T901 Voltage Controlled Oscillator Bank Clone

Introduction

Well - this was quite a hard piece of work. Although a lot of people had successfully cloned the 901 oascillator modules before me I was quite nervous when I started the project. Two strange components (2N2646 and CA3019) and one of the oldest Moog circuit designs (beginning of the 60ies) weakend my belief in a successful project - but finally I got a working oscillator bank - almost.

And - it is not an exact clone. Compared to the original my T901 has some modifcations / improvements on the one hand, and on the other hand some module functions don't work as exact as in the original (see below).

T901 Voltage Controlled Oscillator Bank Clone

Differences to the original

First of all: It is one module with one front panel, not one oscillator controller module and three oscillator modules. In a standard System 55 the oscillators appear as banks in most cases - one controller and a set of oscillators - or as composite module (one controller and one oscillator), but in both cases the oscillator modules don't come alone. So, as I decided to build a "standard" bank I chose the 1 + 3 solution, but with only one front panel as I certainly will never have the need for dividing the modules apart.

Next issue is the pulse width topic. I never understood why Moog put the PW control on the controller module, because pulse width is a feature of pulse generation, so it is an attribute of the oscillator, not of the oscillator controller. With my solution (and the solution of all oscillator manufacturers I know) you can set up the pulse width individually for each oscillator and achieve a higher flexibility and deepness in sound if more than one oscillator work together. So - the pulse width knob migrated from the oscillator controller to the oscillator(s).

Additionally I decided to add pulse width modulation. This I consider as very important. The smooth phasing sound of slowly changing pulse witdh controlled by an LFO is magic - especially if you use special LFO waveforms (listen to sound example below).

Of course these changes concerning pulse width (modulation) resulted in front panel modifications also (see description of human interface below).

Complete Module

Click to enlarge

Frontend / Human Interface

Controller

On top you find a Fixed Control Voltage rotary switch - one position change changes the control voltage base for one volt. In the original module this should increase or decrease the output frequency about one octave, but in my clone the output is changed about a third only. But a change in an incoming real control voltage (from front panel CV input (see below) or from internal node connector) of one volt leads to a change of one octave indeed, so my clone operates at 1V/oct in general, but not by switching the front panel selector.
First I tried to solve this mystery, but then I saw the advantage in this behaviour: a possibility of octave part selection in module tuning. So I decided to keep it this way.

The second knob does a fine tuning of +/- one third of an octave. Same behavour as above, a third instead of a complete octave.

And finally the three control inputs. 1V/oct characteristic. These three inputs and the internal node connector are mixed and determine the note / frequency beeing played / created.

Oscillator

On top you find another octave selector. Again, a position change should result in an octave change, but my clone changes the output on certain positions more than one octave, and on other positions less than one octave, depending on the integration capacitor selected by the rotary switch (see circuit description below). To avoid this I would have to do experiments with different capacitor values, but again: why should I do this? Quick changing of octaves in live performances will never be my problem, so I use it as it is, no more effort, in my studio I have enough time to calibrate my modular exactly.

The Frequency Vernier knob changes the ouput octave about +/- 1 oct. Bad side effect: It changes the tracking also! The reason is obvious if you study the original circuit diagram (see circuit description below). I've not understood the concept behind this yet, perhaps someday someone explains it to me.

The Pulse Width pot migrated from the controller module to the oscillator(s), as mentioned above. Little icons on the front panel show the effect of each position to the pulse width.

The Pulse Witdh Modulation input (PWM) sets the pulse width depending on an input voltage. This is something the original missed, so I decided to add it as I consider this as important (see above).

Finally the waveform ouputs. No waveform selector, as in other oscillators. I decided to keep it that way, as I'd run out of front panel space if I'd add an additional knob and an additional output.

Circuit

Schemas: Attention: Clicking a schema means acceptance of disclaimer on page bottom! These schemas are screenshots of my LTSpice simulation file. The following explanations are not complete. On the one hand I didn't want to explain every resistor, and on the other hand I didn't understand everything... So only the key elements are explained.

Please refer to the original schematics for more details / component values.

