Filters For Low-Noise Measurement Preamplifier

[bryman79]

Hello All,

I'm currently in the process of assembling a low-noise measurement preamplifier (LNMP) with three gain stages (each 10x): AD797 > AD797 > LT1206. I'm planning on laying out the first two gain stages similar to the attached circuit taken from the book "Low-Noise Electronic System Design" by Motchenbacher and Connelly. However, since I intend to use this primarily for testing the noise of audio PSUs, I'd like to change the high-pass RC filter values (before the input stage and between the two input stages) to provide an fc of about 10Hz. Using the resistor values in the attached M-C circuit (10M before the input stage and 1M between the two stages), I've calculated cap values to be 1500pF (0.0015uF) and 0.015uF, respectively, to provide each filter with an fc of 10Hz.

Anyway, I've read that AC coupling the input can raise the noise floor of the preamplifer (see the first paragraph on page 3 of the Audioexpress article here: http://www.audioxpress.com/magsdirx/.../colin2764.pdf). So my question is how does the size of the AC coupling capacitor effect the noise floor? In other words, can or should I adjust the value of the caps that I've calculated to reduce the noise added to the circuit?

Thanks,
Bryan

[jcx]

AD745 is a fet input device, AD797 is a high bias bjt with 2.5 orders of magnitude more input current noise

"noise impedance", en/in is <500 Ohm for the 797, >100KOhm for the 745

for measuring low impedance ps use the 797 with much lower feedback, DC shunt R and big coupling C

op amp input current noise creates a noise V when it flows in the source impedance - for ps measurement with AD797 <100 Ohm DC shunt R could be justified


the 797 is much faster, with higher open loop gain than the AD745 and can be externally decompensated so 100x gain per stage is quite reasonable

[Mooly]

If you are measuring noise on DC rails you might be advised to have some form of input protection. Connecting the input to a rail above or below the opamp supply (even if AC coupled) will degrade, if not damage the first device input stages.

The problem with caps and noise is that they can physically pick up interference due to their shape/construction.
If you don't need high input impedance FET's are not the best choice, something like an NE5534 would be much quieter at low impedances.
Cummulative effects of DC offset will be significant with cascaded high gain stages unless trimmed.

[Samuel Groner]

A low-noise measurement preamplifier (LNMP) with three gain stages (each 10x): AD797 > AD797 > LT1206.

Not a good plan. It is very difficult to make a 10x gain stage with low voltage noise because of the feedback network contribution. You'd better use a single 60 dB stage with a single AD797. The bandwidth should still suffice for audio frequency range measurements (otherwise use the decompensation pin) and distortion doesn't matter for that application.

I'd set the -3 dB point for the HP filter to about 1 Hz to give good accuracy at 10 Hz. The impedance of the network should be low to keep opamp current noise and thermal resistor noise contributions insignificant at low frequencies. I'd use e.g. 1k6 and 100 uF.

Samuel

[AndrewT]

I think three *10 stages will be quieter than a single *1000 stage.
The posted schematic is using two *33 stages for the same total gain and this was done to reduce the noise that would come from a single stage amplifier.

I suspect, with the correct chips/opamps in each stage, that the three stage will be quieter than the two stage.

[Samuel Groner]

The posted schematic is using two *33 stages for the same total gain and this was done to reduce the noise that would come from a single stage amplifier.

No. It's done for bandwidth reasons. Have you read the book the schematic is copied from?

I think three *10 stages will be quieter than a single *1000 stage.

I suspect, with the correct chips/opamps in each stage, that the three stage will be quieter than the two stage.

A second/third stage can't help adding noise--do the math, the difference can be several dBs.

Samuel

[AndrewT]

Originally Posted by Samuel Groner

Have you read the book the schematic is copied from?

no.

[Samuel Groner]

It is excellent reading. I've got the first edition which IMHO offers more insight into discrete design.

Samuel

[bryman79]

Thanks for the input.

Most all LNMP circuits using the AD797 use it in inverting mode. Which mode is preferable (inverting or non-inverting) for the AD797 in this application?

JCX mentioned using a bigger coupling capacitor. Is this to reduce noise or to adjust the fc?

Finally, with regard to Samuel's suggestion, if I use a single AD797 in non-inverting mode to provide 100X gain (bandwidth reduced to 100kHz using a cap in parallel with the feedback resistor), would external compensation be beneficial or necessary?

Thanks,
Bryan

[Conrad Hoffman]

Remember that signal to noise performance is set at the input and nothing you do beyond that can improve it.