Author Topic: Semiconductor early aging effects on performance  (Read 1864 times)

0 Members and 1 Guest are viewing this topic.

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Semiconductor early aging effects on performance
« on: June 16, 2018, 12:27:55 am »
I got the idea of trying to use some cheap LED-phototransistor distance sensors for high resolution measurement.  Because of the application, absolute accuracy is not required, but stability is critical.  This is my precision square checker thread.

I've seen a 25% change in phototransistor current with a constant LED current over the course of two days.  Initially I thought this was temperature sensitive gain related, but it appears now that might not be the case.  Temperature changes do not seem to be driving the change, but total operating time does.  The initial change is a quite rapid drop in current of several uV/hr gradually becoming a uV in several hours.

I had no success finding anything that looked plausibly related with google.

There is a lot of discussion of this in the context of voltage references, but I don't find anything on other devices.

I've got a number of devices to test, but I can only test one at a time and as each test will take a few days, I thought I should do a bit of reading while I wait.

So can anyone point to relevant literature?
 

Offline chris_leyson

  • Super Contributor
  • ***
  • Posts: 1541
  • Country: wales
Re: Semiconductor early aging effects on performance
« Reply #1 on: June 16, 2018, 02:10:14 am »
You would have to driving the LED at way over it's published maximum current rating to get that sort of drift, expect complete failue in a week or two if that's the case. LEDs don't age that fast unless you overdrive them. Maybe think about differential measurement techniques if you want stability.
« Last Edit: June 16, 2018, 02:21:47 am by chris_leyson »
 

Offline Conrad Hoffman

  • Super Contributor
  • ***
  • Posts: 1928
  • Country: us
    • The Messy Basement
Re: Semiconductor early aging effects on performance
« Reply #2 on: June 16, 2018, 03:37:42 am »
I trust nothing. If it were me, I'd break the thing in half and measure the LED with a photodiode. Those are very stable, in fact I wish they made photo interruptors with them. Then at least you'll know which side is the problem.
 

Offline martinr33

  • Frequent Contributor
  • **
  • Posts: 363
  • Country: us
Re: Semiconductor early aging effects on performance
« Reply #3 on: June 16, 2018, 04:13:52 am »
Just checking - this is a red LED or other monochromatic device.
    - some phosphors degrade quite quickly, esp. on older parts.

And it is a 5mm part, and running at less than 20mA.
   - overdriving causes a rapid drop in brightness
   - 3mm parts seem to struggle at 20mA.

Also, a brief short across the limiting resistor could cause that kind of shift. To Chris' point, once these things are damaged, they keep going downhill.
 

Offline IconicPCB

  • Super Contributor
  • ***
  • Posts: 1527
  • Country: au
Re: Semiconductor early aging effects on performance
« Reply #4 on: June 16, 2018, 05:18:44 am »
down like a LED balloon?
 

Offline DC1MC

  • Super Contributor
  • ***
  • Posts: 1882
  • Country: de
Re: Semiconductor early aging effects on performance
« Reply #5 on: June 16, 2018, 09:12:27 am »
Hi, is the change reversible, it was not clear for me if after pausing the LED, it comes back to the original parameters ?

Also, compared with the data sheet Imax, how big is in % your driving current ?

And, last but not the least, you're sure that the constant current feed trough the LED is realyy constat ?

EDIT:
- Similar experiment: https://electronics.stackexchange.com/questions/261313/led-intensity-changes-with-time
It's clear that to have a constant optical output, the LED has to be temperature stabilized and not very strongly driven to avoid other long term efects (LED rot).
Hey, Huges even patented this in '87
https://patents.google.com/patent/US5013971


 Cheers,
 DC1MC
« Last Edit: June 16, 2018, 09:18:31 am by DC1MC »
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Semiconductor early aging effects on performance
« Reply #6 on: June 16, 2018, 12:40:26 pm »
I'm beginning to suspect that it is a transient effect of  thermal stresses caused by soldering. The device is being operated well inside of the maximums.  I'm running at 20% of the maximum LED current of 50 mA and at 30% of the Vceo maximum.

It's been quite stable now for over 12 hrs at 99-101 uA.  A hair dryer will instantly cause it to jump to 105 uA.  But it settles back to the prior value very quickly when the heat is removed.

Conrad, watch out or I'll send you the ones I've broken the leads off to take apart.  They're ~ 0.010 x 0.013.  The thru hole "leads" are just long gull wings and oh too thin.  Just gluing one to a nail nail to mount it has been a chore. 

There are devices with better lenses for very little more which have a longer range.

The tempco is looking to be about +1%/F based on readings 1:30 apart with a 0.9 F temperature rise in the room.  So with temperature correction repeatabiity  below 0.0001" may be possible after all.

