Bricolo said:
Paint them. As I just said in a recent post, I have seen a
recommendation somewhere to cover 1N4148s with black
paint. I have also seen a design in an app note where to
1N4148s should be mounted in close proximity, which I
assume is to guarantee that they get an equal share of
ambient light, although in that case it is most probably
just to ensure they don't get different leakage current
due to light.
jam said:
Excellent work! You have just helped me answer some questions that I have been working on in my design..............
What current did you run the 1N914 at for lowest noise? And have you tried a small signal transistor strapped as a diode?
Well, I think these questions have been hanging in the air all
the time I've been on the forum and never got satisfactoy
answers. Not that I claim this to be the final or "correct"
answers, but hopefully we know a little bit more now.
I suppose you mean the 1N4148? I did all tests at three
currents: 1, 5 and 20 mA. Ordinary diodes showed the
lowest noise at 20 mA.
As I just replied to Pjotr, I thought about trying a BJT as
a diode but assumed it wouldn't differ much from a plain
diode. Pjotrs claim that they can work as zener diodes
in reverse sounds interesting to try though. Hm, I wonder
if they will survive being used like that in the long run, but
maybe they will as long as the current is limited.
Re: What about impedance?
The impedance of the DUT probably matters, but is something
you cannot do anything about. I haven't tried to meause it,
but I would guess all diodes have a low enough dynamic
resistance that it is not the main contributing factor to the
noise. That is just a guess though. The thing one can do
something about is the current sources, but at least I use
the same ones for all devices I test. Anyway, the main point
of the exercise was not to get accurate absolute noise figures,
but to try getting some qualitative information about what
device is best under what circumstances. After all, if you find
that for instance a red LED is the least noisy choice for you,
it doesn't matter that much how noisy it is in absolute figures
since it is the best you can get anyway.
boholm said:Christer:
Another important (IMHO
The impedance of the DUT probably matters, but is something
you cannot do anything about. I haven't tried to meause it,
but I would guess all diodes have a low enough dynamic
resistance that it is not the main contributing factor to the
noise. That is just a guess though. The thing one can do
something about is the current sources, but at least I use
the same ones for all devices I test. Anyway, the main point
of the exercise was not to get accurate absolute noise figures,
but to try getting some qualitative information about what
device is best under what circumstances. After all, if you find
that for instance a red LED is the least noisy choice for you,
it doesn't matter that much how noisy it is in absolute figures
since it is the best you can get anyway.
jam said:
Yes the 1N4148 and the 1N914 are similar. Mr.Pass used a transistor as a zener in the Stasis amplifiers. I have been told that a transistor wired as a diode is quiter but then again, you are the guy with the test set up.....................
It might be worth trying then.
Well I am the one with the test set up right now, but it would
be very useful if someone tried to repeat my experiments to
get a double-check on the results. Empirical results must be
repeatable to be of any value.
have you looked at this paper from Dr. Leach? :
http://users.ece.gatech.edu/~mleach/ece6416/noisepot/tablecon.pdf
diode noise:
http://users.ece.gatech.edu/~mleach/ece6416/noisepot/chap07.pdf
http://users.ece.gatech.edu/~mleach/ece6416/noisepot/tablecon.pdf
diode noise:
http://users.ece.gatech.edu/~mleach/ece6416/noisepot/chap07.pdf
Light sensitivity
Christer,
if I remember correctly the light sensitivity of 1N4148's and the like are discussed briefly in Analog Devices AN-280 "Mixed Signal Circuit Techniques". The issue there was that clamping diodes protecting an instrument amplifiers inputs injected 100/120 Hz noise and AD received claims that PSRR of that particular amp was out of spec (which it was not).
Good work!
/Magnus
Christer,
if I remember correctly the light sensitivity of 1N4148's and the like are discussed briefly in Analog Devices AN-280 "Mixed Signal Circuit Techniques". The issue there was that clamping diodes protecting an instrument amplifiers inputs injected 100/120 Hz noise and AD received claims that PSRR of that particular amp was out of spec (which it was not).
Good work!
/Magnus
Christer said:
Hi Christer,
Yes they do, provided you limit the current so the b-e junction does not overheat. In fact you use the junction in avalanche mode. Have seen the trick in some IC’s for reference purposes but also as a wide band noise source
But nevertheless I am curious about the real values.Regarding impedance. It can be that the dynamic resistance IS the main contributing source of the noise. So it can be interesting to measure also LED’s made for high pulsing applications. These have usually low dynamic resistance.
Cheers
jackinnj said:
Thanks, I have never seen that one and it never came up in
any of my googlings for noise papers earlier. It looks like a
quite comprehensive write-up on the theory of noise.
Something to download for reading on a rainy day perhaps
(hm, we seem to get plenty of this this summer), or rather,
a rainy month.
