I needed to make an amplitude modulator as part of a signal generator design. I took the approach of using an RCA type CA3004 as a variable transconductance amplifier. It worked very nicely.


Figure 1
The CA3004 of fond memory

The transconductance exhibited by Q1 and Q2 in the above figure varies directly and linearly versus their emitter currents as controlled by Q3. The more current via Q3, the higher the transconductances of Q1 and Q2 and the more signal gain that Q1 and Q2 provide. This overall circuit is colloquially called a "long-tailed pair."

Having gotten my circuit working very nicely, word came down from ‘on high’ that there must not be any sole source parts. The CA3004 was sole sourced from RCA so I had to abandon my successful circuit and come up with something else.

Just to note, the ‘on high’ guys were right. The CA3004 was discontinued by RCA approximately a year later. There had been a whole product line of similar devices which included the CA3028 but all except the CA3028 got the ax at the same time. The CA3028 product lifetime was long, but I had backed the wrong horse.

My first attempt to use multi-source parts was to make another long-tailed pair using discrete transistors of type 2N918. That's when I saw why integrated circuits were invented. The 2N918s didn't parametrically match or track each other so the current sharing between the two of them was extremely unstable versus temperature.

The answer was to make a four-transistor differential amplifier, which might be called a "long-tailed quad."

The following circuit isn't the exact design I made because the 2N918 is not modeled in the MultiSim SPICE version I used and also because I haven't actually looked at that schematic in nearly 50 years, so memory has failed me. However, the operating principles are there.


Figure 2
The long-tailed quad

Here, Q1 and Q2 still operate as a differential pair, but their emitters are capacitively coupled to each other instead of being directly connected. Q1 and Q2 each have their own current sources, Q3 and Q4, rather than sharing a single current source.

Those two current sources are pretty much thermally stable and so if Q1 and Q2 are parametrically mismatched versus temperature, it doesn't really matter. From the above circuit, we see the following SPICE results.


Figure 3
Carrier signals with no modulation applied

The carrier signal from source V1 appears differentially at the collectors of Q1 and Q2. Please note that the collector signals are 180° out of phase with respect to each other.


Figure 4
Modulating signals with no carrier

The modulating signal from source V4 shows up in common mode at the collectors of Q1 and Q2. The collector signals for that are at zero degrees phase angle with respect to each other.


Figure 5
Carrier signals with modulation applied

If we take the collector signal of Q1, subtract the collector signal of Q2, and turn everything on, the result is the amplitude modulated carrier with the modulating signal itself getting cancelled out in that subtraction.

This thing really did the job. The carrier frequency range was 10 kHz to 40 MHz while the modulating frequency range was 10 Hz to 50 kHz. The linear relationship of transconductance versus emitter current meant that the amplitude modulation performance was very good and the managerial requirement to use only multiple sourced components was met.

John Dunn is an electronics consultant, and a graduate of The Polytechnic Institute of Brooklyn (BSEE) and of New York University (MSEE).

Related articles: