Let us summarize our most important ideas in a few words.

Philosophy and mission of NDB Pipe Organ Samples

1. We have created this sound sampling project
   • to give artists, studios, companies, individuals, and schools an opportunity for playing and making recordings with the organs we have sampled,
   • to help retain the state or reconstruct the pipe organs we have sampled by donating them most of our income,
   • to make a world-class Hungarian product.
2. In the creation process of NDB, we wanted to
• achieve unbeatable quality and versatility,
• create a new standard in sampling technology.
3. With NDB, we would like to achieve
   • satisfaction of serious artists and composers,
   • a bigger popularity and better understanding of pipe organ music and literature,
   • interest in Hungary and Budapest.
4. We truly think it is important to make authentic recordings with the samples by indicating their electronic origin.

Philosophy of our recordings and sound sampling

1. In cooperation with serious organists and pipe organ experts, we have carefully selected which stops and combinations are the most important to record in order to produce a library, capable of the performance of almost the whole pipe organ literature.
2. We made our recordings with the finest equipment available today. We chose a pair of Neumann U 87 microphones and a custom-built preamplifier. These microphones are widely used in the pro audio, studio and broadcasting fields. We chose to record at 96 kHz with 32 bit depth to capture everything as accurately as possible. DAT recorders were not enough, we plugged the preamp's output directly into a computer. Many hours of measurements, fine-tuning, and discussion with sound engineers lead us to this decision.
3. Recording of every second note on the most characteristic stops, every third note on the other stops in the manuals in about 30 seconds each, and every note or every second note of the pedal notes in full minute length without a loop are second to none. All sound transients, valves and windchest action were recorded and kept precisely.
4. "Noise-free" environment. Our recordings took place late at night to achieve a quiet environment at the cathedral.
5. We honestly think, that noise reduction is important. It is quite risky, though: reducing the noise may destruct or 'sterilize' the sound. Extreme care was taken to avoid this. The reason of the importance of noise reduction is simple. In a sampling library, where every note is stored and played back discretely, a chord played back with noisy samples will contain as much noise compared to the noise level of the single-note sample, as the number of notes the chord contains. Measurements prove, that the noise level is still intolerably high, when the original note, which builds up the chord, contains marginally low level of noise. This is much worse in the case of quiet stops, like the Voix Céleste, in which the signal-to-noise ratio (SNR) is bad.
   Therefore, we have removed (not reduced) the noise of the organ engine from all the samples, while retaining its original quality, and we created a single instrument, in which the organ engine could be heard alone. You may add the organ engine back again to your original recording, if you wish. The engine will be added only once to your soundtrack. This is very important, when you play big chords, or 'staccato' chords at a higher tempo. Listen to a noise reduction example (Fonds 8') here (wma, 224 kB).
6. Post processing while retaining the original sound quality. Post processing is very important to be (and so it was) done at the original sampling rate, at 96kHz/32bit to be as accurate as possible. Removing the noise, while keeping the sound intact took thousands of hours to complete. Those stops we couldn't denoise were not included in the library. Sounds of the valves and windchest behavior remained intact, including in the release samples.
7. The natural reverb of the cathedrals were recorded 'as is', and were cut into release samples. They were also denoised precisely to achieve a real decay into silence, not a decay into engine noise. For the ability of reverberating other instruments (or the organ itself) when creating recordings together with our samples, we have measured the impulse responses of the cathedrals in different mirophone setups, using two commonly known methods: MLS and TSP.
8. Filtering was done to some notes to remove the sounds of passing cars and trams outside the cathedral, which appeared in our recordings at around 20-200 Hz. We repeated the recording of the notes, where these frequencies were important part of their sound spectrum. Higher notes have limited bandwidth, so noise in the low frequencies were simply removed.
9. Here are some answers to frequently asked questions why they were enough.
   Q: Sampling every second or third note result in different tuning than the organ's, doesn't it?
   A: No. It is possible to retune the samples for every note even in the case of having only a few samples in an octave, which would solve the problem of unequal tuning, but these symphonic organs are tuned equally. There would be problems only in recording combinations on unequally tuned organs, which is not our case. A sample count of minimum 5 from an octave proved to sound good enough (please check the demos). The intonation of the pipes does not significantly change in 3 adjacent notes - just think how strange it would sound, if the organ would have a too loud note in the neighborhood of others.
   Q: Transposing the samples will transpose the wind noise too, which is again unnatural.
   A: The wind noise is an airy sound, and is very similar to white noise above approximately 3000 Hz. A white noise has constant amplitude in all of its frequencies, which means that transposing it - shifting its frequencies up or down - will not affect it's sound.
   Q: Transposing the release samples would change the lenght of them, which results in unnatural sound, right?
   A: No, not unnatural. Please see the diagram below why. It shows the T20 curve (ISO 3382) of the cathedral reverb (red curve is the average values of the left and right channel). Transposing a sample 3 half-notes up would shorten its release sample length to 84% of the original, thus shortening it's reverberation time to 84%, but as you can see on the diagram (blue curve), the transposed notes, having higher frequencies than the original, has nearly the same reverb coloration - the curve follows the original reverb with a slightly shorter reverberation time (curves are below the original), so it still sounds very much like the same as if they were played in the cathedral. Of course, the coloration changes a bit, but this difference is really unnoticable. The reverberation time will be slightly shorter: the average deviance is 383,72 ms when transposing a sample 3 half-notes up, and 261,401 ms when transposing 2 half-notes up. The greatest deviation is in the 250 Hz band (795,343 ms). As you can see, it is better to have more samples (green curve).
   
   The conclusion is that notes between C3 and A4 will have a slightly shorter reverb in their first harmonics than the in reality. This is another important reason why the convolution reverbs sound more natural than release samples.
   
   The T20 curve is derived from the –5 dB and –25 dB section of the decay curve, which is derived from the measured impulse response of the cathedral. From each corresponding slope, T20 is calculated as the time to reach -60 dB of the original level.