I've wanted to make a couple of mechanical modules. One is a Raymond Scott Circle Machine and the other is a Hammond Vibrato Scanner. I completed my mechanical Hammond vibrato scanner and also a Jürgen Haible electronic scanner chorus/vibrato module so it was time to work on the Circle Machine.
I don't know a lot about the circle machine other than the brightness of lamps and the speed of the rotation was the basis for the sequencing operation. There is some additional information of the Circle Machine at the Encyclotronic Electronic Music Archive.
I only had one obscure small photo that is no longer on the web. I obtained this photo taken by Raymond Scott, and provided courtesy of the Raymond Scott Archives, many years after building my Circle machine. It appears the bulbs are "read" from the side. I'm not sure how ambient light is dealt with nor the spacing between bulbs. In listening to various demos of the Circle Machine, the sounds are always staccato so I suspect there is some type of "gating" function that defines when the output is valid. It is truly a work of art.
Photo courtesy of the Raymond Scott Archives
I installed my Circle Machine, PSIM, and the PSIM Display in my 10U expansion rack cabinet to have a self-contained sequencer. I modified the software for my PSIM which frees up my ComputerVoltageSource module for other applications. This expansion rack cabinet sits on a 24" tall garage over my analog delay reel-reel as my right side wing.
Here is the PSIM, PSIM Display, and Circle Machine in the expansion rack cabinet.
This project is referenced on various web sites and blogs:
- Muffwiggler Forum (my posting)
- Raymond Scott part 1 and part 2
- Steelberry Clones
Circle Machine Construction
I built this Circle Machine using 16 lamps and rheostats, motor, armature-sensor, and slip-disc. The Hammond vibrato scanner has an excellent slip-disc that I wanted to use so I mounted the round base on a 5Ux5U panel along with the lamps, rheostats, and control jacks. The motor and lamps require 1.25A of +12V so I built this in a separate desktop wood enclosure using an external power supply.
I had to make a separate bracket to mount the slip-disc pickup since the vibrato scanner cover would enclose the inner workings of the Circle Machine. I removed three of the vanes and the counterbalance weights to lighten the armature. The lower vane supports the CdS sensor and the upper vane is used for the photo interrupter sensor.
I used my ComputerVoltageSource to generate the clocks for the stepper motor and to generate the output control voltage. Three cables are required between the Circle Machine and the ComputerVoltageSource. The program samples the CdS sensor "on the fly" when it is over the correct lamp and inverts and scales it to generate the control voltage. I quantize the output and have a Tune control which adjusts the sequence +/- 12 semitones.
The Sequence Length control sets the sequence to 4, 8, or 16 steps. I simply skip reading the in-between lamps on the shorter sequences. Since the rate does not change, the tempo drops for the shorter sequences (8 at 1/2 and 4 at 1/4 tempo).
ComputerVoltageSource Circle Machine program
|In-2||Sequence Length (4-16)||Out-2||5 mS Trigger (per lamp)|
|In-3||Tune (+/- 12 semitones)||Out-3||Gate (lamp 1 only)|
|In-4||CdS sensor||Aux||Stepper motor clock|
|Start||Home sensor||Stop||Quantize (on/off)|
This video shows the completed Circle Machine in operation. I use a voltage control from an LFO to drive the sequence length. The changing of the sequence length adds an interesting effect as the shorter sequences occur at 1/2 and 1/4 tempo. The CV output is driving my Living VCO which is tuned to disharmonious notes, and then run through a low pass filter. The envelope for the low pass filter is controlled by the Circle Machine trigger. The LCD displays the step number, sequence length, and note number (if quantized) along with the sequence length CV (displayed at the end of line 1).
This video shows another demo of the Circle Machine. I again used a voltage control from an LFO to drive the sequence length. The CV output is driving my Living VCO which is run through my Resonant Lopass Gate, 3.35" tape delays, and a spring reverb. The envelope for the Lopass Gate is controlled by the Circle Machine trigger.
