aboutabout.html
blogblog/blog.html
gallerygallery/gallery.html
projectsprojects.html
 

Stereolith

At the start of my electrical engineering degree, I knew I wanted my senior design project to be something that fused aesthetics and electronics. Steam punk really took off during these years, and had an obvious appeal. The nixie clock craze also went through a huge boom phase. While reading about all of the nixie projects out there, I came across magic-eye tubes. These are indicator lamps housed in vacuum tubes. I knew that these would be a huge inspiration for my senior design project. And I knew exactly how I wanted to use them...

My passion for tube amps started several years ago when I started playing guitar. I asked a guitar guru I respected what kind of amp I should get. He suggested a Fender Blues Jr. It’s a small, 12” 15 watt, open back amp. When I asked why, he replied, “Because nothing sounds like tubes.”  At first I wasn’t so sure. It took a while to warm up, and the tone was a little unpredictable. But, once everything was setup, the tone was impeccable. I convinced the other guitarist in my band that this was how we were going to get the sound we were looking for. For several months we played on our new amps and thought little of the tube tone we were using. At a party we went to, there were some other musicians (all of which were far more qualified than me) with some really nice gear. However, when they started playing, I looked at my fellow guitarist and we knew what the other was thinking. These expensive transistor amps can’t hold a candle to the tone of a tube amp. We were hooked.


There is a lot of debate in the audio world about how tube amps can really sound that much different. I’ve found one paper that attributes it to the harmonics present after amplification. I’ve heard transistor amp setups I like and some I dislike. On the other hand, I’ve heard tube amps I like, and some I don’t.


That all said, I wanted to build a tube amp. But, the magic-eye tubes were the real kicker. We would use 10(!) of them to provide a 10-band spectrum analyzer of the audio signal being played. For me, this seemed like enough of a project. However, one of my team mates felt we could also build a wireless, touch-screen remote that would control volume and a 10-band parametric equalizer. To me this seemed like too much for our plate. But, we had an extremely driven team of talented people. Through some weekend meetings and long nights, we were able to meet all of our specifications.





This is, unfortunately the only picture I have of the finished project. We stayed up until 4a.m. searching for ground issues the night before our presentation day. 3 hours later I was back at school with a suit on trying to field questions from engineers of all disciplines. Somewhere in all that, I forgot to take photos. Anyhow, here’s a breakdown of the technical aspects of the project. The following text is mostly quoted from our group’s final report.






The block diagram above shows the signal flow of the entire project. Notice there is a remote, and an enclosure that houses all of the signal processing, amplification and spectral display.


Technical Background:


Vacuum tubes were used for switching and amplifying in electronics before the invention of solid-state transistors. At the heart of the vacuum tube is a heater, or filament. The heater is just that; a piece of metal through which current is passed in order to heat the metal. This is what causes tubes to “glow”. It also liberates electrons from the surface of the metal. Once electrons are freed from the surface of the negative side of the tube (or cathode), they fall back to the cathode. In order to conduct current, a large voltage must be present to pull electrons away from the cathode. This high voltage (on the order of 250v to 400v) is applied to the positive side of the tube, known as the plate. With these two connections in mind, the tube can be thought of as a diode. Electrons can be liberated from the cathode and flow out the plate, but cannot be forced from the plate back into the cathode. To control current flow, an additional connection must be made. The current-controlling connection is called the grid or screen. It is a grid of metal placed between the cathode and anode. By varying the voltage on the grid, some electrons are repelled and do not make the entire journey to the plate. Using these three connections, a tube can be used like any transistor where a small voltage controls a large current.






Tube Amplifier:


The design for the tube amplifier is based on the “Compact Hi-Fi Amplifier” from Melvin Leibowitz in the June 1961 issue of Electronics world. In 2008 the circuit was modernized by Bruce Heran. He implemented a voltage regulator IC to serve as a constant current source for the output stage.





The amp consists of two stages; a driver stage and an output stage.  The driver stage (consisting of the network surrounding the left two tubes in the schematic) of the amplifier does little power amplification. Rather, its purpose is to amplify small audio signals (less than a volt) to a swing large enough to drive the output stage. The driver stage is a common push-pull arrangement.


