Precision Voltage Switch and Integrator

In the development of the Model III EMS, I like to develop individual circuit designs, to prove them out before connecting them to a larger circuit design or sub-system.

In this case, I'm working on making a new VCO with the LM324 op amp. This device is really useful for a lot of reasons. One is that in today's world of supply-chain shortages, the LM324 is one of the most available ICs there is. It's not a high-performance op amp, but it is good enough, when used within its limits.

In particular, the VCO I'm interested in has a top frequency of about 8 kHz. Because a higher frequency than that is not musically usable. Based on this limitation, and in keeping the integrator waveform excursion well under ΔV < 5V, the LM324 can be part of a Triangle Wave Generator (TWG) with very high pitch accuracy up to 8 kHz. This is because the LM324 as a 1 MHz op amp with a 0.5V/μS Slew Rate has very little integrator rate error voltage with a signal swing of ΔV = 5/3 volts, the voltage comparison range of the trigger circuit in a LMC555N timer operated with a total of 5V across Vcc and GND. 

The circuit below is a precision voltage switch tied to an integrator. It's not the TWG generator loop, just the core of it, so that it can be tested separately.

To make a voltage-controlled TWG, some form of balanced but opposing currents or voltages must be created, so as to get the integrator to oscillate between two preset voltage limits. There are a myriad of ways to do this, but I found one that just involves a DC op amp (a LM324!) and an ultrafast switching transistor, a 2N7000.

Precision Voltage Switch and Integrator

This circuit was developed for an earlier VCO effort, in which the a balanced set of control voltages was created from an input (negative) control voltage -Vc. This earlier circuit was done the way it was because when using a fast VMOS transistor to switch voltages, VHF frequency charge injection was going into the op amp integrator terminals. It produced integration onset waveforms that had ripples and switching transients for some 10s of nanoseconds. Obviously, it would be unlikely to hear such a thing, but I still didn't like it. The new circuit came about as a means of isolating the junction capacitance and charge transfer of a device like the 2N7000. 

The 2N7000 drives a current bridge that is comprised with resistors. At all times in the circuit, a -2X precision op amp inverter outputs the input -Vc voltage as +2Vc. When the switch is "off" (e.g. Vg = 0V), the R/2R resistor network results at a current +I that is proportional to the |Vc| into the virtual ground, or summing junction of the integrator op amp. This happens because the resistor network opposes -Vc with +2Vc, across equal values of resistors tied to the virtual ground: 2R and R + R.

When the switch is "on" (Vg > Vgs), the R + R junction is shorted to ground. Because the op amp is also at (virtual) ground, no current flows across the resistor between the op amp and the 2N7000 switch. As a result, all of the current comes from the -Vc input, which is -I.

So the switch selects either +I or -I into the integrator, allowing the generation of alternating voltage sweeps at rates set precisely by -Vc. Notice that the -2X precision inverter does not switch! Only the current from it's resistively-coupled output is switched, via the 2N7000. The 2N7000 device switching these very modest DC currents is extraordinarly fast for audio: Tpd ≤ 75 ns. Also, the R + R network isolates all of the 2N7000 switching capacitance away from the op amps, there is no deleterious impact on the integrator or the DC amplifier. Extremely clean current switching is all that happens.

By use of 0.1% R and 2R resistors (10.0K and 20.0K, respectively), the circuit can generate precision equal time sweep voltages from one control voltage input. No trimming is required either. The resistor bridge is actually much simpler than a diode bridge driven by complementary current sources, which would require BJT or other current mirrors. Even a very low cost op amp like the LM324 is capable of precision gain.

The fact that a 2X voltage is produced does limit the full range of control voltage input. For example, an input range of -5.000V ≤ Vc ≤ 0.000V would require that the DC op amp produce a maximum of +10.000V. This could be scaled up somewhat, perhaps to -Vc = -6.000V. But in principle, this is not a problem for a VCO, because the circuit preceeding this is likely to be an antilog generator, and the scaling can be whatever required. Also, that preceding stage provides a low-impedance source input for the resistor bridge. All part of the plan for a VCO!

Some examplary scope fotos illustrating how this precision voltage switch circuit drives an op amp integrator now follow. The control voltage -Vc was provided by one channel of a SDG 1032X signal generator, and the gate voltage Vg > Vgs was provided as a 0-4V square wave by the second channel of the signal generator.

Basic I/O, with a 50 Hz gate signal (Vg) and Control Voltage (Vc) = -1.000V.

Zooming in to the sweep down, with -Vc = -1.000V.

The down sweep changes ΔT precisely in half, with -Vc = -2.000V.

-Vc = -3.000V.

-Vc = -4.000V.

-Vc = -5.000V.



Comments