This is how a low-noise instrumentation amplifier with nanovolt sensitivity is built!

The AD8428 low-noise instrumentation amplifier provides an accurate gain of 2000 and has all the features necessary to solve these problems. The AD8428 features 5 ppm/°C maximum gain drift, 0.3 μV/°C maximum offset voltage drift, 140 dB minimum CMRR to 60 Hz (120 dB minimum to 50 kHz), 130 dB minimum PSRR, and 3.5 MHz bandwidth for low power applications. Flat measurement system. Notable is the device's 1.3 nV/Hz voltage noise (1 kHz) and 40 nV pp noise (0.1 Hz to 10 Hz) performance, resulting in high signal-to-noise ratio at very small signals. Two additional pins allow designers to change gain or add filters to reduce noise bandwidth. These filter pins also provide a unique method of reducing noise.

Figure 1 shows a circuit configuration that further reduces system noise. The input and filter pins of the four AD8428s are short-circuited to each other to reduce the noise to half of the original value. The output of either instrumentation amplifier can be used to maintain low output impedance. This circuit can be expanded to reduce noise by a factor of the square root of the number of amplifiers used.

Figure 1. Noise reduction circuit using four AD8428 instrumentation amplifiers.

Each AD8428 produces a typical referred-to-input (RTI) spectral noise of 1.3 nV/Hz, which is uncorrelated with the noise produced by the other amplifiers. Uncorrelated noise sources add to the filter pins in a root sum of squares (RSS) fashion. On the other hand, the input signal is positively correlated. Each AD8428 generates the same voltage at the filter pin in response to the signal, so connecting multiple AD8428s does not change the voltage and the gain remains at 2000.

Analysis of a simplified version of the circuit in Figure 2 shows that connecting two AD8428s in this manner reduces noise by a factor of 2. The noise of each AD8428 can be modeled at the +IN pin. To determine the total noise, the inputs can be grounded and superposition used to combine the noise sources.

Noise source en1 is amplified by a differential gain of 200 and reaches the output of preamplifier A1. For this part of the analysis, the output of preamplifier A2 is noise-free when the inputs are grounded. The 6 kΩ/6 kΩ resistor divider between each output of preamplifier A1 and the corresponding output of preamplifier A2 can be replaced by a Thevenin equivalent circuit: half the noise voltage at the output of preamplifier A1 and a 3 kΩ series resistor. This is the mechanism for noise reduction. The complete nodal analysis shows that the output voltage in response to _{en1 is 1000 × en1} . Due to symmetry, the output voltage in response to the noise voltage en2 _is 1000 × _en2 . Both _en1 and _en2 have amplitudes equal to _en and will add as RSS, resulting in a total output noise of 1414 × _en .

Figure 2. Simplified circuit model for noise analysis

In order to translate this back to the input, the gain must be verified. Assume that a differential signal VIN is applied between +INPUT and –INPUT. The differential voltage at the output of the first stage of A1 is equal to VIN × 200. The same voltage appears at the output of preamplifier A2, so no divided signal goes into the 6 kΩ/6 kΩ divider, and node analysis shows that the output is VIN × 2000. Therefore, the total voltage noise RTI is en _× 1414/2000, which is equivalent to en _/ 2. Using the AD8428's typical noise density of 1.3 nV/Hz, the resulting noise density of the two amplifier configuration is approximately 0.92 nV/Hz.

With the additional amplifier, the impedance at the filter pins changes, further reducing noise. For example, using four AD8428s as shown in Figure 1, the 6 kΩ resistor from the preamplifier output to the filter pin would be followed by three 6 kΩ resistors to the output of each noiseless preamplifier. This effectively creates a 6 kΩ/2 kΩ resistor divider, dividing the noise by four. Therefore, as predicted, the total noise of the four amplifiers is equal to en/2.

SPICE circuit simulation, while not a substitute for prototyping, is useful as a first step in validating such circuit ideas. To verify this circuit, you can use the ADIsimPE simulator and the AD8428 SPICE macromodel to simulate the circuit performance when two devices are connected in parallel. The simulation results in Figure 3 show that the circuit behaves as expected: a gain of 2000 and a 30% noise reduction.

Figure 3. SPICE simulation results

The complete circuit consisting of four AD8428s was measured on the bench. The RTI noise spectral density was measured to be 0.7 nV/Hz (1 kHz) with 25 nV pp from 0.1 Hz to 10 Hz. This is lower than the noise of many nanovoltmeters. The measured noise spectrum and peak-to-peak noise are shown in Figure 4 and Figure 5, respectively.

Figure 4. Measured voltage noise spectrum for the circuit in Figure 1.

Figure 5. Measured 0.1 Hz to 10 Hz RTI noise for the circuit in Figure 1