Altronics "Powerhouse" 600W Inverter.


The front panel labelling is misleading. The inverter requires wiring and component modifications to select between 12V and 24V input. It is not automatic as might be implied.

This inverter was available as a kit or a made up version from the Australian electronics parts retailer, Altronics. It was developed in 1996, but the origins go way back to the Electronics Australia 300W inverter from February 1979, which was their first solid state design. For readers outside of Australasia, "Electronics Australia" was one of two electronics magazines which existed at the time. Electronics Australia subsequently produced several more inverters along this design, each being an improvement over its predecessor. One such design is this one.


From the Altronics 1997 catalog.

Altronics produced a few more designs themselves. The two versions shown in the ad above, along with a 300W model, were not described in Electronics Australia, probably because the magazine was by this time in the stages of being 'dumbed down' prior to its death in 2001. These two inverters were an improvement on the December 1987 and February 1992 designs, incorporating a microprocessor and "phase correction" circuitry.

Chronology of Electronics Australia Solid State Inverters.

Prior to 1979, the last inverter described in Electronics Australia (previously called "Radio, Television & Hobbies") was a 15W vibrator type for shavers in February 1960.



The Altronics K6776 600W Inverter.
This inverter came to me from an HRSA meeting for the low price of $25. Somewhat cheaper than the original $535! Circa 2019 when I bought it, the technology was outdated, with inverters of this power having used switchmode designs since the early 2000's.  It was unassembled and in the original box. These days, a 600W inverter is a fraction of the cost, and considerably smaller and lighter. There is a tradeoff, however, and that is such designs tend to be fragile and easily damaged by anything besides a resistive load.

Circuit Design.


Circuit of the control board.

The heart of the circuit is an MC68HCO5P9 microprocessor which replaces all the CMOS IC's and op-amps used previously. The microprocessor generates the anti-phase 50Hz drive signals for the switching MOSFETs, provides the voltage regulation circuitry, and incorporates the auto-start and under/over voltage detection circuitry. The timebase is crystal controlled, unlike some previous models.
Over temperature sensing is provided, as with all the previous 300W/600W/1200W designs. However, electronic current sensing for overload protection is not used. This is achieved simply by using a circuit breaker.
Depending on the DIP switch settings, choice of transorbs, and the transformer connections, the same circuit caters for both 12V and 24V operation. The 1200W inverter is the same design, except has double the number of switching MOSFETs, and of course a larger transformer.

Regulation.
The inverter is a modified square wave design. Regulation is achieved by adjusting the pulse width. Feedback is provided by a 12V winding on the power transformer - the voltage here being proportional to the voltage induced in the 240V output winding.
The feedback voltage is rectified and adjusted by RV2, prior to being fed into pin 16 of the micro. It's compared to an internal voltage reference, and if the output voltage low, pulse width (duty cycle) is increased to compensate, and vice versa.

Auto-Start.
The auto-start circuit is based on that used since the September 1985 design, based on a reader submitted circuit in the February 1981 "Circuit and Design Ideas" section of EA. A bridge rectifier is connected between the load and transformer secondary. The DC output of the rectifier is short circuited, which results in 1.4Vp-p across its AC terminals when the inverter is operating, and a with a load drawing current. At the same time, a small current is fed from the 12V battery supply into one of the AC terminals of the rectifier, superimposing 0.7V on it. When the load is connected, the DC path allows this 0.7V to come back into the other AC terminal which is used to sense whether the inverter should start.
In more detail, R10 (22k) feeds 545uA into the upper right diode of the bridge. When the load connected has a DC path, 0.7V appears at C9 (47uF). Because there's AC present once the inverter starts, D6 rectifies this, and charges C10. Once a voltage appears across C10, the micro starts the inverter, by generating the drive voltages for the switching MOSFETs. C9 is used to prevent stray capacitance in connecting leads from causing the auto-start to run. This might happen because with enough capacitance between cable conductors, enough current will keep flowing to keep the auto-start circuit activated. C9 shunts the AC to earth, and ensures only DC coming back from the load will activate the auto-start.

24V supply.
Described later for the "phase correction" circuitry, a supply of around 12V greater than the battery voltage is required. This is generated by an audio transformer T1. It's fed by a  approximately 12Vp-p 1.3kHz square wave. The rectified secondary voltage is connected in series with the 12V supply so that it adds, producing about 24V.

Over Temperature Sensing.
An LM334Z is mounted on the heatsink, and if the MOSFETs should run excessively hot, as with too high of a load connected, the voltage at pin 18 of the micro will rise sufficiently to cause the micro to shut off the inverter. This can be adjusted by RV1.

