Kundo Electronique Motor Repair

Main Index

Clocks Index

This was the first article in the horology section of the ESP site, and describes the repair of a Kundo (Kieninger and Obergfell) electronically switched electro-mechanical clock.  The material here was finally updated in April 2014 (one can't rush these things after all).

To the left is the circuit diagram and the movement of a typical Kundo electric (Electronique) clock is shown on the right.  These clocks use a pendulum that is kicked into continuous oscillation by a very simple circuit.  The basic schematic is shown, but note that there are many variations.  The circuit shown is the original version that I was unable to use after repairing the 'motor' coils, because the transistor was faulty.  The transistor used is PNP (as are most common germanium transistors).  These clocks always used a germanium transistor, and it is thought by many that silicon transistors simply won't work.

This is not necessarily true, as I have replaced the germanium transistor with silicon in the once faulty Kundo movement, and it works just fine - albeit with some fairly pronounced caveats.  Unfortunately, the original coil was open circuit as well, so the clock required a complete rewind of the coil.  This is not a fun job with 0.0635mm wire (#42 AWG/B&S), as the wire is so fine that it is difficult to avoid breakage - especially with several thousand turns.  Nevertheless, the movement (pictured on the right) now works very well, and only awaits a case, face and hands (all of which were missing when I acquired it ... the hands you see are only temporary!).

Although the original did indeed use a germanium transistor, I replaced it with a BC549 silicon device after rewinding the coils - mainly to find out if it would work (the battery polarity must be reversed if an NPN transistor is used).  Although fitting the circuitry is fiddly, I also tried an AC128 germanium transistor - bad move!  As I discovered (after reassembling the motor), the cases of the AC128 transistors I have are steel (not aluminium as I had assumed) and are of course magnetic.  This causes the pendulum to deflect alarmingly.  The original transistor used a black painted glass case - these were common in the early days of English and European transistors, but are now unobtainable.

There is no doubt that the circuit works much better with a germanium transistor (ignoring the pendulum deflection which makes it unusable), but I don't have any of the old glass case types (such as the original TF65 pictured below the schematic), so I reverted to the silicon device for a while.  The drive used now is external, because a silicon transistor is far too temperamental to use in such a simplified circuit.

Since the movement is a 6 jewel type and is of good quality, it is actually worth the effort of restoring it IMO.

I did take a few measurements when I first used the silicon transistor.  These clocks are real power misers - at least with a silicon transistor.  The measured impulse duration is only 78ms, and with a peak current of 1.7mA, the average current drain is less than 200µA.  the power is approximately 300µW ... yes, microwatts.  The pendulum rate is 3 beats/second (1.5 complete swings per second).  The pendulum is pulsed once per complete swing (every 667ms).  The direction that the pendulum is impulsed depends on the coil and magnet directions - it may occur on either left-to-right or right-to-left swings.

A magnet is buried inside the curved section of the pendulum, and that induces a tiny current into the base of the transistor via the outer coil.  Once the transistor draws collector current (the red dot shows the collector), the current in the inner coil induces more current into the outer coil (connected to the base), turning the transistor on more.  This continues until the current can increase no further, at which point the transistor turns off completely (and rather abruptly - the 5.1k resistor is fitted to limit the maximum back EMF voltage swing which otherwise can easily reach 20 or 30 Volts).  The impulse repels the magnet, pushing it away.

Clock Motor Components

Coil Bobbin Detail

Transistor Installed

Above, you can see the components of the clock's motor system.  The curved section of the pendulum houses a magnet (seen peeping out from the right side).  The coil bobbin holds the circuitry and the coils.  The coil itself is seen in the centre image.  I colour coded the wires when I rewound the coil so I'd know which lead went where.  Unfortunately, the thinnest coloured wire I had available is still much too thick, but it works.

The transistor mounting is shown on the right.  The hole is designed to take a germanium device (same diameter, but much longer than the silicon transistor used as a test).  Wiring is laid along the channels in the bobbin, and there is actually plenty of room.  The 5.1k resistor isn't shown, but it is tucked into a relatively large cavity in the bobbin.  Almost a year after having repaired the motor, as you can see I haven't progressed very far, and had forgotten the exact wiring (the circuit is so simple, of course I'll remember it ... or so I though at the time).

Power is transferred to the movement by means of a tiny pawl that is connected to the pendulum, and this drives a ratchet wheel with a detent to prevent it from turning backwards as the pawl moves from left to right.  A photo of a typical Kundo drive system is shown on the left.  Unlike a traditional weight or spring driven movement, the motion is geared down by the motion train, with each successive wheel turning slower than the one driving it.  Only a tiny amount of the pendulum's energy is lost at each right-to-left swing which advances the wheel, so the electric 'motor' unit only needs to replace that amount of energy to keep the clock running.  No power switch is provided or needed.  Simply locking the pendulum in place prevents the battery from being discharged, since the circuit needs the magnetic impulse to do anything at all (although this may not be the case with a germanium transistor, because they have relatively high leakage).

