Non-HP linear supplies (plus a Norand 15-volt switching supply):
Note inability of HP 3561A FFT analyzer to resolve spurs in the 1-10 kHz decade; this may be fixable with segment-table optimizations. There's also an unexplained 10-dB discontinuity in the HP 6263B's trace at 100 kHz, which was repeatable with that supply but not with any of the others.
While the three-terminal and Darlington regulators were tested with a resistive load, the bench supplies were tested driving a 12-volt relay winding, so some of the low-frequency spurious content may be due to stray AC field response on the part of the relay coil. Coupling to the HP 11848A's 50-ohm baseband noise input was achieved via parallel 0.15 uF tantalum/22 uF aluminum electrolytic capacitors, followed by a 20 dB attenuator (HP 355D). The resulting 3-dB corner frequency is 120 Hz.
Graph scales are in dB below 1 volt RMS, normalized to 1 Hz bandwidth. For example, -150 dBv/Hz equals 1 / 10^(150/20) volts, or 32 nv/Hz. To generalize this figure, multiply by the square root of the bandwidth; -150 dBv/Hz of Gaussian noise at 1 kHz would correspond to approximately 1 uv/root-Hz.
The green trace below represents the measurement floor with the DC block/relay coil disconnected from any supplies. Again, 20 dB of dynamic range is lost due to the attenuator: