Why did Ge transistors persist as output-stages?

[G6Tanuki]

I'm talking here about the way in low-power applications the likes of the AC187/188 or the AD161/162 pairs of PNP/NPN push-pull audio output transistors continued being specified up until the mid-1980s despite the rest of the transistors used with them in the radio/tape-player/whatever having transitioned to silicon a decade or so earlier?

The higher-power amplifiers with non-complementary-symmetry finals all went silicon-throughout quite quickly, with the coming of the likes of the 2N3055 and the "TIP" plastic-power transistors from TI, along with the Motorola '40xxx' range. You never found silicon low-level stages and phase-inverter driving a pair of Germanium OC35s or AD149s as the finals.

Why did this happen? Was there some good technical reason to stick with Ge or was it just rampant conservatism by equipment designers?

[paulsherwin]

I think it took a while for Si complementary pairs with a bit of grunt to come down in price. The Roberts R606MB switched from Ge to Si in 1976 or 7, in the middle of the production run.

[dave_n_t]

It could also be to do with the lower saturation and B-E voltages. That enabled (theoretically, at least) slightly higher power output from the same battery voltage and speaker impedance.

(I recall a receiver circuit in 'Smithy and Dick' which used a ZN414 front end, and a complementary AD161-AD162 output pair. The whole thing was powered from a type 800 3V battery. Smithy averred that if there was any sign of thermal runaway, the battery would probably run out of juice before any harm was done to the trannies , even though they weren't heatsinked).

Dave

[kalee20]

Yes I recall that article too. The AD161/162 is a reasonably chunky pair, in metal can package, MD17c I think, like a dwarf TO3.

But, why germanium power stages? Agree that the lower base-emitter voltage would allow a bit more power from a limited supply voltage. But against that, germanium gives up at a much lower temperature than silicon, exactly what's not wanted for the power stage!

[turretslug]

Did these middling-power Ge devices have better current gain than equivalent Si devices, at least for a while? In a high-powered mains-supplied amplifier, it wouldn't matter too much if your Si output structure itself effectively required a small power amplifier to drive it, whereas for a posh (i.e. respectably powerful) portable radio etc. device, it would help battery life if the few watts output stage could be driven with much the same essentially voltage amplification circuitry as the previous generation's few hundred milliwatt-type output stage. Just a thought....

[Synchrodyne]

I’d be inclined to look at relative economics and to some extent inertia as the reasons for the delayed change to silicon output devices in “non-hi fi” domestic equipment.

The putative lower distortion of Ge output devices as compared with Si in otherwise similar simpler circuits was less likely to be a factor in the initial choice of output devices for such equipment, but it may have provided a rationale for not making an early change to Si.

Originally Posted by G6Tanuki

The higher-power amplifiers with non-complementary-symmetry finals all went silicon-throughout quite quickly, with the coming of the likes of the 2N3055 and the "TIP" plastic-power transistors from TI, along with the Motorola '40xxx' range. You never found silicon low-level stages and phase-inverter driving a pair of Germanium OC35s or AD149s as the finals.

And they were interesting times. Certainly in virtually all respects, not the least robustness, Si output devices were preferable to Ge for hi-fi equipment and similar. But their adoption shone the light on the elephant in the room, which was the inherent asymmetry in the Lin quasi-complementary (QC) output circuit, which in turn was a major cause of crossover distortion and the so-called “transistor sound”. I think – although I am not completely sure – that Si devices were worse than Ge because their transfer functions had tighter curves near the origin, with a marked difference between the Darlington pair and the conjugate pair, the latter being tighter than the former. The asymptotic slopes of the Darlington and conjugate pairs were different too, although in that aspect Ge and Si might not have been similar. Thus with this background, there were several approaches to the adoption of Si output devices, including:

1. Use the standard Lin circuit, pile on the NFB and look the other way.
2. Avoidance by opting for transformer drive (e.g. Rogers, Bailey (1966)).
3. Avoidance by adopting Class A (e.g. Richard Allan/Sugden and Linsley Hood (1969))
4. Develop symmetrical QC output circuitry.
5. Await the arrival of suitable Si PNP output devices and use a fully complementary output. (Bailey (1968) was an early example, and I think was the basis for the Radford SCA30.)

Re item (4), Quad seems to have been the first to do that with its output triples in the Quad 303 of 1967. The initial breakthrough having been made, simpler symmetrical QC circuits followed in 1969, with Shaw including a power diode in the conjugate pair and Baxandall using a signal diode. The latter I think became the norm for QC outputs.

Not all agreed that the fully complementary circuit was as symmetrical as it promised to be, particularly at higher frequencies. As I recall, Vereker was in that camp, and chose what was basically the Baxandall circuit for the Naim power amplifiers.

