VCO's for homebuilt synths

gstopp at gstopp at
Fri Aug 11 00:59:20 CEST 1995

     Hi all,
     Warning - this is a long post!
     I have done some research into VCO's lately to see if I could come up 
     with a VCO design for homebuilt synthesizers with the best tradeoff 
     between two major factors:
     1. Performance - tracking, stability, range, features
     2. "Buildability" - cost, number of components, number of adjustments
        needed, availability of components
     Here are the designs I've tried:
     1. Curtis chip
     2. National App note #299
     3. Electronotes sawtooth oscillator
     4. Electronotes tri-square oscillator
     My requirements for the topics within the two major factors are as follows:
     Tracking - the ideal VCO should have exponential V-F response at 1 
     volt per octave. It should be able to be tuned beat-free to a fixed 
     reference tone, and then tuned in octaves against that reference tone 
     over the entire musical range (say 15hz - 15khz), without any beating 
     at any octave. Range of tolerance for real-world VCOs would be beating 
     to the reference at a period of not more than 4 or 5 seconds.
     Stability - the ideal VCO should not drift in frequency over time, 
     regardless of environmental conditions. For real-world VCOs, the 
     closer to ideal the better.
     Range - the VCO should be able to maintain oscillation down to a 
     frequency between 1/2 and 1/10 hz. The highest frequency should be 
     well above the audio range, at least to 40 khz.
     Features - the VCO should have, as a minimum, the following features:
        * Sawtooth waveform, -5 to +5, rising or falling 
        * Triangle waveform, -5 to +5
        * Pulse waveform, -5 to +5, with adjustable and VC width
        The VCO may have the following as well:
        * Sine wave output, -5 to +5
        * Sawtooth waveform, -5 to +5, the inverse of the one above 
        * Dedicated Square wave output, -5 to +5
        * Sync input, hard or soft or both
        * Linear FM input
     Cost - the VCO should cost between $10 and $25 for all components. 
     This does not include the circuit board, which is up to the builder.
     Number of components - of course the fewer components the better as 
     far as ease of assembly is concerned.
     Number of adjustments needed - the VCO must have a volts per octave 
     trimpot. Additional trimpots are optional as needed. Some designs need 
     a hi-frequency track trimmer for optimum tracking. Trimmers in the 
     waveform converters are always a good thing to provide accurate wave 
     shapes and amplitudes.
     Availability of components - this is a big issue, do be discussed 
     after the VCOs are described below:
     1. Curtis CEM VCO
     Major parts: CEM 3340.
     This design works well using the schematic that comes in the data sheet or 
     in the docs that come with the kit from PAIA. I include this here because 
     I have used it in the past with good results; however, I do not have the 
     bench test specs handy so I will trust that most people know about this 
     one already. I think the main reason to go with a discrete design rather 
     than this chip is to overcome any current or future availability issues. 
     Perhaps somebody can update me here - are these things hard to get these 
     days? Are they still in production? I have a few but I keep them as spares 
     for my OB-8, not as project parts.
     2. National App note #299 VCO
     Major parts: CA3046, LM329, 2N2222A, 2N2907A, LF412.
     This design is basically a single op-amp VCO with additional circuitry 
     to provide an elevated chip temperature for temperature stability. The 
     theory here is that since the exponential converter response is tied 
     to the ambient temperature (unless specifically engineered not to with 
     special components), then if the temperature of the converter is kept 
     high enough such that the environment will never match it, it will not 
     drift since it doesn't change over time. Therefore no tempco device is 
     required, eliminating that particular component procurement headache. 
     There is a temperature stable zener diode in this design as well.
     On paper the design needs some additional circuitry to be brought up 
     to the minimum VCO requirements. There is no true input summing node 
     so this needs to be implemented with a unity-gain summing inverter 
     followed by another inverter to maintain overall polarity (2 op-amps). 
     Also the sawtooth needs to be level-shifted to comply with -5 to +5 
     output waveform standards, and a pulse waveshaper needs to be added. 
     There is no provision for a sync input, but there may be a simple way 
     to add this by finding a point at which an externally applied edge 
     causes lowering of the switching threshold and therefore early 
     sawtooth discharge. There is no linear FM input.
     When I built this design I used all components as specified with the 
     following deviations:
     a. LF412 FET op-amp replaced with TL082 FET op-amp
     b. LM329 6.9v zener replaced with standard 6.8v zener
     c. 4700pf Polystyrene cap replaced with 4700pf polyester cap
     d. Input divider of 10.5k/250 ohms replaced with 10k/1k trim/220 ohms
     The VCO provided a good sawtooth waveform between 40 hz and 7 khz. It 
     did not oscillate outside of this range - the waveform falls apart and 
     the VCO stops. Also I could not get 1 volt per octave response over a 
     range of more than 1.5 octave; my circuit did not have true exponential 
     response. I could not trim it for proper operation. Why did this not 
     function per the App note's description of "exponential conformity 
     within 0.25% from 20 hz to 15 khz"? Let me go over the differences 
