Clockwork germanium

A retro version of Yet Another (Discrete) Clock, with vintage parts

Similar projects worth following
Imagine you're in the ealy 20th century, with the old technology but with all the knowledge of today. What would you do ?

* Resistors and capacitors are well known and industrially manufactured
* Piezo-electricity was studied by Pierre & Jacques Curie in 1880. In the 1920s, Pierce nailed oscillators and in 1927, the first quartz-synched clock was built by Warren Marrison and J.W. Horton at Bell Labs.
* Point-contact diodes are from the 1900s too
* The LED effect was first observed in a point-contact diode in 1907. Actual production began in the 1960s.
* The bipolar transistor was discovered in 1947 (more or less simultaneously at Bell Labs and in France by Germans)

Let's say we're in the 1950s and gotten seed funding. Making a clock would be a logical application despite the higher parts cost compared to radios. Logic gates are more tolerant to leakage, low gain and fab variations. Indeed, early computers were built out of low-grade transistors!

This is the "neovintage" version of #Yet Another (Discrete) Clock using original old parts as much as possible "for research purpose".

I (@Yann Guidon / YGDES) receive help from two contributors who have more experience with Germanium than me: @Alexander Shabarshin and @jaromir.sukuba. This is invaluable to decypher russian parts and cyrillic data sheets or circuits, thank you guys ! Alexander writes some of the logs, and he also started his own germanium clock project using circuits we develop here.

Expect power consumption to be higher than the MOSFET version, but fortunately, the invention of the transfer resistor is saving us the insane amounts needed to power the heating elements of vacuum tubes. Relays don't need high voltages either ;-)

Some design constraints

  • old/vintage parts wherever possible and whenever available
  • 4.5V power supply from alkaline batteries (so power consumption should be kept low and supply varies between 4V and 5V))
  • PNP Germanium transistors is the major type, though NPN is really required at some places
  • Low frequency Crystal oscillator as discrete clock source (I got down to 2500 cycles/s !)


The clock is made of a similar design as the #Yet Another (Discrete) Clock but the technology is a bit different so one single module will not fit all. We mostly need

  • Divide-by-two cell for the frequency divider (similar to )
  • Flip-flop cell for the Johnson counters (2, 3 and 5 stages). No schematic defined/found yet but it might be adapted from the 4-T of this 50 years old circuit:

The Johnson counters directly drive the LEDs:

This diagram must be corrected because it works well for MOSFETs but Ge BJTs behave differently, the previous sentence means "there is no 7-segments decoder".


The chosen clock source is a 18KHz quartz resonator in a glass tube. Others, slower, are considered, depending on availability. Total: 39 FlipFlops (the initial estimate).


1. 1st approach
2. Testing Germanium transistors
3. Crystal oscillator
4. Germanium diodes
5. 2nd approach
6. Testing Germanium transistors (2nd part)
7. Crystal Oscillator (Germanium Edition)
8. Triggering Germanium
9. Vintage transistor book
10. Got Quartz !
11. Reverse-engineering vintage quartz resonators
12. It just works
13. More ingredients to cook germanium
14. PCB received
15. New inventory
16. Easier frequency division
17. Another interesting BJT DFF circuit
18. More bistables
19. Even more bistables !
20. Even more bistables: plot twist !
21. More bistables...
22. I don't know why it works
23. D-FF without metastability
24. First shift register
25. Germanium is not for ever...
26. Towards a smaller shift register
27. Pulser-sequencer
28. Reduction by mirroring
29. Germanium lemonade
30. More inspiration

  • 400 × OC70 Vintage Germanium PNP transistor

  • More inspiration

    Yann Guidon / YGDES01/20/2021 at 01:56 0 comments

    As mentioned in #Bipolar Discrete UART  I am in awe when watching this fabulous build :

    The most interesting part is at 15:50 and I have examined and even simulated this circuit at An even better SCR-based sequencer. This is an impressive design that inspires me again because I want to turn it into a Johnson counter, which is a special case of shift register, which is also something needed for the #Bipolar Discrete UART ...

    The name of the game is minimalism ! As noted before in 25. Germanium is not for ever..., the number of available vintage parts has dramatically shrunk and their price has exploded, so the fewer parts, the better !

    The design that Leo chose has 1P1N per count: that's 20 transistors per digit, where a Johnson counter would use only 1P1N×5=10 total, which would use only 48 transistors for H:M:S.

    Transforming the SCR latch sequencer into a shift register is not trivial but it should be possible. This would be also a great thing for the UART, because there would be no "slave latch". Looking at the schematic/diagram, which looks like a transistorized Dekatron, the clock pulse is driving ALL SCR through one shared diode, and "current steering" (something similar to the effect of a differential pair) will send the most current to one of the branches, which is pre-selected by the charge of a capacitor.

