Coming back again to velocity scaling...
Thanks to @aureliopenna, who recorded both his NS2 and the (latest, better working) NS3, I've now been able to compare the behaviour of the NS2 and the NS3. So with apologies for this long post:
Keyboard Velocity Scaling
For Nord (and many other) keyboards, there are several mechamisms involved in velocity scaling.
It starts with
how the velocity of the individual keys is converted to a (MIDI) signal. This part is quite difficult to measure and compare, since it would require an ability to reproduce identical velocities and some way of measuring this. We can also expect that there are differences between the 3 keyboard versions (SW, HP and HA) and between generations (NS, NS2, NS3) or types (NS, NE, NP).
I have earlier done some simple tests that clarified that if the NS3 is set up to "hard" extern MIDI mode, then sending the same MIDI signal back to the keyboard seems to reproduce the exact same sound as it does internally. So we can maybe take the MIDI signal in "hard" mode as a representation of how the keyboard represents velocities internally.
Note that the KBD TOUCH setting does NOT impact the MIDI that is sent.
Also note that Nord calibrates the keys at the factory, and each keyboard has an individual adjustment curve for each key, in order to create a uniform response across the keyboard. If your keyboard is in error (unlikely though), this curve may be incorrect -- but I've heard of such an error.
Engine Velocity Scaling
Next, for the Piano section, there is the KBD TOUCH (0, 1, 2 or 3) setting, which determines how the sound engine responds to incoming MIDI (from the internal keyboard or an external keyboard).
There are two aspects of the sound that are impacted by the velocity: 1)
Which sample is being played and 2)
at what level. The KBD TOUCH parameter governs this mapping.
In the graph below, I have mapped out the relative peak amplitude at each MIDI level for the NS2 (red) and the NS3 (blue), in this case playing a middle F from the Bright Grand -- both at KBD TOUCH 0. The dots mark the transition from one sample-base to the next.
As can be seen, the curves are somewhat different. In theory, since the mapping of actual key velocity to MIDI is unknown and can be different between instruments (as pr. above), the keyboards may not respond to key velocity in the manner shown here. What is shown is how they would react if you play them from the same controller, not necessarily how they respond to you playing them!
These curves not only change with KBD TOUCH, they also look quite differently for different samples (there can be more or less individual samples and the scaling can be very different from instrument to instrument) -- in particular, EP look very different from an acoustic. It appears as if Nord is taking great care in creating and adjusting these curves, tuning primarily (I think) for the optimal key-sound connection we all look for. There are also differences in the total dynamic level between samples, and there are generational differences. My old NE3 has a very different curve when playing the same sample, and it has a larger (maybe too large) dynamic range from pp to ff. My NS3HA, on the other hand, has the exact same curve as the one shown here from @aureliopennas NS3C.
Between the NS2 and NS3, my observations are as follows:
The NS2 plays with no amplitude scaling at MIDI levels 0 up to 21, and only starts to increase in amplitude hereafter. It plays the same sample up until MIDI level 56. The curve is somewhat linear from levels 21 up to around 75, but has a flat shoulder at higher velocities. It is also quite uneven at higher velocities (the loudest samples are not scaled as well as they should be and have errors around 4 dB), and can even drop a bit with increasing MIDI levels.
The NS3 starts to respond to velocity increases already from MIDI level 8 or so, and by the time the NS2 changes to its 2nd sample, the NS3 has gone smoothly through to its 4th sample. The curve continues to scale in a logarithmic way overall, and is more smooth than for the NS2. It appears that the NS3 better uses the MIDI range and offers a more smooth scaling than the NS2; the loudest samples are also better scaled with errors less than 1 dB.
If you are used to playing the NS2, the NS3 may take a little "getting used to" (if the key velocity to MIDI mapping is the same or if you maybe play the NS3 remotely from an NS2 with a similar translations as the "hard" setting); it will basically respond quicker to increased velocity.
However, the NS3 will go through the same levels and dynamic range, and as the sample transitions are spread out wider (from MIDI levels 31 to 116 on the NS3 instead of from 56 to 110) it has the potential to offer better granularity than the NS2: You should be able to master a more nuanced playing better on the NS3 as it responds nicely to a larger incoming velocity range.
Incidentally, note that the first 4 sample transition points appear at roughly the same output amplitude between the NS2 and NS3 here.
In any case, the NS3 is obviously showing clear signs of a more fine-tuned engine.
Sample Similarities
Next, let's look at the similarities between the samples played at each MIDI velocity. I'll spare you for the details (it involves a bit of statistical/signal processing math like cross correlation), but the graph below shows how similar a tone is to another if they were scaled to the same amplitude. Lighter values show identical tones, and darker shows larger differences. Click on the image to show the details better.
The bottom graph shows how a tone from the NS3 relate to another tone at some other (or the same) velocity. It is easy to see how many samples are used and at what MIDI velocity it changes, and it is also easy to see that the larger the velocity difference, the more different (darker) the tones are, as expected. The NS3 shows clear signs of very well-controlled scaling; there are hardly any differences between different output levels of the same sample.
The top right graph shows the same relationship for the NS2. As can be seen, the transition between different samples happens in a much smaller MIDI range. It is also visible that the amplitude scaling is not as smooth; there are more (digital) artefacts.
Now, the most interesting graph is the one at the top left corner. It shows how similar the output is from the NS2 (on the Y axis) and the NS3 (on the X axis). Since the lightest areas are similarly light as in the two other graphs, the outputs from the NS2 and NS3 are very similar when they play the same sound. In fact, the difference between the NS2 and the NS3 playing the same sample is much smaller than the difference between either instrument playing two adjacent velocity samples!
There can still be subtle differences not visible here, though. And some may prefer a different velocity-to-sound response than what is offered by either instrument -- and as it depends on all factors mentioned above, it is difficult to say anything conclusively based on the factual measurements included here whether YOU like one better than the other.
But my overall conclusion is that the NS3 potentially offer better control of the pp to ff range (it responds better to a wider MIDI range -- which I think is consistent with what @analogika has reported), and it has a much better and more carefully tuned sound engine with fewer digital/scaling artefacts. But the NS2 and NS3 both play the same samples, and any difference between them (when playing the same sample) is smaller than differences between adjacent velocity or key samples.
Thanks for following this long (if you did)
Comments and questions are of course welcome.