Current-driving a powerful IR-illuminator array

[John Fields]

Ok I'm back after some actual lab-work

I studied the current-source design with an NPN with an emitter
resistor (Re) to GND and a string of 5 leds between Vcc (12V) and the
collector. I tried to make the dimensions so that Ve is around 500 mV
(to leave space for a Vce of over 1 V and the voltage drops across the
LED's which is around 2.1 V per LED at the currents I'm interested in),
so for a LED current of Ic=250mA, I chose Re=2.2 ohm. Given the drop
over the base-emitter port of the NPN of 0.7V, I'd have to have a Vb of
1.2 V.

Somewhere here I realised that at a beta of say 30-40, I would have to
support an Ib of around 10 mA (I actually measured this), which is
quite much. The trigger source is 5V and I have difficulties of getting
those 5V down to the required 1.2V. A resistor-based voltage-divider is
not good enough since choosing small R forces a too heavy load on the
trigger buffer, and choosing a high R makes the 15 mA create a too big
voltage drop (pulling the NPN out of the "easy" feedback model). So I
switched the transistor to a darlington NPN with beta > 750 giving an
Ib of about 0.2 mA (measured). Still this did not create a sane
configuration with the voltage divider model (the 0.2 mA creates a too
big voltage drop nonetheless).

I also tried various combinations of 1N4148 strings to drop the
voltage, but it does not work as I think it does (I get voltage drops
of only 530mV over each 1N4148 can that be right?). I tried both simply
putting the diodes in a series from the trig buffer (5V) down to the
transistor base. I also tried a resistor from the trig buffer to the
base, and a diode string from the base to GND, but this configuration
didn't work as expected either

I'd be happy to have some thoughts on this seemingly simple circuit
Perhaps adding a voltage follower (another buffer) at the input port to
support the higher currents is enough ? I'll try that in the lab I
guess...

---
Here's the LTSPICE circuit file for a circuit that works:

Version 4
SHEET 1 1308 680
WIRE -848 416 -848 352
WIRE -848 560 -848 496
WIRE -736 352 -848 352
WIRE -736 416 -736 352
WIRE -736 560 -848 560
WIRE -736 560 -736 496
WIRE -672 352 -736 352
WIRE -544 352 -592 352
WIRE -480 -160 -480 -208
WIRE -480 -32 -480 -80
WIRE -480 304 -480 -32
WIRE -480 560 -736 560
WIRE -480 560 -480 400
WIRE -432 -32 -480 -32
WIRE -304 -32 -352 -32
WIRE -240 -208 -480 -208
WIRE -240 -80 -240 -208
WIRE -240 48 -240 16
WIRE -240 176 -240 128
WIRE -240 320 -240 176
WIRE -240 560 -480 560
WIRE -240 560 -240 384
WIRE -176 176 -240 176
WIRE -16 176 -96 176
WIRE -16 176 -16 -896
WIRE -16 320 -16 176
WIRE -16 560 -240 560
WIRE -16 560 -16 400
WIRE 96 -736 96 -928
WIRE 96 336 96 144
WIRE 144 -928 96 -928
WIRE 144 -896 -16 -896
WIRE 144 144 96 144
WIRE 144 176 -16 176
WIRE 176 -944 176 -1280
WIRE 176 -512 176 -880
WIRE 176 -208 -240 -208
WIRE 176 128 176 -208
WIRE 176 560 -16 560
WIRE 176 560 176 192
WIRE 272 -912 208 -912
WIRE 272 160 208 160
WIRE 432 -912 352 -912
WIRE 432 160 352 160
WIRE 496 -960 496 -1008
WIRE 496 -736 96 -736
WIRE 496 -736 496 -864
WIRE 496 -688 496 -736
WIRE 496 -512 176 -512
WIRE 496 -512 496 -608
WIRE 496 -464 496 -512
WIRE 496 112 496 64
WIRE 496 336 96 336
WIRE 496 336 496 208
WIRE 496 384 496 336
WIRE 496 560 176 560
WIRE 496 560 496 464
WIRE 496 608 496 560
WIRE 576 -1008 496 -1008
WIRE 576 64 496 64
WIRE 704 -1008 656 -1008
WIRE 704 64 656 64
WIRE 832 -1008 784 -1008
WIRE 832 64 784 64
WIRE 960 -1008 912 -1008
WIRE 960 64 912 64
WIRE 1088 -1008 1040 -1008
WIRE 1088 64 1040 64
WIRE 1248 -1280 176 -1280
WIRE 1248 -1008 1168 -1008
WIRE 1248 -1008 1248 -1280
WIRE 1248 -208 176 -208
WIRE 1248 -208 1248 -1008
WIRE 1248 64 1168 64
WIRE 1248 64 1248 -208
WIRE 1248 304 1248 64
WIRE 1248 560 496 560
WIRE 1248 560 1248 384
FLAG 496 608 0
FLAG 496 -464 0
SYMBOL Opamps\\LT1006 176 96 R0
SYMATTR InstName U1
SYMBOL res 368 144 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R1
SYMATTR Value 2700
SYMBOL res 512 480 R180
WINDOW 0 36 76 Left 0
WINDOW 3 36 40 Left 0
SYMATTR InstName R2
SYMATTR Value 1
SYMBOL npn 432 112 R0
SYMATTR InstName Q1
SYMATTR Value 2N4401
SYMBOL res 672 48 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R3
SYMATTR Value 8.4
SYMBOL res 800 48 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R4
SYMATTR Value 8.4
SYMBOL res 928 48 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R5
SYMATTR Value 8.4
SYMBOL res 1056 48 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R6
SYMATTR Value 8.4
SYMBOL res 1184 48 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R7
SYMATTR Value 8.4
SYMBOL voltage 1248 288 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V1
SYMATTR Value 12
SYMBOL res -32 304 R0
SYMATTR InstName R8
SYMATTR Value 240
SYMBOL res -80 160 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R9
SYMATTR Value 2200
SYMBOL res -256 32 R0
SYMATTR InstName R10
SYMATTR Value 3300
SYMBOL pnp -304 16 M180
SYMATTR InstName Q2
SYMATTR Value 2N4403
SYMBOL res -336 -48 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R11
SYMATTR Value 10k
SYMBOL res -496 -176 R0
SYMATTR InstName R12
SYMATTR Value 10k
SYMBOL References\\LT1009 -240 352 R0
SYMATTR InstName U2
SYMBOL npn -544 304 R0
SYMATTR InstName Q3
SYMATTR Value 2N3904
SYMBOL res -752 400 R0
SYMATTR InstName R13
SYMATTR Value 10k
SYMBOL voltage -848 400 R0
WINDOW 3 -3 207 Left 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value PULSE(0 5 0 1e-6 1e-6 .001 .01 10)
SYMBOL res -576 336 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R14
SYMATTR Value 1k
SYMBOL Opamps\\LT1006 176 -976 R0
SYMATTR InstName U3
SYMBOL res 368 -928 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R15
SYMATTR Value 2700
SYMBOL res 512 -592 R180
WINDOW 0 36 76 Left 0
WINDOW 3 36 40 Left 0
SYMATTR InstName R16
SYMATTR Value 1
SYMBOL npn 432 -960 R0
SYMATTR InstName Q4
SYMATTR Value 2N4401
SYMBOL res 672 -1024 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R17
SYMATTR Value 8.4
SYMBOL res 800 -1024 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R18
SYMATTR Value 8.4
SYMBOL res 928 -1024 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R19
SYMATTR Value 8.4
SYMBOL res 1056 -1024 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R20
SYMATTR Value 8.4
SYMBOL res 1184 -1024 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R21
SYMATTR Value 8.4
TEXT -858 648 Left 0 !.tran 0 .1 0


It's basically a switched reference driving (if you want 40 lamps)
eight five-lamp constant current drivers.

With 5 lamps dropping 2.1V each that's 10.5 volts out of 12, which
gives you 1.5 volt of headroom to play with, which the 2N4401's and
current sense resistors soak up nicely.

I've modelled the LEDs as 8.4 ohm resistors, since:

2.1V
R = ------- = 8.4 ohms
0.25A


I've only shown two drivers and their loads, for convenience, but
the other six just connect across the 12V supply, with the + input
of all the opamps going to the reference voltage at the junction of
R8 and R9.

It simulates beautifully, as you'll see when/if you run it.

Post back if you have any questions.

 

[Rich Grise]

I wasn't commenting on the intelligibility, I was commenting on the
top-posting.

I used to be an adamant top-poster, but I've been saved! ;-)

Cheers!
Rich

 

[qrk]

Hi! I want to drive a fairly powerful IR-illuminator array using
high-efficiency IR-LED's (Agilent HSDL-4230 to be specific), which can
support continous currents of 100 mA and peak currents of up to 500 mA.
I want perhaps 40 of these.. and essentially I want to flash them all
in sync to an electronic camera shutter of around 1 ms width, with a
duty-cycle of perhaps 1-to-30. Now when googling around for suitable
circuits, most refer to relatively small power demands, with LEDs that
use a current of only a tenth of this.. both with resistors and with
MAX-circuits etc.

