10/17/2017 at 12:10 •
Today I received an email that I couldn't expect at all. Enjoy !
I enjoyed your hackaday posting of the 2n554 flip-flop circuit. I built the computer described in that Popular Electronics article in high-school in 1964. Those circuits were later reused in my KB/67 (Kalin-Burkhart) project.
Here's a bit more about the history of that flip-flop circuit. I was always curious where that circuit design originated, and did some follow-up. That flip-flop was also featured in the January 1960 issue (page 65) of Electronics Illustrated in an article by Ronald Benrey. He contributed many projects to electronics magazines during that time. His homemade satellite appeared in the October 1958 issue of Popular Electronics.
Looks like the flip-flop circuit is from Performance Tested Transistor Circuits, (page 47) published by Sylvania in 1958. Although that circuit uses a different transistor, 2n307, Benrey substituted the 2n554, which was a common substitution.
Before finding the Sylvania document, I thought perhaps that the circuit may have come from a Motorola publication.There was a 2n554 application bulletin published, which I've not been able to find. But I did find an ad (attached) for that bulletin. At least from that write-up, the flip-flop circuit is not included.
(7% distorsion, yay !)
That demo computer described in the PE article also made the cover of Computers and Automation in Nov 1960.
Very brief description on page 20. One of these units was listed on ebay last year, 2 photos attached.
How awesome is that ?
02/04/2017 at 22:40 •
The last log (More bistables) found some explanations of the weird flip-flop schematic I initially found.
This one comes from an even earlier publication (1961) in Popular Electronics, hosted on archive.org :
From the experience with the 10TFF and looking at vintage circuits, I realise something about edge-triggered flip-flops with only 2 transistors. In the 10TFF, a temporary value is held as a charge on one transistor's gate. This principle is found in the 2BJT circuits with actual capacitors, which are "protected" from the previous stage by a series resistors.
This resistor is critical for the speed and power consumption. The RC constant must be adapted to the system's speed. That's where the 2BJT circuits reach their limit... For the "slow" parts of the clock, this is not a problem and this will save quite a lot of germanium parts, but there are quite a few diodes and other passive parts.
01/18/2017 at 23:46 •
I still feel that the divide-by-two diagram in the "details" page is messy.
I tried to address this (rather critical) issue in Easier frequency division but the principle is not reliable enough.
I just stumbled on a better laid out circuit at http://www.smrcc.org.uk/members/g4ugm/Manuals/wirelessworldcomputer.pdf
The old (1968?) book also explains how the whole thing works.
I faintly remember seeing this sort of circuit in one of @Shaos's russian books but 1) I don't read russian 2) the circuit was pretty messy too...
I'm concerned that the OC70 might not be fast enough to work at 18KHz so I must find a more reliable (ECL-based ?) method for the first stages of the predivider.
I'm puzzled by the circuits of that time : they use 3 power rails (-Vcc, 0V and +Vcc). +Vcc seems to be used only as a "pull-up", maybe to accelerate charge dissipation and recover from saturation. I'll have to substitute with Germanium diodes for Baker clamps...
Now, considering that a ECL latch uses about 7 transistors (a DFF uses 14), the predivider would be too expensive with this approach. However, the circuits close to the resonator will work close to the maximum speed of the transistors...
01/09/2017 at 03:25 •
I've been pretty confused by the 2-transistors latch that is shown on the project details page. I've been looking of a simpler, yet compact DFF circuit that uses few transistors.
Today I've done some research on the subject of #CBJT Logic and I came across this PDF : http://orbit.dtu.dk/fedora/objects/orbit:91667/datastreams/file_0b85d918-7d44-40c2-b7c7-a4332b0ce2b6/content
"Figure 5 shows the realization of a triggered bistable circuit with complementary transistors. The set-reset function is here performed by a symmetrical transistor;
if A is 0 volts (true value, "1"), the symmetrical transistor will work as an emitter follower and, by a pulse at its base, set the bistable circuit;
if A is not 0 volts (false value, "0"), the transistor will work as a collector follower and, by a pulse at its base, reset the circuit."
The latching cell contains a OC47 and OC141.
OC141 is NPN germanium, OC47 is PNP (like the OC70 I have).
I am confused by the transistor symbols but I can try to reverse-engineer the circuit...
The right side is the latching part so the base loops to the collector of the other... How do I interpret the OC141 with 4 pins ? Well the right and left are probably the same node...
Creating a latch cell is pretty easy and does not depend on the polarity of the transistor so that's not critical.
What I'm after is a way to change/force the value with the least parts possible and the OC141 on the left does just that. This is the most interesting part !
Let's notice the two capacitors (of identical values: 4.7nF) and the resistor divider (10K, 10K) on the A input.
The capacitor on A stores the data's charge, while the series input resistor isolates it from the source circuit (probably to reduce data leakage while it changes from a simultaneous clock pulse).
The base is clearly (from the text) connected to the resistor divider 22K/2K, energised by the clock signal, through the series capacitor.
The point is clearly how they use a NPN to work as a "pass gate", while the input value is held in the capacitor. The clock capacitor has a similar value so both discharges are simultaneous.
Now, the awaken @esot.eric will notice that when A is low, B is high and a rising pulse appears on T, then the pass transistor is ... reverse biased ? Current will flow from B to the input capacitor (but not A because of the resistor).
Does that remind any Eric of a "almost functioning" circuit with a mistaken transistor ?
Whatever the case, it's very interesting because each DFF uses only 3 transistors, no diode (though I'll add one for the reset), and the circuit can be tuned for other voltage rails. It shouldn't be hard to modify it for an all-OC70 design.
With "only" 3 transistors, plus 2 to drive the outputs, the whole clock system requires something like 39DFF×5=195 transistors. Add some more for housekeeping (oscillator, buffers, drivers, decoders...) and this might reach 250 transistors, which is a desired outcome. The "pass trick" exposed above might be the detail that makes this whole project realistic.
05/23/2016 at 04:10 •
The OC70 is a "lazy" transistor and the 8KHz oscillator is close to its maximum working frequency. Yet this frequency must be divided because the digital counters need "time" to work. I've been naturally thinking of an analog circuit that oscillates at 1/2 or 1/3 of the input frequency, and gets synchronised to the input frequency, which would be the reverse of an "overtone" frequency multiplier.
Well this idea has been studied for a century already, as mentioned in http://www.leapsecond.com/pages/marrison/
This procedure was reversed by Hull and Clapp72, who discovered that the fundamental frequency could be controlled by coupling the high-frequency source directly into the circuit of the multivibrator. This, in fact, is a general property of any oscillator in which the operating cycle involves a non-linear current-voltage characteristic, being most pronounced in those of the relaxation type. Van der Pol and Van der Mark in 1927 reported on some experiments on "frequency demultiplication" using gas tube relaxation oscillators73. The multivibrator is, in effect, a relatively stable relaxation oscillator74, and with slight modification has been used extensively as the frequency-reducing element in quartz-controlled time and frequency standards throughout the world.
One serious difficulty with the multivibrator type of submultiple generator has been that, if the input fails or falls below a critical level, it will continue to deliver an output which, of course, will not hen have the expected frequency. Certain variables in the circuit, such as tube aging, may cause a similar result. With this in view, a general method for frequency conversion has been developed by R. L. Miller75, in which the existence of an output depends directly on the presence of the control input. The basic, idea involved in this, now known as regenerative modulation, was anticipated by J. W. Horton in 191976 but had not been developed prior to Miller's investigations. The circuit of a regenerative modulator in its simplest form as a frequency divider of ratio "two" is shown in Fig. 12.
Fig. 12--Frequency divider for ratio TWO employing regenerative modulation.
So once again, reinventing the clock with old parts lets me discover century-old methods. I have seen a few relaxation oscillators used as frequency dividers and they look more practical than flip-flops, can run faster and use less parts.
It is possible to do a relaxation oscillator by combining a PNP and NPN transistor to create a SCR such as the "SUS" 2N4989:
So it was a very fortunate idea that I got some OC139 (NPN Ge) !
For the Zener, a LED will also work, I suppose ;-)
The other benefit of a relaxation oscillator is the possible cross-coupling with the Xtal oscillator to "help" or "kick" it into oscillation.
My initial idea was more using an astable multivibrator. In fact that's what Injection-Locked Frequency Dividers (ILFD, discovered at Crystal oscillator with MOSFETs (new episode)) do, by modulating the operating voltage of the multivibrator with another transistor. The advantage of a multivibrator is that it uses only one kind of transistors so it will not use my short supply of 0C139.
05/22/2016 at 19:46 •
As time progresses, I (Yann) am turning more toward European sources of vintage transistors. Honestly, I bought the first Germanium transistors (the soviet МП13Б and МП26А) as "double diodes" for the reset signals and only later I thought about using them for switching purposes.
