< Back to IRCAM Forum

“Tune” parameter in PD Antescofo~ v.0.8

Hi,
preparing for an imminently upcoming performance of a piece that has been upgraded to now rely on Antescofo polyphonic score following in an environment rather disturbed by electronic sound, I am making all possible efforts to having the detection engine run as smoothly as possible.

Unfortunately, the Antescofo~ v0.8-master object for PD seems not to respond to the “tune” parameter (i.e. “tune 442.0” sent to antescofo-mess).
(Other parameters work, like “suivi”, “pedal”, “analysis”, etc.)

Is this a known limitation, or could there be anything wrong with my setup?

(I was thinking that by tuning the detection engine of Antescofo as close as possible to the actual tuning of the performance instrument, it would be possible to reduce the “variance” parameter, and thus, to reduce the possibility of detection errors due to electronically produced (microtonal) pitches that are close to the pitches played by the performer.)

Regards,
Christian Jaksjø

Christian,

This is probably due to FLOAT/DOUBLE issues that are not handled well in PureData. I will send you a link to a new version of the PD distribution (which we hope to release soon) which should address this issue, privately.

However, reading your message I highly doubt that the Tune parameter will solve your problem! If you are experiencing “systematic” behavior problems with the listening machine, then you should check the following steps:

  • Make sure that your SCORE is correct: A lot of time, even when listening is working perfectly, there are slight errors in the score transcription that leads to uncertainty… for example big silences that are missing in the score, or CHORDs that are notated as NOTEs or TRILLs, etc. Double-check your Antescofo score in places (and around places) where the following is problematic.

  • Calibration: Make sure you send enough audio energy to Antescofo. We usually separate Antescofo’s audio input from other DSP inputs. We made the Calibration specifically for this… . The calibration value, once the musician is ‘playing’ should be >0.7 and <0.5 on silence (or other instruments playing).

  • Watch for Sampling Rate: It has happened to me before that nothing was working due to the fact that Max/Pd was living in the 44K world and the sound card on 48Khz sampling world!!! Check this out… . A great way to test this is to ask your musician to play an A4 (midi 69) and watch the 2nd output of calibration (which gives the pitch probability for A4). If it goes >0.5 then you’re fine. If not, then there’s an issue with input audio!

Any comments from other users would be welcome.

Arshia,
First of all, thanks for your swift response on this manner. As soon as time permits, I will take a look at the new version to see if this issue is solved.

In order to clarify: I am


    not
experiencing a problem; alltogether, the listening machine operates to a high degree of presicion, and everything works!
(I’ve already edited the score in order to enhance the success rate of detection, checked calibration, audio parameters, etc.)

Still, the particular acoustical situation of the piece in question coupled with practical requirements regarding the upcoming performance situation (limited time, unknown tuning of the instrument, other unknown acoustical factors, etc.), calls for all possible measures to insure a fluent performance of the piece. What I’m after is a small adjustment to the better, not a bulding block in order to make the system work (at all).

Christian,

Sorry for the misunderstanding!

Then you are just on the right track: The tuning parameter usually does not evade the follower per se, but will improve latency and precision of detection (at least in my own experience) especially for high and low pitches.

Another parameter, depending on the instrument, that can improve is the harmlist which was useful in version before 0.6 where the follower was not robust enough. For example, we used to tell Antescofo (after some spectral analysis) which harmonics where the most important ona vibraphone to improve performance. Since v0.8 those pieces have just worked without any tweaking. But knowing this is always useful.

Good luck with the performance.

ps: You can announce your event on this website by creating an Event on the Event page: http://forumnet.ircam.fr/forumnet-events/