Why is the overcurrent sense signal being so heavily filtered in LLC?

[cupoftea]

Why, on pg 14 of the following LLC design App Note, are they calling for the over-current sense signal to be filtered so heavily that the ripple on it is just 20% of the average value? Ripple here should be as much as possible….its the overcurrent signal…only the noise should be filtered off.

Would you agree?

LLC app note

LTspice sim attached, with much less filtering

 

[Easy peasy]

In short - no - having 20% ripple seems appropriate - remember - engineers who work on generating app notes have never put a real psu into production.

 

[dick_freebird]

Pulse-by-pulse current limit needs a fast / unfiltered
sense, but this appears to have a timeout / dwell
setting for more of a "sustained overcurrent" /
"slow blow" behavior? Maybe you want to ignore a
single or short transient because of load attributes,
but want to do a thermal-timescale protection response?

 

[cupoftea]

In short - no - having 20% ripple seems appropriate

...So in that case, do you say that doing current sense with a currnt sense transformer here would be inappropriate here?...because AYK, that would give 100% ripple.

ie, the attached is LTspice sims of the capacitive currnt sense, and the current sense done with current sense transformer for comparison....whats wrong with the CST one?..if anything...ie, for use with the ICE2HS01G?

Or do you believe that ICE2HS01G cacnnot be used with a currnt sense transformer?

ICE2HS01G

https://www.infineon.com/dgdl/Infineon-ICE2HS01G-DS-v02_01-en.pdf?fileId=db3a30432a40a650012a458289712b4c

--- Updated Jun 21, 2022 ---

Pulse-by-pulse current limit needs a fast / unfiltered
sense, but this appears to have a timeout / dwell
setting for more of a "sustained overcurrent" /
"slow blow" behavior? Maybe you want to ignore a
single or short transient because of load attributes,
but want to do a thermal-timescale protection response?

...Thanks, i guess thats what the Apps guys attend... though for our application we want fast pulsebypulse trip...and the ICE2HS01G can facilitate that

 

[mrinalmani]

Hi
I have not gone through the entire datasheet of this LLC controller but I think there should be different levels of OCP (Over Current Protection). With TI LLC controller that I have been using, there are different levels.
I think they are referring to is OCP level-1 which probably is RMS over current protection and not cycle-by-cycle peak current protection.
Regards
Mrinal Mani

 

[cupoftea]

Thanks, for the ICE2HS01G, the overcurrent comparator trips whenever the sense signal goes above 0.8v for level 1. There isnt an RMS sense for ICE2H...the CL pin records the peak of the CS pin. (at least thats what seems likely from the lack of datasheet explanation). The VMC comparator is shown tripping on CS in the blok diag of pg 9...but this is surely incorrect...they must mean "CL"(?)

 

[D.A.(Tony)Stewart]

How "heavily filtered" is 20%? (RC=) T= 0.2 f ?

Is this for a delayed thermal response or to prevent a premature cutoff to a step load given the storage delay time of the output cap.

 

[cupoftea]

Thanks, but i think the current in a LLC can climb ruinously high very quickly (within two switching cycles) and so instantaneous current cut-off is required.....hence filtering is bad for Currnt sense.....unless its that bit of filtering to get rid of /reduce the noise into the sense pin.

The attached LT spice sim of an LLC shows a short cct on the output at 6ms....you see how quickly the voltage on the resonant caps builds up ruinously high......and the LLC controller needs therefore instant overcurrent detection and shutdown. So filtering heavily the sense signal is not appropriate.

 

[FvM]

In contrast to your conclusion, I think the simulation shows that 20% ripple filter gives a useful tradeoff between short circuit protection and temporary peak current capability:

 

[cupoftea]

Thanks, I must admit I still don’t see great benefit from filtering the current sense to 20% of what it was unfiltered. Sorry I should have said, the ICE2H LLC controller has an overload function whereby it doesn’t go into overload protection until after a settable delay anyway...however, yes the fsw will increase when 0.8V is breached…….as the attached LTspice shows…….it doesn’t look to be worthwhile to filter heavily the current sense……the waveform of LLC is effectively filtered anyway by the resonant cap and inductor…….the voltage on the resonant caps builds up really quickly in short circuit…..and we want to be able to alleviate this quickly.

I see what you mean about allowing overloads, but we don’t want to the cust overloading it anyway.

It seems, I don’t know if you agree, that a disadvantage of LLC is that it doesn’t offer tight pulse by pulse current limit, so it seems controversial to make current limit even slower than what it offers.

Its that res cap voltage that causes the worries.