5V Supply.
The micro requires a 5V supply which comes from a low dropout LM2931-5 regulator. There is a pre-regulator based around Q15 which is a P-channel MOSFET. It provides a 12V input to the 5V regulator and MOSFET drivers, and is only really required for the 24V version. Basically, Q15 conducts because of the gate voltage supplied by the 1k (R1) gate resistor. If the output rises above 12.6V, ZD1 conducts, turning on Q2, thus turning off Q1 and removing gate voltage to Q15. It can be seen that when the inverter is run off 12V, Q15 is always saturated.
 


Circuit of the power board.

Transformer Switching.
Since the September 1985 design, all the high power inverters have used toroidal transformers. These are vastly more efficient and have better regulation than the conventional E-I laminated types. They are also lighter.
From the December 1987 designs and onwards, the transformer has been switched by MOSFETs. These have a lower voltage drop compared to bipolar transistors, as well as being available in higher current ratings. Thus, the heatsink runs cooler, and the inverter is more efficient. Aside from that, MOSFETs are voltage driven, which makes for simple driving circuitry since no current is required. Bipolar transistors, on the other hand, require a comparatively considerable base current.
In this inverter, there are parallel sets of seven IRF540's for each side of the transformer primary. 33V transorbs conduct if the back EMF becomes excessive, causing the MOSFETs to turn on, shunting the excess voltage to earth. Because of the gate to source capacitance and fast switching speed required, the gates are fed from push-pull emitter follower circuits comprised of Q10/Q11 and Q13/Q14 on the control board. If the MOSFETs were fed directly from the micro, there's a risk that the current output would not be enough to overcome the gate capacitance, which would result in a slower switching speed. The inverter would still work, but with the slow switching speed would cause the MOSFETs to operate in their linear mode for longer than desirable. At the high currents at which this inverter operates, the MOSFETs would run hotter, as well as cause low efficiency.

Phase Correction.

New to the inverter designs is this so called "Phase Correction Circuit". A more accurate name would be "Active Back EMF Suppression", since that's what it actually does. Phase correction infers correcting for the power factor of inductive loads, which this circuit does not do.
Instead, low power factor loads are simply run as is, and this circuit suppresses the back EMF caused by their inductive loading. (Although a capacitive load also causes a low power factor, most low power factor loads are inductive). To see how this circuit works, note the bridge rectifier connected to the transformer primary (BR10). Only two diodes are used, and these connect via a parallel set of MOSFETs, Q131 and Q133, back to the transformer centre-tap. These MOSFETs are driven by an out of phase voltage derived from the drive voltage, which drives the main switching MOSFETs. The out of phase signal is derived by Q129. The upshot of all this is that the transformer primary is short circuited when the main switching MOSFETs are not conducting. Thus, any back EMF (which occurs during the off time) is short circuited.

Wiring diagram. The 240V wiring and transformer are not actually shown on the circuit diagrams.

An attempt is made to filter the 240V output with a toroidal core in the output wiring and a 0.01uF capacitor. A 275V metal oxide varistor is also included to suppress high voltage spikes. The 0.01uF across the 12V input looks doubtful as to its effectiveness.



Building the Kit.


Parts of the kit prior to assembly.

As packaged, the transformer had already been mounted to the chassis, which in turn was screwed and bolted to the bottom of the plastic case. This makes sense given that such a weighty transformer can't be left loose in the box.
The assembly was not particularly difficult, although missing parts did slow things down. The control board was assembled first and was straightforward. The PCB calls for three links (actually zero ohm resistors). However, the parts list shows only two, which is what was supplied. One of the links was therefore simply a component lead off cut.
The LEDs and power switch are mounted on the PCB which then mounts on the chassis with plastic standoffs. The LEDs did not quite line up with the holes in the front panel, so it was necessary to bend their leads to the side slightly. Interestingly, the DIP switch settings on the board show 36V and 48V operation, as well as for 12V and 24V. Yet, these higher voltage input inverters, if they ever existed, were never advertised.

The biggest difficulty was with the power board and the MOSFETs. The MOSFETs are meant to be insulated from the heatsink bracket with silicone strips extending the full length. Unfortunately, these were missing, and they are not easily available (i.e. can't just go to Jaycar or Altronics to get them), and ordering on the internet would take a long time (ebay - China), or I'd pay excessive postage (Australia). Over the years, I've collected the silicone insulators from things I've pulled apart and I had sufficient to use. But since each individual MOSFET had to be insulated, since I didn't have the strips, it was difficult to keep the insulators in place while tightening the clamps. Added to this was the PCB was warped because of having to add solder to the MOSFET tracks. In the end, I made up small end brackets to secure the PCB against the heatsink bracket. This worked well, except I broke my 9BA tap twice in threading the holes for the two small brackets. I used a small amount of heatsink compound to secure each silicone insulator behind the MOSFET so that they would stay in position until the clamping strips were secured. Eventually, it all went together OK.

