I spent the day measuring some LEDs and got higher noise voltages and especially
high corner frequencies.
Setup: FFT analyzer is an Agilent 89441A vector signal analyzer, test object is in an
Al box, preamp also, they are connected with semi rigid coax and put into a large
aluminium box that has BNC feedthroughs to the FFT analyzer. The object and
the preamp are only connected with the semi rigid; every other connection would
form a loop and make a lot of interference appear. They are isolated by books
that are a good foundation in the figurative sense also: Motchenbacher-Conelly:
Low Noise Design and The Art Of Electronics, 3rd ed. Get them.
The preamp is powered by another 10 NiMH cells.
I have written a program that runs under Linux and controls the 89441A over the
network. It does one FFT per decade, corrects for the varying noise bandwidth
and ties the 7 decades together to a single log/log plot.
0 dB in the plots is a noise density of 1 nV/sqrt Hz or -180dBV, the voltage noise
density of the preamp is 220 pV/sqrt Hz or abt. -193 dBV. 1nV/sqrt Hz is only
slightly worse than what a single AD797 or LT1028 can achieve. The preamp
that is used averages over 20 such op amps. I have described that already.
In the setup picture, the traces are: A 47R wire wound resistor at the amplifier
input, the 47R resistor dc fed from 10 NiMH cells via 2KOhm wire resistor.
You can see that the DC bias does not introduce additional noise. It will be
required for the bias in the real device tests.
The last trace is the amplifier with input shorted; the horizontal part is 220 pV/sqrt Hz.
The next plot is the LEDs, Blue = Osram Blue TOP LED (Osram=ex-Infineon,
ex-ex-Siemens) LBT676.
red= Osram Top LED LST676 633 nm
It is possible that I have swapped the 1K and the 2K cases. Anyway, the
amount of current seems to make quite a difference.
TLRU1008 = Toshiba, red, really tiny.
While I was at it, I also tested some "Zeners". I now have no time to re-test the 3V3.
It is from a tape from a different source than the others. If the results hold true, then
methinks I have found a new friend. I must re-test that, but no time left today.
It seems that the LEDs can be quite good at high frequencies, but then, filtering is
cheap there. I am impressed by the BZX84. It kinda rocks my view of the world.
More investigation required.
regards, Gerhard
Attachments
Last edited:
Man, You've got a sound card. That and an AD797. Just measure it and decide for
yourself; why repeat the pollution of some self-declared audio gods. It just
costs a weekend to make the transition from a believer to a man with knowledge.
regards, Gerhard
That i do when *I* need.
Taking new ideas when i can find some (trying to keep an open mind) and veryfying the assumptions behind by myself.
My remarks were more for having a nice atmosphere in this club. Making measurements just to take over in a controvery seems to me a lost of time.
(I don't say that for your last measurements, and I thank-you for sharing them)
Could you say more? I know that the 1/f region is steeper than expected,
but I don't have much influence on that. When I generate noise that is flat
down to DC ( R noise and 50 or 100 MHz wideband amplifier +50 dB and Schottky
ring mixer for down conversion using R&S SMHU or HP8662A as LO) --> this is
faithfully reproduced as FLAT in my plots..
So the steep rise really must come from the test objects.
That must have been what caused my excess noise in the string of 16 HLMP-6000 parts: I used Motchenbacher and Fitchen, and Art of Electronics 2nd ed. as supports. Hopelessly outdated.
You guys are never going to get J.C. to agree that the effect is meaningless, because his basic philosophy is that everything matters.... even the stuff others can prove does not.
Speaking of optocouplers, has anyone tested an IR emitter for noise yet? They used to be prevalent in optos, although I think the development of red superbrights may have displaced them. The giveaway is the forward drop being lower for the IR ones.
By comparison to the curve for the TLSU1008, I'd think magenta is correct (1k also).
The slopes blend into each other, whenever you have a slope softening with decreasing frequency after rising at a higher rate there is some Lorentzian component of the noise. I have been staring at these for 40 years.
Gerhard,
You also showed something else of interest. The zener diode data sheet shows what they call differential resistance. It does not correlate with your noise measurments.
In the past I have looked at I/V curves and chosen my voltage references based on the lowest slope (least equivalent series resistance model.) This turned out to be a string of 1N4148 diodes.
1N4148 diodes are a bit unusual in that they have some gold doping to increase the switching speed although it increases reverse leakage, that is not an issue here.
You also showed something else of interest. The zener diode data sheet shows what they call differential resistance. It does not correlate with your noise measurments.
In the past I have looked at I/V curves and chosen my voltage references based on the lowest slope (least equivalent series resistance model.) This turned out to be a string of 1N4148 diodes.
1N4148 diodes are a bit unusual in that they have some gold doping to increase the switching speed although it increases reverse leakage, that is not an issue here.
Thanks guys for putting forth a solution for my LED indicator. I already knew this of course, but in all fairness, my best designs run so warm, and are designed for 24 hour operation, so they don't even have an OFF/ON switch, so an indicator light is just not mandatory, because touching the chassis shows whether it is operating or not.
These are wonderful and cheap we use lots and they never seem to break. My latest DIY uses one coupled with parts from AD (823 is best) or TI etc.
Black tape won't block IR totally but 1/4" Aluminum does.
Just flipped through chapter 12 on LEDs/lasers...didn't see anything on collection, it's all emission-related. Haven't read ch 13.
- Status
- Not open for further replies.