Retrotechtacular: Old Transistors

1024px-Regency_transistor_radio
Regency TR-1

Transistors have come a long way. Like everything else electronic, they’ve become both better and cheaper. According to a recent IEEE article, a transistor cost about $8 in today’s money back in the 1960’s. Consider the Regency TR-1, the first transistor radio from TI and IDEA. In late 1954, the four-transistor device went on sale for $49.95. That doesn’t sound like much until you realize that in 1954, this was equivalent to about $441 (a new car cost about $1,700 and a copy of life magazine cost 20 cents). Even at that price, they sold about 150,000 radios.

Part of the reason the transistors cost so much was that production costs were high. But another reason is that yields were poor. In some cases, 4 out of 5 of the devices were not usable. The transistors were not that good even when they did work. The first transistors were germanium which has high leakage and worse thermal properties than silicon.

socketEarly transistors were subject to damage from soldering, so it was common to use an alligator clip or a specific heat sink clip to prevent heat from reaching the transistor during construction. Some gear even used sockets which also allowed the quick substitution of devices, just like the tubes they replaced.

When the economics of transistors changed, it made a lot of things practical. For example, a common piece of gear used to be a transistor tester, like the Heathkit IT-121 in the video below. If you pulled an $8 part out of a socket, you’d want to test it before you spent more money on a replacement. Of course, if you had a curve tracer, that was even better because you could measure the device parameters which were probably more subject to change than a modern device.

Of course, germanium to silicon is only one improvement made over the years. The FET is a fundamentally different kind of transistor that has many desirable properties and, of course, integrating hundreds or even thousands of transistors on one integrated circuit revolutionized electronics of all types. Transistors got better. Parameters become less variable and yields increased. Maximum frequency rises and power handling capacity increases. Devices just keep getting better. And cheaper.

A Brief History of Transistors

The path from vacuum tube to the Regency TR-1 was a twisted one. Everyone knew the disadvantages of tubes: fragile, power hungry, and physically large, although smaller and lower-power tubes would start to appear towards the end of their reign. In 1925 a Canadian physicist patented a FET but failed to publicize it. Beyond that, mass production of semiconductor material was unknown at the time. A German inventor patented a similar device in 1934 that didn’t take off, either.

Replica-of-first-transistor
Replica of the First Transistor

Bell labs researchers worked with germanium and actually understood how to make “point contact” transistors and FETs in 1947. However, Bell’s lawyers found the earlier patents and elected to pursue the conventional transistor patent that would lead to the inventors (John Bardeen, Walter Brattain, and William Shockley) winning the Nobel prize in 1956.

Two Germans working for a Westinghouse subsidiary in Paris independently developed a point contact transistor in 1948. It would be 1954 before silicon transistors became practical. The MOSFET didn’t appear until 1959.

Of course, even these major milestones are subject to incremental improvements. The V channel for MOSFETs, for example, opened the door for FETs to be true power devices, able to switch currents required for motors and other high current devices.

Transistors Today

It is possible to build a lot of things today without ever handling a discrete transistor. However, a modern integrated circuit can contain enormous numbers of devices. A 1959 PDP-1 had about 2,800 discrete transistors in it. Even an 8088, the original IBM PC CPU, had 29,000 devices–over ten times the PDP-1. Some large chips today have billions of devices onboard.

Small transistors that work well are throw away components today. A quick check of a general purpose transistor price shows under 20 cents per device and if you are willing to buy more than one at a time, that price goes down to about 10 cents each, and can go down to 3 cents or less with any quantity. For about $80, you can get 2,500 transistors which ought to be a lifetime supply for most of us. At that point, testers and sockets make less sense.

By the way, we’ve talked about early transistor radios before. Rufus Turner published plans for a crystal radio with a three-transistor audio amplifier back in 1950. Germanium transistors still find use in certain applications where their low forward voltage and in things like fuzz pedals where there is a subjective difference in sound quality provided by germanium transistors.

Photo credits:

Regency TR-1 by [Gregory F. Maxwell] GFDL 1.2

Transistor socket by [ArnoldReinhold]  CC-BY-SA-3.0

 

22 thoughts on “Retrotechtacular: Old Transistors

  1. One thing I’ve always wondered about a common transistor such as the 2N3904:
    Although the characteristics are the same, is a 2N3904 made in 2016 the same physical die size of a 2N3904 made in 1977? Is it made from the same mask or is it an entirely different device but which has the same characteristics.
    Anyone any knowledge/comments?