Controller

The controller part consists of two subunits - a control voltage adder and an exponentiator. The control voltage adder is a "standard" pre-opamp Moog unit which one can find in several modules, e. g. the 904 filter modules.
All incoming CVs, the internal node, the tuning pot output and the output of the octave selection are mixed and feed the differential amplifier tranisitor pair Q1 and Q2. This differential amplifier contains a low pass filter (R9 and C2) to clean up the CV.
Another differential pair (Q3 and Q4) forms the next amplifier stage, and the collector follower Q5 forms the output stage of the mixer. The overall amplification is determined by P1(Scale) which is placed in the feedback branch of the circuit. C1 suppresses self oscillation, I assume, but I am not sure. The High Compensation stage (R15, R16, P2, R17) non linearizes the scale behaviour at the upper end, I assume. In my clone this had no effect at all.
The CV mixer section forms a current sink for the second part of the controller, the exponentiator. The resulting voltage drop is amplified by another two differential amplifier stages (Q6, Q7, Q8, Q9). The bases of the first differential amplifier (Q6, Q7) pass two diode arrays (CA3019). These diode arrays change their inner resistance depending on their temperature. After a heating time of about 10 minutes the CA3019 diode arrays have reached working temperature, which they keep independantly from the temperature of the module surrounding. The oscillator controller works stable then.
At the base of Q10 (which is the output and feedback buffer of the exponentiator) one can measure an expontial falling voltage by changing the input CVs, meaning one more volt at CV input means dividing the output voltage of the exponentiator by 2. And this voltage controles the oscillator (see below, point "A").

Oscillator

The oscillator is controlled by the falling voltage at point "A", the base of Q7 (107 in my spice simulation file), which is a PNP transistor that forms a current source for the saw core consisting of a 2n2646 unijunction transistor and a capacity which is selectable by the front panel octave switch. The capacity (C2 to C7) are loaded up to a level the unijunction transistor "fires" and discharges the capacity immediately. This goes so fast that you can't see the reset time on an oscilloscope. The resulting output voltage is leveled and linearized by Q8 and Q9 and forms a saw.
Now back to the Frequency Vernier problem. As mentioned above it also changes tracking! The reason for this behaviour is the fact that the two halfs of the stereo pot of the Vernier control do different things. One regulates / attenuates the loading current of the saw core. That is ok. The other half works as voltage divider and adds a voltage (!) to the tracking node. The reason is not clear to me, but the effect is that from a certain position of the Vernier knob the tracking changes and lowers the octave spreading.
Waveshaper: The square wave shaper is connected to this (raw) saw via R29. When the base of Q3 reaches +.7V Q3 opens and the collector falls down, when the saw starts again the collector goes up und so on. This resulting square wave is inverted and amplified by Q2 and output buffered and leveled by Q1 and R34 - R36.
The base of Q3 is also input for the pulse width modulation input voltage. Any voltage fed in here manipulates the time Q3 opens and initiates the square cycle. For PWM I added a Moog style opamp free CV mixer to the oscillator circuit. I copied it from the 904A low pass filter circuit. The PW pot voltage and the PWM input voltage are mixed, low pass filtered and feed a scalable two stage differential amplifier and a buffer transistor

LTSpice simulation plot of PWM

The triangle wave shaper is connected to the (raw) saw signal via R17. As long as the voltage of the rising ramp of the saw at the base of Q6 is lower than the collector of Q6, the collector is an inverted representative of the base. But when the base voltgae climbs higher, the base collector relationship works as a diode in forward direction (avalanche breakdown). The voltage marches in opposite direction, this time following the base, not inverting it. The result is - of course - a triangle signal.
This signal is buffered and leveled by Q5, R23 and R24.
The sine waveshaper followes the triangle signal via R25. The architecture can be found in many oscillators types. It works by shaping the sharp edges of the triangle by the (smooth) transition of diodes from conduct to non conduct and vice versa. This is done by D3 to D6. The resulting sine is buffered and leveled by Q4 and R26 to R28.

Obligatory waveform pictures

Saw (Click to enlarge)

Triangle (Click to enlarge)

Sine (Click to enlarge)

Pulse (Click to enlarge)

Two oscillators show the slightly different waveforms below. The reason is a voltage drop caused by the waveshapers following the saw. The effect does not affect the pulse output though. Anyway, I decided to keep it that way, because the little "platform" in the wave cycle adds a smooth little taste of bitter sweet chocolate to the waveform.