A common mistake is to assume that a battery (or 'cell' to be exact) will last a long time without any current drain, but they will eventually leak and corrode the battery housing.  Standard zinc-carbon cells (which were described as 'leakproof' for a while) are marginally better than alkaline cells - these will leak if left in place, and the damage can be considerable.  If a battery clock is not being used, remove the battery.

A small amount of damping is commonly provided to prevent excessive pendulum over-swing.  This is provided by a brass ring (the mounting bracket for the coil, and the centre brass sleeve inside the coil bobbin), which forms a shorted turn for the magnet as it passes through.  Unfortunately, the centre sleeve of the repaired coil assembly had been damaged before I got to it, so it is incomplete - hence, the system presently does have a bit more power than is needed (all 300µW of it).

It is generally recognised that a pendulum should be driven when at the centre of its swing (when vertical and at maximum velocity) for best timekeeping, and the Kundo style of motor does this quite well.  The drive is delivered as the (inner) end of the magnet passes through the coil, and it is quite close to the centre of the curved bar.  Timekeeping seems to be quite good once the clock is regulated properly.

I have since had to perform another coil rewind on another Kundo clock using the same motor unit.  This time, I used the counter on my coil winder to count the number of turns.  The base (sensing) coil has 5,000 turns, and the collector (power) winding has 3,000 turns (give or take a few).  The second repaired unit uses the original TF65 germanium transistor, and refused to work with silicon.  The only real difference is the magnet strength - the first one I did has a much stronger magnet, although in theory, the second one should have worked with a silicon device regardless.

One thing that I was able to do this time was determine the likely cause of the coil failure.  With such a small amount of power and with very low voltage at all times, these coils should last forever, but this is obviously not the case.  It looks like the insulation for the coil termination wires leaches some chemical that attacks the wiring.  There were several patches of green gunge on the windings, right next to where there was contact with the insulation.  Unfortunately, any break that occurs will never be where you can get to it, so this makes it necessary to just remove the old coils completely and start again.  With both motors, I attempted to remove just the outer (faulty) winding, but damage to the inner winding seems inevitable in the process.  It might be possible for someone with close to infinite patience, but that's not me.

Please Note:   I have been advised that some of the early PVC insulation contained plasticisers that may have used PCBs (polychlorinated biphenyls), and that the green gunge referred to above may contain some of this toxic and carcinogenic material.  Care is advised - wear disposable gloves and do not attempt to solder through the gunge.  In quantity, waste material should be disposed of at an approved toxic waste handling centre, however this might be overkill for a few milligrams of material.  Ensure that the waste material is well wrapped before disposal.

For easy reference, here are the coil details (as close as I can get them at least - if anyone has more exact info, please let me know).

Coil	Turns	Location
Sense	5,000	Inside (wound first)
Drive	3,000	Outside (wound second)

Wire size is approximately 0.0635mm or #42 AWG/B&S for both coils.  Be careful, because such fine wire is very easily broken unless your coil winder has a very steady speed.  With so many turns, it's tempting to increase the speed of the winder, but that greatly increases the chance that the wire will break.  At around 60 RPM (1 turn per second) it will take 1 hour and 23 minutes to wind 5,000 turns.  This is a good speed if you are very patient, but I expect that most people will get bored rather quickly, and will run the winder faster that that (I know that I did).  )

Several other (mainly German) manufacturers made somewhat similar clocks at around the same time period as the Kundo.  One of those is Junghans, and the drive circuit is shown below.  Unfortunately, I can't verify this as being the exact circuit used, as there's very little accurate information available for these clocks.  There are a couple of different versions around, but this appears to be at least sensible.  It might be possible to modify a Kundo with an intact coil but faulty transistor to use this circuit, and this is offered as a suggestion only.  I cannot verify that it will work, but since you can't get germanium transistors in anything other than a steel case, it's worth a try.  Expect to have to experiment though!

Junghans Single Transistor Drive Circuit

One characteristic of all of these early clock motor circuits is their frugal use of power.  If the pendulum isn't swinging, no impulse is delivered to the transistor, so it remains turned off.  That means that the only current drawn is the leakage current of the transistor.  We ignore this with silicon devices, but germanium transistors can have a few microamps of leakage current, even when turned off.  They are still very frugal though, so a 1.5V cell will normally last for around one year.

