If you're not 100% satisfied with the harmonizer in your Yamaha, here's something that I noticed on their previous-generation models that bears mentioning...

Yamaha's chord recognition tends to have a bit of a hair trigger. If you play in "AI Fingered" (or any mode other than "Fingered" or "Fingered on Bass" which require a minimum of three notes), the arranger begins to decode the chord as soon as you press the first note. If you press the additional notes needed to form the complete chord even 0.15 seconds late, you will see false chords and very short false notes in the recorded MIDI. Now you might not HEAR them, because Yamaha is also very good and sneaky about using portamento to bend the original notes to what's needed for the chord. BUT-- if the harmonizer is also being driven by the chord channel, it may be receiving those premature, spurious note-on messages, and it might also be using portamento to correct your vocals, adversely affecting the tracking, sound quality, etc.

Wait, there's more... for years people on this forum have bragged about the TC-Helicon built into Korgs. Now I don't sing, but I always wondered whether it was really the TC-Helicon technology, superior integration, or the fact that Korgs "batch" their chord recognition input! Yes, through experimentation I've observed that my Pa800 waits about 0.2 seconds from when the first key is pressed before acknowledging a chord. There is a definite "window," albeit brief, to press additional notes or even release incorrect ones before the arranger signals the new chord. There are almost never false notes in the Korg's recorded MIDI. I find the Korg less responsive to play, and you definitely have to lead the beat by a little more to make your chord changes timely. But I can see where this approach would provide more positive input to a vocal harmonizer.

As I've stated, these observations were made on previous-generation arrangers. Perhaps both companies have since changed their approach, or maybe I'm just tilting at windmills. Your thoughts?