It appears I have an EMI issue as moving the lead from the 34401A measuring the phototransistor current caused a 2 uA jump.  Which goes away if I move it back to where it was.

 

Offline DC1MC

  • Super Contributor
  • ***
  • Posts: 1882
  • Country: de
Re: Semiconductor early aging effects on performance
« Reply #7 on: June 16, 2018, 01:14:06 pm »
Considering the shown test fixture and some later ideas I have, here are some more considerations that may help:

 - About soldering: I think that with or without soldering, the fully plastic 2ct leds will have no long term stability, the plastic will shift, change properties and so on, and affect the results, both on short or long term.
My suggestion is to search for a batch of older glass lenses leds, US is practically flooded with hi-quality military and aerospace surplus and something like this should be available at less ferocious prices:
https://www.ebay.de/itm/1-Stuck-1N6266-General-Electric-GAS-Infrarot-LED-A10-2255/221670551615

 - If you want an absolute low cost, use a socket, don't solder anything on the led pins, just make a socket from a good quality DIP8 with machined pins and adapt it to your fixture, this way you can test led with minimal thermal and mechanical stress and quickly replace them.
One other, a bit out of the box, solution is to use a power led on a ceramic substrate, that will be less prone to thermal stress due to soldering and the whole package should be more stable.
Like these:
https://www.ebay.de/itm/10-Stuck-1W-940NM-Infrarot-IR-LED-1-Watt/173357756157

About temperature, there is no way to ignore it, either you ovenize the LED or put a good thermistor on it and do a calibration table.

 Cheers,
 DC1MC
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Semiconductor early aging effects on performance
« Reply #8 on: June 16, 2018, 03:27:37 pm »
Long term stability really doesn't matter for the precision square testing application.  So long as it is stable over the course of a few hours, that is sufficient as it is the relative readings that matter.  An oven is not practical in this application, but a thermistor bonded to the sensor should be satisfactory.

I stumbled across these by accident while making up a Digikey order, so I thought I'd get a few to play with.  I'm still trying to find a high resolution linear optical sensor.  But there are better options for senosrs which have longer range and are still low cost. 

A friend of mine looked into these sorts of sensors a number of years ago when he was building an instrument for making multicomponent measurements of propagation velocities at seismic frequencies at an oil company research lab.  For engineering reasons, he chose contact sensing even though he would have preferred non-contact sensing.

Most such measurements are made at 0.8-1.2 MHz which is much higher than seismic at 0.1-250 Hz.  So far as I know, only two low frequency instruments have ever been built, though there might a third now.  The really hard part is fixing the leaks in the pressure vessel which has to go over 15,000 psi.
 

Offline Smith

  • Frequent Contributor
  • **
  • Posts: 376
  • Country: 00
Re: Semiconductor early aging effects on performance
« Reply #9 on: June 17, 2018, 07:57:12 am »
I have done some experiments and characterisations of power leds for measurement setups.

We used powerleds on an lower current, around 1mA from a stable diy 8V 1mA current source. I characterised the startup light level of these leds for 30 sec per led, for a few hundred leds. The outcome was amazing: some went to their final state within half a second, some never got to their final state because the light level kept going up/down.  Some reached their final level after 10-20 seconds with a logarithmic rise. A couple of them where extremely noisy, I mean 10-20% noise. I measured these over and over again, they kept giving me the same results.

Before anyone asks I just replaced the leds from a mounting we made, everything else was kept the same.
Trying is the first step towards failure
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: Semiconductor early aging effects on performance
« Reply #10 on: June 17, 2018, 11:17:05 am »
At university they had a system to measure thermal expansion with an optical type sensor of a similar simple LED - photodiode setup. For good resolution it was using a relatively thin slit to open or close, not just simple distance.  AFAIR the LED was temperature stabilized to get a long time stable intensity. The detector is way less critical. However the more obvious choice for a precision detector is a photo-diode and not a phototransistor.

Especially with small LEDs there is a chance that there can be fluctuations similar to popcorn noise, as a single defect can have a significant contribution.
 

Offline Smith

  • Frequent Contributor
  • **
  • Posts: 376
  • Country: 00
Re: Semiconductor early aging effects on performance
« Reply #11 on: June 17, 2018, 11:48:57 am »
I have made a type of light level calibrator with a led a few years ago. It had a selector with multiple light levels from a few millulux to about 50 lux. The external head wich contained the led also contained a temp sensor and a resistor as a heater element. It was controlled analog with an comparator and a integrator. The led was kept at about 40-45 degrees celcius. It has been calibrated multiple times, and has never been adjusted since I as far as I know.
Trying is the first step towards failure
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf