Help with SMPS current sensing

[Pierre]

Hello all.

I am starting to design an offline SMPS, based on half-bridge topology, using a ETD39 core with 1 primary and 2 main secondaries + 1 aux. secondary, intended to give about +/-55V at 500W and 15V for the housekeeping.

I will try voltage-mode regulation with a simple SG3525.
I have seen some designs, mainly in PC power supplies, that implement overcurrent protection (and some kind of cycle-by-cycle current control as the SG3525 permits), but they always use a small current sense transformer as the sensing element.
Is it possible to use a small valued power resistor?

For example, a 0.1ohm/5W resistor should be enough to activate a transistor with a 6A current flowing through it, with a loss of only 3.6W in the worst case.

Is this loss the unique cause for the general tendence to avoid the resistor or is there any other cause for using a more expensive transformer?

Thanks!

[Dem]

No, You can not use resistor at this topology, because both sides of primary are floating... You need a current transformer, but it's not a big deal - just 50 turns of wire on 10-15mm ferrite toroid core for secondary. and for primary 1 turn - just pass "cold" side of primary via the core. You can find a lot of such tr-rs in old SMPS. The secondary should be connected to bridge (4x1N4148) and 20 Ohm load - it will provide (Iout/50 )*20= 400mV/A. Additional R-C filter is desired to remove overshoots.

[Pierre]

Thanks for the quick response.

Best regards,
Pierre

[Pierre]

However, I still don't understand what's the problem with a sense resistor. I assume that it will only "read" current half cycle, when the lower mosfet is on.
My proposal is to put the resistor from the low side mosfet source to gnd.
Shouldn't this be enough for rough overload protection? It worked very well for me in a Class-D power amplifier.

Pierre

[Eva]

The problem with resistor current sensing arises from its lack of mains isolation except when the control circuit is already on the mains side

Current-transformer sensing solves the issolation issue

Also remember that asymetric pulse-by-pulse current limiting drives the transformer into saturation and destroys the switches. Inmediate shutdown is required to avoid that

[Pierre]

Thanks, Eva.
In my case the control circuitry is already on the mains side, but thinking about it twice, asymmetric current control can be dangerous as you said, unless I only use voltage-mode control and use the current sensing signal externally to activate the shutdown pin of the controller with a latch.

Using voltage-mode only can be ok?

[Dem]

My proposal is to put the resistor from the low side mosfet source to gnd.
Shouldn't this be enough for rough overload protection? It worked very well for me in a Class-D power amplifier.

You can use it for average-power overload protection, in such case it's better to place Your sensor in the negative rail of the H-bridge, but in Your 1st post You asked about cycle-by-cycle current control - in this case You have use current tr-r, because You can't control 3525 with low-side pulses only.
If average-power protection is enough for You (it's slow and not so reliable) - to my opinion You can't use 3525 built-in protection as-is (in overcurrent mode output pulses can get very asymmetric) - You will need to design some trigger-like protection mechanism with LM393 etc.,
if it DIY project and not mass-production - take 20 minutes and wind 50 turns ...

[Pierre]

I think DEM has answered my question.
Only a thought more: why do you say average-power detection is slow? I think that it will detect any overcurrent (although only in the negative leg of the h-bridge, ok) and can trigger shutdown with a fast latch very quickly.

[N-Channel]

adding to Eva's comments, it also gives only "half" the picture, as you will only be sensing the current through the lower MOSFET, and not through both. You need to sense the current through both MOSFETs or you will end up with an imbalance, leading to "walking' of the core toward saturation, and destroying one or both MOSFETs in the process.

BTW, even though you are using the SG3525 voltage-mode chip, you can implement a crude cycle-by-cycle current limiting by taking the output from the current-sensing transformer to a signal-conditioning op-amp and then on to pin 10 (SHUTDOWN) of the 3525. Whenever the output from the op-amp exceeds pin 10's threshhold (which, I believe, is 1.0V), if the duration is short enough, the SS capacitor will not discharge enough to trigger a new softstart cycle, enabling you to realize cycle-by-cycle current limiting this way.

Please keep us posted of your progress ans post pics if/when available.

Steve

[Dem]

why do you say average-power detection is slow?

I didn't say that it's slow, I mean that YOU must make it slow to be average - for example Your amplifier "normal" consumption is 100W, You want to shutdown Your supply if awerage power will more than 150W during 2 seconds (someone connected 4Ohm load instead of 8) - than You shutdown Your supply for 10 seconds. Average protection will not protect against short-circuit...
But if You make it fast (comparable with pulse width) - You will go to asymmetries and will get a lot of "goods" that N-Channel described above...