40W continuous or 50W for short periods at 240V AC is provided by this inverter from a 12V DC supply.

Continuing from the other articles on this site about vibrator power supplies, the inverter described here is a practical example of this method of DC voltage conversion.
Whilst most vibrator power supplies are used for DC to DC conversion; e.g.; for a  car radio, there is nothing to stop one using the AC at transformer secondary directly. With the right turns ratio, it then becomes possible to provide AC mains voltage from a battery.

Background.
This inverter actually goes back to 1996. I built it from scratch, including the aluminium cabinet. The inspiration largely came from a Radio & Hobbies article describing such an inverter in April 1951 and January 1957. The R&H design was based around a Ferguson VT146 vibrator transformer and an Oak V6606 or V6612 vibrator. By configuring the transformer primary connections and selecting the appropriate vibrator, the design was suitable for 6 or 12V. Unfortunately, the VT146 transformer being an uncommon item, meant that a substitute would be in order. From previous experience, I knew that an ordinary 9-0-9V to 240V power transformer would be suitable, and so an A&R transformer of about 60VA rating was modified, by having what was originally the secondary, rewound for 18V centre tap. It so happened that I recall seeing a VT146 for sale back in the mid 80's at ACE Radio in Marrickville, but by the time I got around to building this inverter, that business had long since disappeared.

Unfortunately at the time, I knew far less about vibrator power supplies than I do now, and while the inverter worked, the vibrator was not being run under the best conditions. It appeared that the transformer did not have enough inductance, and some difficulty was had eliminating sparking at the vibrator contacts. Furthermore, I did not really know how to select the correct timing capacitor.
During one of the times this inverter was used in a car to power a CD player and one valve amplifier, the vibrator had partially come out of its socket. This resulted in only the contacts on one side working; i.e. it was working in half wave. The result was damaged contacts.

The contacts on the left side have been overheated from half wave operation.

Some years later, having learned the virtues of toroidal transformers, the inverter was rebuilt using an Altronics M5109 80VA toroid. This seemed to work better than the previous laminated transformer, but still the inverter didn't really work as it should, and I sensed the vibrator operation was still not as it should be. I still wasn't aware of the finer points of vibrator power supply design, and the inverter was put into the too hard basket.

And so the present day, with a much greater knowledge of vibrator operating conditions, it was time to finally fix this inverter. In this article I will explain step by step how the new inverter design came about.

Design Begins.
First thing to do is look at the vibrator characteristics. As I wished to use an ordinary radio type of the highest power, this narrowed down the choice to an Oak V6612. This is a dual-interrupter type with a UX-6 base. Oak vibrators have been described here.
The contacts can be wired in parallel, but for better current sharing are used singly with two identical primary windings. 12V Oak non-synchronous car radio type vibrators have a contact rating of 4A. To use one of these (e.g., V5123) limits the input power to 4 x 12V = 48W. In view of transformer efficiency, this results in about 30W for the maximum continuous output power. I have built many inverters of this kind for low power devices with excellent results, but as I have a collection of dual interrupter types, it makes sense to use one and have a more powerful inverter.
Oak radio type vibrators operate at 100c/s, which means any appliance plugged into the inverter will be operating at twice the normal frequency. In many instances this is not problematic, but this, along with the non-sinusoidal waveform, does restrict the types of appliances suitable.

Data for the Oak/MSP dual interrupter vibrators.

Dual Interrupter Vibrator.
Dual interrupter vibrators are manufactured identically to ordinary synchronous types, in that there are two sets of contacts; one for primary switching, and the other for secondary rectification. One may assume these could be paralleled to increase the ratings. However, with the synchronous type, the secondary contacts are designed to close just after the primary contacts close, and open again just before they open. This means while the contact current rating can be doubled while they are actually closed, the actual switching current rating is still the same as one set of contacts.
Nevertheless, where a synchronous vibrator is used in a non-synchronous application, it is worthwhile paralleling the unused contacts because they still make a useful contribution.