The output stage is a modern implementation designed by Heran. He has named the circuit a “Self-Inverting Push-Pull” stage, or SIPP.  The voltage regulator is set in a feedback loop to provide a constant 125mA through the two push-pull driver tubes. A potentiometer between the two tubes allows them to be biased equally. The 1ohm resistors on either side of the potentiometer allow for current calculation in each branch. At DC, both tubes conduct the same current. This means there is no change in current in the transformer primary, so there is no change in output voltage on the transformer secondary. If the signal on the grid of the first tube causes less current to flow, the constant current source will force the other tube to conduct more current. This causes a change in current flow in the transformer and is coupled to the output.


The transformer serves to isolate the circuit from the output while also matching the load. Typical loudspeakers have a DC resistance of 8 ohms. The output stage of the circuit is around 8k ohms. If the speaker were connected directly to the output of the circuit, the gain would be diminished by the impedance mis-match. Maximal power transfer can be had by matching the loudspeaker impedance through the transformer.


Minimizing distortion is paramount in design for high fidelity audio equipment.  Both driver-stage 6SL7 tubes and KT77 output-stage tubes were chosen for their low levels of distortion. Along with careful grounding techniques, and low noise resistors in the signal chain, high fidelity audio reproduction is ensured.



Equalizer:


An equalizer is a circuit that allows a change in the frequency response of a system. It is to say, that frequencies can be boosted or attenuated.  Modifying the frequency response in an audio system correlates to a different sounding output.  The equalizer for this project consists of ten frequency bands capable +/-12dB modification.  The frequency bands are centered at 31.25 Hz, 62.5 Hz, 125 Hz, 250 Hz, 500 Hz, 1 kHz, 2 kHz, 4 kHz, 8 kHz, and 16 kHz.  The topology consists of summing circuitry and ten multiple feedback band pass filters which can be seen in the images below.


Input and Summing Circuitry:




Multiple Feedback Bandpass Filter:




By changing the values of the resistors and capacitors in the multiple feedback band pass filters, one can control the center frequency of the filter.  To have the equalizer controlled from a remote, digital potentiometers are used in the filters; the potentiometer is responsible for boosting or cutting the magnitudes of the signal at that frequency. 


The equalizer is implemented on a custom-made PCB and utilizing Texas Instruments OPA1612 audio op amp with 0.000015% THD.  Also, high precision resistors and capacitors with accuracy ratings of less than 2% were used to make sure the band pass filters were operating over their intended frequency ranges.


Spectrum Analyzer:


A discrete Fourier transform (DFT) transforms a discrete, finite-length signal from the time domain to a discrete sequence in the frequency domain. This is very useful in signal processing because it allows one to calculate the frequency components of their signal. There are many ways to implement a DFT. Quick, efficient algorithms called fast Fourier transforms (FFT) are one of the most common. The methods used to compute the FFT are often very complex and have deep roots in mathematical theory. A spectrum analyzer employs an FFT algorithm to break down a signal, and visually display the frequency content of a signal.


A spectrum analyzer shows a visual representation of the frequency content of a signal. For the project, a 10-band spectrum analyzer is implemented using EM84 tuning-indicator tubes. These 10 bands correspond to the 10 filter bands of the graphic equalizer. The analyzed signal is coming in from an outside music source (i.e. CD player, iPod, etc.). This signal is generally in the range of 50-100mV.  The first step in converting the time domain audio signal to its frequency representation is sending the signal through an automatic gain control circuit (AGC). The AGC ensures that the input to the next stage will be roughly 3.4V peak-peak no matter what the level of the input signal. After the AGC, the signal goes into the on-board 12-bit analog to digital converter  (ADC) on the microcontroller. The ADC turns the analog signal into a digital one, which can be signal processed. Once the signal is digitized, the Fourier transform of the signal can be computed. The Fourier transform takes the signal from the time domain to the frequency domain. There are many ways to implement this, an FFT, or fast Fourier transform, algorithm was chosen. This algorithm was chosen because, like the name says, it’s fast. A relatively quick method was needed to ensure at least 30 transforms per second could be performed.  After the FFT is taken, the signal is broken up into 10 spectral bins between 20Hz and 20kHz. Each of these bins will store the magnitude value in that spectral range. This value will then be normalized, and used to determine the duty cycle for the pulse width modulation (PWM) output of the microcontroller. The 10 PWM outputs (corresponding to the 10 spectral bins) will then each be filtered to get a DC voltage output. This voltage output will be between 0 and 3.4V. This, however, is not enough to drive the indicator tubes used for the display. The tubes need 0 to -21V for full operation. To account for this, driver stage is used, where each PWM output is sent through an inverting op-amp, with a gain of -7V/V. This gives the voltage output needed.


