About this project:

Several years after building the current-linear modulator for high-power LEDs around Luxeon III LEDs, very high-power devices - called "Phlatlights"^tm produced by Luminus Devices became available - namely the CBT-54.

Compared to the Luxon IIIs that had been successfully used for long-distance optical communications of over 273 km, the specifications of the Phlatlight devices were very impressive:  They could operate with good efficiency with currents of over 2.5 amps/mm².  With a die size of 5.4mm², the LED could safely handle a continuous current of at least 8.5 amps when properly heat-sinked, providing unprecedented levels of light output from a fairly small source area. 

Upon obtaining several of these devices I ran some tests and concluded to my satisfaction that the claims were true.  For optical communications, however, it takes more than just lighting up an LED:  One has to modulate it, but what was a "safe" amount?  The spec sheets showed that they could be driven to 8.5 amps continuously and to 13.5 amps with a 50% duty cycle - but other than a graph on the data sheet that went slightly above 15 amps, there was no mention of a "maximum peak" current.

From one of the documents that I'd found on the Luminus web site (alas, since removed...) I spotted a note that they'd been doing long-term endurance testing of Phlatlight modules at continuous currents of 2.5 amps/mm² and had observed no failures - the same current density as 13.5 amp figure quoted in the data sheet for 50% duty cycle drives.  Bolting a device to a heat sink, I did careful testing at approximately 22 amps and found that, with an "infinite" heat sink, and operated the device for 30 seconds while noting the temperature reported by the CBT-54's on-board thermistor.  After being relieved that the device survived just fine, I crunched the numbers and determined that even being overdriven, the calculated LED die temperature was below the "maximum" noted in the specification sheet (125C for the red emitters.)

Comment:

• I did manage to destroy one CBT-54 by accidentally connecting it to a 2200uF capacitor that was charged to 13.5 volts, causing all of the bond wires connecting the cathode of the LED to be burned open.  The total amount of peak current was unknown, but apparently well over 20 amps!
  

This told me what I needed to know:  The devices were going to survive a resting current of 8-9 amps with peaks to 18-22 amps at 100% modulation, a fact that has been field-tested with hours of operation.

Now, what does one use to drive an LED to 22 amps and do so efficiently so that there was some hope of being able to run the modulator from "luggable" batteries?  A new modulator had to be designed.

In order to linearly modulate the intensity of an LED it is necessary to vary the amount of current flowing through the device rather than the voltage across it.  It is fortunate that the luminous output of an LED is linearly proportional to the current flowing through it:  At higher currents, the "current versus light output" curve "flattens" a little bit, but this results on only a very small amount (a few percent at most - see the sidebar below) of distortion and is unnoticeable in voice communications.

Figure 1:
Schematic of the high-compliance current sink, used as a high-power LED modulator.
Click on the image for a larger version.

Notes:

• When you use an LED other than what is mentioned below, please take into account the current ratings of the devices that you plan to use.

• The circuits shown on this page are quite adaptable:  With the techniques described below, I have constructed modulators for LEDs with current ranges from just a few milliamps to 10's of amps!

• Please heed the comments on the use of the LM324 for these circuits - see the sidebar, below.

• The CBT-54 was discontinued in 2009 or 2010.  The equivalent device is the PT-54.  A smaller device with the same current density is the CBT-40 which is arguably better-suited to point-to-point optical communications.
  

First, a bit of explanation as to how the basic circuit works.

The "Precision Current Sink":

The basic current sink circuit may be seen in Figure 1.  This circuit uses a single section of an operational amplifier "wrapped around" a transistor and using a current sense resistor and it works thusly:

• Assume that 1 volt is applied to the non-inverting input of the op amp at Vin.

• If the voltage across Rsense is lower than 1 volt, the output voltage of the op amp is increased, turning FET - Q1 - on until enough current flows through Rsense to achieve a voltage drop equal to that of Vin.  Ohm's law tells us that if Rsense is equal to 1 ohm, we will achieve a drop of exactly one volt when the current through Rsense is one amp.  Likewise, if the current is higher than one amp the op amp will reduce the drive to Q1 to reduce the voltage drop across Rsense until it is one volt.

• As the voltage applied to Vin changes, the op amp will change the drive to make the current through Rsense proportional to Vin.  Because the drain current through Q1 is the same as the source current (ignoring the minute amount of gate-source leakage) we know that the current through the LED's current will be the same as that of Rsense.

