Dual-emitter transistors

Pieter-Tjerk de Boer, PA3FWM web@pa3fwm.nl

(This is an adapted version of part of an article I wrote for the Dutch amateur radio magazine Electron, January 2024.)

A previous installment of this series was about unusual schematic symbols for transistors. While searching these symbols in old databooks, I also stumbled on something else: the RCA 3746 transistor, which does not have just one, but two emitters. Of course, I wanted to know more about that, and mostly thanks to archive.org and worldradiohistory.com I gathered some more information about that RCA transistor, about the behaviour of transistors when connected the wrong way around, and about so-called 'chopper transistors'.

The RCA 3746 transistor

The figure shows a schematic incorporating the '3746' transistor by RCA. This transistor has two emitters and is used here in grounded-base configuration, so it has two inputs. One emitter is used as the input for an HF signal coming from T1 (a ferrite antenna). The other input is connected via T2 to the transistor's collector, making it an oscillator. Together this is a self-oscillating mixer.

In those days (1960) transistors were expensive and it was common practice to use a single transistor simultaneously for the oscillator and mixer functions in a radio receiver. That could also be done with 'normal' transistors having only one emitter. But a difficulty is that the oscillator may stop working in the presence of strong input signals, if the AGC (automatic gain control) reduces the gain too much. Using the double-emitter transistor solves that problem, by having separate inputs for both signals, as explained by the (few) publications about this transistor (such as [1], [2], [3]).

Soon out of production

RCA announced their double-emitter transistor at a conference of the Institute of Radio Engineers (IRE, predecessor of the IEEE) in March 1960 [1], but without a type number. The earliest report of the typenumber that I could find was in a German magazine (which is odd as RCA is an American company), namely Funkschau of December 1960 [2]. An extensive datasheet of no less than 8 pages, dated February 1961, is in [3], from which I also copied the above figure. That figure is the same as what was published in Funkschau and some other places, apart from some factors of 10 in the capacitances: apparently someone had trouble converting between pF, nF and µF...

But... in that same loose-leafed databook [3] we also find an addendum from June 1961, in which the transistor is listed as 'discontinued', without successor! So the transistor has been available for at most a year, but probably only half a year. Nowadays people complain that semiconductor products quickly go out of production, but apparently that already happened back then as well. We can only speculate about why: perhaps it was too expensive, too unreliable, or perhaps the circuit didn't work as well as hoped. I have not found any advertisements offering the transistor, nor other schematics using it. So, I even wonder whether it was ever actually for sale... and if so, if anyone still has one?

B.t.w., it took until 1963 until the Dutch publisher Muiderkring got the news, because their databook [4] still shows a schematic with the 3746 (again the same one shown above); in a sense that's a good thing, because it was in this booklet that I 'discovered' this transistor.

Others

RCA wasn't the only one trying 'strange' transistor variations in those days. RCA didn't even patent the dual-emitter transistor themselves, only some circuits using it, including the one above, which they patented litterally the day before the IRE presentation.

Someone else working on such transistors was W3NHO, who graduated on research into them at the university of Pennsylvania. His MSc thesis [5] reveals that he worked with transistors made by Philco. In his transistor the emitters were not equal; the extra emitter served as a 'control electrode' to control the amplification without adverse effect on the large-signal behaviour. And more people and companies were trying such transistor variants, as witnessed by several patents. But the RCA 3746 seems to have been the only actual product.

Reversing a transistor

Let's go back to 'normal' transistors with only one emitter. Because of their N/P/N or P/N/P structure, with the base in the middle, one might expect that such a transistor works just as well if one interchanges the emitter and the collector. However, that is not the case; although the transistor does work in such a reversed circuit, it has a much smaller current gain. This difference is used by automatic transistor testers to identify the emitter and collector leads.

The gain difference is caused by physical differences between emitter and collector. Usually the emitter is smaller, for the practical reason that most heat is dissipated in the collector. Furthermore, the amount of N- or P-doping in the emitter is larger than in the collector, because in normal use the free electrons (or holes) flow from the emitter through the base to the collector. Those free electrons are present in the emitter due to the doping. Such a 'source' of free electrons is not needed in the collector, so it needs less doping.

