Back when I first became interested in electronics, a simple way to make an oscillator was a circuit called an astable
multivibrator. These got used a lot - the electronics magazines that I used to buy as a teenager were full of them. This
is the form of it
The two transistors are arranged like a flip-flop, with either one or the other conducting. Unlike the flip-flop, where
the conducting transistor will hold the other off, here the additional capacitors mean that after a period the stable
state ends and the circuit will flip (or, maybe, flop) of its own accord.
One of the disadvantages of such an oscillator is that the transistors saturate and there was quite a lot activity from
people trying to come up with alternative designs that would oscillate faster. Another disadvantage was that the
waveforms at the collectors weren't all that good. For the speed side, obviously changing to circuits where the transistors
didn't saturate was the way to go. For the waveform shape issue, several people came up with circuits which had two
transistors in series. This circuit I'm looking at here is by C. F. Ho. I came across it in a Ferranti transistor applications manual from the 1970s.
Here it is in the manual
and here it is in the simulator with meters attached to measure things
At first sight, the circuit looks wonderful; symmetry is often very good in circuit design and this looks like the bee's
knees of circuits (Bat's hats? Cat's corsets? Dog's wotsits?), but appearences can be deceptive as we'll see in a minute.
Note: the component labelled 'IC1 2' isn't a real component and isn't needed for circuit operation, it's an initial
condition for the simulator. Effectively, it gives the circuit a 'kick' at start-up. You'll often find something like
this is necessary to get an oscillator going in simulation (the simulation doesn't have the noise and transients that get
real circuits started).
If I run it in the simulator and plot the charging waveforms of the capacitors, we'll immmediately see one issue with it
VM3 and VM4 are plots of the voltage across each capacitor and you can immediately see that they charge at the same time.
The period of the pulse is then determind by the discharge time.
Nice properties are that there are two outputs - the output can be from either collector - which are complementary (well,
it's a nice property if you have an application where you need two complementary outputs, one of which comes down from
the rail and the other goes up from ground) and the edge rate when it switches is quite good too because there's regeneration
round the loop ("Talkin' 'bout my regeneration" - The Who might have been really big if they'd had me to do the song lyrics
for them) which is where the improvement to the waveform shape comes from.
This is a close up of one edge for the two complementary outputs. The other two traces (the green and red ones) are the
voltages across the two timing capacitors and at this scale hardly seem to be moving.
The horizontal scale is 200nS per division. The edges really move once they get started and the lower one would be fine
for driving 5V CMOS logic. (Ok, perhaps not; who wants a 5V logic system that then needs a 9V supply for the clock?)
Less nice properties are that is sensitive to component values. I found that it was quite fussy about the base resistor
values and if you get it wrong the circuit just sits there and sulks. That in turn limits the charging current so, for
high speed use, the capacitors start to get very small in value.
I don't know if anyone ever did anything with this - perhaps Mr Ho had a specific application for it - but it's an
interesting curiosity to look at with the simulator.
Since the current through the two transistors when they are both on is something like you might use to light up a
reasonably efficient small LED I thought I'd try doing just that. Simply placing the LED between the two transistor
emitters like this suffices.
I can slow down the timing by simply massively increasing the capacitor values; changing the capacitors to 10uF gives a
suitable timing interval between the flashes and the discharge time is just right for the flash time. It's a natural for
operating off of a 9V battery, too - it will happily work down to 6V or so. Now the low current during the charging
period is something of an advantage; here's the current draw running in the simulator - the bulk of the draw from the
battery is only when the LED is lit
One disadvantage is that the charging current is so low that if you used poor quality or old electrolytics you might find
the leakage was of the same order as the charging and it would never get to the switching point.
Anyway, enough theory, you want to see it working. Here it is wired on a breadboard
and finally, here is a video of the LED flashing (because the world doesn't have anything like enough videos of LEDs
flashing - but at least it wasn't an Arduino this time).