I found this paper; although about UV LEDs its conclusions will probably apply to visible LEDs too. Their finding is that LED 1/f light fluctuations are reduced by using a higher resistance drive. The 1/f corner frequency is in the region 100Hz-10kHz; above this shot noise dominates.
There is some correlation between LED light fluctuations and LED voltage (and hence current) fluctuations, so it is unclear to what extent this 1/f could be a problem in LED bias. I am going to hazard a guess that if an LED actually in a circuit does not generate sufficient direct 1/f voltage noise to cause a problem then an LED used to illuminate a gain control element is unlikely to create enough 1/f intermodulation to cause a problem either.
There is some correlation between LED light fluctuations and LED voltage (and hence current) fluctuations, so it is unclear to what extent this 1/f could be a problem in LED bias. I am going to hazard a guess that if an LED actually in a circuit does not generate sufficient direct 1/f voltage noise to cause a problem then an LED used to illuminate a gain control element is unlikely to create enough 1/f intermodulation to cause a problem either.
So constant current drive is optimal for lowest optical output 1/f noise.
100Hz+ is high for corner f, so any slope reduction will have significant effects down at very low frequencies.
There is the answer as I suspected.
Extrapolation and more consequent BS.
Dan.
Would you have preferred that I kept quiet about a paper I found which may support Chris? He could have searched for it and presented it in support of his position; instead it was left to me, who disagrees with him, to potentially undermine my own case. I guess that is how intellectual integrity works.
Incidentally I tried to find info on LDR 1/f noise too, but could not find anything.
Incidentally I tried to find info on LDR 1/f noise too, but could not find anything.
They're not, the reason is the LDR's are very slow to react to led changes, so in a way they are self regulating, and elaborate ways of powering the led's is not needed.
You can see this on the scope using a sine wave, flick the volume up quickly and you can visually see the lag before the sine wave reaches it's new level.
Cheers George
Last edited:
Last edited:
LED 1/f noise will certainly affect the LDR, as the lower frequencies will not be affected by the slow LDR response. The question is whether this will be a significant effect, and how it compares with whatever 1/f noise the LDR has itself. Have you any info on this?
Nope, I got nothing on this just the data sheet below, I doubt very much the bass is effected with the ldr because of it, as listening would show, my system is clean and tight to 20hz. And there nothing to see on the crow either at LF.
Cheers George
OK, no noise data for the optocoupler device.
"Clean and tight to 20Hz" sounds like a frequency response claim, not a subsonic intermodulation claim.
So we don't know if 1/f intermodulation is an issue, so we cannot know if current driving of the LED is an advantage.
"Clean and tight to 20Hz" sounds like a frequency response claim, not a subsonic intermodulation claim.
So we don't know if 1/f intermodulation is an issue, so we cannot know if current driving of the LED is an advantage.
If there is indeed 1/f subsonic intermodulation in LDRs wouldn't there be measurable (and visible on a scope) non-harmonic sidebands up into the sonic range as a result? Would not the absence of such sidebands imply the absence of 1/f noise or at least the absence of 1/f of sufficient magnitude and cut point that it matters audibly?
I realize Chris's claim of current controlled LDRs being sonically better than voltage controlled LDRs remains an open issue, just trying to wrap my head around how 1/f might play a roll in this.
Curious, wouldn't it be a simple matter of cracking open an old one, inserting a wide dynamic response photo diode/transistor and measure the LED response? you'd have real measurements, or am I missing something (as usual).
Sure, you could do that if the objective was to understand the LED as an isolated subcomponent. But since the subject here is overall LDR performance would it not make more sense to just measure the photoreistor's resistance (or impact on a signal across it) as a result of whatever's happening on the LED end? From my perspective there may well be 1/f intermodulation by the LED but if it doesn't impact the photoresistor's behavior I'm indifferent.
Apologies, I'm an engineer by nature, I pull everything apart, study it, understand it, rebuild it.
Agreed, if the LED has no measurable results, it can be discounted, but your debating it, not measuring it, and at the moment, it's still speculation.
That's an assumption I know to be very wrong, if your aim is the best a LDR can do,
You do need sophistication in how the led is powered and to abandon thinking of powering them in a circuit that grounds and has no sensing of the cathode load , The best position i have found for placing the attenuation purpose, is between the shunt LDR cathode and the series LDR cathode, the wiper then to a lower potential, but not the same as signal ground.
LDR's are current driven devices not voltage so current needs to be regulated. The
sophistication should reach such a stage where resistances you use are
internal to semiconductors, and not conventional passives.
I would love to publish a schematic, but that is not the path I am presently taking.
Did this for GE in a pulse oximeter application. All LED's are not created equal they can have sometimes considerable excess noise (manifested optically). There is a rich body of literature in the medical journals much of it free to view. I modified a USB soundcard to go down to DC (heart rate ~1Hz) and took a lot of data trying to get an idea of the SNR that was achievable in the harmonics because there is diagnostic value there. I thought it ironic because they were trying to beat highly trained ears. 🙂
Last edited:
That's very interesting, did you notice any excessive electrical noise from the 'noisy' LED's?
The correlation was not perfect and an excellent point but not good enough to remove it from the optical domain. I'm afraid they were looking for an unachievable SNR.
This is a good free place to start...
http://homepages.rpi.edu/home/70/sawyes/public_html/SPIE_Invited2005.pdf
Last edited:
Thank you for the link.
This is going to set the cat amongst the pigeons, they're playing with optical bias on the other channel...
This is going to set the cat amongst the pigeons, they're playing with optical bias on the other channel...
Unlikely to be visible on a scope. Possibly visible in a very good spectrum analyser.CaptainWatt said:
LDRs are merely light dependent resistors. An LED-LDR optocoupler (which is what I assume you mean by "LDR") can be driven by whatever you wish; the LED is merely a forward-biased diode which happens to be an optical emitter too.Chris Daly said:
No idea what this sentence means.
Nothing is stopping you, apart perhaps from the technical scrutiny it might attract.
That is the paper I found yesterday.scott wurcer said: