Electronic Organ

Fall Senior Design Project EE401

December 3, 2006

by

Matt Jibson, ECE

Dallin Kuzmich, EE

Dawn Dalangin, ME

Prepared to partially fulfill the requirements for EE401

Department of Electrical and Computer Engineering

Colorado State University

Ft Collins, CO 80523

Report Approved:______

Project Advisor

______

Senior Design Coordinator

- i -

ABSTRACT

We are creating an electronically controlled pipe organ, augmented with synthesized (electronically generated) pipe sounds. These synthesized pipe sounds will be produced by analyzing samples from real pipe organs and recreating the harmonic waves. This is a new approach, as most electronics organs today operate by playing back recorded samples. The entire block diagram for the project is shown below (Fig. 1).

Figure 1 Block Diagram

- ii -

TABLE OF CONTENTS

Titlei

Abstractii

Table of Contentsiii

List of Figures and Tablesiv

I. Objectives1

II. Results to Date1

III. Obstacles4

A. Understanding how traditional organs work6

B. Building pipes that would produce accurate pitches6

C. Sizing the blower7

D. Building the wind chest8

E. Adjustment of pipes9

IV. Budget9

V. Plans for Spring Semester10

References12

Acknowledgements12

- iii -

LIST OF FIGURES

Figure 1Block Diagramii

Figure 2Keyboard1

Figure 3FPGA and External Logic Interface2

Figure 4Wind Chest and Pipes2

Figure 5PERT Chart4

Figure 6MOSFET Schematic 16

Figure 7MOSFET Schematic 26

Figure 8Blower7

Figure 9Wind Chest Measurements8

LIST OF TABLES

Table 1Purchased Parts and Cost9

Table 2Donated Parts and Estimated Cost10

- iv -

I. Objectives
Our overall objectives are to have a fully working one rank pipe organ augmented with electronically synthesized pipes. The full manual will comprise 61 individual keys and 49 pipes made up of 5 octaves. The synthesized sound we will use will follow a new approach. Instead of playing pre-recorded sounds like most electronic organs today, we will be analyzing recordings from CSU's world famous Casavant organ and recreating them from the frequency domain. This will be done by harmonic analysis of individual pipe recordings. We anticipate the year long budget not to exceed three hundred dollars.

Fall semester objectives include analysis of audio signal properties, procurement and donation of parts, and construction of wind chest. Budget for the fall semester was expected to be over 500 dollars as we needed to acquire various, expensive parts.

II. Results to Date

End of semester results include the set of tasks and which of those tasks were accomplished. The results to date include a working keyboard interfaced with a Xilinx Spartan 3E (Fig. 2) Field Programmable Gate Array (FPGA), successful testing of the FPGA with logic devices (Fig. 3), and completed wind chest with attached pipes and blower, which includes air passage to the pipes through the toe board (Fig. 4).

Figure 2 Keyboard

- 1 -

Figure 3 FPGA and External Logic Interface

Figure 4 Wind chest and Pipes

- 2 -

Currently the FPGA has many working modules. It has a working VGA output that displays a full keyboard. The MUX control and key sensor interface is working. When wired properly, the VGA display correctly shows which keys are depressed. Basic sinusoid generation is working using a Direct Digital Synthesis (DDS) core from Xilinx (although, we have decided to use System Generator instead, and so will

be removing the DDS sinusoid generation).

Procurement of the MUXs, S-R Latches, and MOSFETs was a necessary incurred cost whereas the electric valves, keyboard, and pipes were among the donated parts. Selection of these electronic devices was based primarily on cost and performance characteristics. In order to build the circuits onto breadboards, all parts needed to be dual in line package devices. The valves were tested and found to operate at 14 Volts and vary in current from 0.1 - 0.16 Amps depending on the size valve.

The mechanical aspect of the electronic organ includes the blower, wind chest, and pipes. The idea behind using air flow to produce sound is to keep the traditional methods of using a blower to generate the “wind” which is held in the wind chest. The sound would be produced via pipes that are directly connected to the wind chest.

The objectives earlier mentioned were combined into a PERT (Program Evaluation Review Technique) chart for improved project management and scheduling between all members of the team. The fall semester tasks were coordinated as shown below (Fig. 5).

- 3 -

Figure 5 PERT Chart

III. Obstacles

There were many small problems that arose when using the FPGA. As this was the first FPGA that Matt used outside of a CSU class, many more details of its operation had to be understood. Fortunately, Matt's employers are former employees of Xilinx, and helped in the creation of software and hardware on the FPGA. They were consulted often on technical aspects of its operation. Furthermore, Verilog has a completely different programming paradigm than normal procedural languages (like C, Java). As it is running on custom hardware, all modules operate in parallel. Due to this problem, what typically happened when a new module was to be implemented was: 1) specs and docs would be consulted regarding the new component (e.g., VGA controller, D/A converter, PS/2 port), 2) an implementation would be written in

- 4 -

Verilog, tested, and found to not work at all or as expected, 3) searching the Internet for working Verilog, implementation and custom tweaking until it works as desired.

One module that has not performed as expected is the PS/2 interface. Some code was downloaded (after trying to write a custom implementation that failed) and hooked in to the build. During testing, it appeared that the FPGA was not receiving data from the mouse (which did have power). To fix this problem more time is need to experiment with the hardware.

Our keyboard has also been a constant source of problems. It is old, and many of the connections are broken or unreliable. Also, we are using 7400-style chips, which are bulky. The many wires we have had to make sometimes break, causing surprising problems to creep up. We have found that the fix for these problems is to constantly monitor all connections, and double check them when we are having unexpected problems.
One issue that arose during the testing stage of the MOSFETs was the switching of the direct electric valves in which our initial approach of selecting a load line for a DC quiescent operating point and corresponding drain current. Referencing the output characteristics of the n-channel enhancement mode power transistor we added an external drain resistor only to discover that the device was already in an 'on' state meaning it was conducting a drain current without an applied gate voltage. More interestingly we found out that by applying a gate voltage above threshold the transistor would turn 'off' and cease to conduct between the drain and source terminals, a depletion mode device characteristic. The problem was brought before the team and Dr. Eads. Matt pointed out this was working as expected, and Dr. Eads suggested a reconfiguration of our schematic by removing the drain resistor and connecting the valve in series with the power supply and drain terminal. Following this plan our results were favorable in that the valve could successfully switch 'on' with relatively low gate voltage and thus operate in the enhancement mode. Shown below are the schematics of the transistor and electric valve before (Figure 6) and after (Figure 7) collaboration with Dr. Eads.

- 5 -

Figure 6 MOSFET 1 Figure 7 MOSFET 2

Understanding how traditional organs work
Organs built today are left to professional organ builders. The challenge with this project was to try to build something that is a learned trade (that takes several years) in the course of one semester. Most of the initial research was conducted simply by reading many organ-building books found at the library. This research helped to narrow down what would be necessary for our scope of the project.
Direct observation was also a key component to understanding how organs are built. A few organs found in the music building at Colorado State University were inspected.One of the best resources was Charles Ruggles, an organ builder located in Conifer, Colorado. Originally, it was thought that a reservoir (to hold wind) and bellows (to regulate wind pressure) would also have to be built. After meeting with Charles in Conifer on September 24th, it was determined that the only components necessary would be the blower, pipes, and wind chest.
Building pipes that would produce accurate pitches
The first design of this project was to build 13 wooden flue pipes. Wood was determined to be the easiest to machine and build. The challenge behind building pipes is that everything affects the pitch, sound, and tone. Factors such as length, diameter, mouth (this is the slit on the side of the pipe) size and location have to be taken into account.
Fortunately, meeting with Charles also proved beneficial for this task. He donated a set of 49 wooden flue pipes (two are actually missing, totaling 47). He also provided charts and graphs to determine what the

