Mensurix

[Stephen Birkett]

With the recent resurgence of the P12 tuning threads, I re-visited 
some of the old posts and came across some about the Mensurix 
software. This reminded me of some concerns I had had at the time on 
the subject of applied predictive models ....

Predicting the behaviour of a real physical system with an applied 
model is a huge can of worms. Any computer savvy person can write 
software that "incorporates a physical model" and give it a fancy 
trademark name. For instance, when Bernhard writes:
>There is also the new Mensurix 5 on the market what is the only 
>scale program available that uses a physical model of the string, to 
>get more realistic sound relevant information for scaling.  (the 
>physical model is build automatically from the string measures of 
>the scale)
this really doesn't say anything at all about the validity of the 
predictions derived from that software.

Commercial trumpet blowing is ok and necessary up to a point when 
trying to sell something. And some level of product secrecy is 
especially important for marketing software. With models, though, you 
can't afford to put out a black box and expect people to take it on 
faith. Secrecy has to be tempered by the damping [now there's some 
mixed piano metaphors] it places on serious discussion of technical 
issues and serious evaluation. Transparency is needed before a model 
can expect to achieve any sort of reputation as a useful predictor of 
reality.

Now I've never seen any reference to what physical model of the 
string-hammer interaction is being used in this Mensurix software. Is 
this information available somewhere? A perfectly respectable model 
could very well be put together in a couple of hours using a simple 
nonlinear spring and the old Suzuki modal transmission line for the 
string (I showed exactly that sort of model at the PTG-Dearborn way 
back). It would seem pretty sophisticated, especially packaged in a 
nice interface, and give complicated looking predictions that 
wouldn't really be much use at all for practical design. An improved 
model could be put together with a bit more effort and the string 
bashed using  a Stulov felt hammer according to a Donald Hall hammer 
string interaction. This one would look super-sophisticated, 
especially with the Stulov hysteresis, but it wouldn't inspire me 
with any more confidence as a practical design tool than the simple 
nonlinear spring. Perhaps Bernhard has some new unpublished model in 
the Mensurix software? But we need details or we're in the dark.

The other side of this is validation which is the only way for a 
model to earn confidence. This is touched on a little by the spectral 
predictions referenced in the Mensurix advertising - predicted vs 
real spectra - but the lack of details makes it impossible to assess 
those claims. And spectral predictions are notoriously imprecise as 
objective measurements of what strings actually sound like 
(representing a  very wide paintbrush.) This is not to say I don't 
believe the predicted vs real spectra, just that it's impossible to 
make a judgement of the model with the information provided. The acid 
test would be objective, preferably independent, scientific testing 
of the model predictions in a variety of areas, especially things 
like "strike point optimization" and "contact ratio", which are 
somewhat intangible concepts incompatible with intuitive evaluation. 
After all the software isn't cheap so some expectations should be 
allowed.

>The time, the hammer is in contact with the string is the main effect of how
>many partials and in what weight they will occur.
>...another important effect on the overtone weighting is the shape (curvature
>and width) of the hammer surface.
>...the contact time itself is affected by the hammer stiffness (complicated
>parameters), the hammer mass, hammer damping loss and hammer shape and the
>strings stiffness, tension and mass.
If there is some new integrated approach to modelling these factors 
and obtaining realistic predictions of string behaviour these results 
would be important, so why not publish them and make them available 
for discussion, instead of concealing them in a sea of 
commercial-eze? If the software is actually based on Stulov+Hall then 
why not say that as these models are well known already?

Stephen

-- 
Dr Stephen Birkett, Associate Professor
Department of Systems Design Engineering
University of Waterloo, Waterloo ON Canada N2L 3G1
Director, Waterloo Piano Systems Group
Associate Member, Piano Technician's Guild

E3 Room 3158
tel: 519-888-4567 Ext. 3792
fax: 519-746-4791
PianoTech Lab Room E3-3160 Ext. 7115
mailto: sbirkett[at]real.uwaterloo.ca
http://real.uwaterloo.ca/~sbirkett