[Home] [EP Action] [Digital Sampling] [Glossary]

A Work in Progress

An approach  for developing useable Digital samples to enhance our PTOS Theatre organs at a reasonable cost.

• Disclaimer
• It works for PTOS.
• It works in four operating theatres and a home installation.
• It may not satisfy you or work for you. It depends on how critical you are of the sounds you hear when the instrument is played.
   
• What it is not!
• It is NOT real pipes.
• It is NOT  Allen or Walker.
• It is NOT an Ahlborn Gallante tone generator.
• It is NOT an attempt to build a digitally sampled organ.
   
• What it is.
  An attempt to add ranks to instruments that are lacking in some of the more necessary sounds used in today’s TO world. The artist has the choice of using them or not. The original organ is still there, available and in pristine condition. BUT we live in this world and time. Given a choice and the resources, we would rather use the real thing. The southeastern US did not have many large instruments. Except for Richmond and Atlanta we have small instruments, 11 ranks and under and very few of those. PTOS also uses digital ranks to fill in the stop keys when adding pipe ranks until the pipes are ready to play.
  
  The link below is an excerpt from a concert played on the Bristol Tennessee Paramount BAL1A Wurlitzer. It will take a little time to download ( ~1 megabytes in .mp3 64K format). The original 11 ranks of Wurlitzer have not been modified in any way. The largest reed on the original organ is a Tuba Horn on 15”. I put the track here to afford the opportunity of hearing the instrument augmented  by a digitally sampled Post Horn 16’-8’, Trumpet 8’, Solo String Celeste, 16’-8’ Oboe Horn, 8’ Quintadena, 8’ Horn Diapason. 16’ Pedal Violone and 16’ Tibia Pedal extension. Allow the file to download then click the arrow in the window below. The square will stop the play and the parallel vertical lines will pause playback.

Click on the icon below to load and play in the Windows Media Player an .mp3 clip from the theme of  “The Phantom of the Opera” recorded by Walter Stony during his concert at the Bristol Paramount Center in October of 2000.

• An Overview of the system.
• The Tone Generator/PA System
• Console control is from the Uniflex 2000 (DOSRelay or ETS) MIDI feature - All five of the organs owned and/or maintained by PTOS have Uniflex systems
• The Uniflex system can treat a MIDI sound as a rank of pipes.  Each MIDI rank can be assigned to a separate channel. That channel can then be modified by assigning MIDI controllers and program changes
• Sound (rank) production is from Alesis QSR Synthesizer Modules.- Three of our organs have two QSR’s and  2 organs have one. The QSR has a MIDI mode that allows assignment of programs (digital ranks) to a specific channel. This dovetails very nicely to the method used by Uniflex to play the rank.
• Sound reinforcement is from Fender KXR200 or JBL EON15 Keyboard Amplifier systems. - Both units are acceptable but the Fender has better speaker response for Flutes and Tibias. Unfortunately, The Fender KXR200 is no longer available.
   
• Feasibility  Digital recording equipment available to sample at a reasonable price. Tape, minidisk and direct to PC capabilities exist. Some of the more interesting and expensive are the muli-track units such as the Alesis ADAT and its cousins.
• Commercially available Synthesizer modules (Alesis QSR) available to do the job.
• 64 note polyphonic.- It is interesting to note that the first Allen organs were at the best 10 note polyphonic then it would release the first in notes played. Polyphony refers to the number of notes a synthesizer can play before it runs out. Most synthesizers have an algorithm to chose which notes to stop if new notes are needed. Usually there are several algorithms to chose from,  each with a different penalty to pay.. The Theatre Organ presents a quite a challenge to the polyphony limit since it is unified and can be intra and inter manually coupled. The maximum number of full keyboard ranks on any QSR that we have at present is 6. Pedals are normally 1 note at a time and so are Chimes. 
• Each voice can be controlled on a separate channel.
• Ability to add additional wave sounds to the synth.
• Reasonably priced  (<$500 ) but also requires an 8Meg PCMCIA card to load our digitally sampled ranks ($64 as of 5/8/2002).
• Commercially available PA equipment reasonably priced (~$500).
   
• How do we get samples
• Where?
• Other PTOS organs.
• Other organs whose owners will allow us to sample.
   
• Tools required
• A DAT recorder (Sony TCD-D8, Sony RM300 or other DAT recorder).
• DAT tapes.
• Microphones minimum, Stereo(Audio Technica 822/825) or 2 Single AKG C1000S).
   
• A process for gathering samples - Most of my successful samples have been recorded in the chamber. The quieter the chamber, the better the sample. Blower rumble is a major consideration because it permeates the total TO environment and usually shows a very broad spike below 120 Hz. It is especially troublesome below C2(8’ C) because when you remove the rumble you can also remove some of the fundamental of the pipe
   

Samples of signal to noise ratio - Dulciana C2 (8’ C)

Noise Sample is at -18db; Dulciana wave shape is at - 3db

FFT display of the noise portion - notice the very broad band at 128Hz and the short spike at 338Hz

The list below is an overview of the process I use to sample pipes.

