Thermal shutdown LCS at near zero load
We are working on a design with the LCS700HG.
Input voltage is 400Vdc, output voltage 12Vdc/4A maximum.
Under some low load circumstances (approx. 50-150mA) the controller heats itself, causing it to shutdown thermally.
We have noticed that when this happens the burst frequency is different from when situations where it doesn't occur. The situation seems to start when a small transient in the load current occurs, from zero load to approx. 100mA.
This power supply is in a redundant, hot swappable, system. The LCS is on the board without heatsink (max. load 48W).
Is this a common problem or are we doing something wrong? If it's common, what can we do to solve it?
Comments
We tried lowering the soft start cap, going from 220nF to 100nF and then to 47nF. We don't see any difference in heating up. The problem is that it doesn't happen every time, but intermittent. What we think to see is not that the switching frequency changes, but the burst frquency. With this I mean the interval between switching off and back on is different. I added some plots from the oscilloscope, may that helps.
The way to make the problem come out:
- power up with zero load
- slowly increase load to 100 mA
What is the setting of your upper burst threshold frequency?
Does the frequency at zero-load startup dip below this value?
The upper burst threshold burst freq. is around 360kHZ
The zero load freq. is sometimes around 600kHz and sometimes it is lower than the 600kHz.
When it is 600kHz the controller will heat it self up, and below 600kHz it keeps on working.
Exactly.
The solution is to make sure at zero-load power up the frequency dips below 360 kHz at some point.
If it never hits 360 kHz, it will burst at 600+ kHz instead of at 360 kHz. This is why it heats up.
Hello PowerTechnics and PI-Tucker,
I have exactly the same Problem. Starting into zero or low output power load of a fiew mA, the LCS700 stays in startup mode, bursting at ~500 kHz.
Also, as PowerTechnics stated, reducing CStart stepwise from 470n to zero has no effect on this behaviour.
You write "The solution is to make sure at zero-load power up the frequency dips below 360 kHz at some point." - so you are saying that the LCS can be convinced to leave Startup somehow. Do you have any further suggestions for us how to do that?
All the best,
Andreas
Another observation:
1.: Starting into low load, bursting at fmax
2.: Disable the LCS by pulling OV/UV low
3.: Wait a fiew seconds
4.: Releasing OV/UV
Most often the chip exits the startup mode (but not always), and correctly bursts at BMfStart to BMfStop.
SOLVED
Just realised that this is a very old thread - anyway:
Added a R-C filter (Approx. 100K, 1µF) before the UV/OV pin of the LCS to delay the startup for some 100 ms.
Leaves startup Fmax and goes to Burst-Mode with very small load (~ 1W) reliably.
This is because at startup with zero load, the frequency never dips down below the upper burst threshold frequency. The controller subsequently doesn't burst at its normal burst threshold, and instead bursts at Fmax. In this condition with 100-150 mA load the ZVS slew rate is so slow that the MOSFET turn-on event at the end of the deadtime truncates it, so you don't get complete ZVS, so the device gets hot.
The solution is to reduce the soft-start capacitor value so the output rise time is faster, forcing the frequency during startup to drop lower, and to dip to the upper burst threshold frequency. Watch the primary current during startup up as youreduce the capacitance. You don't want it so high you need to raise the current limit very high.