In the C source file "tlv320aic3x" in the function "aic3x_set_power", when the argument for power is 0, the hardware registers are saved to the Linux computer as follows: regcache_cache_only(aic3x->regmap, true); And then later when "aic3x_set_power" is called with power set to 1, the previous values are then restored back to their hardware registers: regcache_cache_only(aic3x->regmap, false); regcache_sync(aic3x->regmap); When the device is powered back on, the previous setting for microphone bias (MICBIAS) is lost. If it was previously set to 2 volts or 2.5 volts, then when it's powered back on it becomes Off. I'm not very familiar with audio codec chips and how they work, but I think maybe the order of function calls in the "else" branch might be incorrect. Right now the code looks like this: /* * Do soft reset to this codec instance in order to clear * possible VDD leakage currents in case the supply regulators * remain on */ snd_soc_component_write(component, AIC3X_RESET, SOFT_RESET); regcache_mark_dirty(aic3x->regmap); aic3x->power = 0; /* HW writes are needless when bias is off */ regcache_cache_only(aic3x->regmap, true); ret = regulator_bulk_disable(ARRAY_SIZE(aic3x->supplies), aic3x->supplies); However I think that the call to "regcache_cache_only" should come before the soft reset takeS place. The setting for MICBIAS is lost at this line: snd_soc_component_write(component, AIC3X_RESET, SOFT_RESET); And so I think that the call to "regcache_cache_only" should be moved up above "snd_soc_component_write", as follows: /* * Do soft reset to this codec instance in order to clear * possible VDD leakage currents in case the supply regulators * remain on */ regcache_cache_only(aic3x->regmap, true); snd_soc_component_write(component, AIC3X_RESET, SOFT_RESET); regcache_mark_dirty(aic3x->regmap); aic3x->power = 0; ret = regulator_bulk_disable(ARRAY_SIZE(aic3x->supplies), aic3x->supplies);