TLV9154-Q1: The reason of why PSRR is defined as "±"

Hi Kazuki-san,

Kazuki Kuramochi said:

I suppose that the definition of PSRR is ΔVDD(+/-)/ΔVos.

Yes, this is one of PSRR definition in op amps. There are other PSRR definition, where I am referring to the general PSRR definition, 20log(ΔVsupply/ΔVout). The latter PSRR definition is typically used in power supplies and LDOs etc.. 

PSRR is uV/V or dB are small signal analysis, where small signal in AC (e.g. ripple, noise or perturbation) on the power supply rail is injected into DC supply. You are correct that it should have been expressed in ± bidirectional amplitude. The perturbation AC signal is typically sinusoidal and it has ± directional associated with the small signal injection. So when it is expressed in uV/V or mV/V, then ±uV/V or  ±mV/V are correct usage. The linear expression in ±1uV/V means that 1V of supply ripple cause ±1uV ripple at the output. 

When it comes to dB expression, the ±uV/V terms are calculated in absolute ratio, since negative log is undefined. In dB, it is expressing the ratio of amplitude, not signed voltage. If the injected perturbation signal is in ±uV/V term, the ratio remains positive or unchanged, where no sign is expressed in ratio of magnitude. 

The datasheet is getting a bit sloppy when the measured unit in dB from instrument, and someone is trying to express in uV/V. Often one drops ± sign in the expression. But it is generally understood that ripple voltage, noise or perturbation signals are bidirectional in small signal analysis, and it should have been expressed in ±uV/V magnitude in the linear scale. 

If you have other questions, please let me know. 

Best,

Raymond