A spectrum analyzer with a frequency BW wider than the BW of the LDO is usually used to measure the noise of the LDO. Similar to the LDO, the spectrum analyzer has internal noise referred to as the analyzer noise floor (sensitivity). The noise floor (technically called the displayed average noise level–DANL) of the analyzer is a key specification in the analyzer datasheet over the frequency BW of the analyzer. Measuring and notating the noise floor of the analyzer is critical before making any noise measurement, as explained in the References [2] section.

To measure the noise floor for any analyzer, simply run frequency sweeps with the lowest setting of resolution bandwidth (RBW) available for the analyzer at multiple averages across the desired noise measurement BW. The measured noise curve (noise floor) is the lowest noise the analyzer can resolve over that BW. By knowing this noise level, a determination can be made regarding using such an analyzer to measure the noise of the LDO (DUT noise).

To maintain a noise measurement accuracy greater than 95% for the BW operation of the LDO, the noise curve of the LDO must be 10dB higher than the noise floor curve of the analyzer, as shown in Figure 1-2.

Figure 1-2 Typical LDO Noise Versus Typical Spectrum Analyzer Noise Floor

This 10dB (or 3.16V/V) of gain between the analyzer and the LDO noise achieves the 95% accuracy by combining the curves as two uncorrelated noise sources. If one source (the LDO) is 3.16 (V/V) higher than the other source (the analyzer), then the resulting total noise is more than 95% dominated by the higher source, as demonstrated in Equation 1:

The choice for 95% accuracy is reasonable for any noise measurement of an LDO, given the random nature of noise phenomena. A margin that is lower than 10dB results in a measurement that is less accurate at the BW roll-off of the LDO, as is obvious in Figure 1-2. When the LDO noise (red curve) in Figure 1-2 approaches the spectral analyzer noise floor (green curve), the noise measurement becomes incorrect and significantly dominated by the spectral analyzer noise. This flat noise is not the noise of the LDO, but rather, a combination of noise from the LDO and analyzer significantly dominated by the noise floor of the analyzer.

The internal circuitry of the analyzer can contribute to the measured DUT noise, however, the contributed noise is at or near the absolute minimum value of noise–174dBm/Hz. This extremely low noise level is the thermal noise level (power); the noise levels of the analyzer and LDO (DUT) are examined in detail in the next sections.

Figure 1-3 shows the noise curve for an ultra-low noise LDO, which is more than 10dB lower than the noise floor of a typical spectrum analyzer. So, the question is—how can this ultra-low noise LDO be measured?

Figure 1-3 Ultra-Low Noise of the LDO Versus the Noise Floor of a Typical Spectrum Analyzer

The solution is relatively simple—an amplifier that can gain-up the noise of the LDO to 10dB (or higher) above the noise floor of the analyzer so we can measure the LDO noise. Few spectrum analyzers have a built-in preamplifier for these types of low-noise signals, but preamplifiers are often gain-limited to less than 40dB and the BW for those preamplifiers often starts at 1kHz or higher.

But what if more gain is necessary and the measurement must start at a lower frequency?

The next section discusses the design requirements for a higher gain amplifier with a BW of 10Hz–10MHz.