    This is this individual branch that must be tweaked, adapted, transformed, so it behaves like a D-flip-flop. Then it will be useful for a wide range of circuits :-)

    I know I should look more carefully at the works on the Shockley diode. More early SCR research can be found at

  • Germanium lemonade

    Yann Guidon / YGDES03/08/2020 at 05:35 0 comments

    "When life gives you lemons, make lemonade" they say. This whole project is about "how did they manage to make lemonade back in the days ?"

    To understand this, I explore the old technology and parts "from back then". In this page I try to evaluate the quality and variance of the 70+ OC139 of my stock.

    They come from two sources and I was curious : do they come from the same batch ? Have they been already binned ? How much do they differ ? Is the leakage so bad ?

    The OC139 seem to be really "vintage", looking at the oxidation of the leads. Date code seems to be around 1961.

    Measuring the leakage is a pretty big deal because it badly affects the gain displayed by simple hFEmeters. Look at for the extended explanations.

    I didn't use my multimeter this time, but a handy little Chinese-made tool that guesses the type of tripole and its characteristics, which I wrote below.

    I bought this device at a local store with the usual markup (welcome to France) but I needed it to test my old capacitors and it has other useful features so I didn't care that much. And it was sold by an old, reputable store :-)

    The gain itself is not a big issue, as long as it's about > 20 and there is no point if it is >40 for this specific application. It's also good to see which parts are real lemons and should not be used in a circuit. For example, too high a leakage will disrupt my circuits. We have totally forgotten to take leakage into account in this century !

    Measuring these parameters is not an exact science since germanium is very sensitive to heat (even from the fingers) and even to light (the black paint can't stop all the light if it's intense enough) so sometimes I had to make several measurements.

    minifux1 :

    hFE Vbe Ice0
    106 210 5
    34 224
    31 204 7
    156 219 1
    193 21 164

    36 232
    29 229
    36 248
    32 132
    30 115 14
    44 167 14
    40 195 7
    32 239
    36 131 1
    120 195 7
    46 215
    86 200 7
    26 213
    38 210
    33 268
    31 219
    48 124 4
    49 195 14
    21 210
    30 200 7
    22 224
    31219 14
    173 161 35
    26 205 5


    hFE Vbe (mV) leak (µA)
    40 126 4
    59 205 7
    42 220
    37 200 14
    23 221 8
    34 214 14
    42 219
    24 232 1100
    26 213 5
    34 229
    34 220 1
    28 216
    26 215
    40 204 7
    ?? 196 200
    79 197 8
    33 124 5
    95 220
    73 155 14
    49 215
    29 233 14
    44 137
    134 214 7
    124 159 8
    28 210 2
    46 229
    220 156 35
    37 211
    31 133
    42 204 14
    46 222 4
    296?? 218 11




    Verdict : one totally rotten lemon, several in bad shape, most are bitter but usable, but some have interesting characteristics and demand more examinations. Is their high gain real or an artefact ? and if they really have such a high gain, can they be used for more noble purposes than digital switching ?

    I also wonder about those parts with Vbe < 200mV.

    I didn't/couldn't check for the gain-bandwidth product (or rise and fall time). That should be for another time and for the high frequency dividers :-D

    I guess about half of them would be used for the intended circuit (the shift registers, I need gains between 20 and 40 and no significant leakage). I might have to get more of them to complete the circuit because there are quite a lot of stages... but they have become so expensive !

    Next step would be to plot the data in a graph and re-bin the stock, according to the various groups I identify in the dot cloud.

    And yes, germanium is fun but silicon is so much better :-D I measure so little relative variations...

    Read more »

  • Reduction by mirroring

    Yann Guidon / YGDES03/02/2020 at 02:28 0 comments

    The previous logs Pulser-sequencer and Towards a smaller shift register try to reduce the complexity (and parts count !) as much as possible but the last circuit was still pretty daunting and looks overkill. I couldn't figure out what was feeling wrong until... I tried to reverse the circuit, and that would save a whole layer !

    So the new version both does the sequencing AND pulls the pass transistor bases. In the real/final version, the only transistor per bit will be a NPN, and it saves a bunch of auxiliary parts.

    I added another transistor (+1D1R) at the start to format the input pulse length and the curves show

    1. very short pulse (it works nicely)
    2. long pulse (it works well too)

    So far the cost is 1T 2D 1C 2R and the pass transistor would have the pulse-shortening circuit...

    So the circuit is completed with the 2K2 pull-down replaced by the pass transistor and its RC cell. Here is the circuit :

    My inventory shows I have a quantity (tbd) of 5K6 and about 100× 2K4. The pulse width ratio is given by the ratio of the capacitor values. Here we have 1µF / 220nf so it's approx. 1/4 to 1/5. This provides a niiice clean edge ! If I find a large quantity of cheap non-polarized capacitors at about 1µF, I'll put 2 or 3 of them in parallel on the sequencer side, and only one on the pass transistor.

    The power consumption is still very low when idle, and it can work for the slow parts : seconds, minutes, hours.

    Note that for the sake of evaluation, the "pass transistor" here is a NPN with 1K to the positive rail but the final circuit will have the polarity reversed and the pass transistor (a PNP OC70) will not be tied to ground but to the other latches. I expect that the sequencer-side OC139 will work well.

