@Jason KanterThanks for posting this. There are some similarities in the process I've come to use, and I have already added some of your ideas.
First, some thoughts on the 3-part tuning itself, which I've been using for about 3 months.
-It's
far and away better than the old 2-part tuning function, designed for poorly scaled pianos. I would hardly ever use that because it only helped a fraction of the time, and even that was hard to predict. And the user didn't have as much control.
-I have interested in trying a Perfect 12th-based tuning style for a couple years now, and Tunelab finally supports it with the 3-part function. Now that I'm several dozen tunings in, I have found it better to use a single-octave style for the bass, as 3:1 tends to make it much too compressed. I use 10:5 as a default for grands and 6:3 for uprights, but I always try a couple options before tuning a piano I haven't done before (more on that below). But in the midrange and treble, I really like how well a 3:1 tuning blends the registers on almost all pianos. To my ears it yields very balanced sound, whereas the original TL tunings based on 6:3 / 4:1 (or similar) would often sound to me more compartmentalized. Think of the conductor from Disney's Fantasia, "Here's the bass, bold yet smooth. Here's the tenor, it's sweet and even. There's a bit of a break between them, but we don't talk about that. And now here's the treble, it's where we hear the Railsback curve go up like thiiiis."
-My wishlist for Tunelab would be to support the selection of where the midrange ends as well as where it starts, along with choosing two different interval types, and a percentage balance between them, for each of the three sections.
Something like this: (I'm imagining a medium-sized American upright, where one interval type in each section isn't quite right. I haven't actually tried this particular style, but you get the idea)
Bass: 8:2 / 6:3 @ 60/40%
Midrange: 4:2 / 3:1 @ 50/50%, from E3 to C5
Treble: 4:1 / 3:1 @ 50/50%
(FWIW, Verituner was doing this 15 years ago. But when my VT100 box died I couldn't afford to replace it, and started using Tunelab in the meantime. I like many things about both of them.)
Jason, your protocol brings me to questions I've had and maybe you or someone here can answer
-Where is the sweet spot for the number and selection of Inharmonicity measurements?
I have been doing A0-A6 for a while, but I'm not satisfied that's the best thing. I like limiting the number of measurements, and using A's (TL measures C's by default, I think, but somehow I feel less like a dumb American by using A's, and it gives my one more mostly usable note to sample). However A0 and A6 can be unreliable notes to measure, depending on how the string or hammer responds. So occasionally I will delete the IH constants on one or both of them, or even change them entirely to a value I think will put the extremes of the piano roughly where I want them, depending on where A1 or A5 are. When A0 is much more than 0.800 it seems to shoot down into the basement unnecessarily, and A6 seems to need to be at somewhere between 3.5 and 6.0. For context, I like to avoid extreme offsets, like C8 higher than +45c (or lower than +30c), and A0 at P6 much lower than -20c to -22c. If TL gives me something outside of that, it tells me that either a better interval choice might be available, or that there's no way the last few notes will sound good anyway, and it might be just as well to have more reasonable string tension across those octaves.
So, lately I've been experimenting with other arrangements of sample notes, such as C1, A1, F2, C3, A3, F4, C5, A5, F6 (or substitute E's, but I'm Jazz guy, so I like Fmajor). This satisfies my irrational need for a musical arrangement of sample notes, and is a middle ground between sampling 7 notes (A's only), and 12+ notes (C and F# in octaves 1-6, plus stringing breaks). Also it seems the lowest sample note is "low enough", C3 is right over the tenor break on a lot of grands, and F6 is high enough to have a significant effect on the high treble but is low enough to be a reliable sample note. I prefer to sample the same notes on all pianos, or as many as possible, because I find it very informative to compare this information and try to synthesize it with what I am hearing.
Other questions I have include, "how close to the break do need to measure to get a usable interpetation of the IH jump" and "how accurate does a pitch raise really need to be." Those are probably best done another time.
------------------------------
Gavin McGraw, RPT
------------------------------