How the ARP 1004 VCO Works

Overview The reason I used the 2600 VCO in my 1004 modules in 2017 was I knew the original 1004 circuit was very difficult to get working. It has a lot of selected resistors (marked SEL) with no values stated, and the calibration section in the ARP service manual is rather limited on actual component setup.

The SEL resistors are needed because most of the resistors are 10% carbon composite, so the values need tidying up. I have used 1% metal film resistors to overcome some of the need for resistor selection. The  AM1045 uses the 1004 VCO, so I had to understand the circuit and make it work! Battle commenced in February 2019, with a AM1004 VCO PCB populated and I used photos of actual ARP 1004 and 1023 module PCB’s to decode SEL resistor values.

Triangle Wave

How the VCO Works The VCO has both a saw and a triangle core, with the saw slaved from the triangle core. The cores use two linked expo generators, the 4001 as a current source and the 4002 as a current sink, they work together and are driven by a CV summer Op Amp (LM301), which is inverted for the 4002 by another Op Amp(LM301). I used the same LM301’s to get the circuit working ,then swapped them out for precision LT1013’s.

The circuit around Q3 and Q4 is an integrator to reset the waveform after the Mica capacitor C11 has discharged and its essential to oscillation. The base of Q4 gets hit with a positive going pulse from CR5/R50 that makes it conduct and reset the capacitor, and when everything goes to zero the transistors go back into cutoff. The 2K2 resistor in this circuit may slow the reset which  might cause the oscillator to drift flat at high frequencies. I will investigate.

The triangle core is reasonably easy to get to work first time with no adjustments, there is a SEL resistor R86 in parallel to the 121K at R85 within the integrator. I found the VCO oscillated with no R86 fitted, and R85 at 120k. Looking at photos R86 is a large value of 680k, I left R85 alone and did not add R86 as the VCO oscillates! To ensure frequency stability I mounted the 100R Panasonic SMD tempcos (which are sadly obsolete) in the 4001 (current source) and 4002  (current sink) sub modules. I potted the sub modules at a later stage, as the oscillator changes frequency if you blow on the open sub modules.

Feed Forward The main Triangle and Sawtooth Op Amps were originally 1339, but replaced with LM301’s. The design uses feed forward compensation capacitors to increase the speed of the Op Amps and thereby improve the waveform shapes at higher frequencies. I have retained this design in the THD version used in the 1045 but removed it from the SMD version which uses 4580 Op Amps.

Square Wave

Square & Pulse The tri-core also produces a nice square wave which can be adjusted for symmetry by the trimmer R93. The pulse wave is also easy to get working and R54 and R55 can be selected but I simply varied the values of R79 and R80 to get the correct 5% pulse width at each end of the PW control. I made a mistake on my prototype and put R76 to +15V, this should be PW modulation CV input, so it simply pushed the pulse width to zero/invisible.

This dumb mistake was easy to correct! There are also two more SEL resistors at R91 and R81 to adjust the waveform height of the Square and Pulse waveforms to exactly 10V. With metal film resistors they are not needed.

Sine Waves The sine wave is generated by dual matched transistor pairs working off the triangle wave. There is one SEL resistor at R66 for adjusting the gain of the LM301 Op Amp and therefore the height of the waveform. I put a THAT300 chip in for the matched transistors but the circuit did not work, no output. With some examination there was a 500 mV sine wave at the input to the LM301 but no output.

Back in 1970 ARP used 1339 Op Amps that needed a 10k resistor to pin 5. When they updated the schematics of the 1045 to use the LM301 they forgot to delete this 10K resistor, which actually kills the sine output when used with a LM301. The PCB layout shows this resistor missing, so I removed it and the sine wave sparked into life! A TL071 Op Amp is used in production models and a 91k resistor at R65 makes 9.8V output level, so there is no need for a SEL resistor. I have used the Rpar AS194 chip in production PCB’s rather than the THAT300, same performance but nice TO5 cans. However the shape of the sine wave is disappointing, it is more like a curved square and there are circuits closer to a pure sine wave in the 2600 FET or Emu Systems OTA designs just a few years later.

Sawtooth Wave

Sawtooth Problems The sawtooth wave proved much harder to get working. There are two SEL resistors attached to the R34 Tempco, R35 and R36, which I  knew from the 1023 schematics adjust “sawtooth amplitude”. I initially did not fit them, no sawtooth just a reset pulse. I then put a 10k trimmer in, and managed to get a saw waveform by trimming, but is was unstable and varied in height with frequency.