I just had a brief look at it and it seems to have no empirical
data which is what I was mainly interested, surprising as it
may sound to some
. I suppose it could be interesting toread the theory and see if it correlates with my measurements.
However, my primary purpose was to get some practical
results as a guide for how to choose voltage references.
More measurements
First, I checked the influence of light for a 1N4148 and for
a red LED. It didn't matter if I had a light bulb on close by
or if I shut it off and put an opaque bin upside down over
the circuit. I also have a fluoroscent tube that might have
been on during some of the measurements, so I checked
if that had any influence. It might have a slight influence,
but it is so small that it would be on the order of 10 nV or
so, and wouldn't really affect any interpretation of the
results.
Now for the new measurements. I've tried TL431 strapped
for its standard voltage, 2.5V, and for 5V. It turned out
quite noisy in both cases, but the noise figures turned out
very consistent. This does not seem like a good choice if
noise is an important issue. I also tried BC549 with the
BE diode forward and reverse biased. In the forward mode
it is possibly slightly less noisy than the other plain diodes
I measured, but this is very close to the resolution of my
test set up, so one should be cautios to draw conclusions
from it. It is at least not worse to use a BC549 and it is
possibly better. In reverse bias it had a voltage drop
comparable to an 8.2V zener diode, but was somewhat
less noisy. It still cannot compete with a string of LEDs, though.
Revised report attached. All new measurements at the
end under heading Miscellaneous.
First, I checked the influence of light for a 1N4148 and for
a red LED. It didn't matter if I had a light bulb on close by
or if I shut it off and put an opaque bin upside down over
the circuit. I also have a fluoroscent tube that might have
been on during some of the measurements, so I checked
if that had any influence. It might have a slight influence,
but it is so small that it would be on the order of 10 nV or
so, and wouldn't really affect any interpretation of the
results.
Now for the new measurements. I've tried TL431 strapped
for its standard voltage, 2.5V, and for 5V. It turned out
quite noisy in both cases, but the noise figures turned out
very consistent. This does not seem like a good choice if
noise is an important issue. I also tried BC549 with the
BE diode forward and reverse biased. In the forward mode
it is possibly slightly less noisy than the other plain diodes
I measured, but this is very close to the resolution of my
test set up, so one should be cautios to draw conclusions
from it. It is at least not worse to use a BC549 and it is
possibly better. In reverse bias it had a voltage drop
comparable to an 8.2V zener diode, but was somewhat
less noisy. It still cannot compete with a string of LEDs, though.
Revised report attached. All new measurements at the
end under heading Miscellaneous.
You better measure the dynamic impedance, bub!
It affects the PSRR if you are using one of these devices as a voltage reference.
Lowest noise and highest PSRR are nowhere near the same point.
Once you know both paramaters, and your design objective, you have to make a choice...................
Jocko
It affects the PSRR if you are using one of these devices as a voltage reference.
Lowest noise and highest PSRR are nowhere near the same point.
Once you know both paramaters, and your design objective, you have to make a choice...................
Jocko
Well, this a measurement for noise only, which is also why I
used batteries.
For PSRR, we want to know the dynamic impedance of both
current source and voltage reference, I think. If we had a
perfect current source then the dynamic impedance of
the voltage reference wouldn't matter as far as I can see
(but maybe I am short-sighted?)
The dynamic resistance of the various voltage references I
tried could at least be estimated from the voltage drop
measurements I did. But as always, there are many more
things one can measure. Every answer always tend to
raise a number of new questions begging for answers.
used batteries.
For PSRR, we want to know the dynamic impedance of both
current source and voltage reference, I think. If we had a
perfect current source then the dynamic impedance of
the voltage reference wouldn't matter as far as I can see
(but maybe I am short-sighted?)
The dynamic resistance of the various voltage references I
tried could at least be estimated from the voltage drop
measurements I did. But as always, there are many more
things one can measure. Every answer always tend to
raise a number of new questions begging for answers.
Final call
I only have the test set up on a solderless breadboard so I
will probably tear it apart soon. Is there anything you think
I absolutely have to measure before taking it down?
I might perhaps rebuild it slightly to do some other
measurements rather than tear it up entirely. I haven't
decided yet. I am primarly thinking of a variable current
source or more current sources to measure noise vs.
current for LEDs since there seems to be some optimal
value.
I only have the test set up on a solderless breadboard so I
will probably tear it apart soon. Is there anything you think
I absolutely have to measure before taking it down?
I might perhaps rebuild it slightly to do some other
measurements rather than tear it up entirely. I haven't
decided yet. I am primarly thinking of a variable current
source or more current sources to measure noise vs.
current for LEDs since there seems to be some optimal
value.