• Cfilt is important in that it provides a low-impedance source current.  Without it, audio would be imposed on the power supply leads and get into other circuits.  Also, there is no guarantee that the power source - be it a battery or power supply - will provide a low impedance path at all frequencies - especially if fairly long power leads are used.  For high-power (>1 watt) LEDs, a recommended value is at least 1000uF.
  

One of the most important things to note about this circuit is that LED and supply voltage is irrelevant - provided, of course, that the supply voltage (V+) is sufficient to overcome the voltage drop across the resistor, LED and the resistance of the FET.  In other words, the voltage drop across the FET could be zero volts or 5 volts and as long as the supply voltage high is enough to overcome the voltage drops across the circuit's components (7 volts would probably be adequate) it doesn't matter!  If you use much more voltage for your power supply (V+) than the LED and drop across Rsense require, you'll be excess power in Q1 as heat:  As long as Q1 is properly rated and heat-sinked, it will work fine, but such operation may be wasteful - particularly if you are using battery power!

To make this circuit practical for use as a linear modulator for speech, several refinements need to be made:

• A fixed DC voltage needs to be applied to Vin to establish an "idling current" for the LED.  This idling current would be set to half that of the expected peak current - a value that would correlate with "100% positive modulation."  For the Luminus CBT-54 and PT-54, this would be 8.5-9 amps and for the CBT-40 and PT-39, this would be 6.25-6.75 amps - values that correlate to 1.6-1.7 amps/mm².

• Modulation needs to be applied.  The voltage at Vin would, in addition to the fixed DC voltage, have the audio modulation applied to it.  In this case, this would be a peak-to-peak voltage twice that of the fixed voltage.
  

• A few additional capacitors need to be added to make the modulator stable under varying current conditions.

Figure 2:
Schematic of the "simplified" high-power LED Linear Modulator.  The circuit depicted is for demonstration purposes and it is recommended that if you wish to build a modulator, construct that depicted in Figure 3 (below) instead.
Click on the image for a larger version.

Circuit description of a simple, high-power modulator:

The modulator shown in Figure 2 is a refinement of that in Figure 1, but with added components to ensure stability and allow easy adjustments of maximum current and idling current.

The main difference is in the current sense resistance in the source of the FET, R8.  Since up to 22 amps are going to be conducted, a lower-value resistor is appropriate to maintain operational headroom and with the 0.1 ohm value shown, we could expect to see 2.2 volts across it at that current.  Another component that was "scaled up" was C6, a 6800uF capacitor that provides instantaneous current during modulation peaks - something that may be difficult to deliver due to resistive losses in wiring from the battery - as well as improving overall circuit stability.

For a "reference", U1 provides a stable, clean voltage source that is used to set the idle currents by biasing line and microphone amplifiers at 5 volts.  There are also to potentiometers:  R5 is used to set the maximum idle current when R3 is set to "full" while R3 is used to adjust the idle current from the maximum value set by R5 all the way down to zero.  The advantage of this scheme is that the applied modulation depth stays constant at all settings of R3 making it very easy to adjust the power level as necessary for the link, to conserve battery power and to run link margin tests.  Note that if operated from a 12 volt source, it is possible for the positive modulation excursions to briefly exceed 100% (e.g. >10 volts at the output of U2c) but being AC-coupled, these peak currents cannot be maintained and are unlikely to ever cause damage to the LED.

Also provided are switches SW1 and SW2.  SW1, when open, effectively disables the LED while SW2 - a momentary push button type - allows easy on/off keying of the LED when SW1 is open:  If a tone-modulated audio source is applied to the audio input, this allows sending of Morse by interrupting the light.

U2c amplifies "line level" audio (from a computer or portable audio player) to the 10 volts peak-to-peak required to achieve 100% modulation while also functioning as a mixer for the audio from U2d, an amplifier that can take inputs from line to microphone level (adjustable using R15) to the needed level.  U2b is a unity-gain buffer that takes a sample of the audio from U2a's input to drive either a headphone or the input of an audio recorder to allow monitoring of what is being applied to the modulator.

Particular attention should be noted to the connection of "Battery V-" to the rest of the modulator circuit.  Because of the currents involved, it is recommended that the high-current components (Q1, R8 and C6) be mounted on a heavy ground plane with the ground of the rest of the circuit (U1, U2 and associated components) be connected in exactly ONE place - preferably very near the grounded end of R8 - to maintain stability.