As noted above, the transistor action is regenerative, so when the transistor starts to conduct, the coupling between the sense and drive coils turns the transistor on harder.  This continues until the current stops changing, as electromagnetic induction requires a change of current flow.  Once the transistor is 'fully on', it promptly turns off again, as there is nothing in the circuit to supply any base current.  Turn-off is also regenerative, so the transistor turns off quickly.

A significant part of the problem with the silicon transistor is that silicon has a 'barrier' voltage of nominally 0.65V, whereas that of germanium is around 0.3V.  Silicon also has a negative temperature coefficient (about 2mV / degree C), so when the weather gets cold the single silicon transistor drive simply stops.  The negative tempco means that as temperature falls, the barrier voltage is increased (and vice versa of course).  The same thing happens with germanium, but is less of a problem because of the lower barrier voltage.  Germanium also has a significant (and comparatively high) leakage current that is highly temperature dependent.

There are two alternative motor designs that are shown below.  They have the advantage of using readily available silicon transistors, and the electronics are mounted outside the coil housing.  This makes it easy to experiment, as the coil assembly doesn't have to be dismantled to make changes.  Some electric clocks using similar switching systems already used just a single coil and an external board with the electronics, and since everything is easily hidden it doesn't affect the appearance at all.

Only one winding is used, so only two wires are needed from the coil housing.  A circuit that is (very) similar to the one I'm using now is based on one described in the Free Pendulum Clock article, but there are a few minor changes needed as the first motor drive I built is much too powerful for the Kundo clock, and must be tamed.  This must be done in a way that gives consistent results, yet uses a very simple design so it's easy to build without a printed circuit board.  This is a challenge as it transpires, but I know it can be done (see below).

Since the single silicon transistor in the original circuit is too temperamental and I have no glass (or aluminium) cased germanium devices available, I reverted to a two transistor driver.  After several attempts to get it to work properly, I eventually settled on the motor drive circuit shown below.  It needs 3V to run, but that's no great hardship.  The drive is consistent, and it seems to be a good overall solution.  Using an external switching circuit also makes a coil rewind far easier, as there is no need to use two coils and rewire to the fiddly internal circuit.

Two Transistor Drive Circuit, Version 1

This circuit only needs a single coil, and mine measures about 1k resistance.  I'm unsure how many turns are used, since the coil was re-wound long ago and I've forgotten, but I think there's around 5,000 turns or so, as I seem to recall that I used the inner (sense) coil from the previous rewind.  The parts aren't critical, and the circuit can be built on a piece of Veroboard or similar prototyping board no more than about 25mm square.

The circuit provides a small bias voltage to the first transistor via R4, and in conjunction with R1 there is about 130mV at the base of Q1.  Not much, but enough to ensure that the small pulse from the coil will turn on Q1, thence Q2, and send a drive pulse to the motor coil via C1.  The average current drain is just over 200µA as shown, and it will run for a very long time on a pair of alkaline 'C' cells.  It flatly refused to run with 1.5V, so 3V was the next best thing.  I could have used germanium transistors instead of silicon, and that would probably improve matters, but I don't like using germanium if it can be avoided.

I experimented with a few different circuits before settling on this one.  The first drive circuit I used was similar, and was based on the circuit shown as Figure 9B in the Free Pendulum Clock article and shown below (with a modification).  Despite my best attempts to reduce the coil drive, it still managed to provide way too much power to the coil, so the pendulum would strike the coil assembly.  So far, the circuit shown above seems to be the best, but only time will tell (pardon the pun). 

Two Transistor Drive Circuit, Version 2

R5 is shown as 'Select On Test', and should be able to reduce the drive power to ensure that the pendulum swing is within reasonable limits.  As a starting value, I'd look at perhaps 560 ohms or so as shown, but it might be as low as 100 ohms or as much as 1k.  This is theoretically a better circuit, as it's more sensitive, can provide more drive, and doesn't always need a biasing resistor (R4).  R4 can also be used in this version if it refuses to be triggered (but at a higher value of 180k up to around 1Meg - again, select as needed).  It can also run from 1.5V, so only needs a single cell.

Unfortunately, none of this is an exact science because the coils may differ depending on when the clock was built, and/ or if it's rewound.  The magnet strength also has a significant effect, and if it's weak you need more drive power and a more sensitive circuit ... or you can 're-charge' it with the Magnet Charger also shown in this section of my site.  However, it would be a big outlay for a single magnet.

I really don't recommend using a car battery to power the coil for recharging a magnet, but it does work if you are extremely careful and promise not to complain to me if you injure or burn yourself, or damage the battery (and yes, I am deadly serious ! ).

Main Index

Clocks Index