- 6 -

dimensions for the missing pipes would be in order to build them to the correct pitches. This greatly reduced the task of building the original 13, narrowing it down to only two pipes.
Sizing the blower
It was determined that the type of blower needed would have to be a centrifugal blower. Air enters the impeller axially, leaves the impeller at high velocity, and can be discharged in high volumes.
The challenge was to first determine if a blower was necessary, versus other means of air flow such as fans or compressors. After research and communicating with several organ builders, it was determined that a blower was necessary for the project due to the fact that fans operate at less than 2 psig, blowers operate between 2 to 10 psig, compressors operate at greater than 10 psig.
For one rank of pipes, a fan would provide too low an air pressure and compressors too high an air pressure. Blowers are mainly used in organs built and existing today. Additionally, after speaking with Charles, common practice is to first use a blower that is of suitable size (which is determined by the number of ranks used) and then build a bellows to regulate the air pressure.
The blower that was purchased, (suggested by another organ builder) would be sufficient for only one rank of pipes. The blower was also chosen because it could be plugged into a standard wall outlet, so that supplying power would not be an issue. The diameter of the outlet 2 – 3/8 inches also fits a standard PVC pipe fitting and would be easy to mount (Figure 8).

Figure 8 Blower- 7 -

Building the wind chest
There are several components to the wind chest that will be addressed in the following order: type of materials used, toe board and rack, and overall dimensions.
To build the wind chest, the issue was to find cheap materials that would be easily machinable and be able to withstand air pressure from the blower. Organs today have completely wooden wind chests. For our project, it was determined that a combination of Plexiglas and wood would be used because of they are easy to machine and inexpensive.
The main challenge with building the toe board and rack was to figure out where the pipes would be placed and how the holes would be machined in order to achieve a flush frontal pipe surface. After measuring several dimensions on each pipe,specific x and y coordinates were calculated. These x and y coordinates were used to machine holes of certain diameters (varied per pipe) into both the toe board and the rack. Each hole of the toe board had to be machined twice in order to have a depth large enough for the pipe, while enabling only a small amount of air flow (Figure 9).

Figure 9 Wind chest Measurements

The overall dimensions of the toe board were determined from the pipe dimensions and the blower dimensions. Since the design was to use two rows of pipes, the length of the wind chest corresponds to half of the total length. When the pipes were laid out side by side, the total length was 73 inches. Add a quarter inch in between for spacing and a total of 7 feet and one inch was calculated. To allow enough room for the

- 8 -

other sides of the wind chest, 8 feet was determined as a sufficient length (only counting one row of pipes). Divided by two, the wind chest length is approximately 4 feet. The largest dimension of the blower was 8.5 inches. The height of the wind chest was used as the given dimension of the Plexiglas (12.5 inches) to allow extra space for mounting.

Adjustment of pipes

When the blower was plugged in and producing air flow, the challenge was to produce an audible sound (speaking) from the pipes. Upon inspection, there were several air leaks that were temporarily sealed with masking tape. (The wind chest is not sealed for this semester, due to future plans of adding additional pipes and valves.) After these air leaks were sealed, several pipes began speaking. However, the problem of adjusting each pipe to produce the desired frequency remains uncertain. The proposed resolution will be to correct the holes in the toe board so that air leakage is reduced as the connection becomes more secure.

IV. Budget

The fall semester budget is listed in the following tables of those parts purchased and donated throughout the course of the term.

PURCHASED PARTS / COST
15 MUX / $16.78
20 S-R LATCH / $20.08
65 MOSFET / $40.30
100 FT 22 GAUGE WIRE / $23.96
WOOD BOARD AND DOWELS / $37.58
BLOWER / $28.15
TOTAL COST / $166.85

Table 1 Purchased Parts Cost

- 9 -

DONATED PARTS / ESTIMATED COST
XILINX SPARTAN 3e STARTER BOARD / $300.00
AEOLIAN SKINNER ELECTRONIC ACTION KEYBOARD / $1000.00
FULL TOE BAORD WITH DIRECT ELECTRIC VALVES / $300.00
47 PIPES / $1000.00
PLEXIGLASS / $10.00
TOTAL ESTIMATED COST / $2610.00