    The cost so far : 1P 1N 2D 2C 5R (11 parts, ignore the 1K pull-up)

  • Pulser-sequencer

    Yann Guidon / YGDES02/29/2020 at 21:48 0 comments

    Remember the last logs : the goal is to use as few transistors as possible to send a string of consecutive non-overlapping pulses to several pass-gates. Here the design uses NPN as primary type but it will be reversed later because PNP are the majority type in Germanium and NPN are getting hard to get...

    The last log Towards a smaller shift register has explored several methods but nothing satisfying yet.

    Today I tried 2 new tricks :

    • Feed the output back to the input
    • use diodes to separate the stages  and hopefully prevent the simulator to amplify spurious activity.

    The first result is there :

    The down and up edges are pretty clean though the output has a sort of slant after the first half of the pulse.

    There is one great advantage here compared to the previous versions : the cell is triggered by a positive-going pulse, and the idle state does not draw any current. This saves power and reduces drift, compared to the almost-always-on earlier version.

    The "cost" is 1P 1N 2C 3D 6R and can still be tuned.

    However when a resistor is replaced by a diode (to make a sort of charge pump) the result is even more interesting because 2 phases appear (source) :

    The trailing edges are not as sharp but they are now significantly separated, there is a short pulse (going to the pass gate) and then the output to the next stage happens.

    Apparently the sharpness of the trailing edges are not a critical parameter : (source)

    I changed all values to 4K7 and it seems to help a bit.

    I wonder if/how this circuit could be simplified...

    Trying to simplify the circuit and going back to the "first principles", I get this sequencer :

    Falstad did some really strange things with the "charge-pump-like" diodes and capacitor, like creating one pulse of the low AND the high going input pulses, so I had to restart from a clean empty page and compare with real circuits (just to be sure).

    The wave shape of this 5x sequencer doesn't look different from the previous one but it is way cleaner than the early attempts. Now I doubt that the feedback and more diodes are required.

    Note that the capacitor "pull-up" resistor must be significantly smaller than the transistor's base resistors because otherwise, the "going-high" front will make a resistor divider equivalent through the capacitor, which reduces the effective voltage range on the divider network and reduce the pulse time.

    The power is kept rather low, with about 5-8mA during activity, and mostly no current when idle. Only one cell is active at a time, except during transitions. The current is only dependent on the above-mentioned pull-up resistor that charges the capacitor before the up-going front when the transistor is released. That can easily be halved with a higher resistor (2K2) and/or a lower supply voltage.

    Adding a simple complementary common-emitter cell creates overlapping pulses :

    Adding a simple RC cell shortens the output pulse. The shape is not as sharp as I'd like but it's a minimal working circuit :

    This saves quite a lot of diodes, compared to the circuit at the top of this page. And the feedback is not even used.

    So it's working, the required power is reasonable, the pulses do not overlap and I can control the pulse widths with the corresponding RC constant...

    I tried to remove the diodes but the sequencer would be much less reliable. This output drivers seem to work without diodes but the trailing edge is not clean.

    However, adding one diode as feedback brings some advantage :

    It prevents the capacitor from discharging, until the PNP's output goes to 2 diode drops. The rising edge is delayed until a safer output level is reached. This prevents overlapping conduction for the rest of the circuit, AND both base resistors can now have the same value (they are 2×10K or 2×4K7 but now this difference is not required).

    This new version also adds a series  diode to the output buffer, to ensure that the voltage...

    Read more »

  • Towards a smaller shift register

    Yann Guidon / YGDES02/28/2020 at 01:58 0 comments

    (see updates at the bottom)


    I seem to have solved the problem of cascading the latest flipflops, using better decoupling (and some diodes ?). Here's the new circuit :

    For the "fast" parts, a pull-down resistor would be required on the base of the pass transistors, to evacuate the base charges and "cut hard".

    There is a better Power On Reset as well, and this system only uses 6 transistors because there is no need to invert the feedback data (just use an inverted value).

    Then the challenge is to properly sequence the pulses to the pass transistors.

    The question here is not to control a master and a slave latch, but a series of consecutive latches ! so there are as many pulses as there are latches. Here I focus on the "slow" part (< 1Hz update) with a string of one-shot pulsers :

    I have really NO IDEA why the last stage emits multiple pulses... The mysteries of Falstad...

    But the principle seems to be realistic : each bit of the shift register uses 2 T for the FF, 1T for the pass transistor and another for the pulse/delay. It's far from perfectly working but there is hope that each bit uses only 4 transistors. I might add some diodes to create a charge pump or something like that...

    The "fast" circuits might need something more elaborate but at least I might have reduced the parts count for the H:M:S circuits ! (5+3+5+3+5+3)×4 (+ some control stuff) amounts to about a hundred transistors of the same polarity.

    I'll try to make it work, then move to the faster parts : I have found a gorgeous vintage resonator at 2K5Hz !

