Repair advice/hints for largish ultrasonic equipment

[Kremmen]

Last week i was approached by a couple of guys from a company that uses ultrasonic equipment as part of their production process.
The issue they have is that the Chinese equipment they use is prone to malfunctioning and they wanted me to look if there was anything to be done about it.
Typically, the only information is that "they do not work". This is not surprising as these guys are not able to analyze what might be the cause.
Also typically, there is no user manual at all, no schematics either, much less a troubleshooting guide. I didn't give them much hope but promised to take it on anyway.
Unfortunately, at this time i do not have a working unit available for comparison. It may come to that but meanwhile this:

The device consists of a hefty transducer assembly weighing roughly 10 kilos, the US driver alone is the size of a Red Bull can. They said the power is on the order of 1-2 kW. The drive voltage is several hundred volts at least (haven't tried to measure that as of yet). The customer says that driver units are individually tuned to specific transducers which could be a clue or then again, maybe not.

The electronics unit is mains powered (230VAC single phase), all unisolated. Looks to be a resonant tank circuit driving the US sender, with an excitation transformer to drive the tank, and with a tuning coil, presumably to achieve resonance.
I started by reversing the schematics out of the physical device. I believe i have got it right, but errors are of course possible. See attached pdf for that.
Also, i have attached some photos of the layout and components for orientation.

What i have found out so far is this (there are 4 units to investigate):
-it is all 80's tech level, so chunky thru hole components, not a microchip in sight. The driver is therefore self oscillating.
-there are no obvious blown components and all that i have measured check ok.
-Especially the driver bridge seems to be OK, transistors check out ok.
-Supply voltages are OK, esp the 325V DC rail on the controller board, and the bridge voltages are all as expected (bridge is 320-160-0V).
-the controller board +25V rail is OK, the voltage is stable. (note: the zener diode type code in the schematic is probably not correct. Voltage over each zener is almost exactly 12.5V).
So at idle everything checks out as far as i can see.

Depressing the front panel Check-button activates the device by energizing relay K1. The points open and release the bridge drive signals from ground. This takes place reliably, and the US driver audibly activates (it makes quite a noise. I have submerged the business end in case it needs some resistance to work against. Probably makes no difference however.)
In less that a second the Overload light comes on and the oscillation stops.

So, the circuit does oscillate meaning there is no catastrophic damage to any component. The above can be repeated arbitrarily many times with identical result.
Next i tried to see what effect the tuning coil has on the oscillation and loading. As expected it does have an effect and quite significant too. Tuning in one direction causes the overload to trigger almost instantaneously. Tuning in the other direction it is possible to avoid triggering the overload but then the system operates more or less erratically. If it starts oscillating the oscillation is intermittent with audible stuttering or, if tuned sufficiently far the oscillation does not start at all. There is a relatively wide overlap where oscillation starts and continues but stutters. However, if you depress the check several times relatively quickly, it will cause an overload. To completely avoid this, the tuning has to be so far from resonance as to not start oscillations or only occasionally.

The overload indicator is triggered and the operating relay released by a SCR (BT169D) that in turn is triggered from 2 sources: either the primary voltage of the excitation transformer (from connector J1 / 2) or bridge drive current measured by transformer T3 on the controller board. I have tried to see which one by measuring the voltages from test points TP3 and TP4 using a regular multimeter but that does not seem to work. Although the voltages at the test points are rectified it could be that the transient is too fast to be captured by a multimeter. If need be, i will connect a scope to see the waveform, but that requires an isolation transformer since i don't wish to blow my scope by accident.

It appears to me the circuit is overly sensitive to the tank tuning. Since some units do operate, the next question is if anyone has practical experience in self oscillating US drivers of this kind and if so, do you or anyone have ideas what might actually be the issue here. Could the circuit be this sensitive to component tolerance variations or e.g. cap aging? (The unit is relatively new, made in 2020 so at least old age cannot be a cause, but bad electrolytic caps + overvoltage might be).

I can provide more information if needed.
Any ideas are very much appreciated.

[aqibi2000]

I was told the US cleaner’s transducers can get damaged if the tank they are installed on does not have the minimum amount of water/fluid. Could this be your case someone is “powering” them up without their minimum load (water).

[Kremmen]

This could be part of the issue or a contributing factor, i am not qualified to say. For that reason i did immerse it in a water bucket. In this case the transducer is nothing like the thing you usually see in a small tank cleaner. The actual emitter is a hefty blunt metal wedge itself weighing at least 3-4 kilos. The cylindrical driver is attached on top of that and the assembly is almost half a meter in height. Somehow i suspect it won't mind much if operated "dry" but who knows.

[Squoip]

Do you have a way to test the transducer? That would be my first look. If the transducer or wires are open, it's not going to work. Since that's the part that can be handled by people, it's the most likely to be broken.

[Kremmen]

Thanks for the suggestion.
This is what I and a technician involved in the case are also now considering the prime suspect.
I have meanwhile made some progress in analyzing and checking the circuit and since it is a self resonant circuit using mostly passive components there is little to break. The electronics seem to work as intended.
Unfortunately there are no specs available for the transducer so I wouldn't know how a broken one could be distinguished from a working unit.
I have requested known good units for comparison and that may cast more light on the situation.