INA128: Low Frequency Noise

Hi Michael,

the INA128 shows a noise current density of 0.9pA/SQRT(Hz) at 10Hz. Across a 6MOhm resistor this creates a noise voltage density of

0.9pA/SQRT(Hz) x 6MOhm = 5400nV/SQRT(Hz)

which is huge compared to the noise voltage density of INA128 of 10nV/SQRT(Hz).

With such a high source impedance you should urgently choose an OPAmp with a low noise current density.

And not to forget the 450nV/SQRT(Hz) thermal noise of a 12MOhm resistance.

Kai

Hi Michael,

As far as I see, you power the inamp from V-=0V and V+=10V, the reference input is set at 5V and the 6MOhm resistors are connected to 0V. Am I right?

If so, then the input common mode range ( > (V-)+2V) seems to be violated, and you can get odd behaviour.

As Kai pointed out, the thermal noise of the resistors and the noise current generates large voltage noise at the differential inputs. This source is shunted by the capacitance, so a lowpass filter is formed whose corner frequency depends on the capacitance.

Zoltan

Adding to other comments, the INA128 linear input range is shown below.

However, it changes with supply voltage, Gain, and Vref BUT you may use Analog Engineer Calculator to determine its linear operation - see below.  In order for your circuit to work, you would need to lift the Vcm to 5V or use dual +/-10V supply.

6M resistor will dominate the noise at low frequency in bipolar input INA's.  See alternative part below with FET input and thus lower current noise.

Hi,

Although the 6M resistors generate large noise, due to the capacitor it is filtered. The thermal noise of a 6M resistor can be considered as a voltage generator with output resistance of 6M. Therefore it is passed through a lowpass filter. The corner frequency is quite low due to the high resistor value. I’ve made a simulation to illustrate this using different capacitor values (and used 5V common mode voltage at the inputs):

Using 18nF the thermal noise is not the dominant source above about 50Hz.

It can be surprising, but according to the above, higher resistor value can even mean smaller noise above a certain frequency. See the plots below (G=47, 1.2nF).

This is, because thermal noise amplitude increases as a square root of the resistance, while the corner frequency is inversely proportional to it. So factor of ten means about 3.1x higher noise, but 10 times lower low pass corner frequency. However, at low frequencies the current noise can dominate, since it is not white noise.

I’ve also made simulation for the original configurations, it looks quite similar to the measurement result:

I hope I could help to clarify the situation.

Best wishes, Zoltan

Just to clarify -- the 6M Ohm resistors are connected to 5V, not 0V, apologies for my unclear sketch.

Thanks for the alternative part numbers this is very helpful

Thanks for this, this looks exactly right. 

Hi Michael,

again, the input noise current density is the major noise source here, not the resistor noise.

I have tried to perform a noise simulation with the TI's external noise voltage and noise current generators which I adapted according to this training video:

https://training.ti.com/ti-precision-labs-op-amps-noise-verifying-noise-model?context=1139747-1139745-14685-1138803-13844

As the below simulation shows the simulated noise current density is slightly higher at 10Hz compared to the curve in the datasheet, but still acceptably precise:

The following simulation shows the noise voltage density at the output "Vn_out" while taking into calculation the resistor noise of a 6M resistor (plus a 0F or 2.4nF cap in parallel), the input noise voltage densitiy of INA128 and the input noise current density of INA128. The gain is still 1V/V:

Then with a gain of 47V/V:

And the accumulated noise is:

For simplicity only one 6M leg was simulated. Provided the input noise currents of both inputs of INA128 are uncorrelated, I would estimate a total noise at the output of your circuit of about 1.414 x 1.78mVrms = 2.5mVrms over the 1Hz...100Hz band. The simulation says that the noise will increase even furtherly when you extend the band below 1Hz.

Kai

Good luck

Hi,
I also thank Kai and Marek for their instructive posts! 
I’ve repeated the simulations using the INA121 you’ve ordered.
Different capacitance values:

Different resistance values:

Comparison of the two inamps using the values of the experiments:

I also wish you good luck!
Zoltan

I received an INA121 in the mail today and repeated my experiment, comparing it to the INA128. The result is below. It matches Zoltan's simulation perfectly. The noise performance is much better, I think this is going to work for my application. Thanks everyone! 

Hi Michael,

these are good news

Kai

Thank you ALL.  I'll close this post.  Happy Holidays!