The voltage of Battery V+ will tremendously affect the overall efficiency of the modulator.  If Battery V+ is 12 volts, a significant (>30 watts) amount of heat will be dissipated by Q1 requiring a good heat sink with 8-10 watts also being dissipated in R8 as well and if portable, battery operation is anticipated, this represents a tremendous amount of wasted power.  If the situation allows for "power to burn" (e.g. line-operated power supply or an adequately-large battery - perhaps tied to a vehicle) then this extra heat might be put to good use to keep dew from forming on the optics or even to keep your hands warm!

Efficiency can be improved by reducing the Battery V+ voltage to the absolute minimum required.  Considering that the red Phlatlight will drop about 2.8 volts at peak current (22 amps) and R8's 0.1 ohms will drop 2.2 volts at this same current.  If we assume that we can reasonably pull 22 amps through the FET (Q1) at an "on" resistance of 0.05 ohm with its drop of 1.1 volts, we can see that we'd neeed only 6.1 volts and far less power would be wasted as heat.  If we drop the value of R8 to 0.025 ohms, we can economize still-further making operation (of the LED portion, at least) from a 6 volt battery practical.  (Note that the rest of the circuit would need higher voltage, but only a few 10's of milliamps would be required which could be provided by four 1.5 volt AA or AAA batteries in series with the 6 volt LED supply.)

Knowing that a high-current source of 6 volts is not particularly convenient, it would be advantageous to be able to derive the "Batt V+" voltage from a 12 volt power source that we (probably) already have, and for that we should use a switching voltage converter.

A high-current switching converter:

Figure 3:
Switching voltage converter for the high-power LED modulator.
Click on the image for a larger version.

See Figure 3

Comments about commercially available, ready-built DC-DC converters:

As with the modulator described above, the high-current components (input capacitors C101-C104, output capacitors C113-C116, the FETs Q101 and Q102 L102 and R103) are assembled using single-point ground techniques - preferably on a solid ground plane.  The rest of the circuit (U101 and related components) may be mounted on a small board nearby, connecting to the main ground plane at only one place, near the output capacitors' ground.

U101, the Linear Technologies LT1339 is a "universal" synchronous switchmode power supply controller whereby Q101 forms the series switch and Q102 acts as a low-loss (synchronous) rectifier in order to achieve >92% efficiency at high currents.   During initial testing, neither Q101 or Q102 had heat sinks, yet they only got slightly warm (but not hot) when the supply's output was loaded to 20 amps!  For the most part, the design of this circuit is very straightforward, taken from the device's data sheet.

One minor departure from the "standard circuit" is that of the circuits surrounding Q103-Q105.   If voltage is applied to the "Green Sense" input, Q103-Q105 are turned on, putting R108 in the circuit and raising the voltage.  This was done to allow both red and green Phlatlight LEDs to be used, with the red units needing just 4.2 volts at the drain of the modulator transistor while the green need a bit more the 6 volts.  (When a green Phlatlight is plugged in, there is a wire that feeds voltage to this point to "auto-sense" which type of LED is connected.)

Other notes:  R103 sets the maximum current to 20-22 amps, C108 sets the switching frequency to the 40-50 kHz area, D104 quenches the effects of Miller capacitance, and R106 sets the voltage when the "Green Sense" input is low (e.g. not active) and should always be adjusted before R108.  Also of note is the presence of R101, a 9 amp self-resetting fuse which, in conjunction with D101 and D102, provides reverse-polarity protection in addition to overcurrent fault protection.

Figure 4:
Top:  Minimized version of the high-power LED modulator.
Bottom:  A version of the high-power LED modulator with voltage and current monitoring.
Click on an image for a larger version.

An enhanced high-current modulator with monitoring:

See Figure 4 for the following discussion.

The circuits in Figure 4 are based on that in Figure 2, but with additional circuitry for amplifying the input audio source as well as monitoring what is actually being sent to the LED with the bottom diagram including a digital panel meter that gives the user an indication of the LED's idle current, modulator's operating voltage and the supply (battery) voltage - all helpful tools when actually operating the unit.

The description below references the lower diagram in Figure 4.

For monitoring voltage and current, a red LED digital panel meter was used to allow it to be seen in total darkness without affecting the dark-adapted eye.  In testing it was noted that the panel meter was almost painfully bright when viewed in total darkness but testing revealed that this meter - based on the venerable ICL7107 - functioned properly down to 3.2 volts with a significant reduction of LED brightness and to this end, D202 and D203 were used to drop the voltage to 3.6-3.8 volts.  Other modules (some not based on the ICL7107) were also tested and not all of them worked properly at much lower than 5 volts.  Depending on your preference, an LCD meter could also be used with the caveat that if you plan to use this at night, having a self-illuminated display is very helpful!  If your meter is too bright for night use and you can't easily lower the brightness any other way, a piece of darkened plastic could be placed in front of the display.

