That's great stuff, Toby. And given that I have a Kronos 88 I have a good side by side comparison for the value in all the work they put into it. 

I bought the Kronos 2 in 2018, but it was actually released in 2014. That's exactly a decade of technological advances that Yamaha has been able to take advantage of, and it makes a huge difference. 

On the Kronos' German Grand, and indeed on the Kronos itself, velocity scaling is nowhere near the M8x. The dynamic range that I can get out of the M makes the Kronos feel like I'm playing an organ in comparison. I'm sure much of this comes down to the depth of their sampling and velocity mapping.

I won't make any assumptions about tone based on technology. The German Grand is not as bright as the Hamburg, and I do love the sound. However, that's probably down to the individual instrument that they sampled and their techniques and decisions while doing so.

Also, I'm not ragging on the Kronos. I have no plans on giving it up and I love the German Grand for its own thing, just as I have a Strat sitting next to a PRS and love each for what they do. But playing the Hamburg on the M8x keybed is an absolute joy. Your research shows just how much goes into making that happen.

And as a software developer by trade, notice how I just artfully skipped right over the part about your analysis software even though I wanted to ask? Because I currently have enough rabbit holes in my life and don't need inspiration to seek out more. I salute your endeavor. 🙂

That's great stuff, Toby.

Thanks - I was really just trying to reinforce that, as a synth, the presets are always necessarily designed for 'as is' use in a performance. The individual sound designers often include things to show off what CAN BE done - those things are sometimes hidden until triggered by a scene, switch or knob.

That Hamburg perf uses the Ribbon Controller - if you accidentally brush that controller you can change the sound without even being aware you did it. Good luck putting it back EXACTLY as it was without actually reloading the performance.

I forgot to mention part two is mostly part one with all elements detuned - Coarse +2, Fine -9. The name 'Hamburg Gr Shifted -' was a hint that something strange was happening.

But - part two is NOT used by default. The default Scene 1 only has keyboard control ON for part 1 while Scene 2 is only part 2. Scenes 3,4 and 5 have it ON for both parts.

So this is a common order of events for users new to these Yamaha Synths

1. load a performance

2. audition it and they love the sound

3. play it themselves and it sounds radically different

Why? The audition might do things (change scenes, twirl knobs, set switches) that exercise ALL of the programmed functionality. Some things an audition does, like use an EXACT velocity as a trigger, are things you can't possibly do as a human.

Hamburg Grand - you load, play and like the performance.

But the question is - what IS the real Hamburg Grand? Is it the scene 1 performance you get when first loading? Is it one of the other scenes?

Is there even a set of parameter values that represent the equivalent of a real Hamburg Grand?

And that is, I think, where the frustration sets in. There is no one answer - it all depends on which set of parameters you are using and EXACTLY how the keys are played in terms of timing and velocity.

And as a software developer by trade, notice how I just artfully skipped right over the part about your analysis software even though I wanted to ask?

It isn't so much analysis at this point as it is exploding the SysEx dump of a performance into meaningful sets of data. The end of the Data List doc has all of the dump table formats but they are all encoded.

So I have code that exports SysEx dumps for all the performances into individual files. You can use an app like MIDI-OX to do it manually one at a time.

Then I am creating Java classes that represent each of the nested data tables.

Here is a raw delimited dump of some element data for part 2 of Hamburg Grand

Elem No.,Wave Select,Group No.,Wave No,Note Low,Note High,Vel Low,Vel High,KeyOn Delay,Coarse Tune,Fine Tune
0,0,0,195,0,91,127,127,0,2,-9
1,0,0,194,0,91,115,126,0,2,-9
2,0,0,193,0,91,103,114,0,2,-9
3,0,0,192,0,91,92,102,0,2,-9
4,0,0,191,0,91,81,91,0,2,-9
5,0,0,190,0,91,71,80,0,2,-9
6,0,0,189,0,91,61,70,0,2,-9
7,0,0,188,0,91,51,60,0,2,-9
8,0,0,187,0,91,40,50,0,2,-9
9,0,0,186,0,91,2,39,0,2,-9
10,0,0,195,92,127,127,127,0,2,-9
11,0,0,194,92,127,115,126,0,2,-9
12,0,0,193,92,127,103,114,0,2,-9
13,0,0,192,92,127,92,102,0,2,-9
14,0,0,191,92,127,81,91,0,2,-9
15,0,0,190,92,127,71,80,0,2,-9
16,0,0,189,92,127,61,70,0,2,-9
17,0,0,188,92,127,51,60,0,2,-9
18,0,0,187,92,127,40,50,0,2,-9
19,0,0,186,92,127,2,39,0,2,-9
20,0,0,196,0,91,1,127,0,0,0

Load that into Excel and everything gets formatted to be readable.

But you can see at a glance that the Note Low values are all 0 (C-2) for the 1st 10 elements and 92 (G#5) for the 2nd 10.

And you can see how the velocity ranges are mapped for each set of 10 elements. So what is that 1st element used for that has a range of 127-127? I have no idea. It must be there for something.

As simple as that is I find it a lot easier to look at that data for both parts at once than moving 1 element at a time using the touchscreen.

I'm experimenting with a simple XML format for the above so you can load it into a browse and get the hierarchy in a collapsed/expandable form.

So I'm more using the software to identify performances that have items of interest to explore later.

Which ones use amplitude scaling? Which ones uses a specific controller, etc?

Wow. Data dumps, and Java scripts. I'm starting to wonder what a "Deep" dive would look like. 

Just also wondering, did you ever manage to figure out how Filters and Envelopes work? 

IIRC that was a bit of a "blocker" for your off-line software tools project.