how do i impliment this constant current source

[unclejed613]

you must realize that there's very little thermal inertia in the semiconductor die itself, especially if the LED is properly heatsinked (the heatsink should have a lot of thermal inertia). what you should do here is first find out what the semiconductor die is made of. each semiconductor has a characteristic change of Vf per degree C of temperature change. once you know this, you can measure the forward voltage with just enough current to turn the LED on, and the heatsink in a liquid bath (preferably water) at 27 degrees C (this is 300 degrees K. which is one reason this particular temperature is used for so many measurements). if the spec sheet shows any measurements at 25 degrees C you can use 25 degrees for comparison against the spec sheet. so for every degree C of change, the forward voltage of the semiconductor changes. for instance, silicon (nobody makes LEDs from silicon, but it's Vf/T characteristic is well known) has a characteristic temperature coefficient of -2mV/deg C. so if the forward voltage is 0.7V at 25C, at 27C it will be 0.696V. at 40C it would be 0.670V. the same effect is present in all semiconductors, each material with a different coefficient, some positive, some (as in the case of silicon) negative, and there may be some very close to zero. so, once you have established a forward voltage as a starting point, you can either look up the coefficient (assuming it's either in the data sheet, or they tell you what material is used and you can look it up elsewhere) or determine it yourself. to determine it yourself, you use the current source to set a current that lights the LED but produces very little heating. then you change the temperature of the water bath, and for several points (such as at 10 degree increments) plot the temperature and the forward voltage. so you can plot the curve yourself, do some math and arrive at the coefficient. to measure in the manner you're describing, you need a very fast way of switching between sources for the LED, because waiting a second or two while you switch wires, wait for your meter to settle and read the meter, is way too long. it would be best if you could have not only a fast switching method, but a sample-and-hold circuit so that you could very quickly grab the reading and take all the time you want to read it.

 

[HATHA]

hi all
thank you unclejed613

(nobody makes LEDs from silicon

yes i know, leds may be GaAs or GaAsP but i really don't know. i don't have any specification of those.

this is my improve(i hope ) design and my new goal is 50µ
A accuracy

 

[HATHA]

you must realize that there's very little thermal inertia in the semiconductor die itself, especially if the LED is properly heatsinked (the heatsink should have a lot of thermal inertia). what you should do here is first find out what the semiconductor die is made of. each semiconductor has a characteristic change of Vf per degree C of temperature change. once you know this, you can measure the forward voltage with just enough current to turn the LED on, and the heatsink in a liquid bath (preferably water) at 27 degrees C (this is 300 degrees K. which is one reason this particular temperature is used for so many measurements). if the spec sheet shows any measurements at 25 degrees C you can use 25 degrees for comparison against the spec sheet. so for every degree C of change, the forward voltage of the semiconductor changes. for instance, silicon (nobody makes LEDs from silicon, but it's Vf/T characteristic is well known) has a characteristic temperature coefficient of -2mV/deg C. so if the forward voltage is 0.7V at 25C, at 27C it will be 0.696V. at 40C it would be 0.670V. the same effect is present in all semiconductors, each material with a different coefficient, some positive, some (as in the case of silicon) negative, and there may be some very close to zero. so, once you have established a forward voltage as a starting point, you can either look up the coefficient (assuming it's either in the data sheet, or they tell you what material is used and you can look it up elsewhere) or determine it yourself. to determine it yourself, you use the current source to set a current that lights the LED but produces very little heating. then you change the temperature of the water bath, and for several points (such as at 10 degree increments) plot the temperature and the forward voltage. so you can plot the curve yourself, do some math and arrive at the coefficient. to measure in the manner you're describing, you need a very fast way of switching between sources for the LED, because waiting a second or two while you switch wires, wait for your meter to settle and read the meter, is way too long. it would be best if you could have not only a fast switching method, but a sample-and-hold circuit so that you could very quickly grab the reading and take all the time you want to read it.

yes i do some thing similar to that. but i don't use water bath instead use control ambient. keep the led on temperature controlled oven until thermal equilibrium happen between led junction and surrounding.

 

[mneary]

Glad to see you sharing some secrets. For this very reason I hate tight design specs to solve an undisclosed problem!

If you know within a range the voltage of the LED, then you don't need the insane specs we've been discussing which are compliant over 0V to almost Vcc. You know the LED voltage within a few mV, and LED forward voltage isn't very sensitive to current. From 10.000 mA to 10.10 mA (1%) Vf of your big LED shouldn't change measurably.

If Vcc is reasonably stable, you should be able to use an ordinary resistor as your "current source". Set Vcc to Vf+10V and use a 1k 1% resistor.

 

[unclejed613]

yeah, reminds me of the conundrum of calibrating torque wrenches for a nuclear weapons lab. the wrenches were numbered 1-to-whatever (so that even the person using the wrenches didn't know what they were set to, they were numbered according to a drawing that said "for these bolts use wrench 1"
), so we call the lab and ask "so what values of torque are these wrenches supposed to be set to?" with the answer "we can tell you that the tolerances are +/-2%, but we can't tell you the torque settings"... i think that one was solved by finding a calibration tech with the required security clearance...

 

[HATHA]

Glad to see you sharing some secrets. For this very reason I hate tight design specs to solve an undisclosed problem!

i thought people don't like to know such thing (some times such pysics boring )

ok fun time is over i hire a programmeble electronic load for my work.
now i have to make some falling edge detector to trigger lab multimeter to make measurement at right instant.

what so ever many many thanks for your advices