This meter is a typical "3-1/2 digit" type with no scaling resistors on its input and as such, will display a full-scale reading of "1999" with an input of 199.9 mV with the position of the decimal point being user-selectable with a jumper.  SW204 selects one of several resistive dividers used to monitor the battery input voltage, the LED voltage, and the LED current as sensed across R206-R209 with R224, R226 and R227 providing scaling for these inputs, respectively.  Because the least significant digit of the meter represents just 100 microvolts, it is important that wiring be done carefully with the ground reference of the low current portions of the modulator being connected to the high current sections in only one place - preferably near the ground end of the current sense resistors R206-R209 - to avoid inaccuracies due to ground loops.  Even with these precautions it is likely that slight negative biases may occur due to slight voltage drops in the connecting wiring so high-value resistors (1 Megohm to 22 Megohms as appropriate) are wired between the wipers of R224, R226 and/or R227 and the +5 volt supply as necessary to offset these effects - which are typically on the order of 100-400 microvolts.  It should be noted that the highest-order digit (the far-left "1") isn't used and the decimal point between the right-most two digits is activated (e.g. battery voltage is displayed as "12.6" which correlates to 12.6 millivolts being applied to the input of the meter.)

Switch SW201 - a center-off SPDT - selects either the internal audio amplifiers (U202c and U202d), an external modulator or neither (center).  The external modulator just happens to be the one described in Figure 4 of the current-linear modulator for high-power LEDs (a signal from pin 7 of U1b of this modulator via of 100 ohm resistor) which is at 5 volts resting current with 10 volts peak-to-peak of audio at 100% modulation.  This external modulator is preferred as it has a built-in audio compressor and tone generator and its use eliminated the need to replicate those circuits.  As with the other modulator, momentary push button switches are available (when SW201 is in the center-off position) to allow on-off keying of the LED.

Comments:

• Use only an LM324 on the circuits shown in Figure 4!  If another op amp is used, it must be either capable of operating down to the negative rail or a negative supply (of at least 2 volts) must be supplied.  The use of a different op amp will likely require that the circuit be modified to provide appropriate compensation so that the driver stage (U202a) is stable (e.g. does not oscillate.)

• The frequency response of the circuit shown in Figure 4 is pretty flat to 12 kHz, dropping to %80 maximum modulation at 15 kHz and down to 50% modulation at 30 kHz although the modulation can be driven higher than this if a bit of nonlinearity can be tolerated.  This frequency response is largely a limitation of the LM324 itself and higher modulation frequencies are possible with faster op amps but be aware that the highly capacitive gate capacitance of the FET and the rather large capacitance of the LED itself - both being properties that conspire against a stable circuit!

• As noted above, the drain voltage of the power FET driving the LED can be much higher than the voltage noted (4-5 volts for red LEDs, 6-7 volts for green and blue) but that this will tremendously increase the amount of heat dissipated in the power FET, requiring an appropriate amount of heat-sinking.
  

The upper-left picture in Figure 5 shows the finished modulator with the LED digital power meter and selector switch that allows the battery voltage, LED voltage or LED current to be displayed.  In the lower-right corner of the front panel may be seen the LED current control which allows adjustment from under 100 milliamps to full current (amps!) without affecting the modulation depth.

The upper-right picture in Figure 5 shows the internals of the modulator with the low-level electronics and metering being mounted in the lid and the high-current portions being built on copper-clad circuit board material using single-point grounding techniques.  What is not shown is RFI protection that was added after the pictures were taken which included putting large snap-on ferrite beads on all leads into and out of the modulator.  This was found to be necessary because of the fact that with the high-power switching converter, "hash" was radiated that decreased the effective sensitivity of a 2 meter handie-talkie placed nearby.  Another effective RFI minimization scheme was simply grounding the heat sink itself to the main, internal ground plane - a tactic which makes sense considering that the switching transistors are bolted to it!  Prior to adding the ferrites, the modulator also exhibited a degree of RF sensitivity where the idle current was affected by a close-by RF source, namely a handie-talkie being used next to the modulator.

