Post-Mortem, Reverse Engineering: Tomzn TOQ7-125/2P Automatic Transfer Switch (& More)

The job of an automatic transfer switch (ATS), put simply, is to let a load be powered from two power sources, selecting the preferred input if it is available and the backup input if it is not. A good ATS will do this quickly enough such that the changeover time is short enough that a majority of appliances won’t even notice the interruption in power. Such switches are extremely useful where you want to have single-input devices be powered from multiple sources, or in my case, where you’d like to use a power station to run your appliances but want some continuity in case they fail unexpectedly or run out of charge mid-operation.

I’ve taken a look at a few ATSes and liked the Tomzn Nightlight-series TOQ7-125/2P quite a bit. In testing, it was relatively consistently fast, while featuring an improved build quality compared to other units including additional insulation and arc-chute features. For switching the (usually much less than) 10A of a regular GPO, it was overkill, but I thought that would be great for longevity and to avoid the persistent contact welding issues that happen on smaller relay contacts.

It’s now approximately two months since I placed the unit into service, operating for my Dad and to optimise our solar sharer offer (SSO) usage with minimal manual intervention from him. Based on our usage patterns, it saw perhaps two cycles a day or sometimes less – about 100 cycles in total by my guess.

But this week, it failed. My Dad yelled at me saying there’d been “a problem”. I was downstairs in my room, not noticing a thing. He said he saw a flash from the box and smelled a burning smell. I put my hand on the ATS “box” I had built – it was unusually warm. Indeed, a burning smell was present and there was no lights on the unit at all.

It was then, I realised, that we just lost both power circuits in the house. But because I’ve “power-station-ified” my house, almost everything was running from batteries at the time (it was peak-time evening), so losing the circuits had little impact aside from knocking the idle smart-plugs offline. After disconnecting the box, I went outside to find that the Power 2 (20A C-curve) breaker and the RCD (40A) breaker had tripped – it would seem that the fault was quite an event! Perhaps, a short that might have peaked well above 40A leading to a failure in discrimination (as the ATS does not have any connection to Earth)?

I decided to find out why, as I do have a few other ATSes kicking about and I’d rather not have such fault-current events if I can avoid it.

Teardown & Post-Mortem

The first thing I noticed was that the box was warm. Even minutes after the event, the external casing of the ATS measured over 50 degrees Celsius. The smell was horrible – that sort of burnt plastic, verging on a hot rubber kind of acrid smell, that isn’t easy to clear from your nostrils.

Nevertheless, I took apart the box – none of my wiring was suspect. After all, I’d built it using spare 2.5mm^2 wire, so it was overspecified as it is.

So let’s take apart the unit bit by bit …

Removing the handle, I examined the contact chamber. Both poles moved freely and the arc chute pieces are visible.

The contacts themselves showed a small level of black soot which is normal due to arcing during switching, but the amount was very small. Instead, I was heartened to see how much of the contact area was clean and flat – almost as if the switch had “bedded in” with its contact mates. The contacts themselves were not to blame.

Opening up the main body, the result of the failure is clear to see. The blue-wrapped solenoid for input B has overheated grossly, resulting in the insulation tape fusing together and blowing up bubbles like a balloon. Something else had burned and left soot over the red insulation of the solenoid wires too.

But that is the result of the failure, but not the cause. In fact, the cause is in the picture above – can you spot it?

A view from the side and it’s clear to me the cause of the failure. Notice the mechanical linkage in the middle, made of folded steel? It’s supposed to “grip” into the solenoid plungers and form a cage around the metal “stem” which actuates the pre-loaded spring mechanism which performs the rapid “clunk” switching.

It seems the steel was not folded enough or made a little too short, resulting in the steel U-shape opening up, releasing the plunger for the blue solenoid. How this leads to the complete failure of the unit and a short circuit across the mains (input B) will be explained a little later. Perhaps this is a one-off issue … or perhaps it might be a design flaw.

A view from the rear shows the PCB attached to the solenoid coil containing a microswitch which performs a very important function – disconnecting the coil once switching has completed. This rides along the plastic axle where indents are made to control the plunger of the microswitch.

The main control board rests inside the top shell and is surprisingly, not intelligent at all. Instead, it’s purely electromechanical, relying on an AC-coil 4PDT relay – rather cool.