    That removes some pressure on the divider's design...

    Update : That might work but the simulator is losing its sanity...

    Sometime it wants to find crazy negative voltages at the collectors of the PNPs.  I'll have to test it on the bench but it looks promising.
    I must also find a way to "shape" the pulses because the "decay" is not very clean. But i'm getting closer to my goal :-)

    Cost : 1 PNP + 1 NPN (so it's the same when the polarity is reversed), overall : 5 transistors per bit of a shift register. For a whole display : that's about 120 + 6 transistors... This also facilitates the setting of the time because each shift register can receive a pulse, that could be from a button, and it's inherently debounced.

    Stay tuned !

    Thanks to Google Image search, I just found this :

    The page shows various enhancements to the system. I'll have to test it to improve my sequencer ! However if this trick works for 2 transistors, I wonder if/how it can work for more...

  • Germanium is not for ever...

    Yann Guidon / YGDES02/25/2020 at 19:35 2 comments

    The last log shows a preliminary shift register using 6 transistors per bit. For early Germanium technology this means that each complete clock would require about 100 PNP and 200 NPN.

    PNP are now in stock but early "black glass" NPN are another story, a sad one : I just found that the stocks have now melted and the prices are increasing significantly ! I have just enough parts to test a few partial circuits...

    If anyone can find affordable stock of OC139, OC140 or OC141 please contact me !

    With the waning stock and perspective of suitable NPNs, I have to go back to the drawing board and finally design a bit cell that

    1. uses as few PNP as possible. Probably not 2 because the complexity is then pushed to more diodes and capacitors.
    2. dumps the concept of master-slave latches, to use fewer parts

    I have met many problems with the classic flip-flop because of the metastabilities but I have two new insights and techniques to introduce :

    1. Adding a simple resistor at the middle of the cell (between both emitters and power) turns the FF into a "modified differential amplifier" and this could drastically reduce the metastabilities.
    2. I now understand better how a bipolar transistor could work as a "pass gate". This is a great trick to reduce the cell size because there would be only one point where data is forced into the FF, instead of 2 usually.

    So there is still a chance to come up with something practical...

    So let's start with a differential amplifier.

    Here I have looped one branch to the other, to provide some feedback and hysteresis, this creates about 500mV of hysteresis @5V.

    From there it is easy to loop the other "leg" back to the first and create a bistable circuit, except for the extra common resistor at the emitter. This will help in many ways, and hint Falstad at the much weaker metastability.

    But it's still metastable : this sim initialises in "both conducting" state. There is some difference between the legs (360mV) but this is not enough to "flip hard".

    The imbalance of the resistors should be higher then. But it doesn't solve the problem because then, the system is not bistable anymore...

    I made a different change to manage the imbalance, by changing the gain :

    the default value is 100 and a 10% variation is "normal". This helps the circuit become bistable again, even though it still initialises in metastable state...

    Finally there is a different method to initialise the circuit with a well-defined state : the following circuit has a RC time delay on one branch so the imbalance occurs only during power-up. The transistors' gains don't matter much now.

    Click here :

    It's still not a solution to the dynamic metastability problem but it opens another interesting prospect !

    A whole block of latches can be reset directly by a complementary transistor that connects all the pull resistors. Previously I wanted to use a diode connected to the base of one of the transistors, but this diode would be connected to a common transistor. Here, I can avoid all the diodes, and the transistor can have the same rating, though it is ON all the time.

    From there a more complete FF can be built :

    It uses a single type/value of resistor, replicated by 6 : here I have chosen 8K2 because I just found a lot of vintage ones in my archives. It could be a bit different, depending on the power supply, speed and other details. Here the cell draws about 350µ @ 5V.

    The power-on-reset cell (3 parts) is shared with all the other FF. The values will change and be tuned later.

    The pass transistor is still the same here, but the trigger values vary considerably :

    • The 1M resistor means that the transistor only needs one potential to polarise the component and select which way the base current will go. The whole base current is used to "upset" the base of the FlipFlop. A higher value than 8K2 on the base of the pass transistor would be possible if more current is available from the source. Though...
    Read more »

  • First shift register

    Yann Guidon / YGDES02/09/2020 at 16:18 0 comments

    OK I wonder how I'll manage to find complementary transistors... But you'all can agree it's a pretty awesome and minimalistic circuit !!!

    The first great thing is there is a single clock line, though it must be driven by a push-pull circuit, with probably a complementary PNP-NPN pair.

    The other great thing is that each slice uses 1×PNP and 1×NPN, plus either 1 PNP or NPN. So each bit uses 6 transistors, 4 diodes and 10 resistors. I can't see how to use fewer...

    The main trick is to use "SCR-style" transparent latches, separated by either PNP or NPN "gates". This particular circuit uses the transistor is a very weird way, where the current flows in either the collector or the emitter, depending on the various voltages. So a single transistor serves in 3 cases : isolate, set or reset ! A couple of diodes is required however to prevent the right/destination from overwriting the left/source.