Visible in this same picture is a small cooling fan that relieves heat from the large toroid visible as well as drawing air past the rear of the heat sink to provide a degree of extra cooling.  Worth mentioning is the presence of a "Fan V+" voltage on the power supply diagram (Figure 3) that provides an "isolated" power supply for both the small on-board fan as well as that on the LED module itself.  Without the 10 ohm resistor and capacitor (R113 and C116) it is likely that the "whine" of the fan will find its way into everything else - including the LED's own modulation!

Figure 5:
Top Left:  View of the front panel of the high-power modulator containing the switching voltage converter and the LED voltage/current metering.
Top Right:  Inside the modulator.  In the bottom of the box (lower) is the switching voltage converter on the right and the modulator transistor and current-sense resistors on the left.  In the lid may be seen the low-level modulator components (U202) and related circuitry.
Bottom Left:  green and red Phlatlight modules mounted to heat sinks with secondary optics.
Bottom Right:  A side view of one of the Phlatlight modules showing the heat sinking and the cooling fan.
Click on an image for a larger version.

The importance of maintaining 100% modulation:

Up to this point "100% modulation" has been mentioned - but not explained.  For the purposes of this discussion, 100% modulation refers to the fact that the LED's current varies about a "resting" current from zero to twice the resting current.  As it turns out, this range represents the maximum that the amplitude-modulated LED can be driven and avoid distortion of the original signal.

If you were to attempt to exceed 100% modulation, the most obvious side-effect would be that some portion of your waveforms would try to go below zero current - something that is clearly impossible - and that portion of the audio would be hard-clipped, causing distortion.  On the "positive" side of the modulation, it may be possible that the current could go above twice the resting current - at least until the op-amp, transistor and/or power supply ran out of voltage/current "swing" at which point it would "hard clip."  As mentioned above, the circuits described, when operated from 12-14 volts - a range that encompasses portable battery operation - can produce around 125% positive modulation.

Practically speaking, however, one can overdrive the audio circuit and cause a significant amount of clipping (perhaps 10-20%) and actually improve intelligibility somewhat under noisy/weak-signal conditions.  While the fidelity of the speech is obviously reduced, the impact on its intelligibility is generally negligible and the fact that the "peak-to-average" ratio is reduced (that is, the "quiet" portions of speech are louder by comparison than the loudest portions) and the overall "speech power" is actually increased and can be better heard when conditions are poor.

Exactly how much "clipping" is "too much"?  That's mostly a matter of taste.  Clearly, if one runs the audio too "hot" then excessive clipping will result in enough audio distortion to reduce intelligibility.  Again, it should be noted that the circuits in Figure 3 and Figure 4, when operated from a 12-14 volt supply, can be run into clipping without much fear of damaging the LED as they intrinsically limit the maximum amount of LED current to a reasonably safe value.  The same cannot be said of the circuit in Figure 2, however!

What about lower than 100% modulation?  If one doesn't fully-modulate the LED, the result is that the amount of audio being conveyed by the lightbeam is reduced and it will sound "quieter."  Where this becomes an issue is where the signal is already weak or competing with noise:  A badly-undermodulated (or "quiet") signal is at a significant disadvantage and is clearly not being used to its full potential.  As a matter of comparison, modulating at only 50% represents a 6dB drop in audio while modulating to just 30% is about 10dB!

In other words, in terms of overall effectiveness it's better to run the audio a bit "hot" and put up with a little bit of distortion than run it too low!


Final comments:

Since its construction in 2009, this modulator - and its twin - have seen use out in the field and aside from the aforementioned RFI problems - which have been mitigated - have worked flawlessly.  Phlatlight modules shown in Figure 5 have been fitted to my Foldable Enclosure and a similar red unit has been also fitted to my First Enclosure giving us a complete working pair of units with which we have conducted a number of experiments - including some in broad daylight!

More pictures will be added in the near future.

---

Return to the KA7OEI Optical communications Index page.

If you have questions or comments concerning the contents of this page, or are interested in this circuit, feel free to contact me using the information at this URL.
Keywords:  Lightbeam communications, light beam, lightbeam, laser beam, modulated light, optical communications, through-the-air optical communications, FSO communications, Free-Space Optical communications, LED communications, laser communications, LED, laser, light-emitting diode, lens, fresnel, fresnel lens, photodiode, photomultiplier, PMT, phototransistor, laser tube, laser diode, high power LED, luxeon, cree, phlatlight, lumileds, modulator, detector