13
\$\begingroup\$

I found computer gold today with this amazing article - it details construction of a 400 germanium transistor-based digital bit-serial computer, circa 1967. Before I get my hopes up of spending about $120 to construct it, I noticed it specifies germanium transistors for construction. I was considering using these as a replacement - but I'm not entirely sure this would be suitable. Are there voltage differences or other electrical issues that would arise, or can I take it straight for face value? thanks for the help!

\$\endgroup\$
  • 14
    \$\begingroup\$ Note that if you don't know what NPN and PNP are, you might want to start with a less ambitious project first. \$\endgroup\$ – Dmitry Grigoryev Aug 11 '17 at 10:35
  • 2
    \$\begingroup\$ Leave a little room for a modern MCU to avoid disappointment should all the home-spun gates go belly up when it overheats. \$\endgroup\$ – Andy aka Aug 11 '17 at 11:03
  • 2
    \$\begingroup\$ Reading the prices of the components (2d for a resistor, 9d for a transistor, making the cost of a NOT gate 1s 3d) sure is a throwback to another era ... before Britain got decimal coinage (s = shilling, d = pence). "Computer gold" indeed! \$\endgroup\$ – Floris Aug 11 '17 at 15:11
  • 1
    \$\begingroup\$ Four and tuppence for an AND gate - bargain! \$\endgroup\$ – J... Aug 11 '17 at 17:47
  • \$\begingroup\$ 1. Resistor values will need to be altered. A competent enthusiast will have no trouble helping you delect initial values and THEN some trial and error will be wise. Selection ONLY by trial and error very unwise as if you don't know why it works when it does you will not know why it doesn't when it sometimes doesn't. 2. Changing to all NPN and swapping pos & neg supplies and reversing all diodes will help. \$\endgroup\$ – Russell McMahon Aug 16 '17 at 3:39
16
\$\begingroup\$

They were specified because it was 1967, and silicon diodes were still nascent technology. Germanium was first discovered and used for diodes and transistors, and were commercially abundant. Silicon transistors started replacing germanium but I guess not quick enough for that article.

As mentioned, you want to use the PNP version like a 2n3907. And keep in mind that while they work in similar manners, germanium diodes have a forward voltage of 0.2~0.4 volts on average, while silicon will have 0.6 to 0.8 volts. So the transistors will not act in the same exact manner.

This site shows 3 problems and solutions for converting a germanium circuit to use silicon. http://www.hawestv.com/transistorize/germanium1.htm for the most part, you will need to change multiple resistor values to get it to work.

\$\endgroup\$
  • 1
    \$\begingroup\$ As a present, I was given two of the 2N2222 MilSpec devices about 1965. \$\endgroup\$ – analogsystemsrf Aug 11 '17 at 4:38
12
\$\begingroup\$

It looks like the transistors in the article are PNP type and your proposed replacement is NPN, so that's not going to work. But I think if you pick a PNP silicon transistor you should be able to get it to work.

Just build and verify samples of the different basic gates (NOT, NOR, OR, and AND are the ones used in the article) and flip-flops before you try to build the whole thing. You may need to tweak some resistor values for the best performance (power vs speed).

\$\endgroup\$
6
\$\begingroup\$

Well, for starters germanium transistors have the opposite polarity from most common silicon transistors, like your 2N3904. So you need to swap the plus and minus of your power supply, plus reverse all diodes.

What strikes me as a bit odd on the diagrams in the article is the use of a dual power supply, with both positive and negative voltages. Also, the amplification factor of the 2N3904 may be different, going into saturation earlier (or later). The NOR gate on page 5 for example may work with only two inputs high instead of one. The flip-flop circuit is also sensitive to the exact values of the resistors. So build a few test circuits and see if they work.