    It would be possible (and possibly cheaper) to use only NPN pass gates with a system of complementary clock signals but the driving circuit is a whole different can of worms. The exercise is left to the implementer.

    The source is getting crazy long... I wonder if all web browser will accept it. It's too long for the YT description though. I might upload the next circuits as files...


    Anyway I didn't find any weird or chaotic behaviour, as in the E&C circuit as before. The circuit would work at other supply voltages, with a little adaptation of the resistor values. You can notice the use of a single value : 2.2K :-)

    This circuit is suitable for example for the UART project. The feedback with an inverted value adds another transistor so a divide-by-two is slightly more complicated and larger.

    Anyway : the metastability issue is solved, the parts count is minimised and I can progress !

    It's now a matter of how I can find complementary pairs in Ge... I have only 2 stocks of silicon complementary transistors : the Russian epitaxials, and the standard cheap BC549/BC550.

    For single-type latches, I might have to study the system described by Ken Shirriff at

    bipolar latch in Intel's 3101 SRAM chip

    Enjoy !

  • D-FF without metastability

    Yann Guidon / YGDES02/08/2020 at 16:49 0 comments

    Exploring the First principles of Flip-Flops was very interesting but going back to the basic/classic "Eccles & Jordan" configuration (with 2× PNP or NPN) proved the weakness of the approach. The previous log I don't know why it works shows that metastability creates countless problems : it is very sensitive to initial conditions, operating conditions and parts values tolerances. I don't want to make a circuit that requires fine tuning and works only at a given frequency !

    Metastability is when both transistors in the pair conduct. This is a third state of the flip-flop and under certain precise conditions, I would get a flip-flap-flop circuit, for who knows what reason. I tried my best to avoid the metastability but it's a systemic, fundamental flaw of this circuit and even adding parts, or using Ge transistors, the problem still lurks somewhere.

    So I went back to some ideas from the earliest log From MUX to Latch and took a second look at the "SCR" approach. And this time I did it right :-)

    The basic SCR-latch has 4 resistors and 2 transistors. The only drawback is the complementary configuration : one PNP and one NPN... OTOH it doesn't suffer from the metastability plague and always powers up/initialises properly in "off" state !

    The power draw is easy to control, by changing the resistors. I have put "no-brainer" values, with 1K the basic for medium currents. By default I choose 3V for the power supply, because I want to power it with a pair of AA batteries. A higher value has been used for the base NPN's pull-up so it is more sensitive to setting but it is not critical. The values will be adjusted for each application, with an appropriate compromise between speed and power consumption. Slower circuits will use higher resistors.

    This circuits is reminiscent of the one discussed at 17. Another interesting BJT DFF circuit so I wanted to try the single-BJT circuit for transferring information from the last stage. Previously I used a different system, where I power the flip-flop selectively but I fear this wouldn't work at high enough speeds. However, a transistor can also be used in controlled ways as an almost bidirectional pass gate !

    It is also reminiscent of the TTL gates input circuits, with the added trick that it is controlled by a signal, and not permanently pulled up by a resistor. And swapping the emitter and collector doesn't seem to make any difference.

    In fact the BJT can be seen as a pair of diodes so the current from the base can go to the collector and/or the emitter (with more or less chance) IF the base is at a high enough potential. With some luck there is even some amplification :-)

    Here the goal is to pull or push current through the base of the slave FF. The direction will depend on the potential of the "source" of current (the master). I found that I had to put a diode to prevent the slave from controlling the master FF...

    Here is the link :-)

    The latch is really transparent when the clock pulse is high so this is easily transformed into a DFF by adding the pass gate at the input of the M/s pair.

    (full source in the description of the video)

    Now is the time to turn it into a clock divider...

    This circuit should it :

    But I'm looking for a trick to perform the inversion, without the extra NPN...

    Anyway, it's already good enough for a simple shift register :-)

    In the last video, I saw that the 3.9K pull-up was not required. But that was not the end of it ! I'm now trying to design another system with 3 PNP and 3 NPN  though it's very tricky but the potential savings are significant.
    Now I wonder where/how I can find enough NPN in germanium...

  • I don't know why it works

    Yann Guidon / YGDES02/07/2020 at 21:52 3 comments

    It works as it should but I don't know why.

    Here is the source code for Falstad :

