Thanks Christer
Sorry if i was too lazy to look it up myself.
So there is an actual gain in using more than one LED to have the same voltage as a Zener even though you add the noises (geometrically).
Ex. 5.6vzener @ 20mA ~1.6uV
5xIRleds@ 20mA sqrt(5)*0.24 ~0.536uV
9dB less noise
I tried previously to find any kind of information on zener noise and the only thing i found was the cryptographic people using it as random generators or trying to get it as broadband as possible. Thanks again!
Sorry if i was too lazy to look it up myself.
So there is an actual gain in using more than one LED to have the same voltage as a Zener even though you add the noises (geometrically).
Ex. 5.6vzener @ 20mA ~1.6uV
5xIRleds@ 20mA sqrt(5)*0.24 ~0.536uV
9dB less noise
I tried previously to find any kind of information on zener noise and the only thing i found was the cryptographic people using it as random generators or trying to get it as broadband as possible. Thanks again!
hjelm said:Thanks Christer
Sorry if i was too lazy to look it up myself.
So there is an actual gain in using more than one LED to have the same voltage as a Zener even though you add the noises (geometrically).
Ex. 5.6vzener @ 20mA ~1.6uV
5xIRleds@ 20mA sqrt(5)*0.24 ~0.536uV
9dB less noise
I tried previously to find any kind of information on zener noise and the only thing i found was the cryptographic people using it as random generators or trying to get it as broadband as possible. Thanks again!
In many cases it seems it would be better to use a string of
LEDs. However, note that higher voltage zeners seem to have
a low noise, so it wouldn't pay off to replace a 12V zener
with LEDs it seems. OTOH these measurements are on a too
small scale to draw any statistical conclusions. I am currently
measuring som old LEDs from the 70's which seem to show
horrible performance, so more LEDs should be tested. Any
volunteers?
I've just measured some old 1970's LEDs from my junkbox
and the results are very different from the brand new ones
I measured before. I don't know if it is that LEDs in general
has changed a lot over the years or if it can actually differ
this much between types/brands/batches?
Attaching measurements for these.
and the results are very different from the brand new ones
I measured before. I don't know if it is that LEDs in general
has changed a lot over the years or if it can actually differ
this much between types/brands/batches?
Attaching measurements for these.
Would there be any point in anyone else measuring something?
Without calibrating your setup anyone else's results would not be comparable to yours.
They would however be showing relative differences between the types the second measurer is measuring so maybe not a waste anyway.
I think there is a possibility that the LEDs of other colors than red had lots of problems in the 70's, seem to remember they were harder to make.
Assume that the noisy red ones are degrading due to old age and that the green and yellow ones are inferior due to worse manufacturing methods back then.
Without calibrating your setup anyone else's results would not be comparable to yours.
They would however be showing relative differences between the types the second measurer is measuring so maybe not a waste anyway.
I think there is a possibility that the LEDs of other colors than red had lots of problems in the 70's, seem to remember they were harder to make.
Assume that the noisy red ones are degrading due to old age and that the green and yellow ones are inferior due to worse manufacturing methods back then.
hjelm said:
That was the point, although I wouldn't say my setup is
entirely uncalibrated. The major uncertainty factors are
the bandwidth of the soundcard (and probably the ADC)
and that I measured the level of the calibration signal with
my scope which has never been calibrated. It is not entirely
off, though. I would think the scope is at most some 5% off,
At least the manufacturing techniques has progessed. There
were nu blue LEDs at all then. There is also a risk that some
of my old LEDs have been treated badly by me, but hardly all
of them. Anyway, I have just come back from Elfa with some
more types of LEDs and zeners, so when I get time to do some
measurements I will post.
If someone else would be interested to contribute with
experiments, there are probably many things that could be
done. For instance, buying a larger number of the same
device, but at different times and/or from different suppliers
to check the variation between devices. It would also be
interesting if someone with access to other brands of LEDs
and zeners tried to repeat my experiments to see if the
results correlate with mine.
Jim Williams at Linear Tech has done several application notes dealing with the measurement of noise -- he uses a "Danish Butter Cookie" tin to isolate the device under test from the "world". It's really worth while perusing the application notes at Linear for the varied discussions of noise measurement.
Williams and others also have schematics for very low noise amplifiers and the high pass and low pass filters to bandwidth limit the measurements.
I doubt that you will find it "used" in Europe, but some of the Tek Analog scopes can measure down to a few uV when equipped with the 7A22N or 5A22N differential amplifiers -- I have a 5A22 in a TEK 5103 and it really works well -- they practically give away these units in the US, but shipping to Europe would be a fortune.
Batteries have noise too!
