Connect with us

LED question Vf rise vs. light output

Discussion in 'Electronic Design' started by Anthony Fremont, May 6, 2007.

Scroll to continue with content
  1. While playing around with a PIC and some LEDs, I was observing some
    rise/fall times on my scope. With a traditional i/o pin -> resistor ->
    LED -> Vss setup, I was PWM'ng an LED with a 25% duty cycle at about 60Hz.
    The rise time (10-90%) at the i/o pin was a bit under 5nS. However, the
    rise time at the LED/resistor junction was much slower at around 70nS. I'm
    thinking this rise is tied to the rise in Tj as the LED comes on. Is the
    brightness ramping up in direct proportion to the slower rise time of the
    voltage accross the LED, or is inductance in the resistor slowing things
  2. Jim Thompson

    Jim Thompson Guest

    I would expect that it's the capacitance of the _forward_ biased
    junction... huge, much larger than predicted by the classic diode
    junction model. I implemented such a modified model to GenRad's
    simulator around 1980 (Fortran :)

    ...Jim Thompson
  3. Hmmm, ok then. I'm using a 1k resistor, a 5V Vcc and the LED's Vf peaks at
    just under 3V. Does an extra 65nS sound right for that? None of this is
    indicative of the LED's output? Is it possible that Vf could peak before
    the brightness?

    Yeah, I've done some Fortran, but very little and quite a while back (~1980)
    Did scads of COBOL back then though.
  4. Jim Thompson

    Jim Thompson Guest

    Presuming the 65ns is the time constant, that's 65pF, not an
    outrageous number... but it's probably time to peak, so capacitance is
    even smaller. Why don't you measure capacitance versus forward
    drop... you'll need a high-grade bridge to do it. You will find that
    capacitance rises sharply with forward bias, then drops abruptly as
    junction current rises.
    Fortran and FAP are the only languages I actually have any formal
    training in... summer of 1960 ;-)

    I took a community college course in Pascal around 1990, but it didn't
    really stick... I was constantly fighting the instructor, who took
    umbrage with my use of flow-charting.

    ...Jim Thompson
  5. Robert

    Robert Guest

    Might take a look (in your abundant free time) at an interpreted Language
    like Python. Freely available and there are Scientific Libraries so you
    don't have to re-code the wheel. Not going through the Compile-Link-Run
    cycle sure speeds things up.

  6. John Larkin

    John Larkin Guest

    Pascal is tedious. By the time you finish all the declarations, you're
    too tired to write the executables.
    The best programmers I know never took any programming courses.
    The slowest PowerBasic compiles I've seen lately are 0.2 seconds, but
    most are 0.1. By the time you take your finger off the F9 key, the app
    is running. And it will run useful 30 MHz FOR loops, something
    interpreters may not do.

  7. I forgot to mention that the fall time at the same junction was about 250nS.
    The PIC actively drives low with the same "force" as it does when going
    I was just poking around, noticed the difference in rise times and was
    curious as to why. I'm more curious as to how fast the LED can ramp up its

    The other day, I was tinkering with an IR photodiode and various remote
    controls lying about. I had the photodiode reverse biased with 5V against a
    1M resistor. IIUC, photodiode response is very fast so I take it that the
    rise time I see on the scope should pretty much resemble the actual
    brightness increase as the LED lights up on each pulse. IIRC, the output
    from the photodiode indicated that the brightness was ramping up over uS not
    nS of time, but I'll have to recheck that later. I thought LEDs were faster
    than that.

    I can't seem to find a loose IR LED lying around so the above mentioned Vf
    tests were done with a blue LED. Of course I can't use the IR photodiode
    with them to compare the Vf rise against the output from a photodiode to see
    how much correlation (if any) there is. I'll keep looking for an IR LED, I
    know I have one here somewhere. Perhaps the remote control has some
    capacitance accross the LED slowing it down.