The dual interrupter vibrator has both sets of contacts adjusted to close and open with the same timing. While this may appear to double the rating, in actual fact it is somewhat less. This is because it is impossible, in practice, to have both sets of contacts open and close at exactly the same time, all the time.
Looking at the data above, the V6612 maximum input current is therefore only 4.5A. This applies to each set of contacts switching its own primary winding. In other words, the transformer has two separate centre-tapped primaries. The reason for doing this is to provide more equal current distribution between the contacts. It should be clear that if one set of contacts has a slightly higher resistance than the other, then most of the current will flow through the lower resistance set. A further improvement in current sharing can be obtained by using two separate identical transformers with their secondaries paralleled.
Another limitation with regards to vibrator current is the input voltage. It will be noted that the current rating falls with higher voltages. This is because contact arcing occurs more easily as the supply voltage is increased.

With that information in hand for our 12V inverter, we are limited to 4.5 x 12V = 54W output, assuming 100% efficiency, and two primary windings. For intermittent duty, the current rating is 80% greater, or 8.1A. This would allow an output of 97W. The definition of "intermittent" is not given, but it would no doubt be something along the lines of two way radio transmitter use.
The transformer used with this inverter has only a single centre-tapped primary so this would de-rate the contact switching current. However, it is a toroidal type and this has much greater efficiency than the conventional E-I laminated type normally used.
On that basis, given that there is only one primary winding, it would probably be safe to specify 40W for the continuous output power rating, with 50W drawn for short periods.

The V6612 type has a 12V driving coil. It is quite in order to use the 6V type V6606 on 12V by means of a 27R 5W resistor in series with the driving coil.
There are of course other more powerful types of vibrator available such as the 50 cycle Van Ruyten types which can provide 100W. These are much larger and have four sets of contacts. They were used in some Australian made inverters from the 1950's up until the early 70's. Another option is the Electronics Laboratories high power vibrators designed for AC inverter use. Inverters using these vibrators are described elsewhere on this site.

Transformer.
It might be assumed that one just uses a 12-0-12V to 240V transformer in reverse. Not so, and sadly many inverter circuits appear on the internet doing just this. Their designers have obviously never tested their circuits properly, because they'll find the output voltage somewhat lower; about 180-200V instead of 240V.
First thing to consider is turns ratio. One would assume that a normal 240V to 12V transformer has a turns ratio of 20:1. In the ideal world it would, but given losses in the transformer, it has to be slightly less than this. The "12V" output is at the transformer's rated current. Run with no load, the voltage may be something like 14V, which means a turns ratio of 17:1.
Now, what happens when this transformer gets used in reverse? Assuming 12V is fed in, 204V comes out. And, that's not taking into account things like transistor saturation voltages and the supply voltage drop between battery and inverter.

For vibrator inverters, another factor comes into play, and that's the dead time between the contacts opening on one side and closing on the other. The transformer is fed with no power during this dead time. For Oak vibrators, the duty cycle is 80%.
In practice, a 9-0-9V : 240V transformer is required to provide the correct 240Vrms output with a vibrator.  I have found over time that some conventional 240V transformers can often a poor performer with vibrators. Some do work well, but some don't. This is evident when the idling current is higher than it should be, along with difficulty eliminating contact arcing, and the inability to obtain the correct vibrator waveform. This is mainly a function of winding inductance and lamination design, causing a high magentising current. This can be particularly problematic when ordinary transformers are used in reverse. Transformers designed for vibrator power supplies do not have these problems.

In the modern world, I have found toroid transformers to be excellent substitutes, and possibly even more efficient than a genuine vibrator transformer. To illustrate this, I did initially use a conventional 240V to 18VCT  laminated transformer for this inverter, but magnetising current was about 1A, and it was impossible to eliminate the contact arcing completely. Changing to an 80VA toroid (Altronics M5109) dropped the magnetising current to an insignificant amount with no arcing, but from time to time the output waveform was unstable.
Having a 50VA transformer (Altronics M5009) available, I tried this and was immediately rewarded with results as they should be.

Basic Vibrator Circuit.