Still, they look like transistors and the performance is not fabulous. I can get away with many quirks but the efforts must be worth it. I want the parts to be and look special, what's the point otherwise ?
So far I think I have enough transistors to start maybe one or two full-scale assemblies, using:
- AF137 and G106T (Telefunken, PNP, 25V, 60mW, β around 60, 35MHz, to be confirmed) as well as AF200 (another 4-pinner)
- AC125V (package similar as above, β>90, 30V, but only 1.7MHz)
- OC70 (black glass, unmarked)
The Telefunken parts look great to design a "numbers processing circuit". I'm not sure what can be done with only 400pc though.
The AC125 are not fast enough for high speed processing but can manage the clock oscillator and prescaler of the clock. Or more. It's germanium-y and can make a nice clock too. The datasheet is pretty complete (if it's really from the same manufacturer).
But the OC70 is the thing for this Germanium Clock project, with its totally odd look and antique characteristics, a blast from the late 50s !
Transition frequency of 15KHz ? Wow that's not fast... So it can only be used for the counter of hundredths of seconds, not closer to the prescaler. I have read somewhere that Mullard's OC70 were used in the first British transistorised computers and digital devices so "it should be ok" and it should suitable for the clock's slow counters.
But the look is just right. If you don't know what it is, you are only left with conjectures:
And this is one of my goals for this clock project : it's really a clock but you have to think beyond your habits :-) It's always good to relearn old things...
Now that I have the parts, I can start the design of the elementary blocks : latch and counters. There are about 40 blocks but I don't know how many transistors will be needed. I don't even know who manufactured these or when. I'll have to test them !
Update: A pair of randomly picked OC70 of this batch can run the Xtal oscillator at 8KHz !
I can better appreciate their characteristics : they are "lazy" with just enough gain but low bandwidth. They can be "boosted" by increasing the power supply voltage though. Oscillation can be sustained at 3V but needs double to startup. The dual-stage amplifier circuit works great and it's important to match the sensor's trimmer to the quartz impedance (the trimmer is set at 16.45K). The other pot is used to center the signal/oscillation in the working range (check with the scope if there is no saturation) and you're done.
What is now missing is:
- Fine frequency tuning (I must have #Hacking a FE-5680B rubidium reference clock working again...)
- sufficiently fast logic for the prescaler (darlington ? non-saturating logic ? higher voltage ? germanium Baker clamp ? sorted gains ?)
BTW, the quartz emits a faint audible whine at 8KHz even when driven with a nice sinewave. Now I understand why 32KHz was chosen, after the first clocks at 8KHz, it was not just a matter of size. Energy gets dissipated through the vibrations and it's proportional to the size. OTOH the very low frequency quartz are large and have a high inertia, which is hard to get moving first, but better stabilises the operation. Conclusion: lowering the amplitude reduces the losses...
04/19/2016 at 01:20 •
I've just received 3 PCB of "Germanium NAND" from oshpark.com
It was drawn in gEDA "pcb" software (images generated by OSHPark service):
and represents this schematics:
Also I got 3rd (1958) and 4th (1959) editions of "Transistor Manual" from General Electric Company with more schematics of computer components utilizing germanium transistors:
04/17/2016 at 01:22 •
More retro stuff received:
And in the quartz it looks like only 3 wires are there out of 7 (some of them were connected together or not connected at all):
It's 2 contacts connected together for 1 side of the crystal and 2 separate contacts on other side of the crystal, so I cut all other wires to keep only 3 important ones...
04/16/2016 at 01:43 •
In the previous episode, Reverse-engineering vintage quartz resonators, we had a lot of questions and a few hints. The best way to be sure was to try them.
I took the previous board (see Crystal Oscillator (Germanium Edition)) and removed the 32768Hz Xtal. Instead I put a 3-pins socket, with ground in the middle. I put the 2 crystals (in succession) on a connector and tested them.
The wristwatch crystal needs quite specific parameters to work correctly.
The russian crystal just worked, oscillating immediately with the parameter of the tiny crystal. Oscillation persists even with both trimmers cranked up to the max (47K Ohms) at 1.5V (though startup was slow). But even then, it just worked, using low gain transistors (hFE=37).
I feared that operating it would be difficult, like requiring more driving energy because of the larger size (higher motional inertia) and other factors. But the tube just "rings". No I can't hear it but it picked up the oscillations much better than the tiny tuning fork, requiring less energy.