When one interchanges the emitter and the collector, the transistor does not just have less current gain, but also a lower saturation voltage. This is shown in the figure, in which the collector-emitter voltage is plotted vertically against the collector-emitter current horizontally, as measured on a BC546B transistor at a base current of 1 mA. That's quite a large base current for such a transistor, so it saturates.

The first thing that is noticeable, is that the lines are approximately straight. A straight-line relationship between voltage and current is typical for a resistor. The slopes of the lines as seen here correspond to about 4 Ω (normal circuit) and 5 Ω (reversed). For a pure resistor, the lines would go through the origin: at 0 mA the voltage should also be 0 mV. However, on the transistor we measure about 3.9 mV (normal) and 0.4 mV (reverse) at 0 mA. Thus, the saturated transistor behaves like a resistor, with a constant voltage in series, and that voltage is much lower in the 'reverse' circuit than in the 'normal' circuit. So if we want to use the transistor as a switch for analog signals, the reverse circuit is better: the extra series voltage is smaller so the transistor behaves more like a normal resistor.

B.t.w.: many years ago I acquired a nice digital capacitance meter, the Meratronik E315A (see [13]). Internally, it's a bridge ciruit which is brought into equilibrium by switching in or out a large number of resistors using transistors. Back then I wondered whether the saturation voltage across those transistors would not introduce inaccuracies. As it turns out, these transistors are in fact connected in 'reverse', presumably indeed because of the lower saturation voltage. The figure shows part of its schematic, where a sine wave signal (of about half a volt) is divided by a factor that can be controlled by four bits (A through D), which connect resistors of 10, 20, 40 and 80 kΩ via the upper transistors to the input, or via the lower transistors to ground. The control current flows through the collectors, not the emitters, because the emitters are connected via the (relatively large) 10 to 80 kΩ resistors, while the collectors see a low resistance to ground, either direct or via the AC voltage source.

Chopper amplifiers

One application of electronic analog switches is in so-called chopper amplifiers. A chopper amplifier is made to amplify a small DC voltage by a large factor, e.g. in a sensitive measurement setup. It is difficult to make a good DC amplifier. For example, the amplifier may easily introduce an offset voltage (which might even vary with temperature), and at very low frequencies transistors produce more noise than at higher frequencies (the so-called 1/f noise). In a chopper amplifier the signal is switched on and off at a high rate, i.e. 'chopped up', so it becomes AC which can be amplifed more easily.

The figure (from [8]) shows a chopper amplifier in which the switch is implemented as a transistor. The 'modulator' transistor periodically shorts the input to ground. In this case, but not always, there is another switching transistor to restore the DC level at the output, and there's feedback via R_F.

Again, we see both transistors connected in reverse: the control current from the square wave generator flows through the B-C junction, not the B-E junction. Of course, this is done because of the lower saturation voltage.

If the saturation voltage of such a reversed transistor is still too high, it can be reduced further by connecting two transistors in series in opposite direction, as shown in the figure. If they are conducting, there is, as noted above, about 0.4 mV between the collector and emitter of each transistor. But looking between both emitters, the C-E voltages are in opposite polarity so they cancel (if both transistors are identical). This forms a switch with even less offset voltage than a single 'reverse' transistor.

Chopper transistors

One way of making both transistors as similar as possible, is making them on a single piece of silicon. Thus we get the so-called 'chopper transistor': a transistor with two emitters, which boils down to both transistors sharing the base and collector. Such chopper transistors have been a common commercial product for quite a while, manufactured by companies such as Sprague, Crystalonics, Texas Instruments and Fairchild, as witnessed by advertisements in electronics magazines.

The left figure, taken from the Sprague 3N117 datasheet, shows how the specifications were measured, and emphasizes how a chopper transistor is used. If one makes a current flow through the base-collector junction, one finds between the emitters a low resistance with very little offset voltage (e.g. 0.1 mV or even less). And without base current, one finds a high resistance between the emitters, just like in a normal transistor without base current.

Note that we've made two conceptual steps. First we let the control current flow through the collector rather than the emitter, and next we merged two such transistors into one. Thus, we went from something that looked very common (a transistor in grounded-emitter circuit), blue in the earlier figure, to the final figure which looks rather unfamiliar.

B.t.w., various people (again) couldn't agree on what schematic symbol to use; the right-part of the figure shows various symbols that I've encountered in various publications.

(Digital) ICs

Perhaps the best-known form of a transistor with multiple emitters is found in digital ICs from the 74xx family. The figure show the schematic of a NAND gate, four of which are in a 7400 chip. At the input is a transistor with two emitters. That transistor can, just like the chopper transistor, be considered as two normal transistors with the bases and collectors connected. The idea is that if one of the inputs is pulled to ground (i.e., connected to logic 0), the transistor conducts and pulls down the collector as well. The three other transistors invert and amplify this, so the output goes high (logic 1). Only if none of the inputs are pulled low (all inputs logic 1), the transistor will not conduct. However, that's not quite true. In fact, in that case, the base-collector diode will conduct, pulling the collector up to the positive power supply; via the other three transistors this will result in the output going low.

Searching the literature and datasheets, one finds more examples of multi-emitter transistors in ICs, for example in the famous 741 opamp. IC designers have more freedom to use 'strange' transistors on their chips, than 'normal' electronics designers who are limited to transistors that are available on the market.

Oblivion?

The RCA 3746 was advertised to replace two (expensive) transistors, but soon transistors became so cheap that that was no longer a selling point. Chopper transistors as analog switches have been surpassed by various FETs, which do this much better. The 7400-series TTL logic family has been replaced by much more efficient and faster CMOS families. And thus one might think that multi-emitter transistors have disappeared into oblivion.

But no, to my surprise the Crystalonics firm still has them for sale! Crystalonics [10] is (nowadays) a small-scale manufacturer of special diodes and transistors, and even supplies to private customers. In amateur radio circles they have become well known for their CP666 FET, used in the AMRAD active antenna [11]. When asked, Crystalonics owner Paul Weinstein told me that they have the chopper transistors in stock and also still manufacture them, as spare parts for (old) aerospace electronics.

References

[1] L. Plus en R.A. Santilli: A New Concept in Transistor Converters, IRE International Convention, New York, 21-24 March 1960
[2] Rolf Spies: Der neue Drift-Field-Transistor RCA 3746 mit doppeltem Emitter, Funkschau, nr. 24, 1960
[3] RCA Tube Handbook HB3, Semiconductor device section, 1961-1962 online
[4] Transistor Handbook, 4th Edition, July 1963; part of: Tube and transistor handbook, 10th revised edition, De Muiderkring.
[5] John Francis Bogusz (W3NHO): The double emitter micro-alloy and double emitter micro-alloy drift transistors; MSc thesis, Univ. Pennsylvania, May 1959
[6] R.L. Bright: Junction transistors used as switches; Tr. Am. Inst. Elec. Eng., part I: Communication and Electronics, 1955.
[7] Type 3N117, 3N118, and 3N119 Duet Dual-Emitter sEPT Transistors; Sprague Engineering Bulletin No.31703, 1965
[8] J. Watson: Semiconductor Circuit Design, 1977
[9] Datasheet FJH131 (= European equivalent of the 7400), Mullard, 1970
[10] https://crystalonics.com/
[11] Frank Gentges, K0BRA: The AMRAD Active LF Antenna, QST 9/2001. online
[12] http://www.ic-ts-histo.de/fad/halbl/3n90/3n90.htm
[13] Technische notities van PA3FWM, Electron 7/2012; online in English

---

Text on this page is copyright 2024, P.T. de Boer, web@pa3fwm.nl .
Republication is only allowed with my explicit permission.