Any vibrator power supply has three inter-dependent components; the vibrator, the transformer, and the timing (buffer) capacitor. The transformer must suit the frequency and duty cycle of the vibrator, and the timing capacitor is determined by the characteristics of both of those. For this reason, components in a vibrator power supply must not be substituted without checking the operating conditions.
The purpose of the timing capacitance is to reverse the transformer primary voltage, when the contacts open, so that the voltage at the contacts is at minimum, at the time when they re-connect. It is at the point when contacts are actually closing that they are most vulnerable because of imperfect connection, and thus sparking is likely. Once closed, however, much higher current can flow without harm. Again, when the contacts start to open, breaking a high current is apt to cause sparking. The object of the timing capacitance, therefore, is to ensure the transformer primary current flows after the contacts have closed, and ceases just before they open. This also prevents point material contact transfer. In a DC circuit where switch contacts are in use, material from one switch contact will eventually migrate to the other. A common example of this is the Kettering ignition system used in cars; one contact of the points will eventually become pitted, while the other contact will build up material in a mirror image. The rough contact surface so formed causes erratic performance.
In the case of a vibrator power supply, transfer of contact material is avoided by correct timing capacitance.
It must be stressed that the timing capacitance is carefully selected, and is not merely "whatever value that stops sparking".

Having selected the vibrator and transformer, it's now time to find the correct timing capacitance. The ideal waveform as seen across the entire transformer primary was obtained thus:

Ideal and practical waveforms for selecting buffer capacitance. Slight vibrator unbalance shows up in these waveforms.

It is possible to calculate the ideal timing capacitance and the formula is given in the Mallory publication available here.
However, it is considerably easier to simply set up the vibrator and transformer and use a decade capacitance box instead. For the V6612 vibrator and M5009 transformer, the ideal timing capacitance is 0.003uF. For the ideal waveform, the slope during the dead time is straight and continuous between the voltage peaks. The ideal value of capacitance also coincides with minimum transformer primary current.
In practice, slightly more capacitance is desirable because of vibrator tolerances - being a mechanical device and adjusted by humans in the factory, there is liable to be slight differences in the finished product. Since anything less than the ideal timing capacitance is very undesirable, because of high peak voltages that might appear in the circuit, one must err on the side of more timing capacitance to cover any likely example of the vibrator type chosen. Furthermore, there is the question of changes in vibrator characteristics over time. Contact gap (and thus dead time) will increase if the contact arms are made of poorly chosen material which deforms under the constant hammering of the contacts. It will also increase if the contact material erodes due to poor design, such as inadequate surface area for the current being switched, or just poor choice of material.
For this reason, there is a "practical" value of timing capacitance; the details of which are described in the above Mallory literature. Essentially, the timing capacitance is increased so the sloping portion becomes straight about 2/3 of the way down the waveform.
Here, the capacitance value determined was 0.006uF. Depending on transformer characteristics, some ringing may be evident in the waveform, and this can be removed by including a damping resistor in series with the buffer capacitance. This can be selected experimentally by choosing a resistor that just eliminates overshoot without rounding off the waveform. For the above waveforms, 525 ohms was found to be the ideal value.

At this point, we have matched the vibrator, transformer, and timing capacitance, and the AC from the transformer secondary can be fed into any full wave rectifier to provide B+ for valve circuitry. To make a practical power supply, other components will be required for filtering out interference both at audio and radio frequencies. That will be entirely dependent on the apparatus powered by the vibrator unit.

Using the vibrator test panel to observe operating conditions of the vibrator and transformer.

Use for an AC Inverter.
The transformer secondary can of course be used to power AC loads such as domestic appliances. Unfortunately, our "ideal" design is no longer anything of the sort. The problem is the variation in loading with different appliances. A simple resistive load such as an incandescent light bulb or a soldering iron stops the timing capacitance doing its job to a lesser or greater degree. This is because the transformer is still loaded during the dead time, and thus the timing capacitance discharges much faster than it normally would. Effectively, the circuit is operating with insufficient timing capacitance. Despite the lack of timing capacitance, high peak voltages are prevented because of the loading, and so the inverter can be used this way. The catch is that the vibrator contacts will now be switching the full current right at opening and closing time. For this reason, vibrator life in an AC inverter might be shorter than in a DC-DC converter.
Then there is the question of low power factor loads. Mostly these are inductive, such as induction motors or fluorescent lamps using iron cored chokes. Capacitive loads include devices where a capacitor is used as a voltage dropper - for example LED lamps for mains operation. Also some appliances might have capacitance across the supply for mains filtering.
In the case of inductive loads, the timing capacitance across the inverter transformer secondary is effectively reduced, causing vibrator contact arcing. It is of course possible to increase the timing capacitance to compensate, but then the inverter can only be used with this load.
For capacitive loads the inverse applies; there is now too much timing capacitance. There may still be too much timing capacitance presented by the load even if all timing capacitance is removed from across the transformer secondary.

