I understand your question perfectly, because I have a PSR-E443 and know exactly what you're talking about.
When you choose a voice, the keyboard's operating system is programmed to automatically change several parameters to specific defaults for that particular voice-- e.g., the voice's octave offset, pan setting, reverb and chorus send levels, etc. The purpose of this is to ensure consistency when selecting a voice-- i.e., so when you select a given voice, it always sounds the same, rather than sounding differently each time due to the settings of the various parameters. The specific parameters which are automatically set this way seem to vary from model to model based upon which parameters are used by the preset voices-- e.g., my YPT-400 (or PSR-E403) doesn't set some of the parameters which do get set on my PSR-E433 and PSR-E443, since none of the preset voices on the YPT-400 make use of those parameters. What I mean is that if a given model doesn't have any preset voices which automatically turn on the Dual option and set the Dual Voice to get a layered sound, then none of the Dual Voice settings (including Dual On/Off) will be automatically changed when you select a voice. Or, if a given model doesn't have any preset voices which automatically turn on the Arpeggio option and set the Arpeggio Type, then none of the Arpeggio-related settings (including Pedal Function) will be automatically changed when you select a voice.
Since the PSR-E443 does have preset voices which use the Arpeggio option, all of its voices are programmed to automatically change all of the Arpeggio-related settings whenever you select a voice, even if the voice you're selecting doesn't use the Arpeggio option. Simply turning the Arpeggio option on or off doesn't affect the other Arpeggio settings (Arpeggio Type, Arpeggio Velocity, Arpeggio Quantize, and Pedal Function), but selecting a voice does. Consequently, the only suggestion I can make is to set up and save Registrations for each of the various voices that you want to use during a performance, and make sure you set all of the Arpeggio-related functions as desired for each voice. Then, when you want to change from one voice to another while playing, use the Registration recall rather than selecting the new voice by its Voice Number, so the Arpeggio-related parameters will be set as desired. Note that this would still change the Pedal Function to Arpeggio Hold if that's how it's set in the Registration, regardless of whether you turn the Arpeggio on or off.