In the Wireless World article on his 1966 design (WW 1966 November p.542ff), Bailey offered both germanium (2N2147) and silicon (2N3716) output variants, the rest of the circuit being Si, wit the following commentary:

"The size of heat sink required for the output transistors depends on the ambient temperature range, the type of output transistor and the type of service considered. For example the most severe test is for germanium transistors tested under full load current conditions into a reactive load and at a high ambient temperature. Rather than deal with large heat sinks for arduous duty, the author feels that it is better to specify silicon output transistors where severe conditions are likely to be encountered. For normal domestic duty into loudspeaker loads the cheaper option of germanium transistors is perfectly satisfactory. In fact germanium transistors usually give far lower distortion due to their better linearity."

In this transitional era, there were one or two commercial designs with Ge outputs in otherwise Si circuitry. Bryan comes to mind as one, but the details escape me right now. So Bailey was not alone in his thinking.

Bailey’s 1966 amplifier was a transformer-drive design, but the reasons for that – avoidance of asymmetry - were better explained in the article (WW 1968 May p.94ff) describing his 1968 fully complementary design.

The Quad 303 circuit was described in WW 1968 April, p.67. There is a more detailed treatment of QC asymmetry and the Quad 303 circuit by Baxandall (in his characteristically lucid style) in Amos (ed), Radio, TV & Audio Technical Reference Book, p.14-9ff. The Shaw diode circuit was described in WW 1969 June p.65ff, and the Baxandall diode in WW 1969 September p.416ff.

With all of this going on in the hi-fi field, with claims and counterclaims, it is not so surprising that the non-hi fi domestic equipment makers stayed put for a while. By happenstance the commonly used fully complementary Ge output would not have looked to bad in the overall performance hierarchy. So waiting until the prices for suitable Si devices came down enough was a reasonable pathway, even if in some cases it was less a deliberate choice than the result of inertia.

Cheers,

[paulsherwin]

I still think the problem for things like transistor radios was the availability of PNP complements at a competitive price. The ancient 2N2222 is perfectly capable of use as a low power output transistor, but the complementary 2N2907 took a lot longer to develop (the number is a giveaway) and was originally expensive.

[Nicklyons2]

For what my two penneth is worth and I'm talking only of low power stuff, small box record players & portable radios: I'm going 50/50 with design conservatism & 'we've got a load already' together with the undoubted advantage of Ge on lower voltage stuff. Remember there were quite a few bigger portable radios which used 3 X 'C' or 'D' cells. The really tiny stuff used 9V PP3 and transformer coupled & outputted OC81 types. 4.5 V with Si would have given a bit less power and would require a lot more careful design in biasing to prevent cross-over distortion.

[Hartley118]

.....If the design ain't broke, why fix it? - particularly if Si was more costly then than Ge.

Martin

['LIVEWIRE?']

One area in which Ge power transistors continued in use well into the 70s was Car radio o/p stages. Radiomobile & Motorola used the ubiquitous AD149 in single-ended amplifiers (Models 1070/737 & 80), and the equally common AD161/162 in push-pull amplifiers (1080/1085/1095, etc.). Similar Ge Transistors were used by Blaupunkt (AD157/158 IIRC), Voxson, Autovox, and others, the latter using a Ge/Silicon o/p pair in their MA758 Radio/Cassette player, which was also marketed by Motorola as their model 252. I presume the reason that these manufacturers continued with Ge. Transistors in audio amplifiers, especially o/p stages for reasons already stated by other posters in this section. Most. IIRC, skipped discrete silicon o/p stages, and went straight to IC audio amps., though there were some exceptions

[Restoration73]

By 1980, Ge power devices were no longer available from UK manufacturers, but new
devices were imported of Indian manufacture e.g. OC35.

[Peter.N.]

The first Ge real power transistor I came across was the OC16 in the hybrid Pye car radio, it did a brilliant job especially for a single ended circuit, the audio quality was as good as I had heard on any valve radio and considerably better than most transistor portables using a pair of OC81s or the like.

Peter

[G6Tanuki]

Some interesting thoughts here: I can see how the lower forward voltage-drop of Ge could be useful in battery-powered transistor radios where you wanted to squeeze the last bit of life out of an old PP3!
[Side-thought: 9V seemed to be the standard voltage - supplied from a "PP" layer-battery - for UK-made transistor radios in the 1960s whereas a lot of Japanese/Hong Kong and some European radios of the time used 4.5 or 6V from a few cylindrical torch-battery-style cells - U2, HP11 etc. . . Did Ever-Ready have some leverage with UK radio designers to 'encourage' the use of their layer-cells, or is there a more-technical reason?]


But why the likes of the AD161/162 persisted in in-car entertainment escapes me: the automotive environment experiences probably the widest range of ambient temperatures most electronics-designers are likely to come across and I'd have thought that a switch to Silicon would have made the whole issue of designing for stable biasing and lower chance of thermal-runaway so much easier.

I'm also wondering about production tolerances: complementary-symmetry Ge output transistors were being sold in 'matched pairs' but I never seem to recall this being the case with Si (though we did get "gain-groupings" like the BC109A/B/C of course). Ge transistors subject to wider production-'spreads' of characteristics? From a production-accountant's perspective the need for 'matched pairs' would only push up costs because I'm sure semiconductor manufacturers wouldn't do this matching for free.