Compromise Design.
In practice, because this inverter provides 100c/s output, it is unsuitable for appliances using capacitive droppers anyway. Induction motor loads are unsuitable for the same reason.
Fluorescent lamps are also not suitable because the choke will have a higher reactance at 100c/s and the lamp will be dim, assuming it can start at all.
As it happens, loads which are not frequency dependent tend to be resistive anyway, or they rectify the mains (e.g. switchmode power supplies) and so only draw current at the voltage peaks.
For other electronic loads where a power transformer is used, some inductance will be added, and this will need to be allowed for in the inverter design.

Normally, the "ideal" or "practical" value of timing capacitance suits resistive loads - it ensures the vibrator runs under ideal conditions with no load. Perhaps because of some characteristic of the transformer used, it was noticed that some contact sparking was evident when the circuit was loaded by a 40W incandescent light bulb. It was necessary to increase the timing capacitance up to about 0.2uF to eliminate sparking. Of course now, if the inverter was run unloaded, the capacitance was excessive. For this reason, it is hard to specify a value of timing capacitance for an AC inverter that suits a wide range of loads. It has to be a compromise.

Loaded with a 40W light bulb, it can be seen the 0.006uF "practical" timing capacitance now has no effect. The bottom waveform is the current waveform at the 12V input.

Electronic Laboratories Patent.
The question of low power factor loads with a general purpose vibrator inverter is interesting. In my vibrator research, I discovered a patent from Electronic Laboratories. This company specialised in vibrator power supplies, but particularly AC inverters. One of their products is described here.

Essentially, the patent (U.S. 2086323) states that for inductive loads the timing capacitance must be increased to compensate. However, this value is then too high for other loads, and the problems of excessive timing capacitance become evident - one being that excessive current flows in the secondary circuit with resultant sparking at the vibrator contacts. E-L's idea was to retain the value of timing capacitance which suits inductive loads, but then by means of a high resistance transformer secondary, reduce the current flowing in the timing circuit to an acceptable level when used with non-inductive loads. The patent mentions that ordinary resistors could be used instead of a resistive secondary winding. One thing not discussed was that a resistive secondary would have to cause poor regulation with regards to output voltage.

E-L's idea was that including resistance in series with the timing capacitor allows a higher value of capacitance to be used. Note that this is not the same as including a damping resistor for the purpose of removing overshoot, even though both circuits are schematically the same.

Now was a good time to try this theory out. Having decided that 0.236uF was going to be the minimum timing capacitance (this value comes from two series connected 0.47uF capacitors as will become clear later), which stops sparking with the inverter fully loaded, it was clear that the vibrator was not happy when the inverter was unloaded. The input peak current was measured at the 12V feed to the primary centre tap to be in excess of 8A.

By introducing resistance in series; "R" in the above diagram, the peak input current was brought down to about 5.3A and the vibrator was happy. It certainly seems E-L's idea has merit. In fact, the inverter was docile over a much greater range of loads than with the conventional circuit - even a small desk fan did not cause any contact sparking despite being a fully inductive load. A suitable value of "R" was determined to be 1.2K.

No load and 40W load waveforms with the modified timing circuit.

With a 0.236uF timing capacitor and 1.2K resistance, the input current is kept at around 5.4A peak with no load. Because of the short duration of the current pulses, the DC input current is much less than this, at 800mA, which includes transformer losses, the vibrator driving coil, and primary damping resistors.
With a 40W resistive load, it can be seen the pulses flatten off considerably and the peak current is not much higher than when unloaded. Note also that the higher timing capacitance shows up as a slope during portion of the dead time.

Rebuilding the Inverter.

New transformer and timing circuit installed.

Circuit of rebuilt inverter.