Would it be crazy to try to get the right current by the old
resistor-in-series trick ? Obviously running 20 amps continously
through some resistors would be crazy but here the duty-cycle is so
high that on average the current is only 20/30 amps..

Is there a better way of say switching the LED array with some
darlington transistors and having an additional circuit that monitors
the current and adjusts the current into the transistors to regulate ?

Regards,
Bjorn

A simple current mirror might work for parallel/series strings of
LEDs. You may need to thermally couple the tansistors to keep matching
reasonable.

 

[John Fields]

---
Geez, what happened?

Seems like I posted and the thread died...

 

[Rich Grise]

Don't take it too seriously, John. I've been killing threads for years.
;-P

Cheers!
Rich

 

[Jim Thompson]

I saw it. Noted that it works... as is usual for your posts.

So I left it to the dweebs to ask dumb questions ;-)

...Jim Thompson

 

[John Fields]

I saw it. Noted that it works... as is usual for your posts.

---
Damn, Jim, a compliment???

I must have died and gone to heaven...

Thanks!

 

[Jim Thompson]

---
Damn, Jim, a compliment???

I must have died and gone to heaven...

Thanks!

Have I ever criticized your circuits other than professional-level
questions?

...Jim Thompson

 

[John Fields]

Have I ever criticized your circuits other than professional-level
questions?

 

[Jim Thompson]

---
Yup. A while ago you used to denigrate them by referring to them
as "'designs'", with the quotes around them in a sentence.

But we don't need to go there, do we?

Nope.

...Jim Thompson

 

[BW]

I used the idea with the op-amp and power transistor, and it worked
fine! Of course, afterwards, I recognized the circuit, but it was
buried in my brain since it was about 10 yrs ago I studied this.. The
op-amp circuit has the advantage of reducing the dependence on Vbe and
Rb*Ib, and temperature (within sane limits).

I downloaded Eagle, and drew a schematic and made a PCB. Really good
program!

Now, for the next version of this, I'd like to reduce the number of LED
chains by making some kind of step-up power supply which can take the
12 VDC input and load up the 10,000uF capacitors to a higher voltage
during the low duty cycle, so I could put more LED's in each chain. But
for now, this will do fine.

I wonder how high pulsed current these LED's can sustain. In the
datasheet it only says max 500 mA pulses of 100 us duration. I need to
do at least 1 ms duration, but there are no curves showing max current
as a function of pulse-duration (I've seen those curves in other IR
LED's datasheets). So I put it at 300 mA arbitrarily BTW anyone know
any IR LED's that are more powerful than the Agilent HSDL-4230 ?

Regards,

Bjorn

 

[qrk]

I wonder how high pulsed current these LED's can sustain. In the
datasheet it only says max 500 mA pulses of 100 us duration. I need to
do at least 1 ms duration, but there are no curves showing max current
as a function of pulse-duration (I've seen those curves in other IR
LED's datasheets). So I put it at 300 mA arbitrarily BTW anyone know
any IR LED's that are more powerful than the Agilent HSDL-4230 ?

Regards,

Bjorn

I've done some tests with the HSDL-4400 recently. At a 1 kHz rep rate,
on time = 2 us, current = 9 Amps, I ran the LED for 1.8 million pulses
without problem. The same LED went through a battery of tests from <1A
to 9A without degredation. Probably around 5 million flashes total
until I cooked the poor thing above 10 A and a high duty cycle. I
found that you don't want to run these LEDs at much more than 2 Amps
as the efficiency gets pretty bad. 1 Amp seems like a nice compromise
for a 2 us pulse at 0.2% duty cycle. For your parameters you need to
do some tests to see where the intensity starts dropping off due to
heating. Find a PIN photodiode to use as a receiver and start running
some tests.

 

[BW]

I've done some tests with the HSDL-4400 recently. At a 1 kHz rep rate,
on time = 2 us, current = 9 Amps, I ran the LED for 1.8 million pulses
without problem. The same LED went through a battery of tests from <1A
to 9A without degredation. Probably around 5 million flashes total
until I cooked the poor thing above 10 A and a high duty cycle. I
found that you don't want to run these LEDs at much more than 2 Amps
as the efficiency gets pretty bad. 1 Amp seems like a nice compromise
for a 2 us pulse at 0.2% duty cycle. For your parameters you need to
do some tests to see where the intensity starts dropping off due to
heating. Find a PIN photodiode to use as a receiver and start running
some tests.

