Between Discrete & Integrated Circuits

Years ago, Tektronix could not justify an expensive, high-performance semiconductor operation solely for producing competition-defeating ICs for oscilloscopes, and formed a joint venture with Maxim called (appropriately) Maxtek. Maxim then participated in the benefits of access to the “super high frequency” processes that Tek had for making BJTs (bipolar junction transistors). Back then, another jointly-shared technology was more prominent than now, that of hybrid circuits .

In previous decades, the desire to increase integration of scopes at Tektronix led at times to hybrid circuits. These circuits are made by screen-printing and laser trimming thick-film resistors or etching small capacitors on a ceramic substrate, and then surface-mounting components on the substrate, sometimes on both sides. Some are shown below.

Pins are placed on a substrate edge, and the resulting subsystem is used as an integrated circuit. Sometimes hybrids were not much larger than through-hole ICs of the time. The high-frequency performance of ceramic is superior to fiberglass etched circuit-board (ECB) material — having less dC/df , called “hook.” All resistors are fabricated in the same process step, and hybrids in this sense are closer to monolithic construction than board-based circuits.

However, surface-mount technology (SMT) has diminished the benefits of hybrids, though it should not be a forgotten option. The BJTs and chip capacitors on the above hybrids are SMT parts. The only significant difference from SMT ECBs is that the resistors and small capacitors are an integral part of the assembly. Except for some specialized applications, hybrid circuits are at a cost disadvantage compared with SMT ECB technology.

As for ECB technology itself, it need not be very discrete. The use of ICs such as op-amps and ADCs on boards as semi-discrete technology continues to be the mainstream level of integration for non-size-critical applications. Transistor arrays for semi-discrete design offer monolithic circuit properties. Intersil has come out with a second generation of CA3000-series BJT arrays, except the PNPs are also fast. The HFA30xx and HFA31xx arrays have an fT spec of multi-GHz. Additionally, Mosfets are available in arrays beginning with the (slow) CMOS logic family part, the 4007UB: three Mosfets of each polarity in configurations suitable for implementing bipolar current mirrors.

Arrays have some of the advantages of monolithic circuitry in a discrete context. For those engineers who like to optimize their designs, finding the right op-amp is not the solution to every circuit design problem. Sometimes, a superior circuit can result from the careful and clever application of a few transistors, especially when matched and thermally connected in arrays.

The old RCA CA3000 series of BJT arrays included the CA3096, a 3-NPN, 2-PNP array with matched NPNs and matched, lateral (low-β) PNPs. The PNPs are slow; fT = 6 MHz, and the NPNs are like 2N3904s, with about 300 MHz fT . These parts have been discontinued by Intersil because over the years it's been selling off a large inventory inherited from RCA. The CA3046 5 NPN array was used extensively in the Tek 2205 20MHz oscilloscope. Although analog scopes are fading, the entire transistor circuitry of the 2205 vertical amplifier looks as if it could be integrated into one present-day IC. Indeed, with a few capacitors hanging on external pins, a many-pin package might contain the entire amplifier and time-base part of this oscilloscope. A bird's-eye view of the Tek circuit diagram of the vertical amplifier output stages is given below.

For a better view of this schematic, click here. (Source: Tektronix)

For a better view of this schematic, click here.
(Source: Tektronix)

U130 (upper middle) is a CA3102, two diff-pairs with BJT emitter current sources, for constructing transconductance multipliers. (The Intersil HFA3101 is similar, with internally cross-connected collectors.) The CA3102 BJTs have fT values of 1.35GHz. This kind of variable-gain amplifier depends on matched p-n junctions, and the array is a semi-discrete design solution that otherwise would require an ASIC or manually matched BJTs — an unattractive alternative.

Transistor arrays are a possible design familiarization path to full integration. In the 1980s, there was a brief flurry of arrays available for discrete prototyping of what were intended to become semi-custom integrated circuits. Such arrays are useful building blocks, and they prepare the engineering mind for the notion of putting more and more on the array, eventually resulting in a monolithic circuit.

What is least integrated about the amplifier are the switches. (The instrument settings or configuration problem is addressed in the instrument-on-a-chip series; see below.) Configuration switching requires additional creative design to integrate, but with high-performance analog switches, both of the slow, high-current relay type and fast, low-voltage type, they reduce or eliminate the need for electromechanical relays.

What were your experiences with hybrid devices — both designing them and using them?

Related posts:

11 comments on “Between Discrete & Integrated Circuits

  1. nathandavidson
    September 17, 2018

    I am actually going to ask a very silly question – but this is because as much as I know about analog systems, I'm constantly fumbling with my chips and figuring out which way is up. I mean clearly there is a right way and a wrong way of going things, but can somebody, for the love of God, try to invent something that's usable on both ends without having to worry about which way it to proper orientation?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.