Topic: pianoteq and ambiophonics
As I said in a previous post, I have always prefered headphones while playing with pianoteq instead of loudspeakers.
I hinted then about "ambiophonics" which is a name given to methods for crosstalk cancellation between speakers, the absence of crosstalk being the main reason why headphones are more realistic.
This post is concerned with people playing at home and not in a live performance on stage. In that case, ambiophonics is really of no use.
The main technique, as explained in detail here :
http://www.ambiophonics.org/Ambiofiles.htm
is to reposition speakers closer together (at a 20 degree angle instead of the usual equilateral triangle) and using DSP to effectively remove the crosstalk in real time.
If you want to be convinced of the effectiveness, you can download from the site some examples pre-processed with the RACE (Recursive Ambiophonic Crosstalk Eliminator) algorithm. This test is easy to do with cheap "computer speakers" that can be easily moved. The resulting widening and "out of speaker" sound even with low-fi equipment may surprise you.
I have used both methods described on the site with the AudioMulch software with some interesting results and some limitations also.
Since AudioMulch can load VST plugins, it is possible to play pianoteq in real-time while also producing crosstalk-cancelled output.
I have loaded an image (pianoteq_audiomulch.jpg) to show the setup.
The interesting part is that there is a real widening of audio image if speakers are properly positioned which I try to demonstrate with an uploaded example.
The files are :
pb_deb.mid - a performance of Debussy's La cathédrale engloutie which I downloaded from Classical Archives.
pb_deb.fxp - the setting which I used to record the first 3 minutes or so of that midi file with pianoteq.
pb_deb_RACE.mp3 the resulting pre-processed file that MUST be listened to with the proper speaker positioning.
The less interesting part is, although the CPU demands are very low (on my setup anyway) there is a slight increase in latency which makes the playing more sluggish.
There is also a gain (volume) loss and subtle changes to the attack of the piano which may be acceptable or not.
The RACE method is less demanding (but a bit less realistic) than the BACCH one (which uses convolution) for real-time playing, but the BACCH technique can be extended to more than 2 speakers (which I have not tried)
In conclusion, this is an interesting experiment with some good and some bad, you be the judge.
This could contain suggestions for improving future versions of pianoteq's speaker output.