    Formal training????? What's that? ;-)
    You're not missing anything, nobody uses PASCAL for anything real. Real
    work is done with assembler. ;-) I never was much on flowcharts myself,
    way too much work.
  8. Robert

    Robert Guest

    They claim the Library routines are quite fast.

    Wouldn't know.

  9. John Larkin

    John Larkin Guest

    As far as I can tell, Python usually compiles (the terminology is
    often "compiles") to virtual machine byte code. That's better than
    actual interpreting, but is still a lot slower than machine
    instructions. A lot of our stuff is realtime, for test sets and such,
    so we want fast execution and bare-metal hardware access, which is not
    the classic "computing" environment.

    A lot of Pascal was like that, compiling for a fictitious P-machine.
    Microsoft did this for PDS Basic and I think still does for Visual
    Basic. Java works like this too, the advantage being that the runtime
    system can seriously protect the OS from the code if the code is not
    truly executable.

  10. John Larkin

    John Larkin Guest

    The rise is just a portion of the rc exponential... the diode clamps
    at a volt or so out of a 5-volt charging curve. On the fall, you see
    the full exponential, all the way down to zero.
    That will *very* slow. Dump it into 50 ohms or so.

  11. Paul Mathews

    Paul Mathews Guest

    Depending on photodiode area, expect from 1pF to 1000pF or more, even
    reverse biased. It's actually quite common to have time constants in
    the 10s of microseconds with a 1 MegOhm load. It's extremely easy to
    check: use 500K and see if things speed up a factor or 2. If so, it's
    all about your photodiode load.
    Paul Mathews
  12. john jardine

    john jardine Guest

    That glorious F9!
    Wrote a PB prog last night for a customer. It looks for exact matches of
    contiguous code segments in ROM dumps from two different products. All found
    code runs then stored in a third file plus some other statistical info
    accrued during the searching.
    From sitting down it took a whopping 2 hours and 60 lines.
    My last 'real' PB prog was over a year ago but being Basic it takes little
    time to come up to speed. A quick check on the "binary file open" syntax and
    I was in. A few dozen F9's later and voila!, finished.
    A far, far cry from the C compiler I watched last week, that chuntered away
    to itself for an eternity as it compiled and linked and did other strange
    stuff that only "modern" compilers can and must do.
    Difference is, I'm looking for a result and not to secure my career.

  13. The blue LED is clamping the voltage at just under 3V max with a couple of
    mA of current. When the PIC pin is brought low, it seems to me that I
    should see the 3V fall to zero in the same amount of time. In fact, I see
    the 3V drop to 0V in about 3-4 times the amount of time. This is why I
    thought that I might be looking at heating/cooling factors.

    This seems odd to me and doesn't really match my experiments. I have the
    diode on top of the resistor with the diode reverse biased by the 5V supply.
    At the junction of the photodiode and resistor, I am taking my measurements.
    Under ambient conditions here in the house, I get an "idle" voltage of about
    60mV. When the remote control is activated, I get pulses that peak at 5V
    naturally with the remote control about 1' away.

    I don't see why the large resistor would be slowing down the first (rising)
    edge of the pulse. I can understand that it would slow down the falling
    edge though since the photodiode needs to bleed the 5V off thru it. I was
    under the impression that the resistor I chose would set the "gain" of the
    photodiode and this is indeed what seems to happen. If I lower the
    resistance to say 100R, the sensitivity plummets to the point that I can
    barely get 100mV out of it and that is with the remote control jammed
    against the photodiode. The rise time is even worse than before. BTW, I
    was actually using a 100K not a 1M resistor with the photodiode. A 1M
    resistor increases the gain of the photodiode substantially and results in
    an even faster response, even though the voltage peaks are much higher. It
    also increases the "idle" voltage to about 300mV at the junction of the
    photodiode and the 1M resistor supporting the bigger resistor = bigger gain