Williams and others also have schematics for very low noise amplifiers and the high pass and low pass filters to bandwidth limit the measurements.
I doubt that you will find it "used" in Europe, but some of the Tek Analog scopes can measure down to a few uV when equipped with the 7A22N or 5A22N differential amplifiers -- I have a 5A22 in a TEK 5103 and it really works well -- they practically give away these units in the US, but shipping to Europe would be a fortune.
Batteries have noise too!
Regarding diodes
I did check a 1N4148 in my preliminary tests and it was very
quiet. I forgot to include it in the real test suite. I will measure
some more LEDs and some more zener voltages when I get
time, and I'll try to throw in some ordinary diodes too as well
as TL431.
BTW, as can be seen in my schematic the CCSs use two
1N4148 and the CCSs seem to contribute very little noise.
Note however, that the noise of the CCS gets lumped into
the noise figure, as does the amplifier noise, but these are
at least the same for all DUTs.
Jackinnij,
thanks for the hints but I seem to get good enough results
for my purposes. I am at or very close to the theoretical
limt already and I can obvoiously distinguish between
components by measuring their noise. The main purpose
of the experiment was to get relative noise figures to try
extracting qualitative results, and that seems possible to
do (although the sample sizes are too small to get any
real statistical significance). Of course, if there is
anybody having better equipment and is willing to repeat
my experiments, that would be most welcome.
I did check a 1N4148 in my preliminary tests and it was very
quiet. I forgot to include it in the real test suite. I will measure
some more LEDs and some more zener voltages when I get
time, and I'll try to throw in some ordinary diodes too as well
as TL431.
BTW, as can be seen in my schematic the CCSs use two
1N4148 and the CCSs seem to contribute very little noise.
Note however, that the noise of the CCS gets lumped into
the noise figure, as does the amplifier noise, but these are
at least the same for all DUTs.
Jackinnij,
thanks for the hints but I seem to get good enough results
for my purposes. I am at or very close to the theoretical
limt already and I can obvoiously distinguish between
components by measuring their noise. The main purpose
of the experiment was to get relative noise figures to try
extracting qualitative results, and that seems possible to
do (although the sample sizes are too small to get any
real statistical significance). Of course, if there is
anybody having better equipment and is willing to repeat
my experiments, that would be most welcome.
Speaking of 1N4148, I notice on the data sheet that the 1n4448, which is siomilar in most ways has about 1/2 the capacitance. I wonder if that relates to relative noise in any way.
Also, I'm not trying to make work for anyone but also wonder about consitancy (voltage drop and noise) unit-to-unit in specific LEDs. I wonder because the device is designed to produce light and the use as a voltage reference is only incidental.
Also, I'm not trying to make work for anyone but also wonder about consitancy (voltage drop and noise) unit-to-unit in specific LEDs. I wonder because the device is designed to produce light and the use as a voltage reference is only incidental.
Extended report w. more measurements
OK I admit it, I am crazy. I have spent many hours tonight
doing more noise measurements.
First, forget that update I posted using old junkbox LEDs.
The results seem to have been erroneous due to a battery
going flat. Since I bought some more types of LEDs and
zeners today, I used these and skipped the old LEDs
altogether. An extended report is attached.
Although the data sample set is still quite small and doesn't
warrant any statistically significant results, there seem to
be certain consistencies in the results that one may perhaps
dare to generalize from.
* It seems not to matter much what LED one uses as long as
it is the same colour (I didn't check any high-power ones,
though).
* If you can get the desired voltage drop with one single
diode or LED (excluding blue LEDs) then it seems usually
best to use that single device. Special requirements for
the current may affect the choice, though. For higher voltages
up to or just below 12V, string together a suitable number
of red, yellow or green LEDs. From 12V upwards, use a
zener diode (or several if required).
* Avoid blue LEDs.
* Ordinary diodes, IR diodes and blue LEDs seem to get less
noisy the hotter they run, ie. the more current you use
(don't fry them, though).
* Red yellow and possibly green LEDs seem to have an
optimal current that minimizes the noise. This test was
not set up so it could properly find that optimum.
* The data does not support any conclusions whether
high-power zeners are more or less noisy than low-power
ones.
OK I admit it, I am crazy. I have spent many hours tonight
doing more noise measurements.
First, forget that update I posted using old junkbox LEDs.
The results seem to have been erroneous due to a battery
going flat. Since I bought some more types of LEDs and
zeners today, I used these and skipped the old LEDs
altogether. An extended report is attached.
Although the data sample set is still quite small and doesn't
warrant any statistically significant results, there seem to
be certain consistencies in the results that one may perhaps
dare to generalize from.
