low frequency measurements of an LED-LED optocoupler
OK; to enter the fray.
Rather than account for the a.c. signals coupling and producing potential errors, I found a makeshift optocoupler I made a while back that consists of two similar standard red 3mm radial-lead LEDs in a piece of shrink tubing. I did some d.c. measurements.
I put 7.36mA into the emitter. I measure a ~short-circuit (meter burden 997 ohms) photocurrent of 1.06uA in the detector LED.
I found that to see 200uV of emitter forward drop change in the ~1.70V required a current increase of about 19.2uA (I could use a higher-res voltmeter). So the 1.06uA of strongly-illuminated photocurrent (presuming that the quantum efficiency is as good at the forward bias as at zero bias) would be expected to produce a change of about 200/[19.2/1.06], or about 11uV.
So if the light source were of a similarly high flux and modulated on and off, for a 1mA current generator with 1V across the emitter resistor, the random noise contributions for a 30 ohm rbb' transistor would be tallied as about e sub n of 712pV/sq rt Hz for the transistor, hence producing a fluctuation in the collector current of 712fA/sq rt Hz; base current shot noise at least of 894fA/sq rt Hz for a beta of 400, contributing to the output current noise. Supposing the LED has a noiseless current drive, its voltage noise density will be in the vicinity of 400pV/sq rt Hz, hence accounting for another current noise contribution of 400fA/sq rt Hz. But the thermal noise current of the 1k resistor will dominate at 4.07pA/sq rt Hz. Gathering all this together and skipping 1/f components we have an expected random current noise density in the 1mA of 4.25pA/sq rt Hz.
The effect of a on-off modulated intense light source of the experiment would be a peak-to-peak signal of 11uV/1k or 11nA; if a square wave that would be 5.5nA rms. If the random noise above is white and the bandwidth 22kHz, its contribution to the total would be 630pA rms. If the photonic stimulus from hell is as described above, it could be detected.
But that is an extreme case for illumination and especially for modulation of the illumination. And it's easy to keep the LED in the dark.
OK; to enter the fray.
Rather than account for the a.c. signals coupling and producing potential errors, I found a makeshift optocoupler I made a while back that consists of two similar standard red 3mm radial-lead LEDs in a piece of shrink tubing. I did some d.c. measurements.
I put 7.36mA into the emitter. I measure a ~short-circuit (meter burden 997 ohms) photocurrent of 1.06uA in the detector LED.
I found that to see 200uV of emitter forward drop change in the ~1.70V required a current increase of about 19.2uA (I could use a higher-res voltmeter). So the 1.06uA of strongly-illuminated photocurrent (presuming that the quantum efficiency is as good at the forward bias as at zero bias) would be expected to produce a change of about 200/[19.2/1.06], or about 11uV.
So if the light source were of a similarly high flux and modulated on and off, for a 1mA current generator with 1V across the emitter resistor, the random noise contributions for a 30 ohm rbb' transistor would be tallied as about e sub n of 712pV/sq rt Hz for the transistor, hence producing a fluctuation in the collector current of 712fA/sq rt Hz; base current shot noise at least of 894fA/sq rt Hz for a beta of 400, contributing to the output current noise. Supposing the LED has a noiseless current drive, its voltage noise density will be in the vicinity of 400pV/sq rt Hz, hence accounting for another current noise contribution of 400fA/sq rt Hz. But the thermal noise current of the 1k resistor will dominate at 4.07pA/sq rt Hz. Gathering all this together and skipping 1/f components we have an expected random current noise density in the 1mA of 4.25pA/sq rt Hz.
The effect of a on-off modulated intense light source of the experiment would be a peak-to-peak signal of 11uV/1k or 11nA; if a square wave that would be 5.5nA rms. If the random noise above is white and the bandwidth 22kHz, its contribution to the total would be 630pA rms. If the photonic stimulus from hell is as described above, it could be detected.
But that is an extreme case for illumination and especially for modulation of the illumination. And it's easy to keep the LED in the dark.
Last edited:
Yeah, and the virtually-outlawed incandescents do have plenty of flux, but certainly not a very deep modulation at double the mains frequency. If the device is in a showroom with a spotlight on it, and the indicator lamp is red or green (blues are practically zero QE for red light), and the indicator lamp is also a bias source for some open-loop amplifier stage, one could suppose it might create a detectable level of hum. A lot of assumptions to get a very-worst-case scenario.
The question raised was how much effect LED photosensitivity has on current sources. SY opined there was none based on his tests with a strobe. Others took measurements and found a small amount. I tried a stobe and got a larger signal. Not exactly a surprise.
The effect seems to be at least 40 dB down for example if you use an LED as a power indicator and reference, going from dark to sunlight.
That should not be an issue in almost any practical circuit. The exception would be in a current load for a high gain microphone or phono preamp.
So effect is real, and a nice bit of trivia.
But SY is still arguing. HE IS WRONG. Everyone else sees a small level of sensitivity even in forward biased diodes. I also got a peak at a 1N4148 that showed a bit of sensitivity. So Gerhard has a good idea, a bit of black paint is a good safeguard.
Last edited:
Finally !
I was laughing all the day long with this fight about the sex of the angels.
Well they do like those zero PSRR circuits.
In the case of my phono stage, the photocurrents are injected right at the noninverting input, so yeah, zero PSRR by definition. Note that even though the top of the preamp is perforated and plenty of light gets in, there's not even a smidgen of audible hum from the room lights. So even after 90dB or so of amplification, nada.
Is the strobe's pulsewidth on the order of a few hundred nanoseconds? Otherwise, that looks very highpassed - wouldn't something more rectangular be expected (like you'd get with an optocoupler?). What makes the trace so impulsive?
- Status
- Not open for further replies.