    I did some more basic tinkering and was surprised to see what the rise and
    fall times for an unterminated 1K resistor hanging on the PIC pin was
    compared to what you see at the pin (<5nS). The resistor leads (untrimmed)
    added more than 35nS to the rise time. I guess there really are
    consequences to using the solderless breadboard and long leads.
    ;-) An unterminated 100R resistor had a lesser effect, adding only about

    I picked up a few different IR LEDs today and noticed substantially faster
    rise times for the Vf vs. the blue LEDs I was using. Of course Vf only
    peaks at 1V now instead of 3V, but the slope of the rise is still steeper
    with the IR LED. The rise time of the output of the photodiode is still
    much slower than that of the Vf of the LED. This seems to indicate that the
    Vf rise I'm seeing is due to capacitance of the LED and not the ramping up
    of the brightness. If the output of the photodiode is to be believed, then
    the brightness of the LED increases much more slowly than the forward
    current of the LED (like uS instead of nS). With a small enough current
    limiting resistor, I can turn the LED on (in terms of getting current
    flowing thru it) in as little as 7 or 8 nS including the 5nS that the PIC
    takes to start with. Apparently, the light emissions don't ramp up nearly
    so fast.

    I found this interesting: Increasing the reverse bias to 12V on the
    photodiode resulted in no change in dV/dT vs. the 5V bias.
  14. Jim Yanik

    Jim Yanik Guest

    can your logic IC source the current the LED requires?
  15. With the 1K resistor, easily. With the 100R resistor, not legally. ;-)

    Just so everything is clear, I got two things going on but each one has
    interesting aspects.
    1) A PIC driving an LED at 25% duty cycle for detection by an IR photodiode.
    I noticed a slower voltage drop rise across the LED vs. the actual rise time
    of the driving pin. That seems to mostly be due to the actual resistor
    itself and not the LED so it's quickly losing its appeal.

    2) IR photodiode "on top of" a 1M resistor to ground. The photodiode is
    reverse biased so that I'm using it in "photoconductive" vs. photovoltaic

    I'm trying to determine how fast an IR LED illuminates vs. the rise in
    current flow thru the LED. IOW, trying to measure the lag, if any. I
    understand that LEDs can be quite fast in response and I was trying to see
    if I could actually measure the lag/rise-time in response of different IR
    LEDs. So far it's not working out too well. Either the LEDs illuminate
    much slower than the current flow or my photo-diode is not nearly fast
    enough to measure.

    I'm using a Zetex BPW41D.
  16. Winfield

    Winfield Guest

    A reasonable first guess, perhaps, but under closer scrutiny not well
    explaining the scene. One clue to resolve the issue: if the LED
    emission is filtered with a narrowband filter at the spec'd
    appropriate to the bandgap of the LED junction, the desired very fast
    light risetime and falltime is observed. All the "slow light" is at
    wavelengths. Unfortunately, this slow light is often the source of a
    significant fraction of the LED's overall emission.
  17. Jim Thompson

    Jim Thompson Guest

    Naaah! The "slow" risetime is "at the LED/resistor junction"... an
    electrical issue, not a "slow light" issue.

    ...Jim Thompson
  18. Guest

    Since this is just an experiment versus real life, i.e. you could have
    a really inefficient circuit, why not experiment with a current source
    feeding the LED, with pulsing done by shunting the current with a N-
    fet. LEDs are really current devices, and the pic output plus resistor
    isn't a very ideal current source. To be a bit more specific, get a
    bench supply of say 20V (pretty common). If you want 20ma in the LED,
    then use a 1kohm resistor in series with the LED. The mosfet source
    and LED cathode go to ground. The mosdet drain goes to the anode of
    the diode. Drive the gate with the pic.

    With a current source, you can estimate the capacitance directly from
    the rise time.
  19. Paul Mathews

    Paul Mathews Guest

    I encourage you once again to do the obvious: reduce the load
    resistance and see what happens.
    Paul Mathews
  20. Jim Yanik

    Jim Yanik Guest

    doesn't the junctionC of the LED act with the series R as a low-pass
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day