Preamp noise

[Mechanic]

Hi Guys,

So, I'm having a bit of a discussion with a friend regarding amplifier noise figures.
As a general rule we want a low noise preamp, but which bands of noise figure are actually important for PI detectors?

I say that the 0.001 to say 3 or 4 hz is not that important as this noise should be cancelled by the sampling, just as earth field is cancelled.

That being said, if an amplifier has a good noise figure down to 10hz, then in general the lower band of noise is also good.
Many amps with a good advertised noise floor start to get noisy under 100hz.

Bear in mind that I am referring to Minelab type front ends with low input resistance <10ohms and amplifier noise is dominant.
Replacing the ad797 with an Ne4435 definitely makes the detector more noisy.

Cheers Mick

[Altra]

Bear in mind that I am referring to Minelab type front ends with low input resistance <10ohms and amplifier noise is dominant.
Replacing the ad797 with an Ne4435 definitely makes the detector more noisy.

Cheers Mick

Hi Mick, when you say noisy can you describe it? Is it a threshold wobble, or snap crackle pop?

I just got a copy of Douglas Self's "Small Signal Audio Design 2nd ed." Great book, has quite a bit of info on different high end op-amps and noise. Design of super low noise preamps. Good read.

One source

https://www.ebay.com/itm/Small-Signa...4AAOSwgX9ct-bH

[green]

Hi Guys,

So, I'm having a bit of a discussion with a friend regarding amplifier noise figures.
As a general rule we want a low noise preamp, but which bands of noise figure are actually important for PI detectors?

I say that the 0.001 to say 3 or 4 hz is not that important as this noise should be cancelled by the sampling, just as earth field is cancelled.

That being said, if an amplifier has a good noise figure down to 10hz, then in general the lower band of noise is also good.
Many amps with a good advertised noise floor start to get noisy under 100hz.

Bear in mind that I am referring to Minelab type front ends with low input resistance <10ohms and amplifier noise is dominant.
Replacing the ad797 with an Ne4435 definitely makes the detector more noisy.

Cheers Mick

I agree, the integrator and adding an inverted sample does a good job of cancelling noise below 1/2 the sample rate.The signal aliases near the sampling rate and above and doesn't cancel noise. Sample time does some filtering at high frequency.

Including a spice sim. Change frequency V3 or sample time V4 and V5 to see effect. Maybe there is a better way?

[Carl-NC]

Correct, an EF subtraction sample (or bipolar operation) largely cancels preamp flicker noise. Time-domain noise sims are difficult unless you have a simulator that specifically has that feature (aka periodic noise analysis) but I've found a trick that works pretty well. I use Easy-Spice (a variant of SIMetrix) which has a definable transfer function block; I use this to create a sinusoidal sweep generator and then do a (very time-consuming) time domain sim. A comparison of a half-wave demod (green) & full-wave demod (red) is shown in the attachment. These would correspond to "no EF sample" vs "EF sample."

[green]

Correct, an EF subtraction sample (or bipolar operation) largely cancels preamp flicker noise. Time-domain noise sims are difficult unless you have a simulator that specifically has that feature (aka periodic noise analysis) but I've found a trick that works pretty well. I use Easy-Spice (a variant of SIMetrix) which has a definable transfer function block; I use this to create a sinusoidal sweep generator and then do a (very time-consuming) time domain sim. A comparison of a half-wave demod (green) & full-wave demod (red) is shown in the attachment. These would correspond to "no EF sample" vs "EF sample."

[ATTACH]46563[/ATTACH]

Probably looking at it wrong, why doesn't it alias at 20kHz?

[Carl-NC]

Probably looking at it wrong, why doesn't it alias at 20kHz?

It's not a sampling circuit, it's an averaging (integration) circuit. So there is a null at 20kHz. Alias peaks are at 30k, 50k, etc.

[green]

It's not a sampling circuit, it's an averaging (integration) circuit. So there is a null at 20kHz. Alias peaks are at 30k, 50k, etc.

I have thought my circuit reply #3 simulates a normal PI integrator. Null or constant amplitude out at sample frequency and multiples of(1kHz in example). Aliases above 1/2 sample frequency, 995, 1005, 9995 and 10005Hz in = 5Hz out. Am I missing something?

Trying to figure what your circuit is doing. What does fclk=10kHz mean? How does your circuit alias if it doesn't sample. Are you charting RMS amplitude out vs constant amplitude frequency in or something else?

[Carl-NC]

My sim is for a VLF-style demod, where the switches are 50% duty cycle. Therefore it is averaging all the time. fclk is the demod clock, output is the absolute value of the demod voltage which is why it is only positive. At fin=10kHz (synchronous) the peak is 0.636 volts, which is 2/PI, exactly as it should be.

In your PI sim, you have discrete samples and (I believe) the spacing is going to affect the aliasing. I'd have to chew on that a bit to be sure.

[green]

My sim is for a VLF-style demod, where the switches are 50% duty cycle. Therefore it is averaging all the time. fclk is the demod clock, output is the absolute value of the demod voltage which is why it is only positive. At fin=10kHz (synchronous) the peak is 0.636 volts, which is 2/PI, exactly as it should be.

In your PI sim, you have discrete samples and (I believe) the spacing is going to affect the aliasing. I'd have to chew on that a bit to be sure.

Thanks for the reply. I'll think about it some more.

[Mechanic]

Hi Carl,

Thanks for that, it confirms what I suspected.

Hmmm!

Altra, it is like random small false signals heard in the threshold occasionally.

Cheers Mick

[green]

Hi Guys,

So, I'm having a bit of a discussion with a friend regarding amplifier noise figures.
As a general rule we want a low noise preamp, but which bands of noise figure are actually important for PI detectors?

I say that the 0.001 to say 3 or 4 hz is not that important as this noise should be cancelled by the sampling, just as earth field is cancelled.

That being said, if an amplifier has a good noise figure down to 10hz, then in general the lower band of noise is also good.
Many amps with a good advertised noise floor start to get noisy under 100hz.

Bear in mind that I am referring to Minelab type front ends with low input resistance <10ohms and amplifier noise is dominant.
Replacing the ad797 with an Ne4435 definitely makes the detector more noisy.

Cheers Mick

Which band is the most important for a PI? 1k to 10k, 10k to 100k or 100k to 1M. Some other band?

Carl, thanks for the explanation (switch closed half the time). Was thinking should alias multiples of the sampling rate 10kHz. For 10kHz sample rate switch is closed for whole cycle at 20kHz in so cancels.

[Altra]

Altra, it is like random small false signals heard in the threshold occasionally.

Cheers Mick

Thanks