After may hours I worked out that I had put a PNP transistor into the 4001 when a NPN was needed. A quick change in transistor and this corrected the varying amplitude. Wow, progress! By measuring the trimmer resistance I could determine the resistor values. I also looked at all the 1004/1023 module PCB photos on the Internet to see what ARP had fitted and the values. R36 is 470k and R35 is empty. I also discovered that the 680pF capacitor was a Mica type in a recent 1023 replica, so I used a polystyrene type.

With the sawtooth working, we can turn to the adjustment trimmer at R41 which varies the amplitude of the sawtooth wave, so there is no need for a trimmer at R35, we just need a fixed value resistor. The sawtooth works well, although there is a spike at the bottom of the reset (the images are at around 100Hz).

Expo Generators There are two SEL resistors R37 and R38 attached to the 4001 and 4002 to link the reference trim pins to the voltage rails. I initially set these to 100k as there is a 1M5 resistor in series in each module. I guess they used these resistors to compensate for the inaccuracy of the 1M5 resistor, because like all the resistors in the 4001 and 4002 they are wide tolerance 10% carbon composites.

The new AM sub modules use 1% metal film resistors, so this problem goes away. In fact looking at photos, ARP simply fitted a wire link at R37 and R38, so it can’t have been too critical! I have also gone with a wire link in the prototype (100k worked fine as well), and production PCB’s don’t have these resistors at all.

High Frequency Trim The VCO has a high frequency trimmer at R59, which was a bit of a surprise as its not marked as such on the schematics and I thought the 1004 had high frequency tracking issues. Luckily the trimmer at R59 is described in the calibration instructions in the schematics, so we can now trim the VCO at higher frequencies like 4kHz to 8kHz.  I used a multi turn cermet trimmer. There is also a capacitor SEL at C5 which I need to investigate when I setup the VCO, I have not fitted it.

Calibration We are not finished yet! There are two SEL resistors at R11 and R13 for adjusting the range of the Coarse and Fine frequency pots. Whilst metal film resistors are used, this enables fine tuning of the range of these pots. Rather than adding more components I have selected exactly the right resistor values. Back in 1970 1% precision resistors were expensive and 10%  carbon composite cheap, this has reversed 50 years later. There is also the usual trimmers in the LM301 based CV summer to set up the frequency range and an accurate 1V per octave tracking. Frequency trim at R1 and V/OCT trim at R6. I have used multi turn cermet trimmers, a nice upgrade from the original open frame single turn versions.

Low Frequency Mode The 1004 VCO has a low frequency mode which is 547x slower than the normal frequency, and this ratio is determined by the capacitors in the cores.  The Triangle and Saw cores are also at a fixed ratio of 3.45:1 in both high and low modes. The main core capacitors are 1% silver mica (330pF and 1300pF) when in normal frequency mode. In low frequency mode polycarbonate capacitors are switched into the circuit in parallel to the existing Mica capacitors used; triangle core has 560nF+150nF and the saw core has 150nF+56nF in parallel. There is also a small value ceramic disk capacitor in the sawtooth core that has to be selected to give the correct 546:1 ratio, it is 47pF. However its worth measuring the capacitors before they are fitted and adjusting this capacitor to get the exact ratio. I have used polypropylene 1% capacitors, as polycarbonate are pretty much obsolete.

Frequency Stability The initial tests against my reference Emu Systems VCO were positive with little drift, especially when the 1004 VCO is open and cold with no sub module potting. I will do some careful tests once the sub modules have been potted. The precision 10V rails were defeated in the prototype and 15V used to ensure the VCO worked. The next step is to revert to the 10V rails for the frequency pots/trimmer and possibly the 4001 and 4002 reference voltages, along with using LT1013 Op Amps in the CV summers and JFET Op Amps replacing the LM301’s.

Conclusion Overall I am impressed by the engineering quality of this VCO that was created in 1970 with basic electronics, standard signal transistors and the new but slow LM301 Op Amp. Whilst not up there with the revolutionary Dave Rossum 2200 VCO from a few years later, the 1004 had to work without fast Op Amps like the LM318. From a personal perspective at last I have the 1004 VCO working and calibrated! With 10 hand selected resistors and 1 capacitor, plus 5 trimmers, double that and more for the 1023 dual VCO, it is no wonder this was an expensive VCO to build. It retailed at US$490 in 1970 which is over £2000 today. The use of metal film resistors and trimmers has reduced the need for any hand selection of components and modern parts have brought the price down to more like £200.

Build Outcomes The proven ARP 1004 VCO goes straight into the three AM1045 modules I built and completed in February – March 2020, Check out the 1045 page for more info – here.

Further Notes Using MAT12 matched transistors in the 4001 worked in the 1006 VCF but they did not work in the VCO, for unknown reasons with a very high oscillation observed on sawtooth output. A SMD version of the VCO works very well using PMP4201 & 5201 matched transistors.

Copyright AMSynths 2024