It seems that things escalated quite quickly. The coil likely was stuck on because it had escaped the mechanical linkage, which would have drawn excessive current. This appears to have caused some traces to vaporise – almost acting as a fuse, if it had not caused the generation of plasma. My best guess is that the plasma bridged the two power input contacts on the terminal block from the rear (uncoated joints), causing a flash-over short across the mains that likely caused the breaker to trip, leaving the black soot mark. After all, the terminal block pin pitch is around 3.5mm and the slots likely didn’t do anything to help when the trace next to one of the terminals was the one that vaporised.

It seems very likely that the other part of the trace that survived might have done so, but just barely, as it might have dissipated its heat slightly better.

The carnage is on the other side of the PCB as well – with one trace on this side delaminating from the board and shedding most of its solder mask.

The solenoid coil is driven by AC, so to make it work, they’ve used a bridge rectifier. As with such loads driven by relays, there is a risk of contact welding, so it appears they’ve used a MOV to limit voltage spikes albeit without any additional protection (it’s only transiently-powered anyway).

I’m not sure if it’s a result of the over-heating but the microswitch for the blue coil is a little too far out to properly actuate when it should – this would cause failure as well, but this might have been collateral damage rather than the initiating cause.

Reverse Engineering & Mechanism of Operation

It seemed remarkably simple, so I decided to get KiCad out and draw up a schematic for how it all works. I can’t guarantee the schematic is perfectly accurate, but I believe it’s close enough to illustrate its simplicity.

Put simply, the 4PDT relay is the brains of the operation. Two of the poles are simply dedicated to the A and B status outputs – if you’re not using them (as I was not), then half the relay is essentially redundant. The other two poles switch the actives from each input A and B, while the neutrals are always hard-wired through to the coil boards. The manual/auto slide-switch merely disconnects the switched actives to prevent the switch from “fighting” a manual selection.

When power is applied to A, it energises the relay, causing the relay to flip which disconnects the active from P2-B to JB and applies the active from P1-A to JA. If power is lost from A, then the reverse happens. This explains why, in testing for power consumption, the A input consumes more power than B – the B input is essentially “passive” and is chosen whenever power is available on B simply because of the “normal” power-less state of the relay. The A input consumes more power because it is the power needed to keep this mains input relay latched.

Both inputs also consume power to light their status LEDs – these follow a simple LED + resistor combination which many advocates will say “should not work” but does because it’s such a high resistance that the reverse current and, therefore, voltage is strictly limited. The indicators don’t need to be particularly bright and it saves a diode too … so that explains the ~0.4W of input B at steady state.

The fun happens on a change, as this applies mains to a corresponding coil board. This results in the (3A-rated) bridge rectifier rectifying mains to create a ripply DC which is applied directly to the coil which measures around 24 ohms which suggests a current up to 10A could flow in steady state. In normal operation, however, this flow should happen for such a short time (a cycle or two) that perhaps full current is not reached and thermal mass keeps everything happy. But this should produce a very strong impulse pull on the switching mechanism to cause it to flip positions, in the process, rotating the axle and pushing down on the MX11 microswitch which then cuts out the circuit entirely. Finally, the MOV appears to be there simply to clamp any voltage spikes to ensure the main relay and microswitch contacts don’t weld.

Therefore, when everything works, the coils are only active for a fraction until the switch is completed, cutting power and letting the coils rest. This also explains why the manual warns that low mains voltage could cause the coil to burn – the solenoid may not develop enough “pull” to complete the switch and will remain energised until it burns up. But there are more reasons for the solenoid to burn, such as:

• A broken linkage (most likely, in this case) where the solenoid pulls in faithfully, but is tugging on nothing and thus the switch never changes position and the coil cut-out never actuates.
• A misaligned microswitch (less likely, as it may be a consequence of deformation due to heating) which means that the axle doesn’t push down on the microswitch plunger to cut-out the power.
• The mechanism seizing or becoming mechanically stuck in such a way switching does not fully complete (not a factor in this case).
• Someone manually attempting to switch the mechanism against its will (that’s why there’s a warning about this on the product).