    $ 3 0.000001 0.34903429574618416 58 3 43
    R 696 16 696 -16 0 0 40 4 0 0 0.5
    w 808 352 808 384 1
    w 808 192 808 176 0
    w 584 176 584 192 0
    w 664 96 592 96 3
    w 808 288 808 256 0
    w 728 176 808 176 0
    w 384 64 384 136 3
    w 696 112 696 176 1
    w 584 176 664 176 2
    w 584 288 584 320 0
    w 584 256 584 288 0
    w 664 288 584 288 3
    w 808 320 808 288 0
    w 728 288 808 288 3
    w 672 336 728 288 0
    w 720 336 664 288 1
    r 808 256 808 192 0 1000
    r 584 256 584 192 0 1000
    t 616 336 584 336 0 1 0.5772267829335654 0.639069922813361 50
    t 776 336 808 336 0 1 -1.1769217530853389 0.06184313995994596 50
    g 808 384 808 416 0
    g 584 376 584 416 0
    w 696 16 696 32 1
    w 696 32 696 80 0
    t 664 96 696 96 0 1 -0.05293572317914741 0.6386489173065311 200
    r 384 64 464 64 0 100000
    w 496 80 496 96 0
    t 464 64 496 64 0 -1 -0.5078879880313893 -0.5343748422129604 100
    w 496 32 496 48 0
    w 696 32 1160 32 0
    r 496 96 592 96 0 1000
    w 376 448 376 304 0
    d 376 304 440 304 2 1N4148
    d 376 256 456 288 2 1N4148
    w 496 304 552 304 1
    w 552 304 648 304 0
    w 696 176 728 176 0
    w 696 176 664 176 0
    r 776 336 720 336 0 1000
    r 672 336 616 336 0 1000
    w 648 304 672 336 0
    r 512 288 456 288 0 200
    r 496 304 440 304 0 200
    w 512 288 584 288 1
    w 976 288 1048 288 1
    r 960 304 904 304 0 200
    r 976 288 920 288 0 200
    w 1112 304 1136 336 0
    r 1136 336 1080 336 0 1000
    r 1240 336 1184 336 0 1000
    w 1160 176 1128 176 0
    w 1160 176 1192 176 0
    w 1016 304 1112 304 0
    w 960 304 1016 304 1
    d 840 256 920 288 2 1N4148
    d 840 304 904 304 2 1N4148
    g 1048 384 1048 416 0
    g 1272 384 1272 416 0
    t 1240 336 1272 336 0 1 -0.5571041332660905 0.05744512506309504 50
    t 1080 336 1048 336 0 1 0.5113871189769901 0.5688322439565177 50
    r 1048 256 1048 192 0 1000
    r 1272 256 1272 192 0 1000
    w 1184 336 1128 288 0
    w 1136 336 1192 288 0
    w 1192 288 1272 288 3
    w 1272 320 1272 288 0
    w 1128 288 1048 288 3
    w 1048 256 1048 288 0
    w 1048 288 1048 320 0
    w 1048 176 1128 176 2
    w 1192 176 1272 176 0
    w 1272 288 1272 256 0
    w 1048 176 1048 192 0
    w 1272 192 1272 176 0
    w 1048 352 1048 384 1
    w 1272 352 1272 384 1
    w 1160 32 1160 80 0
    t 1128 96 1160 96 0 1 -3.9999999998995 -0.3359971915536067 200
    w 1128 96 1064 96 3
    r 968 96 1064 96 0 1000
    w 1160 112 1160 176 1
    w 376 448 1112 448 0
    w 384 136 968 136 0
    w 968 136 968 96 0
    w 664 288 704 256 0
    w 840 256 704 256 0
    w 808 288 840 304 0
    w 1112 448 1184 336 0
    w 1136 336 1160 472 0
    w 1160 472 352 472 0
    w 352 472 352 256 0
    w 352 256 376 256 0
    w 496 32 696 32 0
    R 384 136 336 136 0 2 10000 2 2 0 0.5
    w 584 352 584 376 1
    o 94 2 0 4355 5 0.00009765625 0 2 94 3
    o 14 2 0 4355 1.25 0.0015625 0 2 14 3
    o 85 2 0 4355 1.25 0.0015625 0 2 85 3
    o 53 2 0 4355 1.25 0.0015625 0 2 53 3
    o 67 2 0 4355 1.25 0.003125 0 2 67 3

     But when I change anything, it breaks... Worse, it can also create weird chaotic oscillations in some uncontrolled cases.

    Metastability (when both transistors in a pair are ON) is a big problem as well... So I played with a capacitor to create some tiny imbalance but it was not the most efficient method.

  • More bistables...

    Yann Guidon / YGDES02/03/2020 at 02:16 0 comments

    @jaromir.sukuba  sent more more vintage germanium yummies and the cogs and wheels in my head resumed spinning again !

    In the log From MUX to Latch, I started investigating other topologies for flip-flops and came up with a cell that was interesting but I wondered if it was optimal. After all the transistors have crawled upon the Earth for something like an eternity now and I've seen many circuits explored in the history, in this project...

    The 2-transistor system has been extensively covered by this page and has always been lacking in a way or another. Slow, too many passive parts, tricky to setup... I want something with the fewest parts possible, stable, no capacitor and able to run fast. Even with lousy Ge trannies.

    I was still not satisfied with the complex circuit I devised. The transistor count was ok but there are too many diodes for my taste. However the design of this flip-flop gave me more insights into what was good and what was to be avoided.

    So I restarted back to the basics and used Falstad's online interactive simulator to build a cell from the basic principles ! And the result is quite unexpected...