• Record the rank name before you start.. - It’s amazing how much some strings sound like reeds when you start playing them back.
• Record at 48KHz. - This adds a little safety factor to reduce potential aliasing problems. Aliasing is a phenomenon that occurs when the frequency being sampled is more than half of the sampling frequency... It causes the frequency being recorded to be folded back and added to the  material being recorded and causes distortion
• Be sure all pipes are playing and reasonably in tune. - This is more a convenience to you when you start selecting your samples. It helps you to get an idea of the rolloff both in loudness and frequency that occurs in the rank as the frequency increases
• Get as close to the rank you are sampling as possible(S/N ratio). - I don’t mean that we are really interested in the speech defects in the pipe (unless you trying to record a chiffy flute!!). I normally record Wurlitzers by placing the mike in the center of the chest. That means the no pipe is much more than 4’ away on each side. I try to record the offsets separately.
• Avoid or eliminate echo in the chamber if you can. - I haven’t figured out how to do it yet but then our chambers are small and the echo has not been insurmountable. I expect a light weight Styrofoam sheet around the outside of the chest would help.. I’ll have to try to see if it buys anything. (Maybe that’s why Allen uses an anechoic room!!)
• Be sure when you record that the pipe sound is at least 18db over the noise in the chamber. - The greater the Signal to Noise ratio, the better sample you will have because you wont have to use as much noise reduction when you prepare the sample for looping . I have not been able to prove it but I expect that the noise reduction methods do affect the harmonic structure to some extent if you are not careful.
• Record ALL PIPES - You may need additional samples. I had a recent experience with a virgin Wulritzer Dulciana where the samples of 4 of the pipes I had chosen had a 3rd harmonic content that was greater than the fundamental. To my knowledge, these pipes had never been “improved”. There are .certain things I can think of that could cause a situation like this 1) that’s the way the pipes were voiced, 2) reverberations / echoes in the chamber added to cause the situation, 3) the microphones peaked for some reason at that frequency.  I went through the 3 octaves and did a FFT on each pipe and it turns out that the 4 pipes that I had chosen where the only ones that had the third harmonic anomaly
• Record 4 seconds on, one second off for each note. Leave a 4 second interval for dumb pipes so you can spot it when you play it back!. This gives you enough time to find a good steady state portion of the pipe wave to use for looping. The one second interval between wave shapes give enough space to get a good noise sample and also an undisturbed stop wave shape for the previous pipe and start wave shape for the next one.
   
• What do we do with the Samples
• What result do we want.
• Fact-- We can’t make sample as long as organist can play . It’s been tried by using tape loops and other methods.. The basic problem is reentrancy and it occurs whether it’s a tape loop or a programming loop. In either case, the amount of wave stored is a function of how much memory is available to store it. Once you figure that out then you have to figure out how to stop  the sound. and still get the stop characteristics.
• Looping is an art.
• Noise and  echoes in chambers makes looping difficult.
   
• Some decisions.
• All samples are recorded in stereo but converted to MONO. I use stereo to record because the echo and noise characteristics are different in each chamber and for each rank. Having two samples for each pipe allows for a much greater choice when I go to find the sample I want to use. I play it back in MONO because the chamber opening is small enough to in effect reduce even the pipes to a monophonic voice. In our case at the Paramount, the architect goofed and took 2 feet away from us in both chambers.  We fit the organ in by putting the chests in perpendicular to the swell shades.
• I take untremmed samples and use QSR trems. It works OK for most ranks but maybe not for a Tibia. Looping a tremmed sample will take much more space on the memory card. Good tremulation on a Theatre organ occurs around 6 Hz. That means that the loop now has to be built around a much longer sample than without trem. The other concern  I have is that if the trem is built into the wave shape I would have to obtain a new sample and reloop the whole rank if I want to change any trem characteristics.
  
  Recent information received from a friend has identified another problem using tremmed samples. If you stretch a specific sample across several notes, the trem rate also changes as the playback unit resamples the sound. This problem can be mostly avoided by using individual samples for each note which gets us back into the problem of how many instruments(ranks) can be put on an 8 meg card. (see next bullet.)
• I use two samples per octave (C and F#). With only 8 meg of sample room. I want to keep as many samples on one memory card as possible. The synthesizer actually “resamples” the wave files as it plays each key. That means that it changes the oscillator speed that drives the wave sample for each note. As it does this, the timber of the voice changes slightly. The farther away the resampling gets form the base frequency, the more the timbre changes. Some sounds are affected more than others. This is where it is important to be able to hear the voice as you are building the rank. You may have to spread a sample out more than 6 notes (normally C to F#, G to B) to make the next sample voice to the previous or following sample to get a good match..
• No stop wave form is carried with the sample. The best I can tell, Artisan layers two samples to make one. The QSR can layer up to 4 voices/samples in each program (patch). The first sound is the start wave envelope and loop sample and the second sound is the stop envelope sample As part of the ADSR (Attack, Decay, Sustain, Release) envelop controls, the QSR can play the ADS part using Sound 1  then play Sound 2 for the Release portion. The problem with using layers in the QSR is that it uses up a polyphony note for each layer If 2 layers are used and you play one note, the QSR uses 2 notes of the 64 note polyphony to play the note. It you have 4 layers then it uses 4 of the polyphony notes. You can see that would not take too long to run out of notes.
  