In this case, my best guess is that the solenoid would have drawn up to 10A steady state, but this would not be for long as the heating would likely cause the insulation to break down and adjacent coils would short together, increasing the current so much that the traces vaporised, forming a plasma that shorted out the mains connection at the solder joints behind the terminal block and tripping the circuit breaker, ending the sequence of events and the life of this ATS.

Conclusion

While I had high hopes for the Tomzn NightLight-series TOQ7-125/2P, one of my units has failed in around 100 cycles switching a leisurely <10A load, knocking out the power circuit (20A) in my house and tripping a 40A RCD in the process. While somewhat scary in a way, I would still consider this failure a “positive” in that it wasn’t an arc between power sources (thus my power station escaped unscathed).

The failure appears most likely due to the metal mechanical linkage between both solenoids being too weak, causing the linkage to fail and a solenoid plunger to pull loose from the linkage. After this, the simple electromechanical circuit behaved as it was wired to do so, applying power to the solenoid coil continuously until it burned up, likely shorting out in the process, causing tracks on the control PCB to vaporise which may have formed a conductive plasma that shorted out the mains input at the solder joints on the back-side of the PCB. The event stopped when the 20A C-curve circuit breaker for the house removed power from the circuit entirely, but may not have persisted much longer if left to naturally extinguish.

Whether this is a pattern or a one-off defect is unknown. Rather than send it back, I used it as a learning experience, seeing how the circuitry worked. It’s remarkably electromechanical in nature, with no intelligence at all. But it is unfortunate, as ATSes are usually used to improve reliability, that this ATS failed so soon and contributed to a loss of power event.

Bonus: Hawd HDQ3-125E/2P Test & A Crazy Idea

This would have been a short post, if it were not for the fact that my AliExpress hunting led me to buy some more questionable ATSes. This one is branded Hawd, looks vaguely like my Earu ATS but was slightly cheaper.

As usual, I decided to build it into a box to make it usable “in-the-home”.

The top of the unit looks similar to the Earu and peeking under the black plate reveals bare swinging contacts with no arc chutes – the Tomzn guys are right when they claim their unit is an “upgraded” design.

To make it easy to use, two IEC inputs ….

… and a double GPO on the output. I had some high hopes that this would be similar to my Earu unit but …

… it’s not as fast. A to B switches happen in 20-66ms, which is quite a bit slower than the Tomzn and long enough to make some of my electronics reset.

On the other side, B-A switches happen in 10-24ms, so faster, but still not so fast. Ideally, an ATS should be fast in both directions and preferably equally so.

In another stroke of craziness, I decided to try building myself a four-input power arbiter box that would select from multiple inputs. It wouldn’t be fast, cascading a fast ATS with two slower ATSes …

I chose a larger box for this, but without knowing that the rail set-up in this box was not ideal requiring some modification. This cascaded set-up will result in multiple switching “drop-outs” but the idea would be that Priority 4 would be used for mains, with three power-station inputs that would be selected in sequence. Between each “slow” selection by the TOQ9 (which behaves practically identically to my Earu 63A units), the faster TOQ7 would substitute mains for a short period, maintaining some continuity.

It’s a heavy, noisy and perhaps not very practical behemoth. But I wanted to see what would happen and if this would be useful enough in practice.

Four inputs, each IEC – hand-drilled. I got to say, the more I do this, the better I get …

… and the double GPO output as well. I tried this out with a few power stations running my front-loader washing machine on an atypical blistering 90 degrees C hot wash cycle (we’re usually a cold-wash family, but some things do need to be sanitised from time-to-time). This consumed so much energy that all three power stations (~1kWh each) had been mostly consumed. The switching was not lighting fast, but the washing machine is “outage tolerant” – short outages result in the display going to “—-” for under 0.5s and the unit continuing as if nothing happened, longer outages apparently result in the washing machine resuming the cycle but perhaps delaying slightly. It was good enough to get my clothes washed, which I’d say is a success …

Definitely not something I’d encourage other people to try … but I’m a bit crazy, so that’s just how I roll …

In the meantime, without any better ATSes to go with, I’ve got another two Tomzn TOQ7’s on the way … one to replace this blown one and one for a spare. But I’m starting to have a feeling that I might regret this … perhaps they too will fail, perhaps this is overkill and perhaps the claimed cycle life (2,000 cycles full-load electrical, 5,000 cycles mechanical) is a bit optimistic given the violent nature of the switching.