2

I'm looking to use an LED in a control application. (feedback via LED to photodiode.)

I found a few questions that didn't fit.

I'd like to make it all linear, no PWM.
(I worry about the led turn on/off time with PWM, I can average, but that's slower. Even with pwm having a linear 'current' control knob might help.)

What frequency you ask?; I guess 1-10 MHz, for a start, but faster is better. :) I dream of finding an LED whose light output (in mW) is linear with current, from say 1 uA to 1 mA. Is there some search phrase I can look for? Searching for leds on the web is a fire house. I know this is kinda a shopping question, Is there a meta-shopping type question? How to shop?

Let me edit: Yes a linear opto-coupler would be even better. If it was fast. > 1MHz.

Transistor
  • 168,990
  • 12
  • 186
  • 385
George Herold
  • 4,754
  • 1
  • 18
  • 25
  • 1
    XY problem. http://www.ixysic.com/home/pdfs.nsf/www/AN-107.pdf/$file/AN-107.pdf – τεκ Oct 31 '17 at 00:28
  • 3
    I think you are looking for "Linear Opto-couplers" example http://www.vishay.com/docs/83622/il300.pdf. Those have dual matched detectors. You use one on the primary side in the feedback loop. See the examples in the data sheet. – Trevor_G Oct 31 '17 at 01:10
  • do you actually need linearly controlled light somewhere or only a linear optocoupled signal path? –  Oct 31 '17 at 01:13
  • 1
    How linear is your "linear"? Generally, light output is nearly linear with current in all LEDs. I think this thread can give you few angles on the subject, https://electronics.stackexchange.com/q/17528/117785 – Ale..chenski Oct 31 '17 at 01:28
  • @τεκ, that would be nice if faster. – George Herold Oct 31 '17 at 01:29
  • @AliChen, thanks I'll look at kingbright. how linear? I want to use in a photodiode TIA instead of feedback R. I'll do a current drive for the led but the more linear the better. – George Herold Oct 31 '17 at 01:35
  • I have never measured light output at low currents such as 1uA, but I have measured backlight brightness over a range of around 1mA to 20mA or so, and found it to be linear enough that any nonlinearity was lost in the uncertainty of my luminance meter. I seem to remember that LED's have a "knee" somewhere where they become less linear or the slope changes. But I think that is at a higher current, far above 1mA. – user57037 Oct 31 '17 at 02:20
  • 1
    I think there is also a way to linearize the light by adding a second sensor on the same side of the board as the LED. – user57037 Oct 31 '17 at 02:42
  • From experience: buy 1000 LEDs from a single batch from a single supplier then only about 50 of them will be stable -- if you hold them at a stable temperature and feed them a very stable current. They will still need to be calibrated. But a few will be stable. The others you might as well throw in the garbage can. (This includes the use of a calibrated Keithley 6221 for accurate currents.) LEDs just drift. I don't think 'stability' or 'precision' of light output relative to current input is on any spec sheet. Will be interested in what you find. (Three orders of linearity? Hah.) – jonk Oct 31 '17 at 07:39
  • @jonk, I'm not sure why only 50 are useful. A little drift in output should be OK. It's in a feedback loop. – George Herold Oct 31 '17 at 15:24

1 Answers1

1

Since Diodes and transistor hFE are both inherently non-linear , one way to make it ultralinear is to use a dual version and compensate one side using linear error correction feedback on the other side( sophisticated method).

Optoisolators are inherently nonlinear may have a 3:1 gain tolerance in output range for any input current. Expecting a linear response on anything more requires expertise in the secondary opto loop using error correction with an Op Amp.

One needs to have reasonable specs and tolerances for linearity and bandwidth then solve the technical issues. Since you have neither no rational solution can be given, other than a book on opto linear design.

LED dynamic range may be up to 3 decades but very non-linear and bandwidth constraints are severe.

  • 1st you need a voltage to current feedback converter with OA
  • 2nd you need a 2ndary transfer function error correction circuit design with low gain. (<1)

EDIT

Fortunately the experts in opto design Sharp (now licensed to Vishay) have such a solution with dual emitters (LED) for high gain linearity and stability.

  • 0.01 % servo linearity
  • Wide bandwidth, > 200 kHz
  • High gain stability, ± 0.005 %/°C typically

    Detectors are very stable and low error and well matched while Emitters are high deviation thus only 1 emitter uses 2 detectors, 1 for voltage to current feedback and the other detector for output to current to voltage sense resistive load.

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
  • I don't know why LED's should be inherently non-linear (with current drive). But in practice I agree they are. The compensation trick for the LED is something I'll think about. I'd like to do something like this... http://aip.scitation.org/doi/full/10.1063/1.4931042 Note: they compensate for the LED non-linearity by using a 3rd TIA for the output. – George Herold Oct 31 '17 at 15:21