This document is presented as a working paper for those interested in sampling pipe organs for the purpose of creating a sample set. The same basice apply to creating a soundfont for jOrgan-Fluidsynth usage, for Hauptwerk sample set, and for GrandOrgue sample sets.
Wet – recording of the pipe sound including as much of the acoustic of the building as possible
Dry – recording of the pipe sound with as little of the acoustic of the building as possible
Staccato Release – recording of the pipe sound in a Wet environment with a very fast note on to note off phase. This captures the difference in reverb between a note played staccato verses one held for a longer period of time
Set your audio input device and recording software to record at the highest possible sample rate and bit depth (ideally 24-bit 96khz and higher, 48khz is acceptable).
Test a few mic positions for each of the pipe chambers, making notes of the levels of the lowest octaves of the loudest rank and the softest rank in the chamber. If you are recording in Stereo, then select a position that allows a good stereo spread to be captured, as well as good levels of all the ranks in the chamber. Do some research on Stereo recording positions, as the angle and distance between the microphones can make quite a difference.
Wet sampling records the acoustic of the building and so will use the Omnidirectional recording pattern. The microphone position will need to be at least 3 meters from the pipe chamber so that the reverb of the building can be captured.
Dry sampling eliminates as much of the acoustic of the building and so will use the Cardioid (directional) recording pattern. The microphone position will either be at the swell shade entrance or at a closer position within the pipe chamber.
Test your audio equipment: listen for electrical interference and grounding (hum) noises.
Sampling should be done when ambient noises (traffic, etc.) are at their lowest. Many sample sets are recorded between midnight and 4am! If this is not possible for you then listen carefully while you record so that you can re-record a note if a bus honks its horn in the distance during your recording session.
A single microphone position per pipe chamber is recommended, as this allows the best natural voicing to be captured as part of the sample set.
If needed, different gain settings can be used for quieter vs. louder stops, but make a record of the gain levels used.
You will need to record between 7 to 12 seconds of the sound of each pipe.
Allow enough time between each note for the echo/reverb to completely die away (give 3 or more seconds more between each note after you have heard all reverb die away as the electronics are more sensitive that your ears)
The Noise Reduction software needs noise profiles in order to remove as much of the noise from the samples as is desirable. Before each octave of samples, record at least 10 seconds of ‘just noise’ between each octave. This allows us to create a unique noise profile for each octave of the rank that is specific to the room at that temperature and that barometric pressure, etc.
Say (out loud) the name of the rank and whether it is tremmed or not at the beginning of each rank being recorded (this way you will not forget what that recording is of).
ANY NOISES THAT YOU MAKE WILL BE RECORDED AS WELL…. So make sure that you shift your position on the organ bench when a note is NOT being recorded… if you sneeze while a note is being recorded, record that note again, etc. The same goes for traffic noise, building creaks and thumps...
When recording a rank, first record the rank without Tremulant, and then record it again with the Tremulant activated. Recording the non-trem and the trem in a close a time period as possible results in better tuning options during the processing stage.
Mixtures and compound stops (like the Cornet V) will need longer samples as the complex interaction of the tones takes longer to ‘settle’. So you will need to record between 12 to 15 seconds of each note for these Stops.
As the acoustic of the building is captured during the recording of each pipe, added realism is required in the sample set to mimic the difference in ‘reverb’ between a note that is held for a very short period of time and one held for a longer period of time. Three release types are therefore recorded as part of the recording session for a Wet sample set. A very quick jab of the note (50-100ms) (then let all the reverb die away), then a more controlled staccato that is held slightly longer (200-250ms) (then let all the reverb die away again), and then the final 7 to 12 second duration.
For a truly Dry sample set one can move the position of the microphone for each pipe, setting the microphone close to the mouth of the pipe so that all the ‘speech’ of the pipe is captured. This is labour intensive and will only be used by high end HW users, so the return on investment is lower. Therefore a couple of mic positions within the pipe chamber can be used.
Semi-Dry recordings have been used very successfully to create both Dry and Wet versions of an organ installation. These recordings are made about 1/2 a meter from the swell shades and contain just enough of the acoustic of the chamber that they sound 'present', but without the larger acoustic of the church/hall/theater. An Impulse Response recording is then made of the acoustic (the mic is positioned whereever the best reverb is heard, and then a loud bang from a starter piston or popped balloon is recorded, allowing the full reverb/echo to sound and at least 5 seconds of silence).
Voicing
A ‘voicing recording’ of the whole organ needs to be made. A single recording session with the microphone placed in a sweet spot in the theatre/hall/church where each rank’s ‘C’ notes are individually recorded (just a couple of seconds each). Say the name of the rank, record the Cs, then go on with the next rank. For each division, play a middle C based chord using the softest stop, and sequentially add ranks, saying the name of each rank added, until you reach the full organ for that division. Finally play a well-known piece of music.
This recording will be used as a reference point for adjusting the samples in Hauptwerk
In order to get an accurate reproduction of the swell variation in a chamber, record a note from each octave with the Swell open, and then the same notes with the Swell shades closed.
Having experimented with mic positions I have chosen a single position for this pipe chamber that allows for a great capture of all the ranks, from the loudest to the softest (in this case it is just outside the swell boards with the swell open)
I then run a quick test of the audio equipment and computer to make sure that things are as they should be
I start the recording session of the first rank, say the name of the rank (Concert Flute) out loud and mention how many pipes are in the rank (97 pipes).
I play the first (lowest) note and count (in my head) to twelve. I let go of the note, hear the release and the reverb fade out, count to three (again silently in my head), and then proceed to play the second note….. etc. until the whole rank has been recorded. Now I turn the tremulant on, and I then say the name of the rank again - out loud, this time saying “Tremmed”, and I mention how many pipes of the rank are tremmed (some trems only start in the 2nd or 3rd octave of pipes). I then proceed to record the lowest note and count (in my head) to ten. I let the reverb fade out, count to three (again, in my head), and then proceed to play the second note….. etc. until the whole rank with tremulant has been recorded.
I then end that recording session and save the file. I then quickly listen to the recording to make sure it was successful. After that I start a new recording session for the next rank….