And oh, replace the neon indicators with LEDs; much safer :)

\$\endgroup\$
  • 2
    \$\begingroup\$ There are plenty of small signal general purpose silicon PNP transistors available, though the price/performance NPN silicon may be better. But the OP will have to consider redesigning the entire circuit because of the different base-emitter voltages (0.3-0.4V for Ge, 0.6-0.7V for Si) and changing the supply polarity at the same time might not be a big deal. \$\endgroup\$ – alephzero Aug 11 '17 at 9:23
  • 6
    \$\begingroup\$ "germanium transistors have the opposite polarity from most common silicon transistors"???????? \$\endgroup\$ – Andy aka Aug 11 '17 at 11:00
  • \$\begingroup\$ @Andyaka I'm open to suggestions for a better term.... \$\endgroup\$ – JvO Aug 11 '17 at 18:33
  • \$\begingroup\$ There ARE germanium NPN transistors out there. Like the NTE-103/103a for example. They do tend to draw a lot more current than silicon NPN transistors and have lower impedance + less stable q-point. \$\endgroup\$ – Mr X Jan 25 at 19:24
2
\$\begingroup\$

Germanium transistors have very different properties from silicon. The Tandy Radio Shack "75-in-one" and "150-in-one" electronics kits sold in the 1970s had germanium PNP transistors and one silicon NPN. The germanium transistors had rather "mushy" performance characteristics compared to silicon ones, but on the flip side they could operate off lower voltages. One of the projects, for example, was an audio oscillator that could run off a 0.6 volt solar cell--something that wouldn't work with silicon transistors.

\$\endgroup\$
2
\$\begingroup\$

You posted this a year ago so I don't know if you are still interested. Hopefully you have got it all figured out by now but I submit my answer for the benefit of anybody else who chances upon this string.

The project is quite historic and I remember it being published in the Wireless World way back in 1967 when I was studying the then cutting edge subject of electronics (with lots of valves!) At the time Wireless World was the premier magazine for electronic design and had many cutting edge articles. Perhaps one of the most famous was Arthur C. Clarke's proposal and calculations for using fixed orbit satellites. If you wish to learn more about computing I would suggest that you look for a much more modern design. However, if you are interested in the history of computing, this would be just the job!

The main difference between germanium and silicon transistors in switching circuits, be they either PNP or NPN transistors is that VBE for small germanium is about 0.3 volts whilst those of silicon are about 0.7 volts. Also germanium are more sensitive to heat than silicon and can end up in thermal runaway and destroy themselves. Silicon are much more robust thermally and that is why they are still used (my goodness, 50 years later!!) and germanium have been relegated to the junk box or perhaps very specialist uses which I am not aware of.

As to your question, looking at fig 3, 4 and 5 on page 5 of the article, I think you could replace the PNP germanium transistors directly with a small silicon PNP transistor such as BC557, 2N3906, BC328-25, or BC640, or any other cheap small signal PNP silicon transistor, without any changes to the rest of the circuit what-so-ever. I'm sure you could also change the 1S130 silicon diodes in the AND and the comparator circuits with a more available silicon 1N914 or similar.

The whole point of a digital transistor circuit is to drive the transistor into saturation, so usually the base resistor is calculated to allow 10 times the Ibe to do this, so is pretty small in the first place and a change of 0.4 VBE is not going to make much difference to the value of the resistors involved. Aiding this saturation is the fact that the gain of silicon transistors is a factor of 10 or more better than the vintage germanium.

The only thing that would worry me is that most silicon transistors have a limit of reverse VBE of about 5V. In the monostable circuit of fig 9, C2 will drive the base of Tr2 into reverse bias by almost the value of the negative supply. VBE reverse max for most silicon transistors is about 5 V, so limiting the supplies to 5 V would deal with this. Over 5V then you could use a 1N914 diode or similar across the base emitter of Tr2 to stop this. Cathode to 0V and anode to the base.

Try the simple ccts and see if they work. Not much to lose with the price of transistors nowadays.

\$\endgroup\$
  • \$\begingroup\$ Germanium is used extensively in optics due to its higher electron mobility. It also is proving useful for high speed/low voltage devices. The reason it fell out of favor is due to A)scarcity compared to silicon and B) the fact that they haven't figured out how to fabricate it into ICs yet and C) it is more heat sensitive and is not suitable for high(er) voltage devices. \$\endgroup\$ – Mr X Jan 25 at 19:27
0
\$\begingroup\$

Yes, the type of transistor matters, if you decide to replace the original design, you will need at least to re-compute the surrounding resistors and capacitors.

The reason is that different transistors have different parameters, for example the threshold voltage may be different, different responsiveness, resistance, etc. (and probably these parameters are not called like this in English :) )

My recommendation is to start building gates / memory bits, test them in isolation, observe the effect of heat and then start integrating them into a computer.

You may consider buying integrated circuits already providing the gates / memory.

Also, this is really a hobby/didactic endeavor, whatever you achieve will be performed ten times faster and more reliable by a $1 micro-controller.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.