I found the wording describing the wiring between the MOSFET sources and the battery negative terminal hard to follow, and had to read it numerous times to understand it. Part of the confusion is that for each of the two groups of MOSFETs, the source wires are two different lengths. With the high currents that could flow, one wire would carry more current than the other.

The lugs for the transformer primary wires did not fit. The transformer has multiple primary windings, connected in parallel for 12V, or in series for 24V. Problem was that the lugs were too small to accept two of the primary wires. Fortunately, I did have some which fitted. The parts list should have specified yellow lugs, not the blue ones supplied. Incidentally, there were insufficient nuts supplied to connect the transformer lugs to the power board. These are 4mm, which I had, so did not cause a problem.

The next hurdle was the short length of cable from the circuit breaker to the positive battery terminal. Again, the lug was too small and it was impossible to push the conductors into the lug. I had to cut a slot down the body of the lug and spread it apart to get the wire to fit.

Only one earth lug was supplied (and no screws for it, or any of the missing two lugs). Luckily, I've got a good stock of parts, and was able to proceed, wiring the earths up from the back panel, chassis, and power point.

The instructions state (and show) three turns of the live and neutral wires passing through the ferrite core, prior to connection to the power point. Not only was this impossible, because the brown wire from the transformer was simply not long enough, neither was the piece of blue wire intended for the neutral connection back to the control board. As such, I could only wind one turn of each through the core. I don't see it as particularly critical and have not worried about it.
A minor point is that if the supplied solder wire was meant to be enough to assemble the kit, it was about a length too short. I used it all up and was using my own stock by the time I got to the connections of the transformer lugs.

Supplied with the kit is a cable with battery clips, intended for use when the inverter is used in a portable situation. It appears to be a set of car jumper leads cut in half, so that one set makes up two inverter kits. As with most car jumper leads, the thick (10mm diameter) insulation implies a high current cable, but is usually quite deceiving. In fact, the internal conductor is only 4mm diameter. At 50A, there would have to be significant voltage drop.

With the inverter finally assembled, the micro was plugged into its socket, and the inverter powered up on a 2A current limited power supply. The green LED came on, and the 50Hz buzz could be heard from the transformer. It appeared all was working, and so the output voltage was adjusted with a true RMS meter, and the temperature sensor was calibrated by measuring its voltage against the ambient temperature (the instructions show how to calculate this).


Completed inverter.



Performance.
With a 10A power supply, a 60W incandescent lamp was tried, and it worked as it should. Waveforms were observed at the drains of one set of MOSFETs:


No load waveform.


60W resistive load.


15W fluorescent lamp. Note the short spike at the leading edge.

A VHF receiver was tried and no RFI was evident. A couple of 15W fluorescent lamps were tried (both without power factor correction) and the inverter was able to drive them.
A slight wavering of the duty cycle is evident, with a shimmering visible with incandescent lamps. This appears to be something to do with the voltage regulation circuitry.
My only criticism of the kit is that for $535, you'd want all the parts to be present. And, you'd want the lugs to fit the cables. There seemed to be some stinginess with the hookup wire supplied. Apart from the blue neutral wire being too short (if the specified three turns through the toroid were included), there was insufficient length of the green earth wire. How one was meant to fit three turns of the transformer secondary wire through the toroid, when it too wasn't long enough, has me curious. I wouldn't want to be in the position of someone in an isolated place who had no stock of parts to draw on, should the kit be missing something.
The instructions were a black and white photocopy, and as a result, the photos of the inverter construction were almost useless. To be fair, it was produced in the days before cheap colour printing, but still for the price and complexity of the kit, they should have provided clear colour photos - or at least photos with a greyscale.


The car jumper leads supplied.

Out of curiosity, I tried the supplied car jumper lead cable with the inverter connected to a 100Ah AGM battery. The inverter load was a 550W toaster. Surprisingly, it worked a lot better than I thought. There was about 0.7V drop between the battery terminals and the inverter. Since the cable is only about 1.5m long, the short length would help.
The cable did run slightly warm, but full 240V output was maintained. A couple of 20W fluorescent lamps were then tested; one being a conventional switch start type, and the other being a rapid start type. These both worked normally. A medium size bar freezer also appeared to work normally. Despite the instructions claiming that a 3/4hp drill press could be powered off the inverter, mine ran for about two seconds before the circuit breaker tripped.

In summary, the inverter is a good design, and my only reservation would be if the micro failed - it would be necessary to substitute it with conventional circuitry; not difficult, but still tedious.



Home