* It seems not to matter much what LED one uses as long as
it is the same colour (I didn't check any high-power ones,
though).
* If you can get the desired voltage drop with one single
diode or LED (excluding blue LEDs) then it seems usually
best to use that single device. Special requirements for
the current may affect the choice, though. For higher voltages
up to or just below 12V, string together a suitable number
of red, yellow or green LEDs. From 12V upwards, use a
zener diode (or several if required).
* Avoid blue LEDs.
* Ordinary diodes, IR diodes and blue LEDs seem to get less
noisy the hotter they run, ie. the more current you use
(don't fry them, though).
* Red yellow and possibly green LEDs seem to have an
optimal current that minimizes the noise. This test was
not set up so it could properly find that optimum.
* The data does not support any conclusions whether
high-power zeners are more or less noisy than low-power
ones.
Put the leds in the dark and measure again.
LED's will generate a small voltage if there is ambient light falling on the junction (actually quite useful as an opto-isolator if you face a pair of narrow-focus leds toward each other) and I was thinking that the light might affect the noise figure, particularly if it is fluorescent light with it's 2x mains frequency modulation.
LED's will generate a small voltage if there is ambient light falling on the junction (actually quite useful as an opto-isolator if you face a pair of narrow-focus leds toward each other) and I was thinking that the light might affect the noise figure, particularly if it is fluorescent light with it's 2x mains frequency modulation.
Nice job Christer
I didn’t read this whole lengthy thread so maybe it is already mentioned. When I compare your results of the less noisy LED with the less noisy zener I find app. 0.3 uV for the LED and the same 0.3 uV for the 12V zener.
However, given the 1.5V across the LED and the 11.5V across the zener, I come to a normalised noise voltage of app. 8 times lower for the 12V zener. I think it is more appropriate to normalise all noise voltages to 1V zener voltages for a fair comparison.
Btw maybe you can measure also a reverse biased b-e junction of a BC559, at 1 mA that gives app 6.8V.
Cheers
I didn’t read this whole lengthy thread so maybe it is already mentioned. When I compare your results of the less noisy LED with the less noisy zener I find app. 0.3 uV for the LED and the same 0.3 uV for the 12V zener.
However, given the 1.5V across the LED and the 11.5V across the zener, I come to a normalised noise voltage of app. 8 times lower for the 12V zener. I think it is more appropriate to normalise all noise voltages to 1V zener voltages for a fair comparison.
Btw maybe you can measure also a reverse biased b-e junction of a BC559, at 1 mA that gives app 6.8V.
Cheers
Re: Put the leds in the dark and measure again.
Yes, right. That is one of probably several things I forgot.
Actually, even glass encapsulated diodes, like 1N4148, are
light sensitive. I have read recommendations to paint them
with black paint somewhere. Maybe one should do that with
LEDs also. Hm, I think the zeners I used are glass encapsulated
too.
I won't rerun all the measurements, it is taking to much time
and batteries, but I should do a test in and out of darkness
for one or a few LEDs and zeners, I guess.
Circlotron said:
Yes, right. That is one of probably several things I forgot.
Actually, even glass encapsulated diodes, like 1N4148, are
light sensitive. I have read recommendations to paint them
with black paint somewhere. Maybe one should do that with
LEDs also. Hm, I think the zeners I used are glass encapsulated
too.
I won't rerun all the measurements, it is taking to much time
and batteries, but I should do a test in and out of darkness
for one or a few LEDs and zeners, I guess.
Pjotr said:Nice job Christer
I didn’t read this whole lengthy thread so maybe it is already mentioned. When I compare your results of the less noisy LED with the less noisy zener I find app. 0.3 uV for the LED and the same 0.3 uV for the 12V zener.
However, given the 1.5V across the LED and the 11.5V across the zener, I come to a normalised noise voltage of app. 8 times lower for the 12V zener. I think it is more appropriate to normalise all noise voltages to 1V zener voltages for a fair comparison.
Btw maybe you can measure also a reverse biased b-e junction of a BC559, at 1 mA that gives app 6.8V.
Cheers
There would be a point in normalising the noise figures as
you say, however, I present the measurements just as
raw data. For a given application you will anyway have to
figure out which component of combination of components
gives the least noise for the voltage drop you need. As I
pointed out in the "recommendations" from 12V and probably
upwards, zeners seem the best, but below that one or
several LEDs.
I didn't know about the zener behaviour of BC559. I thought
about testing it forward biased, but thought it would probably
just be like any other diode. I might do a test reversed, but I
have to limit the tests. It is taking a lot of time and I don't
want to spend a fortune on batteries.