  I am working on technique to simulate rank stop wave with the Sound Forge plugins. I use the fade plugin to make a template of the stop wave shape of the samples. Usually they are fairly characteristic for a given rank so they will be the same basic shape for the notes in the rank. The differences I have seen in the samples seem to be a function of the length of the pipe. This method would give the characteristic of the envelop but would not contain the frequency changes found as a pipe stops playing
   

The Hardware tools required.
The PC requirements I have defined here are based on the equipment that I found after a good bit of trial and error. I do not have the resources to investigate other potential systems. The Sound Blaster Live is  the most flexible sound board I have found for doing the things that I like to do. The fact that you can load and play new sets of instruments makes it of great use to the experimenter. The freeware Vienna Studio program provided by SB is very convenient to use to manufacture a new rank (instrument they call it) As usual, each sound board has it’s own hardware system. SB uses the EMU1000 chip which uses the .sf2 type files to store new voices and uses .wave samples to manufacture new instruments. I don’t know of another sound board that uses the .sf2 format.

I use the Opcode Sonicport to transfer the data from my DAT to the PC  because it has a fiber optics connection. This type of system reduces the generation of hum and noise. Besides - it connects to my PC USB port which makes it easy if install and remove..

I use the Sound Blaster Live card to Voice and regulate the rank. I have an Altec stereo system with a sub woofer connected to the PC to play back the ranks. Once I get the rank to a fairly decent sound, I switch the output of the SB Live over to my stereo system which has better speakers and and an equalizer
 

• .An IBM clone PC.
• Windows 98SE.
• => 400MHz AMD or Intel processor.
• USB Port(s).
• => 30 GBytes HDD.
• Opcode SonicPort (optical system to load Wave forms from DAT to PC).
• Creative Live Sound Board used to hear the wave samples and voice them. Sound Blaster is the only PC Sound Board I have found so far that provides a programming system that will allow you to add wave sampled instruments to a sound generator.
• A good 17’ SVGA Monitor that will be used the to view the wave samples as they are manipulated.
• A reasonably good Audio system to use for regulating and voicing the samples.
• Theatre Organs are inherently noisy. You don’t find out until you sample the pipes how noisy they really are and how much the noise hides bad regulation and voicing.
   
• The Software Tools required -The difference between voicing and regulating samples are not that different from voicing and regulating the analog pipes. The tools enumerated below will be used to do the same basic tasks. It is interesting to see the audio formants of each recorded pipe and how each pipe differs one from the other as to their harmonic train. The Fast Fourier Transform (FFT) in Sound Forge makes this information visible and very obvious.
• Sound Forge 5.0 with Noise Reduction 2.0 - Noise reduction, looping, volume, Paragraphic equalization, pitch and just about any other thing you would want to do to a pipe. This program is the heart of  voicing activity. If your PC is fast enough you can work in real time. As an example, you can actually do noise reduction, listen to it and if you don’t like it restore your file to its original state. The same is true of the all of the functions that you might want to use.  The looping capability in Sound forge is the easiest to use I have found. It is also one of the most important functions you will have to use.because it defines the files base frequency and steady state sound when a note is held. The output from Sound Forge will be used by all the programs that follow.
• Vienna 2.3 (Sound Blaster’s free download ) - Makes ranks (instruments) from individual wave samples.
• AWave (Shareware) - Converts .wav files to .aiff Mac files required by QSR. This program also has looping and minor editing capability but is not as easy to use as SF5.0. It can also generate instruments form individual wave forms . AND -- It is capable of reading and writing just about any digital sound file type available today.
• Sound Bridge 3.0 - QSR free ware to download samples form the PC to the QSR Memory Card
• QSEdit Pro 2.0 (Shareware) - A program to edit sounds/samples in the QSR.. All this can be done at the QSR but it is very labor intensive and error prone
   

Below are four views of a Post Horn wave sample for the Middle C (C4) pipe.

View of the complete C4 Post Horn recorded wave shape defining
 the looped portion

Click on the icon below to load  and play in the Windows Media Player a .mp3 sample of the C4 Post Horn shown above

This view shows the looped area in Zoomed mode to give a better view of the actual wave shape

Four Cycles of the Post Horn C4 pipe to show the detail

Fast Fourier Transform (Spectrograph) of C4 Post Horn Wave
The large peak on the left is the fundamental frequency, C4

Comments to the Web master