The automatic gain control circuit was realized using a THAT4301 dynamics processor IC. This chip was chosen because it is made for audio purposes, so it works for the frequency range of interest. Also, it enables the implementation of an AGC circuit using only one chip, whereas other designs take upward of 4 chips.




An ATMEGA168 microcontroller is used to perform the FFT. Custom FFT code was written by the team, though many libraries are available for FFTs. Unfortunately, this microcontroller only has 6 PWM outputs, and 10 are needed. To solve this problem, a TI TLC5940 PWM chip is used to expand the PWM output channels.. This chip takes serial input, and has the capability to output up to 16 PWM channels. After the FFT is performed, the maximum values are normalized, and used to determine the duty cycle for the PWM output for the corresponding spectral bin.

 

The op-amp driver stage is implemented using OPA453 high voltage op-amps. Using a basic inverting amplifier configuration, the output of the microcontroller (3.4V) is boosted to a level suitable for the tubes (-21V). After the op-amp stage, the 10 signals are each passed through a simple RC filter to get a constant voltage. Finally, the signals are each sent to the input of a tube. The tubes are EM84 tuning indicator tubes. These tubes emit a green glowing bar that varies in length depending on the input voltage (0- to -21V).


Power Supply:


Various voltages are required to supply the electrical energy needed to operate the different components of the amplifier. The design requires taking voltage from a normal wall outlet, (mains or line power) which is nominally 120Vrms, 15A at 60 Hz. To connect the mains to the equipment, an IEC connector was used as the male plug. The cord from an existing power cord was reused. To prevent the transformer from drawing an extremely high amount of current from the wall for any reason, a fast blow fuse is implemented. The power transformer used is the XPWR101 from Edcor. It takes input of 120 Vrms at 60 Hz. On the secondary, there are four different outputs; 240V at 2A, 60V at 0.1A, and two 6.3V outputs at 4A . These are all AC voltages and must be rectified with a full wave bridge rectifier. Being that rectification does not produce a constant DC voltage, smoothing capacitors are used.


Vacuum tubes require high voltages for biasing.  The KT77 output tubes need a 350V supply. This is done by rectifying the 240V secondary from the transformer and passing it through a smoothing (filter) capacitor which is then fed by a resistor to raise the voltage so that 350V can be pulled off the 100 uF capacitor at this point. A separate filter for the 6SL7 driver tubes is then continued off of this circuit (also fed by a resistor) and bringing the final output voltage to 250V on the last 100 uF capacitor in the circuit. Since vaccum tubes draw current non-linearly, the resistances in the 250V circuit had to be determined experimentally, since they are extremely dependent on the loads in the circuit.


There are also two 6.3V secondary windings coming off the transformer which are combined in series to give a 12.6V output. After being rectified, the actual DC voltage is around 17V and is be passed through a series of diodes to supply to the desired 12.6V for the heaters. When designing for the voltage ripple for this part of the supply, designing for a ripple less than 0.1V (using the equation
for full wave rectification) does not apply, nor is it even feasible to carry out due to the huge capacitance that would be needed to accomplish this low of a ripple. This is because there is such a high current being drawn; the full 8A that is supplied from the transformer. Each of the tube’s heaters will draw 0.2A. In the project design there are six audio tubes and ten spectrum analyzer tubes. This will push the secondary to its limit and we must be concerned not to exceed the specified ratings. Since vacuum tubes are extremely tolerant to voltage swings, (so much so that they could be run off of AC voltage) a power smoothing capacitor was chosen such that the voltage ripple is about 1V. This greatly reduced the cost and number of capacitors needed when compared to the 0.1V ripple design.

To provide power to the +/- 24V for the op amps, 2 Murata DC/DC converter chips in parrallel were used. The Murata chips were powered from a  5V LG cell phone charger, which also ran off the wall supply of 120VAC. The Arduino microcontroller boards, and equalizer boards were run of an external +/-15V power supply. In the end, the 60V winding was not used.