Interesting. So what you are saying is that if I hook up a sufficiently
linear (or calibrated) photodetector, I could measure how the LED
behaves when scanning over duty-cycles and/or currents, and where it
starts going non-linear is probably where I don't want to be because it
goes non-linear as an effect of heat buildup at the junction (that's
what kills it right) ?

/Bjorn

 

[Jim Thompson]

Mark,

Has your E-mail address changed?

I sent you an Excel question but have seen no response.

...Jim Thompson

 

[qrk]

Mark,

Has your E-mail address changed?

I sent you an Excel question but have seen no response.

...Jim Thompson

Sent you an email with updated info. Should have been stable for the
past 10 months.

 

[qrk]

Interesting. So what you are saying is that if I hook up a sufficiently
linear (or calibrated) photodetector, I could measure how the LED
behaves when scanning over duty-cycles and/or currents, and where it
starts going non-linear is probably where I don't want to be because it
goes non-linear as an effect of heat buildup at the junction (that's
what kills it right) ?

/Bjorn

Your main concern isn't really non-linearity, but overheating of the
die. But yes, by monitoring the LED voltage, current, and light output
you will come to some sort of conclusion of an appropriate drive
level. Running the LED at currents over 6A you can see the light
output sag with time after about 1 us. Since your using longer
on-times, you need to operate the diode at a saner current level.

 

[BW]

Seems my PCB worked, driving all 50 LEDs pulsed at the required
current, and I didn't even get any fat jitter on the 12VDC feed thanks
to the big honking capacitor I put to handle the excharge. Thanks guys
for the hints now I have something to play with over easter

I did realize however that it's very clumsy to have to have 10 copies
of the same circuit (opamp + resistor + power trans) so I was wondering
if you who have suggested a higher voltage to run more LEDs in series
have any hints on what to base a DC-DC converter on (say 12 VDC to
10mA@48VDC on average), a Max circuit perhaps controlling an inductor ?
This would be loading up the big capacitor during duty-cycle off.

Of course, putting more LEDs in a single string means if one dies, the
whole string is caput. But maybe that's not probable, what happens
first to a LED with age ? Does it retain its v-i curve but gets less
light-efficiency, or does it change v-i (potentially breaking the
circuit) ?

Regards,
Bjorn

 

[Fred Bloggs]

Ok I'm back after some actual lab-work

I studied the current-source design with an NPN with an emitter
resistor (Re) to GND and a string of 5 leds between Vcc (12V) and the
collector. I tried to make the dimensions so that Ve is around 500 mV
(to leave space for a Vce of over 1 V and the voltage drops across the
LED's which is around 2.1 V per LED at the currents I'm interested in),
so for a LED current of Ic=250mA, I chose Re=2.2 ohm. Given the drop
over the base-emitter port of the NPN of 0.7V, I'd have to have a Vb of
1.2 V.

Somewhere here I realised that at a beta of say 30-40, I would have to
support an Ib of around 10 mA (I actually measured this), which is
quite much. The trigger source is 5V and I have difficulties of getting
those 5V down to the required 1.2V. A resistor-based voltage-divider is
not good enough since choosing small R forces a too heavy load on the
trigger buffer, and choosing a high R makes the 15 mA create a too big
voltage drop (pulling the NPN out of the "easy" feedback model). So I
switched the transistor to a darlington NPN with beta > 750 giving an
Ib of about 0.2 mA (measured). Still this did not create a sane
configuration with the voltage divider model (the 0.2 mA creates a too
big voltage drop nonetheless).

I also tried various combinations of 1N4148 strings to drop the
voltage, but it does not work as I think it does (I get voltage drops
of only 530mV over each 1N4148 can that be right?). I tried both simply
putting the diodes in a series from the trig buffer (5V) down to the
transistor base. I also tried a resistor from the trig buffer to the
base, and a diode string from the base to GND, but this configuration
didn't work as expected either

I'd be happy to have some thoughts on this seemingly simple circuit
Perhaps adding a voltage follower (another buffer) at the input port to
support the higher currents is enough ? I'll try that in the lab I
guess...

Your main problem is being sent on a wild-goose chase by a well known
newsgroup pest, troll, and ignorant pretentious idiot, "Rich Grease"-
the "dreaded" this and that. You will notice the complete fake offered
no further suggestions to handling your quandary as is usual with that
kind. At 12V you end up with bunches of current source strings, each
requiring a bunch of parts overhead and delivering poor accuracy.

 

[John Larkin]

[...]

Well, he said he was, about 20 lines up. I just wanted to make sure he
didn't size the series resistors based on average current.

John

John, I'm missing something here. How does rms enter here, and how did you
get an rms current of 3.5A?

20 * sqrt(0.03) = 3.4641, just as I expected.

John