I suppose that the lower frequency also contributes to this sensitivity, since the transistors are not ultra-high performance. So the rough MP13B might work as well. A recent planar transistor will fly. I should still investigate the MOSFETs...
One surprising result was that it works better when the "large electrode" is on the driving side, leaving the small electrode to the sensing side. I haven't tried all the 6 combinations/permutations of the pins yet, though, but this would be a very interesting experminent. But first I need a tool/measure to measure the gain, other than the time taken by the oscillations to reach full range.
Whatever the result, my bet on this tube has been a clear win, on every account. It costs more than a dumb wristwatch crystal but it requires less driving energy so it's energy-saving. It looks much better, might have a better temperature stability and it's pretty unique...
We'll see soon how it behaves and how it is tuned.
By popular demand, here is the updated schematic :
R2 and R4 have been increased because the signal saturates fast, and saturation is not good. Actually, the best is to have a small sine, less distorsion and better accuracy.
Due to the higher sensitivity of the crystal, R1 and R3 can be increased as well.
I will have to evaluate the effect of leakage, thus of temperature-induced drift.
To compensate power supply voltage drift, I'll probably add a micropower voltage regulator. No Zener diode because of the tempco and the worse regulation ratio. This is why I try to make it work at a lower voltage than the rest.
04/15/2016 at 16:43 •
The seller provided the following picture but even after receiving the crystals in tubes, many questions remain.
First, is it series or parallel resonance ?
Then what type of cut is it ? https://en.wikipedia.org/wiki/Crystal_oscillator#Crystal_cuts has many choices but many are vague. I can already filter with the frequency range (when provided). The shape of the crystal cut and the plating/electrodes reduce the possibilities further : the oscillation mode is obviously "bending", like a vibraphone bar. From https://upload.wikimedia.org/wikipedia/commons/8/8b/Crystal_modes_multilingual.svg:
The 4 connections are at the "immobile" points of the crystal, to prevent damping, at 1/4 and 3/4 of the length of the crystal (though it looks like 1/6 and 5/6 to me in my tube).
The possible cuts are
- XY (though this fabulous article would not agree)
The electrodes prevent most harmonic/overtone vibration modes because the "middle" electrode (2/3) short-circuits opposite charges. So the frequency and waveform/shape would be quite pure and stable.
The "passport" that @[skaarj] translated (in the comments) mentions -80°C to +80°C operating conditions, which are another hint for the cut. It seems to be symmetric around 0°C, which is unusual. Finding (or measuring???) the temperature response curve would help even more. Does anyone have a climate testing chamber ?
Initially I feared that the frequency stability would be poor but it might be the reverse, this will have to be measured and I suppose that the germanium leakage will be the greatest source of variations. I'll probably have to find/imagine a balanced/symmetrical oscillator circuit using matched pairs...
One more hint would be the internal connexions. Some electrodes are connected to structural elements, hence have higher capacitance. Some output wires are connected in parallel with the structure, as well.
http://www.ieee-uffc.org/main/history-marrison.asp contains a very interesting hint at fig.8 :
(Wait: what is this "bridge-stabilised oscillator" ? I have to look that upThe two bottom circuits use the classical 2-electrodes quartz but the top one has 4 electrodes, more like my tube !
Update: ok, no, it works in series resistance resonance but I don't know if the tubes are rated for series or parallel mode)
I suppose this 4-electrodes crystal works in a different cut/mode but from there, my latest guesses are confirmed and I suppose the following:
- short electrode (1/3) for driving the crystal (with enough drive, but smaller surface if compensated by high gain circuits)
- long electrode (2/3) for collecting the charges and providing feedback
- opposite side electrode (full length, 3/3) is grounded (otherwise, relative to what point do we measure the charges ?)
Or maybe swap 2/3 and 3/3. I'll have to test that... After all there are only 6 combinations :-D
- The quartz bar is 30mm long
- contacts are at 7.5mm from each end
- The short electrode is 8.3mm long
Things add up nicely :-) 7.5×4=30mm, so the oscillation is in bending mode, perpendicular to the electrode platings. Contacts are placed at the point of least motion, to prevent attenuation, at 1/4 and 3/4.
The puzzling part is the 8.3mm of the short electrode, I don't know why this length. The ±3.6 ratio does not correspond to an obvious mathematical relation. It's close to 3.3 and has misled me but maybe the measurement is not precise enough.