This was rather a fun project and my first using a stepper motor. The mass of the armature is sufficient to cause a "bounce" at the end of each step. The CdS sensor "go dark" time is rather quick and the "go light" time is rather slow. This causes the CV to drop out between lamps and settle slowly for each lamp. I chose to sample the CdS sensor to hold the output between lamps. This also enabled me to process and quantize the output. Since the CdS sensor has insufficient time to settle at the operational rotational rates, you have to tune the Circle Machine while running. Static tuning at each lamp position produces a slightly different control voltage since the sensor has time to fully settle.
I built this Circle Machine fun since I thought it would be visually interesting. I didn't intend for this to be a precision sequencer nor a faithful recreation of the Raymond Scott Circle Machine. There are slight issues with absolute repeatability due to ambient room lighting and the CdS sensor settling time. Obvious modifications could include using a higher torque stepper motor or a DC motor and changing the sensor to a photo transistor. However I am pleased with the results as it works well to generate interesting sequences and is fun to watch.
I used a Howard 1-19-4203 stepper motor which will do 3.6º steps and 1.8º half-steps. I built a separate stepper motor control circuit so I just have to provide a clock. After this photo was taken I added a bit more circuitry for the home and CdS sensor.
Stepper motor and control schematic
The L297 stepper motor controller limits the current by chopping the motor drive phase signals. This scope image shows the chopping on all four phase drive signals.
I tried stepping rapidly to each lamp and then pausing. This video demonstrates an 8 step sequence by stepping lamp 1 thru 8 and then returning 180º back to the beginning of the sequence. The time required to return the lamp 1 position was significant so I had to adjust the various step delays to keep an even tempo. I eventually changed to skipping lamps for a 4 or 8 note sequence.
I had to machine the center of the Hammond vibrato scanner base plate to mount the stepper motor. I also had to machine a coupler to adapt the stepper motor shaft to the armature shaft. This photo shows the modified base plate and the armature prior to removing the three vanes and counterbalance weights.
I mounted the CdS photoresistor on the bottom vane of the armature. I used #2174 12 volt 40 mA lamps and pressed them into grommets in the base plate. The lamp brightness diminishes to very dim with a series resistance of ~500 ohms. I characterized the lamp power with a series resistor and the maximum power dissipated is 170 mW. This allowed me to use 500 ohm ½ watt linear taper potentiometers as rheostats to control the brightness of each lamp.
This photo shows the painted base plate with the bracket for the slip-disc pickup and 6 lamps installed.
This photo shows the Hammond vibrato scanner slip disc. It is very well designed using three carbon pads - one to press on the top of the pin and two that slip over the pin and are held with spring tension. This slip-disc was designed to pass audio so it passes a clean sensor signal.
This video shows the scanner base temporarily wired up 6 lamps with fixed resistors and a 225º return to the beginning of the sequence. The photoresistor forms the bottom half of a voltage divider that I scale, offset, and run through a lag circuit to generate the control voltage for a VCO. I later eliminated this external processing by sampling the CdS sensor and processing the output with my ComputerVoltageSource.
This video shows the Hammond vibrato scanner base mounted on a 5U x 5U panel with 8 lamps and 16 potentiometers. It is playing an 8 step octave sequence from C3 to C4. There is a lot of equipment running in the background so the audio is not as clear.
The sensor settling time is about 140 mS which is too long for a 16 note sequence less than 2.25 seconds long. I chose to change the operation of the stepper motor to a constant delay per 1.8° half step and sample the CdS sensor "on the fly". This minimized the bounce and provided a fairly repeatable sensor reading during the transition time. This scope image shows the faster "go dark" and the slower "go light" response times as well as the bounce of the armature.
I used the Circle Machine to generate a sequence for a demonstration of the the Polivoks filter grunge factor. Here is a quick video of the Circle Machine in action.