Table 2 Donated Parts Cost

The budget for the fall semester has been mostly used to fund the procurement of fifteen 16x1 Multiplexers ($16.78), twenty S-R Latches ($20.08), and sixty-five MOSFETs ($40.30), 100 feet of 22 gauge copper wire ($23.96), wood boards and dowels ($37.58), blower ($28.15), totaling $166.85. Our donated parts are a Xilinx Spartan 3e Starter Board ($300), an old Aeolian Skinner electronic-action keyboard ($1000), a full toe board with direct electric valves ($300), 47 pipes from a local organ builder ($1000), and various pieces of Plexiglas ($10).

V. Plans for Spring Semester

Our final deliverable will be a working one manual, no pedal, pipe organ. It will have one full rank of pipes. It will be augmented with at least six ranks of simulated pipes, created dynamically from harmonic analysis of recordings from a real pipe organ. To achieve this result we will need to first wire the logic devices (input/output MUX, S-R Latch) in conjunction with the FPGA, keys, and electric valves in order to begin testing and most likely debugging the controlled response of the organ. Several pipes need to be

- 10 -

constructed to complete the rank and the air blower will need to be connected to the completed wind chestwith electric valves. A small hole in the wind chest will be made to allow the bundled wire from the MOSFETs to connect to the valves which will then be sealed to prevent excess air leakage. The final step will be to implement the synthesized waveforms from the FPGA. We will also continue to ask for donation of parts (wire and power supply) and expertise to see through to the final deliverable.
Matt will continue to work on the FPGA. This will include fixing the PS/2 port, creating a rank control system, putting the stop list on the VGA display with mouse-click toggle, and implementing the harmonic recreation code using System Generator, a software package from Xilinx. This product is a plugin for MATLAB's Simulink. Dallin and Matt will work together using MATLAB analysis tools to create appropriate algorithms that can easily be used in SysGen and integrated into the FPGA.

Dallin will analyze the audio recordings of the Casavant organ in order to construct a sinusoidal function to account for the changing harmonics. This work will be accomplished through an understanding of signal processing and testing procedures of MATLAB software. The specific issue being addressed is how harmonics change over frequency of one rank. That is, in one rank of pipes, the harmonics generated above the main tone differ depending on the frequency of the pipe. For example, a high frequency pipe may have lots of harmonics in relatively similar amplitudes, whereas a low pipe may have clearly defined first through fourth harmonics, but no notable contributions afterwards. In addition to these tasks it will probably be more beneficial to re-wire the keyboard in a fashion less confusing than the bundled mess of wires.

Dawn will finish construction of the two missing pipes and various repairs on the already existing pipes. Other plans include tuning the pipes to equal temperament and constructing a cabinet to house the keyboard.

We anticipate spending $100 for the spring semester. Possibilities include various smaller parts for the wind chest, reconstruction of broken pipes, and more wire and bread boards. The largest part will likely be larger pieces of wood for the cabinet that will house the keyboard and possibly the wind chest.

- 11 -

References

Donahue, Thomas, “The Modern Classical Organ,” McFarland Company, Inc., 1991.

Matley, Jay, and The Staff of Chemical Engineering, “Fluid Moveres: Pumps, Compressors, Fans, and Blowers,” McGraw - Hill Publications, Co., 1979.

Norman, Herert and Norman, H. John, “The Organ Today,” Barrie Books Ltd., 1967.

Whitworth, Reginald, “The Electric Organ,” Musical Opinion Ltd., 1948.

Whitworth, Reginald, “The Student’s Guide to the Organ,” Musical Opinion Ltd., 1935.

Acknowledgements

Dr. Eads – EE Project Advisor

Tom Fischer – FPGA engineering assistant

Charles Ruggles – Organ part donations and expertise

Judson Murphy – Organ part donations

- 12 -