      1. The vendors need some inventory as they make parts in batches, but a large inventory of parts sitting in a warehouse would not be generating profits and they won’t look good on Wall Street. The vendors try not to have excessive inventory and have gone too much in the other direction resulting in parts have long lead time.

        https://www.fairchildsemi.com/products/discretes/bipolar-transistors/small-signal-bjts/2N3904.html
        Package marking: Line 3 -&3 (3-Digit Date Code)

    1. The minimun size of the die is limited by power capacity of the transistor. I doubt that the die of modern 2N3904 is significantly smaller than those of 1977. According to wikipedia this transistor was registered by Motorola in the mid 60’s. It’s maximun Ic being 200mA the junction must be large enough to carry this current without overheating. Nevertheless the manufacturing technology have change now these are made with planar thechnology. The first junctions transistors where junction grown.

      ref: https://en.wikipedia.org/wiki/2N3904
      ref: https://en.wikipedia.org/wiki/Diffusion_transistor#Planar_transistor

    2. The mininum size is limited by the current carrying capacity of the transistor to avoid junction over heating. So it shouldn’t be significantly smaller than the first version.

      1. I have heard stories of MOSFET die shrinks and new parts have different thermal characteristics. This might not be an issue if you use them as a switch as the vendors assume people would do. On the other hand, if you are using a switching part for analog application, better read those datasheets carefully.

      2. Yet old power transistors do indeed have bigger dies then the newer ones…
        In mass production once you sort out the yield problems, the cost comes down to die size, as you can fit more smaller dies on a wafer.
        The net result is that while they both perform exactly the same when you’re within maximum operating parameters set by the manufacturer, the old parts with bigger dies are far more tolerant to current and/or power dissipation overloading.

        some data to back this up:

        picture shows 4 transistors, 3 are 2N3055, one is a KD605.
        2N3055 is specified to max 60V/15A, 115W, Tmax 200°C
        The KD605 is specified to 40V/10A, 70W, Tmax 155°C.
        From left to right, RCA is obvious (fairly old), “GM” is a new chinese bottom tier production (bottom is steel, not copper), 3rd one is not-so-old ST production, last is the KD605, (made by Tesla in Czechoslovakia prob. in the 70-80s)
        current it took to kill them (Uce was between 6-8V):
        RCA – 17A
        ST – 15A
        GM – mysteriously died well before reaching max specified current
        KD605 – not even 20A for over 1 minute (test author could not source more) damaged it…

        Goes to show that “they don’t make ’em like they used to”.

        1. The manufacturing process in Tesla varied quite a lot, so they labelled these after they measured the parameters. There are ones that barely meet the datasheet specifications and there are ones that exceed them more than twice. Also we had state regulated industry here at these times, so when there was order for KD605 they labelled everything that meet at least KD605 requirements KD605.

          And you can even find second grade quality and third grade ones (they have small number two or three punched between the legs). The second grade had small defects on package, etc. The third grade is generally useless as the parameters are way to off. There were also military graded ones that have two crossed daggers on the top of the package…

          So all in all I would say that it depends on greatly on luck with these old transistors.

          1. Most of the older stuff was way overbuilt and as a result of that, excessively expensive…
            My personal experience with KD60X transisotrs (can’t remeber which one exactly, I just grabbed the ones with the biggest current ratings to quickly cobble together a self-oscillating HV power supply) was that they withstood way more abuse then anything else I had lying around.

            Look at the leads of the RCA transistor and the leads to the ST and GM one, the ST collector wire burned (maybe that was deliberate? :D) while the RCA one stayed even when the chip was busted…
            Also note – the RCA and Tesla ones have the leads more or less straight, the newer ones bend them :P

    3. 2N3904 is not a transistor, it *is* a specification number. Modern transistors that are labeled with this specification number must of course comply with the same specification just as the older transistors did.

      Newer manufacturing processes have less variation and smaller die size so newer transistors can behave differently (thermal transfer characteristics for example) but they *will* still comply with the original specification.

      When there is a spec that has “Min”, “Typical”, and “Max” values, a more recently manufactured transistor will be much closer to the typical value specified.

      Because of the larger variation in manufacture, older transistors were manufactured with a lot of leeway to increase yield of conforming components. Consequently, most of the time you could “get away with” running far out of spec. So you could run a 10 Amp transistor at 15 Amps for a short period of time and it would be fine. Newer transistors are more inclined to explode at 10.0000001 Amps – a bit of an exaggeration but you get my drift.

  2. “Some gear even used sockets …” for example the iconic Tektronix 465 and the Dumont 1060 oscilloscopes I have. I learned from an old Tek repairman the standard technique for checking out a wonky scope. Hit it *hard* with a big stick and see if the behavior changes. If it starts working, pull out and reinsert the transistors in that section to scrape the corrosion off the leads. I’ve been strongly tempted to solder them in the sockets, but there are a lot of transistors and I’m not sure if the sockets would stand it.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s