    It's really very basic ! All the resistors have the same value ! And I have avoided diodes ! You can run the sim with this code :

    $ 1 0.000005 2.3728258192205156 60 5 43
    g 128 352 128 384 0
    g 336 352 336 384 0
    t 304 336 336 336 0 1 0.5879810154813296 0.6224917671893107 200
    t 160 336 128 336 0 1 -0.5879810154813296 0.03451075170798105 200
    r 128 256 128 192 0 1000
    r 336 256 336 192 0 1000
    w 256 336 208 288 0
    w 208 336 256 288 0
    w 256 288 336 288 3
    w 336 320 336 288 0
    w 208 288 128 288 3
    w 128 256 128 288 0
    w 128 288 128 320 0
    w 128 176 240 176 0
    R 240 80 240 32 0 0 40 2 0 0 0.5
    w 160 336 208 336 0
    w 256 336 304 336 0
    w 240 112 240 176 1
    w 48 288 16 288 3
    w 240 176 336 176 0
    r 96 96 176 96 0 1000
    r 48 288 128 288 0 1000
    L 16 288 -16 288 0 1 false 0.63 0
    L 96 96 64 96 0 1 false 2 0
    w 336 288 336 256 0
    w 208 96 176 96 3
    t 208 96 240 96 0 1 -0.01041339040851752 0.6145396107402716 200
    w 128 176 128 192 0
    w 336 192 336 176 0

    You can find the classic flip-flop with its two interlocked transistors, you can't make it more simple. I have chosen the version with only 2 resistors on the collectors, not the bases.

    Then there is an emitter-follower that powers the flip-flop. The clock must go all the way up to the +Vcc rail but the driving current is nicely low, and decreases as the beta of the transistor increases. So a single PNP transistor could drive the main clock of a group. There is a 0.6V drop but it doesn't matter much. Swap it for a PNP if you prefer. It's interesting however because it insulates the pair of switching transistors from the supply, we'll see later why. The emitter follower can also be used to tune the supply voltage and current and more of these can be chained together.

    So where does the data come from ? That's the whole trick in fact and it's not obvious...  The principle/idea is :

    • when the FF power is removed, the input data "bootstraps" the feedback loop.
    • when the FF power is back on, it takes over and reinforces the loop, and makes the input data ineffective.

    The solution I have found is apparently a simple resistor...

    I still have some cases where the FF initialises with both transistors passing and I can't see why. Falstad can show some weird transient behaviours but something else is happening and I suppose SPICE would solve or resolve these patterns.

    The resistor liaison seems to work almost well but relies on a very low input impedance behind the resistor, which might not be available, particularly in this circuit where I try to reduce the current as much as possible (because if it works, this circuit would be replicated tens of times !)

    So I'm struggling to get a low impedance output and avoid forbidden states in the FF.

    Anyway, 3 transistors and 3 resistors is a very promising circuit for a latch and I'll try to make it a full DFF !

    The "bad starts"...

    Read more »

View all 30 project logs

Enjoy this project?



add_ocean wrote 09/26/2017 at 21:52 point

Well, let me add some ideas :)

While i studied "yet another discrete clock" project (very neat D-type flip-flops and Johnson counters), i must admit that binary logic is best in that it is very reliable and tolerates components drift but at cost of complexity.

Is actually binary a must? Why not to have "analog" appliance, at least it will be an unusual approach.

This is schematic of divider (from book described supposedly tested designs of amateur builders):

There are oscillator and two stages of divide by 10. It says that each stage can divide by max 7 with ferrite cores, and by max 13 with now obsolete Al-Si-Fe cores (huge 36 mm diameter, low mu, low frequency, high flux. I believe modern iron powder might go just as fine). First coil is 0,33 Henry and next 3,3 Henry. Of course main disadvantages are complicated adjusting tank freq. and need for thermal stability of LC tanks (i believe 2,5% or better in whole temperature range ). Pobably need for careful  temp compensation?

Then, simple ring counter that drives LEDs, no binary D flip flops. More ring stages but much simplier each stage. Schematic from somewhere:

I am so sorry to show schematics that i did not try myself, so please forgive if these have some flaws. :)

  Are you sure? yes | no

Yann Guidon / YGDES wrote 09/26/2017 at 22:18 point

oh, you mean to tune a LC tank to the desired frequency ? That's... not reliable enough in my book :-/

I couldn't make sense of all the circuits, they surprise me a lot and understanding them will take time... But it's a an interesting puzzle/challenge :-D

  Are you sure? yes | no

add_ocean wrote 09/26/2017 at 22:53 point

Putting this design into industry grade device will require high end, rock stable LC tanks, i suppose. And, i see it now, below 20 Hz it will be even less feasible as inductance rises, and not to mention core materials won't work fine at such low freq. :(

  Are you sure? yes | no

Yann Guidon / YGDES wrote 05/10/2016 at 12:41 point
I'm still "sourcing" transistors. Hundreds of AF137, G106T, AC125V, OC602, AF200...

And probably the weirdest : unmarked OC70, that look like it was made yesterday.