We can now see the reason for the strange value of timing capacitance. In order to keep the AC output balanced above earth (to reduce RFI), the timing capacitance is split into two, with the junction of capacitors connected to earth. Likewise, to keep the circuit symmetrical, the series resistor is divided in two. As I didn't have 560R resistors, I used 510R instead.
Because there is about 30V across each resistor, they need to have a 5W rating. Because of the introduction of the series resistors, it could be imagined that the two 0.47uF's are not so effective now for RF bypassing. Indeed this turned out to be true, and the addition of 0.01uF capacitors as shown fixed this. The extra 0.005uF is negligible compared to the existing timing capacitance.
It should also be pointed out that the balanced output connection is also desirable from a safety perspective. If one side of the transformer secondary was earthed inside the inverter, a shock hazard exists under some fault conditions, whereby the 12V input actually becomes live at 240V AC. For example, assume the neutral pin of the socket was earthed. Now assume the inverter is supplying an appliance through a length of flex. Perhaps the flex gets damaged and the wire connected to the live socket pin becomes earthed - e.g. caught in a metal door of a caravan or shed. With the normal public mains supply, a short circuit would occur and the breaker would trip / fuse would blow. But here, the inverter earth would now be 240V above the real earth, and as this earth is shared with the 12V input, the battery terminals would become very dangerous to touch. The casing of the inverter itself would also become live.

Although not really needed, a neon indicator shows the inverter is working. Normally, the buzzing of the vibrator would indicate this, but in a noisy vehicle the buzz won't be heard, and besides, it makes the appearance of the front panel symmetrical.

Front and back labels of inverter.

For the primary circuit, the 12V input feeds the inverter via a 7A fuse; high enough to prevent nuisance blowing, but low enough to protect the vibrator under excessive load.
Because of imperfect regulation, the output voltage will be excessive with low loading. A resistor of 0.33R can be switched in for loads less than about 20W. High power inverters use a transformer with a tapped secondary to perform this function, but it requires a special transformer, and in a low power inverter like this, the loss of efficiency is trivial.

120R resistors are connected across the vibrator contacts. While not strictly necessary, it is good practice. They damp any high voltage pulses in the primary circuit, helping to reduce RF interference, but also to make things slightly easier for the contacts.
The remainder of RF filtering is provided by the 1uF capacitor at the 12V input, and a 1.5uF at the vibrator socket.

Construction.
The cabinet was made out of aluminium, cut on a guillotine and then notched and bent up with a Di-Acro notcher and brake. A louvered punch was used to create the louvres in the cabinet sides. A small subchassis was made to support the six pin vibrator socket. So that the vibrator does not ever fall out of the socket again, a piece of foam rubber was attached to the cabinet lid to keep pressure on the vibrator. The cabinet was sprayed in a hammertone green. Rivnuts were mounted along the edges of the chassis so the lid can be secured with 3mm screws.
The labels were made using the now obsolete "Scotchcal" process. This is a photographic method to produce aluminium adhesive labels, once popular with electronics magazine projects in Australia.
A few metres of twin flex and a two pin polarised plug supply power to the inverter.

Performance.
A selection of light bulb loads was used to check output voltage under varying loads. Input was maintained at 12.6V at the fuse. Output voltages are given for both High and Low switch settings.
 

Load	Input Current	O/P (Low)	O/P (High)
No load	800mA	271V	277V
15W	2A	240V	255V
25W	2.4A	223V	243V
40W	3.8A	201V	231V
50W	4.2A	190V	224V

Filtering Improvement.
While the inverter was electrically very quiet when used with a MW receiver, it was noted that VHF reception did have some interference. It was found the RFI was coming from the 240V output.

Additional filter components are mounted in front of vibrator. The two ceramic 4700pF's can be seen along with one of the chokes.

A simple filter was made up using two toroidal chokes and two 4700pF capacitors and connected at the output socket. The chokes are connected in series with the live and neutral output socket pins, which are each bypassed to earth with the 4700pF's. These components were obtained from a defective switchmode power supply. While larger capacitors could be used, it must be remembered that they add to the timing capacitance. Here, the extra value is 2350pF (the two 4700pF's are in series) which is not significant.
The improvement was quite marked. MW interference was totally eliminated, and on VHF it was reduced to the point of being in the normal atmospheric noise level.

