You are taking about 3 *identical* devices in series. Identical devices wasn't mentioned in the post I replied to. When talking about LEDs as reference voltage, possibly different LEDs have different noise and possibly a certain LED of higher voltage may have lower noise than another certain LED of lower voltage.
So what? Noise voltage still sums as power. Two red LEDs, for example, will have similar Vf and en. And the noise is uncorrelated unless you're invoking miniature demons. So the 3dB s/n advantage is just basic physics.
Again, I'm not following your point- do you understand the difference between voltage and power summations?
Again, I'm not following your point- do you understand the difference between voltage and power summations?
Agree. They don't even have to be three LEDs, it can be a LED, a zener and an integrated ref. Noise (uncorrelated) adds with power, the ref voltages add linearly.
Although, I think that the minimum of 3dB for a pair is approached closer if the noise sources are approximately the same.
jan
Last edited:
That is normally not the case. Most higher voltage references are build from a basic low voltage reference (a so called 'bandgap reference') with a build-in amplifier. So the higher voltages ones have generally higher noise in proportion to the internal gain.
In such a case, a series of lower voltage (and noise) refs have less total noise.
jan
Bandgap references are a poor choice for applications where noise is important, a quite unnecessary degradation.
Buried zener references have as much as 20x better noise performance, and this can be improved further by increasing their working current.
So, Joshua is right - a 10V buried zener, for example, will easily surpass the noise performance of a 1,2V bandgap, despite the higher dc voltage. The difference is that a high quality buried-zener is designed for low-noise, whereas a bandgap is not.
Please read the old Application Notes of Jim Williams (God bless 'is soul) in the Linear Technology AN library, for the full story.
That's not the interesting question. The interesting question is: Where is the money?
oohps, no, not here: on what DC voltage?
More current is easy. Just parallel enough of these thingies.
1 nV/sqrt Hz on a 100 mV source is beginner level. Divide what you have and conquer
if you have something halfway reasonable.
(exaggerating.. 100mV is pathetic.)
1 nV/sqrt Hz on a 10V source requires some thinking, but can be done for less than
€25 on Digi-Key parts. Just remember that 4 AD797s give you 500pV/sqrt Hz and that
that bounds your loop error budget. Spend the rest on your reference. A LT6655
should be a good start, still needs some filtering. Or a string of red LEDs.
OK, make that two ADA4898-2s for the amplifier to meet the cost limit.
If you need more than 15V for your line level input stage, rethink your concept.
Or prepare to buy fat capacitors.
regards, Gerhard
No, it is true for all cases. It doesn't matter if it's a stack of red LEDs, a stack of TL431s, a stack of 12V zeners, whatever you choose. Stacking increases s/n. Period.
You can't wave away physical first principles.
I bet lots of us would be interested in your assessment, like a relative weighting, of the importances of low vs. resistive source impedance from the power supplies to the signal pass stages. Could also be described as optimum bandwidth of the power supply. You've given some practical clues, but a more formal overview would be welcome, if possible.
Thanks very much,
Chris
Thanks very much,
Chris
That's the input-referred noise voltage density spec. Will that be your output noise?
I would assume that you'd double the input current noise density as well, which will make the reference slightly trickier.
- Status
- Not open for further replies.