Apparently, "germanium is a thing" like vinyl LPs, factories seems to be restarted !

Anyway I have enough stock to work on 4T flip-flops.

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 13:32 point

ebay has a lot of similar stuff - I bought a bag of Russian MP-transistors which are the same germanium retro like 2N1xx. They have a nice "feature" - you can use pliers to deform its head and then you can remove it completely - inside it will be a bare P-N-P (or N-P-N) assembly that is LIGHT-SENSITIVE ;)

  Are you sure? yes | no

Yann Guidon / YGDES wrote 04/03/2016 at 13:35 point

OMG ! I will try :-D

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 13:48 point

I remember article from soviet magazine for kids where authors built "solar panel" out of a few dozens opened MP-transistors that generated about 2V on sun ;)

P.S. I found it:

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 13:53 point

This is a table from the same article with experimental data about what single diode (two columns on the left) or single transistor (two columns on the right) can generate from light (voltage and current):

P.S. Full article:
Interesting fact that I didn't remember - author of this article is a woman :)

  Are you sure? yes | no

Yann Guidon / YGDES wrote 04/03/2016 at 14:25 point

Awesome ! Thank you :-) I will certainly reuse this material so I'll need more infos :-)

If you have other "ideas" about them, please share ! For example I found this "logic gate"

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 14:51 point

Sure ;)

Another article series from the same soviet magazine was about building logic elements from those transistors (because they were very popular in Russia of 80s):

Here you can see NAND3 with open collector (блок А), NAND3 (блок Б) and trigger (блок В) that actually constructed from 2 NANDs:

  Are you sure? yes | no

Yann Guidon / YGDES wrote 04/03/2016 at 15:02 point

@Alexander Shabarshin awesome, thank you !

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 15:19 point

And these accidental memories made me curious how fast such NANDs could run ;)

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 18:02 point

OK, I built it :)

It's MP25A p-n-p germanium transistor (similar to 2N189) manufactured in USSR in September 1979 (I got bunch of them from ebay relatively cheap) connected as 2nd circuit from article (NAND element), but instead of one diode V4 I put 3 silicon diodes connected in series (the circuit refused to work with 1 silicon diode as V4) and I use 10K resistors. According to spec MP25A should be able to work with frequencies up to 250 kHz, but this particular circuit has a limit about 10 kHz - see oscillograms for 5,10 and 20 kHz:

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 19:44 point

I put 4 diodes to the base circuit in order to shift threshold lower:

Then I put capacitor 480nF over the diode series and resistor from base to emitter 1K - as per @matseng 

and I got nice 100 kHz :)

P.S. Reducing voltage to 4V actually allowed 200 kHz signal to be processed:

and U-curve is more natural with 4V (threshold closer to U/2):

  Are you sure? yes | no

Yann Guidon / YGDES wrote 04/03/2016 at 23:35 point

It seems we got a 2nd contributor here ;-)

I haven't received my eBay orders yet (it will take a while) so I hope that @Alexander Shabarshin will make a proper log page to describe these experiments :-)

Thanks for your precious time !

  Are you sure? yes | no

matseng wrote 04/03/2016 at 14:51 point

On the other hand any silicon BJT is also light sensitive. I remember cutting of the tops of TO3 encapsulated 2N3055's and using them as foto sensors when I was a teenager back in the late seventies.... I think that they actually generated quite a low of power in bright light.

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 14:54 point

These monster transistors are more difficult to open I assume...

  Are you sure? yes | no

jaromir.sukuba wrote 04/03/2016 at 18:24 point

Oh those yellowish scans from Russian magazines bring back the memories to five issues of Modelist-Konstruktor magazines I bought for 3 Slovak crowns (~= ten eurocents) from my schoolmate - back in 1993 - I was in third grade of elementary school. Funny thing is that I didn't understand a single word, I didn't even understand Russian alphabet. My father learned me it in one Sunday afternoon, so I could somehow read the articles (one in five Russian words resembles somehow equivalent Slovak word). That was the time when I got fascinated by electronics.

The M-K magazine had similar look & feel.

  Are you sure? yes | no

SHAOS wrote 04/03/2016 at 20:29 point

Yes, I remember Modelist-Konstruktor, but I don't remember any article from it :)

  Are you sure? yes | no

jaromir.sukuba wrote 04/03/2016 at 12:08 point

Having a bag of 500+ germanium switching transistors, this looks like the right pointless project for me!

  Are you sure? yes | no

Yann Guidon / YGDES wrote 04/03/2016 at 12:55 point

are they for sale ? :-P

  Are you sure? yes | no

jaromir.sukuba wrote 04/03/2016 at 19:14 point

Don't ask hoarder to sell his sacremental ;-) though you can take a look at part of my collection

I must admit I'm slightly obsessed with germanium components.

  Are you sure? yes | no

Yann Guidon / YGDES wrote 04/04/2016 at 00:17 point

Well then thank you for your kind participation :-D

Would you like to use some of them to help with this project ? :-)

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates