Help with SMPS current sensing

[Pierre]

Without feedback, it worked good, but /as I counted on it/ the output voltage dropped a lot, when I put a big dummyload on the SMPS.

How much is "a lot" for you?
Perhaps you could increase your bus capacitor.

How did you measure efficiency? I have read that the best way is to measure the input power is to multiply the current drawn from the _rectified and filtered_ line voltage (bus) and voltage, simultaneously.

Some people try to measured drawn power by looking at the AC mains rms current and voltage, that's very unaccurate.

Best regards,
Pierre

[Danko]

Originally posted by Pierre


How much is "a lot" for you?
Perhaps you could increase your bus capacitor.

The output voltage was about 2x60V, and it dropped about 15V.
the infput buffer capacitor is 470uF/400V, and paralell with it a 2.2uF/400V WIMA foil capacitor. Big red "brick"
On the output i dont remember huw much capacitance was.
I choose 470uF/400V elko, becouse this has a good price/performance. One step bigger cap was very expensive.

Originally posted by Pierre
How did you measure efficiency? I have read that the best way is to measure the input power is to multiply the current drawn from the _rectified and filtered_ line voltage (bus) and voltage, simultaneously.

I measured the input power exactly like this. I have a Maxwell MX25304 multimeter.
I measured the output power also with the same method.

Originally posted by Pierre

Some people try to measured drawn power by looking at the AC mains rms current and voltage, that's very unaccurate.

If I remember correctly , that could be OK, but only, when it's measured with a true RMS multimeter. But True RMS multimeters are far more expensive, than the normal multimeters.


Danko

[Danko]

Originally posted by N-Channel
Danko-

I think you are referring to several chips from OnSemi.

[...]

Originally posted by N-Channel


2) the Resonant-mode chips are:
a) MC33066 Resonant-mode controller
b) MC33067 Zero-switching Resonant-mode controller.

Yes, i thought one of theese.

Originally posted by N-Channel

BTW, 90-93% efficiency for a 600-700W converter is VERY good. you must be very good in your overall design to achieve that figure the first time!

Post more pics.

Steve

Yes, maybe a little bit too much. Now i'm in high school, i learned some good tricks, techiques from my classmates, teachers, I hope, I can make more accurate measures.

I will post some pictures, when I get home. Now i'm in dormitory. At home I can make pictures of the completed (but untested) circuit.

Danko

[John Hope]

Bonjour Pierre: To return to the original topic, I would recommend using a current transformer + load + rectifier combination as suggested earlier. And I would very strongly recommend using a fast overload protection scheme that uses cycle-by-cycle peak current monitoring, as follows.

The signal from the CT rectifier should be lightly filtered (eg 470pF + 1k) to remove the leading edge spike of the mosfet switching. (Some control chips do this with blanking, but you still want to do it for your independent overcurrent limit). This signal can be fed to the - i/p of a logic comparator with the + input being set to an appropriate dc 'overcurrent reference' value by a resistor divider chain. So the comparator /TRIP o/p will go low on overcurrent. If you have trouble with false trips, and your reference is already a sensible value, increase the RC time constant of your filter.

If you use the IR2110/IR2113 gate driver you can make use of the SD shutdown input. Feed it with the /Q from a D flipflop, with the async /RST being activated by your /TRIP o/p from the comparator. The D input of the f/f goes to Vcc and the f/f is clocked by the upgoing edge of your system reset. Whatever your control chip is, it will have some sort of shutdown input, and this can also be activated by the /Q or the Q o/p of the flipflop, depending on the polarity required. Strictly speaking it won't be necessary to do this, because once the gate driver is shutdown it doesn't matter what the control circuit does, but do it anyway.

So the circuit will wake up tripped, and after the reset interval the IR2110 gate driver (and the control chip) will be activated. Any peak overcurrent condition will cause the gate drive to be removed and it will stay removed until the system is reset, typically when input power is cycled off and on again.
Regards

John

[Pierre]

Thanks for your help, John.
I usually use a 555 timer for shutdown control in my circuits, so once a protection triggers it, it remains in shutdown for a while. I only have to check if it is fast enough, but I think so.

I usually use the current sense input (whatever it comes from) and divide it with resistors to activate a transistor that pulls the TRIG pin of the 555 low so it triggers.

[Workhorse]

Hi PIERRE,

What is the switching frequency of your SMPS......
You could also try using opto-coupled gate drivers if your frequency is below 35KHZ.... which saves lot of isolation problem....

Does anybody ever Built Dedicated Resonant Converter Supply with Philips "TEA1610T" Resonant Mode controller Chip.
It features dedicated resonant controller + half bridge gate driver in one chip solution....
I have seen these chips in supplies upto 5KW


regards,
Kanwar

[Dem]

To the original topic - Half-bridge current sensing:
recently I found in one of commercial supplies Half-bridge current sensor based on measuring voltage ripple on the sentral point of Half-bridge capasitors. Looks interesting, and there is a healthy logic in this - this voltage is really proportional to the Half-bridge current, but I have never heard of such solution before... What do You all think ? Is it used somewhere?

[Pierre]

looks interesting and, above all, simple. But... would it be fast enough?

[John Hope]

The overcurrent output signal should be a rectified and proportional version of the coil current. With this circuit I'm not convinced that the output will respond to coil currents flowing in both directions. Or am I just being exceptionally dense today?

Response time? The rate of current rise is limited by the coil inductance. For a given current I through the coil, C*V = I*T, so
that the change in voltage, V = I*T/C. It looks to me like the cap value is going to have an impact on the response time.