page with my results.

February 21st 2002 Update:

The testing of the Luxeons was not very encouraging.  They have a very wide dispersal pattern and measured 21fc with the light-meter, but they can't be overdriven to increase their brightness.  Also, at 21fc of output, you would need 60 Luxeons to achieve a 600fc strobe flash...at $23 per Luxeon, this is not cost-effective.

Initially, I had hoped to 'overdrive' the Luxeon's to increase their light output (see the Jim Weir's article here for increasing the visual brightness of LEDs).  Since the Luxeons would only need to be 'ON' for about 50-milliseconds out of every second for the strobe (a 5% duty cycle) they would only produce 5% of the heat that a 100% duty cycle would produce.  So, the hope was that I could overdrive the LEDs by 500%, but for a very short period.  Alas, this is not possible with the Luxeons as they are very near their Thermal Run-away Limit operationally, and pulsing high currents, even for short duty-cycles at less than 100-microseconds will damage the device.  The Thermal Run-away Limit is the internal current limit where the LED stops producing light and only produces more heat, leading to a thermal run-away in the die that will eventually damage the LED.

The reason that I started investigating the Luxeons more thoroughly is that I was not happy with the 'Chia-pet' style LED Array as I saw it as too labor intensive to build and too fragile (too many parts to break!).  That and the form-factor issues made me decide that it just was not commercially viable to build.

Since this was the last avenue to explore, I am ending my experiment here.  Read the opening page for my conclusions.

I'll leave the proof of concepts below for your edification (it may benefit someone!)

Version 3 Concept

The CAD Concept PCB

Version 1 Concept

Here is a picture of the Version-1 CAD concept model (the PCB measures approximately 4.5" x 2.5" with the 4.5" axis being parallel to the longitudinal axis of the plane):

Version-2 Concept

The Version-2 Concept takes advantage of a super-thin PCB (8-mils thick!) that can be bent around a form and then soldered to.  The interior form is 0.068" aluminum plate.

Concept Evaluation Spreadsheet

I've also updated the Excel spreadsheet:

Nav_Strobe_Evaluation.xls