The Japanese seemed to switch to Si-everywhere much sooner than the Europeans: I wonder if there's a moral here?

[Radio Wrangler]

The Germanium transistors did the job and the industry was tooled-up around them and the management were loath to invest in anything which seemed unnecessary.

Once silicon devices had got to the point where they could do the job better than the Germanium parts, British manufacturers still didn't change over. They only did once the price became favourable as well.

The Japanese manufacturers were somewhat more forwards-looking. They saw that moving their turnover from Germanium to Silicon would result in cheaper transistors even it the step at the point of change was upwards. When you design a product, you are shooting at a moving target, you have to aim ahead. It looks unnecessary and wasteful to those who don't think ahead. The Japanese made stuff which did what people wanted, was reliable and affordable.

David

[mhennessy]

It takes a brave designer to move away from something that has been successfully put into production 100,000 times! Would you? The AC187/188 or AC128/176 pairs were incredibly successful and very cheap. And easy to put on a heat sink. Whereas the "Lockfit" BC465/464 pair were none of those things

Today, a similar change might be moving from Si to Gan switching elements in a switched-mode power supply. From what I was reading the other day, it's a very exciting development, but still "early days". If my career prospects were at stake, I'd be waiting to see what the competition do

http://www.ti.com/lsds/ti/power-mana...-overview.page

The voltage-efficiency argument is often raised - I remember trying it myself at school (and it fooled my teacher!) - but it's totally academic in the real world. An extra 0.4V peak, perhaps, with a following wind? From a 9V supply, giving 8V pk-pk or 4V peak, we're talking perhaps an extra 10% of peak voltage swing. Sounds like a worthwhile improvement until we put the numbers into a calculator and discover that the difference is a totally inaudible 0.8dB!

Of course, things change once you get beyond 5 or 10 watts. Germanium always struggled here - not helped by designers who were doing their best to feel their way with this new technology while being highly aware of the commercial realities of shipping products.

This fed in another factor: designers knew that germanium transistors were fragile, so generally provided decent heat sinking. When silicon power devices arrived, the perceived wisdom (or indeed, main selling point!) was that they were bomb-proof at high temperatures. So everyone scaled back their heat sinks. And we know what that meant...

[John_BS]

I suspect it may have to do with the Ge devices having a high beta at normal operating currents: the AD161 was spec'ed at 80 to 300 at 500mA, and not many Si types were capable of matching that. So, they could be driven directly from a class A -biased common emitter stage without another intervening pair of trannies operating as "Darlingtons".

John

[Hybrid tellies]

I think back in the 1970's they had problems with Si transistors especially in low powered low voltage transformerless output stages. Si transistors seem to be fragile and frequently went leaky or short circuit. My ITT Colt portable radio used a Si output stage and I had to replace the output and driver transistors every few years. Philips portable radios also had the same problems. It took quite a few years before these problems were overcome and Si output stages became the norm and it was quite interesting to note that Ge output stages, with and without transformers, continued to be used on a lot of far east popular transistor radios well into the 1980's.

[M0XNA Neil]

I have an old (and still working!) Elpico 50W PA which has Germanium TO3 transformer coupled output stage. The chassis looks like it was originally designed for valves and still has the 'HT choke'.

Neil

[dseymo1]

It possibly had something to do with what deals the company buyers could make. A regular order of 'x' hundred well-established devices a week (covering a number of production lines) would be likely to attract a much better discount than a few of a newly-released type.

As regards 'PP' batteries, don't they have a higher like-for-like energy density than individual cells? The more sophisticated manufacturers would have appreciated the 'longer life' marketing opportunity which was perhaps missed by the makers of cheap imported sets.

[G6Tanuki]

Originally Posted by dseymo1

It possibly had something to do with what deals the company buyers could make. A regular order of 'x' hundred well-established devices a week (covering a number of production lines) would be likely to attract a much better discount than a few of a newly-released type.

Up to a point: surely as the industry switched from Ge to Si in the mid/late-1960s [I cite Fairchild's silicon-planar technology as crucial here - they licensed it to plenty of other manufacturers - it gave massive improvements in consistency and yield] the demand for legacy Ge transistors would have fallen away and so the cost of keeping a production facility for these running would have risen...

Originally Posted by dseymo1

As regards 'PP' batteries, don't they have a higher like-for-like energy density than individual cells? The more sophisticated manufacturers would have appreciated the 'longer life' marketing opportunity which was perhaps missed by the makers of cheap imported sets.

I kinda thought it was the other way round - the cylindrical cells of the late-60s onwards having better weight-for-weight/volume-for-volume energy-density. The Ever-Ready "Blue/Gold/Silver Seal" debacle -

http://www.independent.co.uk/news/bu...d-1494225.html

being interesting. Truth is, the "PP" layer-type batteries lost out.