Needless to say, with the transformer only about five years old, it was a bit of a disappointment. I have seen failed toroid transformers before, and it seems that the weakness is caused by the method by which they are mounted. The disc shaped clamp which secures the transformer is actually pressing against the windings. If there is an overlap of the copper wire, the mechanical pressure can push through the enamel insulation, causing shorted turns. So what to do? A new one is currently $32.50 which is OK, considering the amount of time it would take to rewind the failed transformer. But this transformer is unique to Altronics, and it is difficult to buy since there is only one store in Sydney, which is a 2hr return trip to get to. Otherwise, one has to add about $20 if ordered on the internet. Yes, I could rewind it - and I'm no stranger to doing that - but as it's likely it's the 240V winding which has failed, that's a lot more work, since the 9V windings have to be removed first.

In checking if I had any spare M-5009 transformers, I found I did have the original 80VA M-5109 which had been previously in this inverter. I thought it might be a good time to revisit what the problem was with this transformer, and analyse it a bit further.
The M-5109 was temporarily connected into the circuit. It seemed OK, but within a minute or so I was reminded of what the problem was.
The waveform would change every few seconds from being normal, to being quite distorted on one half cycle. At the same time, the tone of the vibrator changed, and the output voltage dropped slightly.
However, it was also found that if the 0.33R resistor was switched in, the waveform remained stable.

My Theory.
As I theorised earlier in this article, it seems the problem is DC causing saturation. The M-5109  being an 80VA transformer, presumably has less turns per volt than the 50VA M-5009, since the 80VA transformer has a larger core.
My theory is that any DC in the winding will cause the 80VA transformer to saturate much sooner, because it has less turns. The 50VA transformer has more leeway because of more turns (and more losses).
Toroids are known for their close to ideal characteristics, and it seems in this case the 80VA transformer is too good!
This theory is supported by the fact when resistance is introduced, the transformer runs more happily. Obviously, with the resistance in series, the DC current will be less through the windings.
I don't profess to be an expert in transformer design, so would welcome any comments.

A look at the waveform showed it was unbalanced, in that the duty cycle for one set of contacts was not the same as the other set. An asymmetrical waveform fed into the transformer will create a DC component. With a vibrator being a mechanical device subject to certain tolerances, it's impossible to guarantee a perfect duty cycle. With the usual E-I laminated transformers which vibrators were designed for, a small amount of DC isn't so problematic.
Let's look at the waveforms with the 'problem' transformer:

Primary waveform with the 80VA transformer and V6612 vibrator operating a 40W lamp.

As can be seen, the upper half cycle is badly distorted. This is the waveform with the inverter feeding a 40W light bulb. The 0.33R resistor is not in circuit. Now, looking at the current waveform, we see how bad this really is:

Current waveform. Peak current on every alternate half cycle is 21A!

The effect of an uneven duty cycle is clearly visible. Each division on the graticule is 5A. We can see about 9A peak on one contact, and 21A on the other! Compare this to the waveform shown earlier for the 50VA transformer, with about 5.4A on both contacts.

Obviously, the vibrator has an uneven duty cycle. But, how good or bad is the 80VA transformer if the vibrator had a more even duty cycle? Since they were to hand, I tried a number of Mallory 1701 vibrators, and the adjustment of these was a lot better than the V6612.
Indeed, the 80VA transformer worked without any complaint.

Current is much lower when the vibrator has a more even duty cycle.

Here we can see that the current peaks are now much closer together in amplitude, at about 8A and 9A peak. Despite the improvement, the current peaks are still higher than with the 50VA transformer. It's possible the timing capacitor circuit is not suited to the 80VA transformer, but this was not experimented with.

While it appears that the 80VA transformer can be made to work, I decided against using it, because even if the vibrator has a perfect duty cycle, this can change over time slightly. What would normally be an acceptable variation in duty cycle is clearly not acceptable with the 80VA transformer.

The conclusion from all this is that higher power toroids are not suited for vibrator power supplies. The Jaycar 20VA transformers work very well, and I've not had any trouble with them. And, with the amount of hours clocked up with the Altronics 50VA transformer, it too seems OK, but at the same time this appears to be the upper limit.
Since the power rating of ordinary radio type vibrators is much less than 50VA, there is no actual need to use a higher rated transformer anyway. The point is; don't use a larger toroid just because one is already to hand. A situation where a larger transformer might be OK is where the turns per volt are higher than normal. For example, a 12-0-12V winding (fed from a 12V supply) should be less critical than 9-0-9V. Of course, it follows then that the 240V winding will provide somewhat less output voltage, but if that is in order, the scheme is quite valid.
Finally, it needs to be pointed out that while the Jaycar 20VA transformers, and the Altronics 30VA and 50VA toroidal transformers are OK, this is no guarantee that those from other sources are OK.

