dklein wrote:bm, very nice! I like how you've taken most of the nasal tones out of the Erard. It sounds much more natural on my system.
Curiously, it seems to have even more of the heavy 'keybed thump' than the original Player on my system, felt via the subwoofer. This is prominent on my system on several of the pianos, mostly the older ones such as the Broadwood, but even the Steinway D in ver 5 had it (mostly gone now, thankfully).
I'm glad you like it. However, there is still a lot of work to do mainly on 2 points:
(1) the bass reproduction is insufficient with a spectrum profile as close to the strings (3cm). I had compensated for this quickly with a wooden equalization, with values not adapted to this model of piano from where a too artificial sound, (a little like the bass of the steinway D before the ptq6 version). In fact, I have to rework the spectrum profile of the 1st partial (1 to 8 for A0 for example) with the data of the sound made at 2.5m from the piano that captures better the resonance of the case. (Typically in the example of A0, the gain must then be increased by 5 then 3 then -2db on the 1st partial - with in return a deactivation of the module "EQUALIZER"). I will add online for those interested the capture of the notes of the real instrument with the pickups at 2.5m (first 5 seconds of each note)
(2) Despite the automatic application of the spectrum profile by comparing the profile of the piano and the real instrument (for each part of each note: maximum level difference in db then division by 4: [rescaling] exists some partial for which this formula does not work, or very inadequately (ex: for A0, the partials N38, N54 and N75, N76 for which, manually I will apply soon a attenuation of approximately 10db), work of this type to make for many notes, especially the bass but not only (eg in the comparison made with the instrument with the sonata 14 of Beethoven, the note Ab3 is particularly unnatural, compared to the real instrument). as in some cases, the limitations of pianoteq modeling help to "improve" the instrument making a large number of spurious frequencies disappear on an old real instrument (when these frequencies are outside the spectrum profile which does not correspond to the partial frequencies proposed by ptq pro).
In any case, my discovery with the construction of this fxp was the need to pre-adjust the length of the strings with ptq pro to match the frequencies of the 30 or 40 first partials of the real instrument with those proposed by pianoteq in the spectrum profile window.
The manual recovery one by one of these frequencies is on the other hand a very painful work before being able to propose to my program the mapping of the maximum levels for each partials between the real instrument and the simulated instrument ...
It should also be noted that beyond a difference of about 10db between the real instrument and the virtual instrument for a given partial, the db levels proposed by pianoteq are to be reworked with an empirical function of greater attenuation of the difference in db (I still have an interpolation job to do to improve this attenuation function to be more faithful).
I finally hope that one day, it may be possible in Pianoteq pro to set a little more finely the balance between partial dynamics, in the sense of "hammer hardness: piano / mezzo / forte" (ideally a spectrum profile for each, if this is compatible with the intrinsic possibilities of parametrization of the mathematical model?). There would (ideally) also be the need for a more precise spectrum spectrum (spectrum profile?) for the hammer noise, and the case) .. assuming that this is possible, remaining the problem of a computer powerful enough to take into account in real time all these parameters - unless they can be pre-calculated ...
Bruno
Last edited by bm (02-04-2019 08:23)