Click to enlarge

Part replacements

I tried to build my module as close to the original as possible, but I had to do some changes due to actual components I use instead of the originals:

All NPN transistors have been replaced by the BC547C type in CV paths and the low noise type BC550C in audio paths.
All PNP transistors have been replaced by the BC557C type in CV paths and the low noise type BC560C in audio paths.
All onboard potentiometer have been replaced by multiturn versions.

Circuit changes

Beside the functional changes described above I changed the following component values:

R7 of controller circuit was changed from 6k8 to 1k. This was necessary because the feedback of the CV mixer section had to be reduced due to higher transistor betas of the BC5xx types.
R14 of controller circuit was changed from 220R to 270R. Same reason as above, higher beta.
R30 of controller circuit was changed from 100k to 47k. Same reason as above, higher beta.
R28 of controller circuit was changed from 6k8 to 2k2. Same reason as above, higher beta.
The tracking resistor 270k of the oscillator circuit (doesn't have a number in the original schema) was replaced by 10k. This was necessary to increase the tracking range. ==> General recommendation: Don't use standart onboard trimmer, use multiturn versions and decrease the connection resistors if exist. This leads to much higher tuning ranges of the specific functions.
R29 of oscillator circuit was changed from 13k to 24k. This was necessary to reduce the voltage drop (saw "platform") caused by the pulse waveshaper.

Boards of the T901 clone

Board dimensions controller: 95 x 95 mm
Upper left: CV mixer.
Upper right: Exponentiator.
Lower left: Transistor matching circuit, power supply.
In contrast to the original concept of the Moog modulars I supplied the board with a +/-15V power supply. Voltage regulators on the board convert it to +12V, -6V and -10V. Benefit of this is a higher stability of the power supply.
Lower right: CV input and fixed CV selection.

Click to enlarge

Board dimensions oscillator 120 x 95 mm
Upper left: Tracking and Frequency Vernier control.
Upper right: Waveshaper and Pulse Width Modulation
Lower side: Power supply and saw core.
In contrast to the original concept of the Moog modulars I supplied the board with a +/-15V power supply. Voltage regulators on the board convert it to +12V, -6V and -10V. Benefit of this is a higher stability of the power supply.

Click to enlarge

Sound examples

Saw Pure saw waveform over 5 octaves. No other module / FX / effect involved.
Triangle Pure triangle waveform over 5 octaves. No other module / FX / effect involved.
Pulse Pure symmetric pulse waveform over 5 octaves. No other module / FX / effect involved.
Sine Pure sine waveform over 5 octaves. No other module / FX / effect involved.
PWM Pure single pulse waveform with pulse width modulation by a sine LFO. No FX / effect involved.
VCO Bank All three oscillators work in a chord. One single pulse waveform with pulse width modulation by a sine LFO and two triangles detuned. Mixed by a TCP3A module. No FX / effect involved.

New: Listen to T901 on Sushonyo Synth Project!

Conclusion

Yes, I just LOVE Moog circuits. These are from the 1960's. Beside the CA3019 diode array no integrated circuits are involved. The temperature stability of the controller is as good as from later (IC based) VCOs, perhaps even better. So what do you really need for a operable VCO? Just a few diodes, transistors, resistors and capacitors. Parts that will be available forever (the CA3019 can be replaced by any diode array, I am sure). That's why this cloning project is a fun project, although some problems have occured in every module so far which I had to solve first, but that keeps it interesting. So do you need opamps or OTAs for waveshaping? No, you don't. Do you need an opamp based comparator for a saw core reset? No, you don't. Just use a unijunction transistor which does the whole job. That is great, that is fantastic. Back to the roots I say. This is what I learned from this clone.

Please send questions or remarks to:
Carsten Toensmann

Home

Disclaimer:
No warranty for completeness or correctness of technical descriptions, schemas or any information provided on www.analog-monster.de/* pages.
Contents of linked pages are in responsibility of the page owners and analog-monster dissociates from them.
All tracks provided on www.analog-monster.de/* pages is in ownership of www.analog.monster.de (except especially marked) and for private use only. All rights reserved.