Repairing the Transformer.
Since Toroidal transformers are not difficult to dismantle, as far as removing the low voltage windings go, I decided to have a look. Just maybe the breakdown was in one of the 9V windings.

Transformer with outer insulation removed. These are the two 9V windings.

9V windings removed. This is the 240V winding.

There were 81 turns for the the 9V windings, which implies about 9 turns per volt. I say "about" because the offload voltage is closer to 10V. The 240V winding would contain about 2160 turns.
Unfortunately, with the two 9V windings removed, the primary still drew excess current. With nothing to lose, the next layer of insulation was removed. Interestingly, the 240V winding is insulated better than I thought. It's surrounded by two plastic cups which interleave
with each other.
When these were removed and the transformer powered up again through a light bulb, the fault was evident. It was not very far in from the start of the winding, and was between this top layer and two adjacent turns on the layer beneath.

• See video of shorted turns https://youtu.be/MUH3WSdaAHo

Back in the inverter, it was like nothing had ever happened, and the 40W light bulb used for testing the load lit up normally, and with a normal current consumption on the 12V supply.
Obviously, the vibrator still needed to be adjusted, but the current waveform was checked before doing so, to see how it compared with the 80V transformer.

With the 50VA transformer and vibrator still out of adjustment, this is the primary current waveform.

Instead of 21A, the current peaked at around 15A. This clearly illustrates the difference between the two transformers.
It was interesting to see the condition of the vibrator after all the use it had. There was a light black coating on the inside surfaces which would be contact material. The contacts themselves were dirty, but with minimal pitting. It would seem the timing capacitor circuit has done its job as well as it can. The vibrator was dismantled and the contacts filed smooth with a points file. They were then finished with 600 grade sandpaper. Since one side of the driver coil contact is attached to the frame of the vibrator, this had to be readjusted because when everything was put back together, the frame position relative to the stack (and thus reed contact) was not exactly as before.
Adjustment of the reed contact has an effect on duty cycle, so a finer adjustment was then made so the current peaks on both half cycles were about the same.

The Effects of Timing (Buffer) Capacitor Value.
The following two waveforms show exactly why a high value of timing capacitor is undesirable, even though it appears to stop contact sparking. As explained previously, with a vibrator power supply operating a range of unknown AC loads, the timing capacitor has to be a compromise. The result is that the timing capacitor ends up being higher than ideal, when the inverter is operating with no load, or with a much lower load than it was designed for. Taking the inverter described in this article as an example, here is the primary current waveform when the inverter is operating with no load:

No load primary current waveform. Peak to peak current is approximately 11A.

Here we see short high current pulses of about 11A. The fact they are of such short duration is why they do not show up when the average current draw of the inverter is measured. However, it is important to note that the vibrator contacts are switching this high current.

• One trap for the unwary therefore, is that measuring the DC current draw of a vibrator power supply is not a satisfactory indication that the timing capacitor is correct. The vibrator contacts may still be exposed to high peak currents.

Primary current waveform with 25W load.

Now we see a dramatic reduction in peak current, down to about 7.2A. Obviously, the vibrator contacts will last a lot longer when the inverter is loaded.

Inverter loaded with 40W.

When the inverter is loaded with 40W, there is only a slight further reduction in peak current. The increase in overall contact current can be seen, which is as expected. It is important to note that these three waveforms were measured by connecting the CRO across the 0.33R resistor, and the current readings are not as accurate as when measured with the Hall effect device used previously.

Articles on vibrator power supplies will usually mention the effect of a too small timing capacitor - i.e., excess peak voltages are developed. Only one article I have seen mentioned the effect of too large timing capacitance. It seemed to imply that the only ill effect was an increase in overall current consumption. Now we can see exactly what happens, in that while the average current might not seem to be out of the ordinary, the vibrator contacts can be exposed to very high peak currents.