Re: Long range IR signal


"Richard Hosking" <richardh@xxxxxxxxxxxx> wrote in message
news:425bc4a7$0$20399$5a62ac22@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> Thanks for this
> At present I have a current source (PNP transistor) from + supply to
> diode to FET logic level switch to ground. Assuming I wanted a 1A pulse
> to the diode at about 5-10% duty cycle at 38 KHz, this would mean a 1-2
> usec pulse.


The carrier frequency (in this case 38kHz) should always be about 50% duty
cycle. Assuming you are using a standard receiver IC, then you will not
likely get any benefit from shrinking the carrier frequency duty cycle. In
fact you may end up with less sensitivity even if you are pulsing the LED(s)
with more current.

Normal receiver ICs expect the carrier to be modulated on and off at some
lower frequency. If you continuously run the LEDs at 38kHz the receiver IC
will eventually lose sensitivity. 38kHz is around the frequency of some
compact fluorescent lamps, so they could easily interefere with your signal
if the receiver IC was sensitive to continuous 38kHz. So instead devices
like the TSOP1238 have automatic gain control elements that reduce
sensitivity if the detected 38kHz light modulation is continuous (like it
would be from fluorescent lamps).

Instead the data needs to be encoded somehow to avoid continuous
transmission. The TSOP1238 data*** suggests that the 38kHz carrier should
be turned on for a minimum of ten 38kHz (50% duty cycle) cycles up to a
maximum of seventy cycles. After that no signal should be transmitted for
at least fourteen cycles of the carrier frequency before another pulse
packet it sent.

Since by the sounds of your situation you don't need complicated data
transfer, I suggest using ten or eleven cycle bursts of 38kHz @ 50% Duty to
minimize LED stress and power consumption. A possible implementation might
be something like follows: Have a free running continuous 38kHz +/-2%
frequency tolerance 50% duty cycle oscillator. Have a switch debounce
circuit for the trigger switch with the output feeding a one shot. The
output of the one shot is programmed to last about ten (or better eleven)
38kHz cycles or in other words about 290us. Have the output of the one shot
AND gated with the 38kHz 50% duty cycle oscillator. The output of the AND
gate drives the LED (through suitable high current LED driver).

The net result is under normal operation the LED is not driven at all. Then
when someone presses the trigger you get one set of ten or eleven LED pulses
each lasting about 13us (with 13us gaps in between), and nothing else after
that (until the trigger is manually pressed again).

This is the basic idea, but feel free to modify it to suit your needs
better.




> I presume you would have to use a fairly fast RF transistor
> with a suitable current rating as the current source with a low ESR
> capacitor (say ?4700uF) across the supply close to the circuit. The FET
> swich should handle the situation OK. I would use the PNP current source
> with emitter resistor of 0.6 ohm and two diodes to base from + supply
> - would this work or is there a better solution?
> What devices would be suitable for this, and what other considerations?


I don't understand your description of your constant current source, so I
can't make comments as to its feasibility, but using a constant current
source for driving the LED seems rather more complicated than needed.
Assuming you have a regulated supply voltage (5V or more would be nice),
what is wrong with using a simple resistor to limit the current?

What is your power supply (batteries? regulated? voltage? current
capability?)? An eleven pulse packet consisting of eleven 13us 1A pulses
requires 143 microcoulombs of charge (Q=I*t). If this is delivered entirely
from a capacitor, then you can approximate the voltage droop for various
capacitances using the formula Q=CV. For example if we allow a 250mV sag
when drawing 143 microcoulombs, then we need 0.000143 = C * 0.25, so a
capacitance C of 572uF. The ESR should be low enough that the output
voltage of the capacitor doesn't sag unreasonably at 1A peak current. If
you use a resistor to limit the current then you can simply make it smaller
to compensate for whatever ESR the capacitor has to enable 1A peak current
pulses. None of this is too precise, but neither does the LED need precise
current levels. Any standard 470uF or 680uF capacitor or so from 5V should
function quite nicely. If the power supply is particularly stiff and easily
able to cope with fast 1A+ pulses with minimal droop, then no extra
capacitance is strictly required (though a small amount may still be desired
for minimizing inductance based overvoltage effects).


> Thanks for any ideas
> Richard


.