Low noise preamp with low value input coupling capacitor

[Bruce G]

A more appropriate amplifier topology allows low noise coupled with a low cutoff frequency whilst employing a relatively low value coupling capacitor. The first circuit employing a feedback inverter has an output offset of a few tens of  a few tens of mV which isnt significant if the following stage is AC coupled.
The second stage uses feedback integration as well to ensure a low output offset.

[TimFox]

If you make C1 too small, you may have excessive low-frequency noise due to the input noise current (admittedly small for JFET) flowing through the high reactance at low frequencies.

[Marco]

AFAICS the offset cancellation integrator is essentially a DC source noise wise, so this has no functional difference in noise with a high pass filter of 22uF and 100Meg ... the noise from the resistor will be relatively high for the first couple of decades near the -3db point.

PS. personally I'd just go for the ultracap in the source approach.

[Kleinstein]

There is no real need for an especially good OP in this circuit. The JFET should already give enough gain, so that something like a NE5532 or OP27 should be enough. It's also possible that the LEDs give more noise than the OP.

The high impedance input stage gives high noise levels near it's cut of. So there likely needs to follow a second high pass filter that sets the actually used lower frequency limit. With this second filter likely coming anyway, there is no need for the second version with extra integration. Also there is no need to make the inverter stage especially high impedance and thus high noise. So it would be better to use 2 more conventional 10 K Resistors.

[MK]

even a tl071 would do for the second op amp. The second version is a two pole filter and the resistor is actually in parallel with the input impedance, so most likely not an issue

[Bruce G]

AFAICS the offset cancellation integrator is essentially a DC source noise wise, so this has no functional difference in noise with a high pass filter of 22uF and 100Meg ... the noise from the resistor will be relatively high for the first couple of decades near the -3db point.

PS. personally I'd just go for the ultracap in the source approach.

Try some analysis rather than speculation you will be surprised to discover that the JFET input voltage  noise dominates down to 10mHz.

Bruce

[Bruce G]

even a tl071 would do for the second op amp. The second version is a two pole filter and the resistor is actually in parallel with the input impedance, so most likely not an issue

Not a good choice as it will noticeably increase the input noise unless the circuit is modified to increase the gain of the JFET input stage.
 Either increase the drain load resistor to 2k (adding 3 LEDs to the string biasing the non inverting input of the following opamp )
or use a low noise active load (this is possible but the compensation needs to be modified).

[Bruce G]

There is no real need for an especially good OP in this circuit. The JFET should already give enough gain, so that something like a NE5532 or OP27 should be enough. It's also possible that the LEDs give more noise than the OP.

The high impedance input stage gives high noise levels near it's cut of. So there likely needs to follow a second high pass filter that sets the actually used lower frequency limit. With this second filter likely coming anyway, there is no need for the second version with extra integration. Also there is no need to make the inverter stage especially high impedance and thus high noise. So it would be better to use 2 more conventional 10 K Resistors.

Try some analysis, the noise of the feedback inverter input and feedback resistors is swamped by the 100M feedback resistor.
LED noise is very low as long as they are shielded from ambient light (usually takes something like a 75W incandescent virtually on to of the LED to make the effect of LED photocurrents detectable) and convection currents. Zeners are noisier as are typical power supplies (even batteries if the current drawn isnt very small).

[Bruce G]

In my humble opinion, OPA827 for the second stage is clearly overkill ($10.12/unit from Digikey at small quantities).

Could you explain the roles of C6+R12 and R2?

R2 is to suppress HF/VHF oscillations a ferrite bead is probably a better choice.
C6 +R12 compensate for the negative input resistance and capacitance of the JFET input stage.
See Tektronix tutorial text on oscilloscope vertical amplifiers.

[MK]

even a tl071 would do for the second op amp. The second version is a two pole filter and the resistor is actually in parallel with the input impedance, so most likely not an issue

Not a good choice as it will noticeably increase the input noise unless the circuit is modified to increase the gain of the JFET input stage.
 Either increase the drain load resistor to 2k (adding 3 LEDs to the string biasing the non inverting input of the following opamp )
or use a low noise active load (this is possible but the compensation needs to be modified).

I am talking about U2 in the second circuit, you have 100M in series with the output AND you are using a 1M input resistor to U2, it certainly does not need to be an opa827

[Bruce G]

PS. personally I'd just go for the ultracap in the source approach.

Been there done that however the ESR of the supercap needs to be very low so that the gain is determined largely by the feedback resistors rather than the ESR of the super cap. If R is reduced to 1 ohm (appropriate if several JFETs are paralled to reduce the input noise) finding a suitable supercap becomes problematic unless one has the space to parallel several super caps.

NB not all supercaps have low esr.

[Bruce G]

even a tl071 would do for the second op amp. The second version is a two pole filter and the resistor is actually in parallel with the input impedance, so most likely not an issue

Not a good choice as it will noticeably increase the input noise unless the circuit is modified to increase the gain of the JFET input stage.
 Either increase the drain load resistor to 2k (adding 3 LEDs to the string biasing the non inverting input of the following opamp )
or use a low noise active load (this is possible but the compensation needs to be modified).

I am talking about U2 in the second circuit, you have 100M in series with the output AND you are using a 1M input resistor to U2, it certainly does not need to be an opa827

Sorry my mistake.
Neither the gain bandwidth nor the noise of U2 are particularly critical.
Only the input bias current and input offset matter in the second version of the circuit.
For the first version lower values for R9 and R10 can be employed with a bipolar input opamp for U2.

[Marco]

Try some analysis rather than speculation you will be surprised to discover that the JFET input voltage  noise dominates down to 10mHz.

I said the first couple of decades, I just didn't notice the -3db point was at 10 microHz. How does it respond to a couple V step signal?

[Marco]

The second stage uses feedback integration as well to ensure a low output offset.

By the way, why do you care about the DC coupled offset at the output? It's not a small magnitude signal anymore and the opamp can source quite a lot of current, so what's wrong with a high pass filter?

It can push the amplifier out of saturation a bit quicker ... but it's trying to push it back through a 100e6 resistor, so the recovery time will still be in the same order of magnitude (of course you wouldn't want it to go much faster, because the source is delivering the current).

[Bruce G]

Try some analysis rather than speculation you will be surprised to discover that the JFET input voltage  noise dominates down to 10mHz.

I said the first couple of decades, I just didn't notice the -3db point was at 10 microHz. How does it respond to a couple V step signal?

3dB point is about 0.008Hz ie 8 milliHz.

[Marco]

3dB point is about 0.008Hz ie 8 milliHz.

As long as it stays linear, so how does it respond to a couple volt step?

[Bruce G]

The second stage uses feedback integration as well to ensure a low output offset.

By the way, why do you care about the DC coupled offset at the output? It's not a small magnitude signal anymore and the opamp can source quite a lot of current, so what's wrong with a high pass filter?

It can push the amplifier out of saturation a bit quicker ... but it's trying to push it back through a 100e6 resistor, so the recovery time will still be in the same order of magnitude (of course you wouldn't want it to go much faster, because the source is delivering the current).

The integrator also adds a 2nd zero to roll off the response to low frequency noise faster than a single pole.
The recovery to input overload as may occur when connecting to the output of a voltage source (power supply, reference or battery) the noise of which is to be measured can sped up by clamping the input whilst the input capacitor charges.