Amplifying 2kHz weak signals (300nv), how to choose opamp?

[Terminator3]

Please give me ideas how to amplify such weak signals? I am thinking about using one expensive opamp with high GBW and low noise, or two opamps. With two opamps first stage is something low noise, an second stage is LM358, used as two series opamps in one package. I want to understand which parameter of opamp can be used to estimate its noise amplitude relative to input signal amplitude.

 

[D.A.(Tony)Stewart]

What is the spectral BW? 0 to 2kHz or 2kHz +/-?

Input noise: 0.65 nV/√Hz for TSH300 is one option.

Hardly expensive @$3 (1pc)

 

[KlausST]

Hi,

My first question is: What do you want to do with the amplified signal?
--> If you just want to measure it's amplitude: try if a lock in amplifier can be used.

***
The next question is: Where does the signal come from, and all the specification for the signal and your application.
Source impedance? Is it a fixed frequency? Do you actively generate this frequency? What other frequencies exist, that need not to be amplified (suppressed)?

****
The rest is guessing:

If you really need the analog waveform, then you need a true amplifier.
Whether to use a single Opamp or two stages depend on your needs.

I imagine you want it to amplify to get a signal of about 2Vpp, then you need a gain of about 1 million.
Oh yes, that's huge.. A single stage won't work..

Klaus

 

[Terminator3]

Signal is 0..2kHz, signal source is schottky diode mixer. Then I use fft with sliding window to analyze relative phase changes for all frequency bins.

 

[KlausST]

Hi,

you say 0..2kHz.
Then the biggest problem will be the 0Hz. To amplify such a small signal with DC stability will be difficult.

Klaus

 

[Terminator3]

There already too high noise near 0 kHz, around 4bins of 512pt fft spectrum. 2kHz/512*4. 16hz maybe. Sampling at 4kHz. Higher frequencies are more important to me.

I want to put two opamps in parallel, then compare amplified signals. If noise is correlated, then it is mixer noise. But if it is independent, then (I hope) it is opamp noise.

 

[D.A.(Tony)Stewart]

• DC input offset is 1mV at DC
• <1nV/Root Hz BW) Noise is rated at 100kHz
• 1/f noise is huge towards DC

You would need to compute Signal to Noise Ratio from signal vs signal and estimate Shannon's Bandwidth required to give > 10 dB S/N ratio. with very low f , and PLL.

I had a similar problem with VLF Doppler design but capture range and noise bandwidth are tradeoffs and p/f mixers introduce edge noise.

There must be a better way.

Read near end of specs for TSH300 on noise.

When NASA receives signals from furthest satellite the signal is so far below noise, that the bit rate must be extremely low <<<1 Hz and RBW equally low with precision 3 stage IF, BPF , PLL
Maybe you consider Diode modulating GaAs VCO and measure spectrum of output.
Why is signal so small?

 

[KlausST]

Hi,

Input noise: 0.65 nV/√Hz for TSH300 is one option.

I see three problems with this AMP.
1) open loop gain of 67dB is a bit low, the circuit needs very high gain.
2) noise is measured at 100kHz. Noise at the desired range of 0...2kHz is much much higher (high 1/f corner frequency). About 1.4 nV/√Hz @ 1kHz, about 5 nV/√Hz @ 100Hz.
3) offset voltage of 0.5mV

***
LT1028:
1) gain: 7 mio = 135dB
2) noise: 0.9 nV/√Hz @ 1kHz, 0,9 nV/√Hz @ 100Hz
3) Offset: 20uV

LME49990:
1) gain: 135 dB
2) noise: 0.9 nV/√Hz @ 1kHz, 1.0 nV/√Hz @ 100Hz.
3) offset: 130uV

***
If you need low offset voltage (drift):
I recommend to use an additional chopper stabilized OPAMP to stabilize one of the above OPAMPs DC offset.

Klaus

 

[FvM]

I want to put two opamps in parallel, then compare amplified signals. If noise is correlated, then it is mixer noise. But if it is independent, then (I hope) it is opamp noise.

Read a bit about theory of random signals and realize that the idea doesn't work.

Using two amplifiers in parallel can reduce the total noise by 3 dB, presumed the source impedance is low enough that the OP input noise current doesn't contribute much to total noise. Otherwise, the parallel circuit might even decrease the signal-to-noise ratio.

Low-noise amplifiers are optimal only for a specific source impedance range. Optimal source impedance for LT1028 is below 1 kohm. You should also consider that gain setting resistors must be very low not to ruin the low amplifier noise voltage.

 

[Warpspeed]

signal source is schottky diode mixer.

Would it be possible to amplify the signal before the mixer ?
That may introduce some significant problems too, but they may be more manegable.