Removing load capacitance causes power supply trip

Thread Starter

SiCEngineer

Joined May 22, 2019
442
I am having a problem with a high voltage power supply. The task is to reduce/remove the output capacitor bank because during an arc the load is destroyed.
I measured the loop gain and phase and it is stable at the nominal load capacitance of around 100nF. Once I attempt to remove the output capacitors, the loop seems to remain stable as there is no oscillations in the feedback chain.
However, when approaching the target output voltage, the power supply will enter a trip and the output voltage collapses.
I am aware that increasing load capacitance can often cause an over current trip to occur at the load, but I am completely dumbfounded as to why removing it would cause a trip as the power supply approaches regulation.
Has anybody ever faced anything similar? I did not change the compensation network as of yet. I do not have a schematic to hand at this moment and I am rather looking for general ideas of how removing load capacitance could result in a fault condition somewhere in the power supply.
Thanks in advance
SIC
 

BobTPH

Joined Jun 5, 2013
8,649
Once I attempt to remove the output capacitors, the loop seems to remain stable as there is no oscillations in the feedback chain.
However, when approaching the target output voltage, the power supply will enter a trip and the output voltage collapses.
In other words it is stable until it isn’t.
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
In other words it is stable until it isn’t.
I understand the point - but I don’t notice any kind of oscillation in either of the output voltage and the feedback voltage. Would you not expect to see some kind of unstable behaviour in the output if that was the problem? Maybe I will try sweeping the frequency response with a lower output voltage to see if the loop is stable first. Thanks for your input.
 

Papabravo

Joined Feb 24, 2006
20,983
What kind of a power supply is this? The reason I ask is that the output capacitance may be a critical feature of the design, and you may have misidentified the root cause of the problem. More eyes looking at the problem may be helpful.
 

BobTPH

Joined Jun 5, 2013
8,649
Instability does not necessarily result in oscillation. In this case, it looks like the departure from expected behavior caused an immediate shutdowb.
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
What kind of a power supply is this? The reason I ask is that the output capacitance may be a critical feature of the design, and you may have misidentified the root cause of the problem. More eyes looking at the problem may be helpful.
It is a continuous wave travelling wave tube design operating in voltage mode control. This particular tube can only handle a very small energy when an arc occurs. Therefore the aim is to minimise the energy through reducing or removing of the output capacitors to ensure the tube isn’t destroyed if it arcs.
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
Instability does not necessarily result in oscillation. In this case, it looks like the departure from expected behavior caused an immediate shutdowb.
Thank you - I did not know that. I am trying to track down which fault has caused the trip but having no luck as of yet. It can’t be an over-voltage protection as the output voltage seems fairly stable until the trip occurs and doesn’t reach a fault level.
 

Papabravo

Joined Feb 24, 2006
20,983
Instability does not necessarily result in oscillation. In this case, it looks like the departure from expected behavior caused an immediate shutdowb.
Any system whose output increases without bound is deemed unstable. If the usual control mechanism is not functional for some reason the safety shutdown is the thing that keeps worse things from happening. We still have no hypothesis on why the original control strategy is not working with the designed output capacitance. Why are you so convinced this is the problem. what is your evidence?
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
Any system whose output increases without bound is deemed unstable. If the usual control mechanism is not functional for some reason the safety shutdown is the thing that keeps worse things from happening. We still have no hypothesis on why the original control strategy is not working with the designed output capacitance. Why are you so convinced this is the problem. what is your evidence?
It’s a fair point but looking at the feedback signal and the output voltage with an oscilloscope there is nothing to indicate the voltage is rising without bounds. The output voltage rises perfectly fine until the supply gets close to the regulation voltage at which point the whole thing will trip and move into standby mode. I increase the input voltage in very small steps and the problem only occurs when within very close proximity to the expected output voltage. For example at 9.9kV there is no problem whatsoever - increase the output voltage to 10kV and the trip occurs almost immediately.

There was nothing wrong with the original control strategy - the supply regulated at zero and full load without concern. The Bode plot showed very conservative cross over frequency (~100x less than the switching frequency) and adequate phase margin across the sweep. However when the tube is connected and an arc occurs, the tube is destroyed because the amount of energy dumped into it is substantially higher than that which it is rated for.
 
Last edited:

Papabravo

Joined Feb 24, 2006
20,983
Ad. understand it, it DID work with the design capacitance and stopped working when the capacitor was removed.
Why would anyone think that removing capacitance would be an effective strategy? I'm at a loss to speculate a reason for this belief. Only magic springs to mind.
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
Why would anyone think that removing capacitance would be an effective strategy? I'm at a loss to speculate a reason for this belief. Only magic springs to mind.
The reason is because the amount of energy in the output of the power supply is related to the level of capacitance. By reducing the capacitance one should be able to minimise the energy during an arc event.
If there are any other ideas for how to achieve this please do discuss them!
 

Pyrex

Joined Feb 16, 2022
236
Hi,
I saw a design of a laboratory power supply unit. Circuit diagrams of two PSU's with 30V output were compared. In the first PSU there's a large capacitor at the output- 100 uF , and a general purpose opamp (741) was used to regulate the voltage. In the second PSU capacitance was much lower- 10 uF , and a fast opamp was used (OP01). And a 10 Ohm resistor was in series with the capacitor.
Imho, try a RC circuit with lower capacitance at the output
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
Hi,
I saw a design of a laboratory power supply unit. Circuit diagrams of two PSU's with 30V output were compared. In the first PSU there's a large capacitor at the output- 100 uF , and a general purpose opamp (741) was used to regulate the voltage. In the second PSU capacitance was much lower- 10 uF , and a fast opamp was used (OP01). And a 10 Ohm resistor was in series with the capacitor.
Imho, try a RC circuit with lower capacitance at the output
Currently we already have an RC network as the load filter. I am considering an additional CLC pi-filter in hopes that the inductor helps limit the rise time of the fault current but we would still have to worry about the last capacitor before the tube…
 

Pyrex

Joined Feb 16, 2022
236
100 nF capacitor at the output...
If the aim is to decrease it, say, to 10 nF, imho, the feedback loop need be much faster. You need the circuit diagram to find what to modify.
As about additional CLC filter, it is not so simple to impement it. CLC will shift the phase, so the system may become instable
 

Thread Starter

SiCEngineer

Joined May 22, 2019
442
100 nF capacitor at the output...
If the aim is to decrease it, say, to 10 nF, imho, the feedback loop need be much faster. You need the circuit diagram to find what to modify.
As about additional CLC filter, it is not so simple to impement it. CLC will shift the phase, so the system may become instable
Another fair point. I didn’t expect we would need to make the control loop faster as it is a continuous wave tube which means it works at a constant full load power with zero transients.
 

Pyrex

Joined Feb 16, 2022
236
Another fair point. I didn’t expect we would need to make the control loop faster as it is a continuous wave tube which means it works at a constant full load power with zero transients.
A small transients do exist in any real life system. And the system is to be designed is such a way to remain stable. Imho, your system may work with much lower capacitor at the output, if feedback components will be modified in a proper way
 
Top