Simpler PI...

[clancy]

Was wondering when you going to thinks you forget current noise.
J1 = (Johnson noise R1)^2/Hz, J2 = (Johnson noise R2)^2/Hz

N1= (measured noise R1)^2, N2=( measured noise R2)^2, N0= (measured noise 0 Ohms)^2
(Amplifier voltage noise at input)/sqrt(Hz) = sqrt{[J1(R2/R1)^2-J2]/[(R2/R1)^2(N1/N0-1)-N2/N0+1]} = 0.65nV/sqrt(Hz). This math independent of gain as should.

[clancy]

BTW, this preamp noise figure poor compared to good low noise opamp because of very high noise current = sqrt([N1(J2^2+V^2)-N2(J1^2+V^2)]/[N2R1^2-N1R2^2]) =7.8pA/sqrt(Hz), V=preamp voltage noise.

[Aziz]

Was wondering when you going to thinks you forget current noise.
J1 = (Johnson noise R1)^2/Hz, J2 = (Johnson noise R2)^2/Hz

N1= (measured noise R1)^2, N2=( measured noise R2)^2, N0= (measured noise 0 Ohms)^2
(Amplifier voltage noise at input)/sqrt(Hz) = sqrt{[J1(R2/R1)^2-J2]/[(R2/R1)^2(N1/N0-1)-N2/N0+1]} = 0.65nV/sqrt(Hz). This math independent of gain as should.

Bravo Clancy,

I'm very impressed! I get the same voltage noise density e_n = 0.649 nV/sqrt(Hz). And current density i_n = 7.835 pA/sqrt(Hz).
The formulas are somewhat f**kn complex (see below). Anyway. I have to check them once again. But I'm satisfied to get the same voltage density result.

Who would have thought, that my first simple measurement method would fail so much? The current noise can't be neglected with resistor noise measurements. I'm happy, that we have now a nice & more accurate measurement method.

Excel-Formula for en:
=WURZEL(_c*_Ri*(_Ri/_Rp - 1) / ((H86*H86-1)*_Ri*_Ri/(_Rp*_Rp) - I86*I86 + 1))*1000000000

Excel Formula for in:
=WURZEL((J86*J86/1000000000000000000*(H86*H86-1) - _c*_Rp)/(_Rp*_Rp))*1000000000000

WURZEL = sqrt()-function
c = 4*k*T

Cheers,
Aziz
PS: Congratulations, you have won that great prize!!!

[Aziz]

But this is a very bad news for the 1k series resistor fraction.
In the example above of i_n = 7.8 pA/sqrt(Hz):
i_n is causing 7.8 pA * 1000 Ohm = 7.8 nV voltage noise. This is all screwing the performance of the ultra-low-noise thing.


Anyway. I'm looking forward to the new set of more accurate measurements.
Cheers,
^sif

[Aziz]

BTW, this preamp noise figure poor compared to good low noise opamp because of very high noise current = sqrt([N1(J2^2+V^2)-N2(J1^2+V^2)]/[N2R1^2-N1R2^2]) =7.8pA/sqrt(Hz), V=preamp voltage noise.

Thanks for providing the current density too. I just saw it. I get the same result.
Aziz

[Aziz]

Hi all,

I have updated my Excel file and have put a few measurements this morning (see below zip-file). So you can check, whether I have calculated it correctly. I have to check it yet. But these noise figures make more sense now!
Ok, up that new goal to 0.5 nV/sqrt(Hz). Oh boy!, this is going to be very very difficult now.

Excel zip-file:
AC-Amplifier-Noise-Measurements-01.zip

Cheers,
Aziz,
onto way to ultra low noise heaven.

[Aziz]

Interesting. The 2x 2N 3904 is performing very well for very low source impedances.
And very cheap (0.04 EUR per BJT).
Again the poor performance of the BC 550C. Forget it! Throw it out of the window. *LOL*


Cheers,
^sif

[clancy]

Thanks Aziz.
Actually, these useful measures because it give measurement bandwidth = 85.5Hz. This mean that your measures in #96 (Shorted-Input-0-Ohm-Gain206.gif), preamp noise be 1.3nV/sqrt(Hz) at 1kHz, and 0.94nV/sqrt(Hz) at -115dB.

[Aziz]

Thanks Aziz.
Actually, these useful measures because it give measurement bandwidth = 85.5Hz. This mean that your measures in #96 (Shorted-Input-0-Ohm-Gain206.gif), preamp noise be 1.3nV/sqrt(Hz) at 1kHz, and 0.94nV/sqrt(Hz) at -115dB.

Hi Clancy,

my latest noise floor level decibel measurements were taken from the flat noise region of the frequency spectrum (averaged over 6 kHz to 30 kHz). I don't look at the low frequency region at the moment. The low frequency region noise is larger of course. I'm focussing to the approaching noise density level (best noise figure). In principle, I could measure the whole frequency spectrum from 50 Hz to 48 kHz with FFT-bin size frequency resolution and calculate the noise density for each.
The flat region is often beginning earlier (>1 kHz). But some transistors/op-amps causing it happen later. I could lower it to 3 kHz but I think, it's not required.
Cheers,
Aziz

[Aziz]

Did we "invent" a new measurement method?

Trivial. I don't think so!
^sif

[Aziz]

Latest Measurements

Hi all,

this is the latest measurement list (see below the Excel zip-file). I have to make more measurements with different collector currents yet (only for some interesting BJT's of course).

AC-Amplifier-Noise-Measurements-04.zip

See, how a parallel of super cheap 2N 3904's excels.


Cheers,
Aziz

[vbeeeks]

Found one quick reference for noise measurements:
http://www.4shared.com/office/ZSM__s...Lowest-No.html

[Aziz]

Found one quick reference for noise measurements:
http://www.4shared.com/office/ZSM__s...Lowest-No.html

Thanks! I couldn't load it so I have found another source:
http://www.filestube.to/5KSStganB10XkxmAT3MJoA
or

http://depositfiles.com/files/wbtg5uoxv

Aziz

[Aziz]

Hi all,

I'm very excited about the 2N 3904!

Total collector current 5.8 mA (seems to be the best value so far). I have added a complementary NPN/PNP BJT driver stage to the output of the op-amp.

Results:
4x 2N 3904 + NE5534A (Gain resistors: 470 Ohm, 2.2 Ohm):
en = 0.376 nV/sqrt(Hz), in = 6.356 pA/sqrt(Hz)

3x 2N 3904 + NE5534A (Gain resistors: 470 Ohm, 2.2 Ohm):
en = 0.398 nV/sqrt(Hz), in = 6.505 pA/sqrt(Hz)

4x 2N 3904 + NE5534A (Gain resistors: 1k Ohm, 2.2 Ohm, more gain):
en = 0.284 nV/sqrt(Hz), in = 6.051 pA/sqrt(Hz)
(Cheap trick of course! *LOL*)

How is that now?


I'll start the PNP prototype soon. Or investigating another NPN BJT. But the MJE 13007 doesn't like paralleling. I can use only one.
Cheers,
^sif

[Davor]

Emitter degeneration is beneficial for 1/f noise. It is a proven fact for oscillators and it can't hurt to have some, provided it does not ruin the noise performance. In your setup emitter is in a feedback path so the emitter impedance is actively lowered by feedback in case of a signal, but not so much for the error (1/f noise etc.)
Current noise is not a problem for low impedance sources, and also the source - a coil - has a finite resistance and hence noise. There is something called "optimum impedance" that occasionally pops here and there, and it is simply ein/iin (voltage noise/current noise) and your input impedance is ideally matched if it is near it.