Converting Amplifier Noise Density into a Measurement Budget

A noise-density value is not the noise voltage that a complete measurement will report. Converting it into a useful budget requires the frequency range, gain and filtering of the actual signal path. Two amplifiers with similar headline values can produce different results in a slow precision channel.

Define the reported quantity

Analog Devices tutorial MT-048 explains the relationship between noise density, integrated RMS noise and equivalent noise bandwidth. It also distinguishes broadband noise from the rising low-frequency region, which matters when the measurement extends toward DC.

State whether the product requirement is RMS output noise, input-referred noise or variation between reported readings. Identify the analog bandwidth, digital filtering, record length and update interval. These descriptions are necessary before a component specification can be compared with a system limit.

Use the correct transfer function

For a flat noise density through an ideal rectangular bandwidth, RMS noise equals density multiplied by the square root of bandwidth. As an illustrative calculation, 10 nV per square root hertz over 10 kHz gives 1 microvolt RMS before gain. A real filter needs its equivalent noise bandwidth or direct integration.

Apply the noise transfer function of each contributing source. The gain experienced by an amplifier’s internal voltage noise may differ from the signal gain in the selected circuit. Include resistor and current-noise contributions separately before combining independent contributions by power.

Avoid hiding the low-frequency requirement

Identify how long the product observes a nominally constant input. A quick oscilloscope capture and a long sequence of precision readings investigate different behavior. Include the low-frequency density information relevant to the observation interval instead of extending a high-frequency spot value down to zero.

Keep periodic interference separate from random-noise estimates. A narrow switching spur can dominate a particular measurement even when integrated broadband noise is acceptable. Record it as a specific spectral component and investigate its coupling path.

Make the bench result comparable

Measure with the same bandwidth and processing assumed by the calculation. Characterize the measurement instrument’s own contribution and use an input condition representative of the source impedance. Save raw data when practical so changed averaging settings can be assessed without repeating the hardware experiment.

Keep the density curves, transfer assumptions and measured result together. Explain any remaining difference before changing parts. A useful noise budget identifies the contributor worth improving and prevents spending effort on a quieter amplifier when another element or an unnecessarily wide measurement bandwidth sets the observed floor.