I am pleased to report what I regard as very pleasing results in the past few days in making up two presets with true recorded samples. In the past I have been forced to resort to making “short hybrid samples” simply because of the inferiority of my recordings. There were two problems with the recordings I had on my computer, reflecting recording sessions which I made with various pipe organs over a period of many years. Those problems were excessive noise (either from the instrument itself or from my recording setup), and quite short notes recorded. I didn’t hold the keys down long enough!
The recordings I have been working with recently still need the use of SPEAR in order to deal with the noise problem. SPEAR deals with it so effectively that I find it necessary to introduce some noise at a low level! The steps involved are quite simple, and the resulting sound is, to my ears, very satisfying - the point being that SPEAR does not lose the random fluctuations in pitch and amplitude of the various harmonic components of the pipe sound. Attack characteristics are also faithfully reproduced, if the user is careful to select the chiff components when using SPEAR. I need to add that the samples are all “dry”, having been recorded quite close to the pipes. There is doubt in my my mind, however, whether SPEAR can be successfully used in this fashion for “wet” samples or when processing release sections of samples.
The other point is that the notes were all held down for 5 seconds or more. By having samples this long, the process of finding good looping points is made much easier. I have now used “Endless WAV” to loop about 60 samples (30 per stop), and although in many cases I had to try a few times before being satisfied, eventually a good pair of markers were forthcoming. And here I am not just talking about eliminating looping “clicks”. To my ears, many samples of recorded notes are marred to some extent by a repetitive or cyclic pattern, mainly in the noise content, which certainly does not sound random. I won’t claim that this effect has been avoided absolutely in the samples I have made, but I do think that it has been kept down to a satisfactory level. (A subjective judgment, of course).
As a former “champion” of short hybrid samples, I now have an enthusiasm for true recorded ones, if the recordings available are of a high enough standard. If this condition is met, then they are now my preference. ;-)
That's great news, John! You description of the trials and tribulations of an organ builder are always fascinating and informative. I look forward to hearing the results of your longer samples, especially since they were close miked.
I am now in a position to add somewhat to my previous post in this thread about using recorded samples of adequate length (about 5 seconds). My admiration for SPEAR as a tool for dealing with noise in the samples has developed greatly, due to hints from Dries (this time I did not have to re-invent the wheel!). I have also discovered the Fade In and Fade Out effects in Audacity, which renders my previous method of using the Envelope Control obsolete. Also, using these effects is much quicker.
A number of additional presets have now been made, and the most recent is a five-rank Cornet stop. The exciting thing here is that it is usually regarded as next to impossible to loop samples made with compound stops satisfactorily, as all the pipes sounding simultaneously, when one key is pressed down, will be slightly out of tune, and so the waveform will never repeat. This problem has been overcome by copying a 1 1/2 second segment near the start of the sample, and cross-fading it "back in" near the end. Phase differences are accommodated easily and quickly (in most cases) by a simple graphical method, without having to resort to those mathematical calculations I described in my tutorials. Although some samples of the Cornet still needed a bit of doctoring, most delivered perfect looping points after one or two tries with Endless WAV. While this method completely eliminates looping clicks, it doesn't necessarily deal with the cyclic "swish, swish" present in many recorded samples, and this is a separate problem which may always have to be tolerated to some extent.
I plan to write a new tutorial on these new methods, and bring the old tutorials up to date.