     between my circuit and the App. note:
     a. Op-amp - this could well be the cause. I could not find the LF412 at
        my local NTE-carrying house and the databooks suggested that the 
        TL082 was pretty close. I will try to get a sample of the LF412.
     b. LM329 - this part is like a normal zener except that it has precise
        temperature stability. It is used in this circuit as a reference 
        voltage for the various functions. The use of a regular zener of 
        similar voltage should only affect long-term stability, not basic 
     c. 4700pf cap - this change also should not affect short-term
     d. Input divider change - this change should not affect operation.
     Conclusions: I have a bad feeling about this circuit. It appears at 
     first to be a special-part-free design; however if it does rely on the 
     specific characteristics of the LF412 then is becomes just as dependent 
     on a specific component as other designs. The extended low and high 
     operating range could not be determined. The parts count for a quality 
     design seems to be about the same as many other designs.
     3. Electronotes sawtooth oscillator
     Major parts: CA3046, KE4359 JFET, CA3130, LM311, 2K tempco resistor.
     This circuit uses standard op-amps with a CA3046 transistor array to 
     provide exponential current into an integrator based on a CA3130 
     CMOS-input op-amp with an LM311 comparator resetting the sawtooth with 
     a JFET across the integrator's cap. There is a 2K +3000ppm/C' tempco 
     resistor and a hi-freq track trimmer in the exponential converter.
     When I built this design I used all components as specified with the 
     following deviations:
     a. CA3130 replaced with TL082
     b. KE4859 JFET replaced with 2N4856
     c. 2K tempco resistor replaced with 2K 5% metal film.
     This guy oscillated but would not go over a couple hundred hz. Also 
     the sawtooth waveform was rounded off at the top. Why? Again let me go 
     over the deviations from the original schematic:
     a. TL082 instead of CA3130 - I'll bet this is the culprit. Local store
        had a hook for CA3130, out of stock, no NTE cross. The CA3130 
        op-amp has really, really high input impedance so given my 
        waveshape and frequency problems this sounds like it.
     b. 2N4856 instead of KE4859 - probably not the problem, I looked up
        the 2N4859 and it has pretty normal N-channel JFET characteristics 
        except for a lower on-resistance than all others plus a higher 
        current rating - both what you would expect for a precision 
        oscillator cap discharge switch. The 2N4856 had the same numbers 
        but a higher breakdown voltage. I may be overlooking something...
     c. Tempco resistor replaced with standard - this would only affect
        long-term stability.
     Conclusions: since this design comes from Electronotes it probably has 
     real good performance, assuming the proper parts are used. I will 
     attempt to get the proper parts. The CA3046 transistor array, the 
     LM311, and the various op-amps are all easily-obtained parts. There is 
     a soft sync input, which looks like it could easily be modified for 
     hard sync. The tempco resistor can be hard to find - more on this 
     4. Electronotes tri-square oscillator.
     Major parts: AD821, CA3080, CA3140, 2K tempco resistor.
     This circuit uses standard op-amps with a matched transistor pair and 
     a 2K +3000ppm/C' tempco resistor in an exponential converter which 
     drives a CA3080 integrator into a TL082 buffer, then into a CA3080 
     schmidt trigger, forming a triangle-square oscillator.
     When I built this design I used all components as specified with the 
     following deviations:
     a. AD821 matched PNP pair replaced with MAT-03 
     b. CA3140 op-amp replaced with TL082
     c. 2K tempco resistor replaced with standard 2K 5% metal film.
     This oscillator came right up. The low frequency goes well below 1 hz, 
     with some assymetry on the waveforms at very low frequencies (up slope 
     <> down slope). High frequency is around 60 Khz with perfect waveshape 
     and amplitude throughout the sweep range. Tracking looked good to the 
     best of my ability to measure it with a DVM and a scope, better tests 
     (by ear) to follow. Short-term and long-term drift and stability have 
     not been measured, pending the arrival of tempco resistors. 
     Possibilities for sync need to be investigated since this is not a 
     traditional sawtooth-discharge design.
     Conclusions: this VCO appears to be quite intolerant to component 
     substitution. Even the PNP pair can be substituted with a couple of 
     2N3906's which can be matched by hand, although this needs to be tried 
     to verify tracking and stability. This would be a good thing 
     considering the cost of the MAT-03.
     ********************* PARTS ROUND-UP *********************************
     CEM 3340:
     Help me out here guys, are these still avaialable? If so then I say that 
     this is a very good option for a homebuilt VCO.
     It's in the catalogs so I would assume no special availability problems.
     Not easy to find on my first try, will dig deeper.
     2N2222A, 2N2907A:
     No problem at all - standard parts.
     My local NTE store had a cross-over, not sure if it works good or not.
     3 different big distributors said "discontinued", not crossed by NTE.
     Easy to find.
     In the local store but not in stock. NTE cross-over is listed.
     In the local store and in stock.
     Easy to find.
     Easy to find.
     Special order from Analog Devices - $8.50 a pop!
     Tempco resistor:
     Found a vendor, getting 20 samples, no pricing yet, 10-12 week lead time.

More information about the Synth-diy mailing list