Block Diagram of Various Voltages used in Project




Wireless Remote:


Rather than use a typical button-press style remote control,  a touch screen is used as the control input.  This option proved to be the better choice, as a programmable touch screen has much more utility than a physically fixed set of buttons.  This allowed the design for the control to be much more flexible, even up to the final stages of production.  The Graphical User Interface (GUI) implemented on the remote is a single screen consisting of several input sections.  Simple sliders, resembling physical mechanical sliding faders found on many audio systems, are be used to obtain values from the user.  Also used are several buttons to bypass the EQ, set preset values to the sliders, and increase or decrease the volume. 


It was decided to use with a simple RF transmitter and receiver.   The RF Link 2400bps receiver and transmitter from Holy Stone Enterprise Co. are used. Both receiver and transmitter operate at 315 MHz.   

A 320x240 OLED pre-constructed touchscreen and microcontroller combination, the Touchshield Slide (TS) is used as the heart of the remote.  The TS was designed by Liquidware, a small open source hardware company.  The TS has limited libraries, but is designed to be compatible with various Arduino development boards.  The only user input obtainable by the TS is the location of a touch in Cartesian coordinates.  In order to make use of our TS as an input, a GUI needed to be designed to convert the touch input from the TS to useable values.


Several input objects are used which direct the user to provide the correct input to the TS.  The sliders are created by first drawing an image on the screen background.





The lines of this background image are then lined up with a fixed axis of movement of the slider knob.  Then, an image representing a knob is drawn over top of these lines.  As the user touches the screen somewhere within the area of the slider, the knob image is drawn in this location.  This process is accomplished in code by first drawing over the pre-existing image with correctly placed lines and pixels to match the original background image.  A new knob image is then drawn at the touched location.  The input is obtained by taking the touched value as a Cartesian coordinate with respect to the axis of the slider.


The wireless remote is entirely self-contained  To do this a case is needed that contains an on/off switch, internal power source, the RF transmitter, the arduino and touch screen.   The remote enclosure is small enough that it can be easily held while using a stylus to manipulate the touch screen.   The case for the wireless remote was made of wood due to ease of assembly, and experience by team member Richards. The case is only large enough to hold the batteries, touch screen, Arduino, transmitter and on/off switch. This ensures a small, ergonomic form factor. The case is made with two boards that have been cut to size and hallowed out with a router to make a compartment for all the components.  The Arduino runs on 6 to 19 volts and the transmitter, from 5 to 12 volts, so a 9 volt source is used.  Two nine volt batteries are used in parallel to increase the run time of the remote.   A hinge and clasp system allows the remote to be opened for battery replacement, or firmware upgrades to the Arduino/TS while maintaining a presentable touchscreen remote when closed.  The case is painted black for aesthetic reasons.



Control System:


The receiver Arduino controls the settings of the equalizer, volume, and a bypass relay for the equalizer.  The individual equalizer bands are controlled by sending values to a digital potentiometer.  The volume, however, is set by a single-turn logarithmic taper potentiometer.  This potentiometer is connected to a motor which is easily controlled by the Arduino.  This approach for the volume ensures the sound quality of a traditional potentiometer, but retains the ability to control the volume electronically.



Enclosure:


The enclosure is a internally designed and hand-made unit.  1/16” steel is necessary to support the weight of the massive transformers. The box cutting, shaping and welding is from in El Paso, Texas by group member Soltero et al. Holes are drilled by group member Oleson (thats me!) in the UN-L Physics Student Machine Shop. The powder coating is by TMCO of Lincoln, NE.


 



Conclusion:

Team Powerhouse, you guys rock. I can’t believe it all worked. There was a point at 3:30am the day before the presentation that we plugged everything in and no sound came out. My stomach flipped. We pulled through.


The thing weighed somewhere around 50 lbs (about 22 kilos for anyone using a logical system of measures). It chowed down about 700watts of power to output around 30 watts of audio. That’s what happens when you have 16 tubes with glowing orange filaments. The beast ran steadily and beautifully until the final hours of the presentation day. Alas, the heat would be its downfall. In the struggle to fit everything inside at 4:00am, there were some adhesives used that couldn’t tolerate the heat for more than six hours of solid running. In an instance, the audio was silenced undramatically. We finished the final hour with no audio. Adrian took the Stereolith home and still has hopes to resurrect it one day.