pulsating Vout problem and questions [first smps project]

[prawncracker]

Hey all! first post!

Getting into this SMPS thing guys and i kinda need help on my first build...
its just a small and simple 13.8v forward converter based on a NCP1200, a UC38xx-ish operating at 60khz.
i think imma use this for charging some backup SLA batteries. core is rewound from an old 9Wish commercial psu.

i built the SMPS fine but i need around 500mA current limit. [around 7W]
the NCP pin3 is current sense input [sensed via source resistance] compared with an internal 1v ref and says to have L.E.B [leading edge blanking]
at 320VDC input, its around 22mA at the primary and a MOSFET source resistance of around 47ohms if my knowledge is sufficient.

the SMPS works fine without the MOSFET source R but when put with 47ohms the output votlage drops to around 6.5V and the LED pulses or dims at one-second intervals. the voltage seems to drop to around 4v and then up to 6.5v but my DMM cant follow fast enough i guess...

i use a 1-2.2ohm R for MOSFET source and it works fine. next value 18ohms then i get pulsing output.
from the MOSFET source R i tried connecting RC lowpass 680ohm + 1nF cap for a 240khz knee/ramp but no avail. what could be wrong?


also id like to shoot these few questions...

currently the voltage feedback is the typical TL431 + EC817 can i use a transistor instead of the TL431? from what i read, a simple R+zener+opto series is not good at regulation.

thank you for your responses and looking forward to more SMPS activities.

 

[gloriel]

as foward conventer usually work with 45-50% duty cyle, Vo will drop around 50% if the output not regulated. 18 ohm to much for Sense. use 1 Ohm

 

[prawncracker]

thanks for the reply. well i thought so... it works flawlessly with 1ohm... i only wanted current limiting functions.

i want to regulate current at 700mA. maybe ill just use the classic LM317 current limiter.

 

[KX36]

Calculating the right components to go around the TL431/opto is not a particularly easy task compared to non-isolating feedback using an opamp error amplifier. it may come in a TO-92 package, but it's not a transistor. This might be the cause of your oscillation, but with no schematic I can only speculate. There could be other major problems.

Lets see for your current sense resistor...
On the back of my envelope, you want output 13.8V on a forward converter so you need a secondary voltage of roughly 32V (minimum; peak of square wave going into the output inductor for a duty cycle of around 0.43) If your minimum mains input voltage is 320Vdc, that's a simple 10:1 transformer ratio. You want to limit the output to 700mA, transformed to the primary is 70mA. If it's like the UC3844, the pulse-by-pulse current limit hits when the voltage across Rsns is 1V, so Rsns=1V/70mA=14.3 ohm. If it was a 500mA secondary limit it would be 20 ohm.

It does seem a lot for a sense resistor compared to bigger SMPS, which is probably what's throwing off gloriel, but that's the way this IC works. Dissipation across the 14.3R resistor at 70mA is only 70mW, so it's not a ridiculous value for Rsns at all.

70mA across 1R is 70mV, which is possibly too low for the IC to even trigger the PWM on, meaning it may well not be regulating at all, but actually just running at its 80% duty cycle limit, which is far too much for a forward converter to avoid transformer saturation, but this looks like a flyback IC. That might be a problem. UC3844 has a 50% duty cycle limit making it probably a better choice for forward converters, or there are other newer ones.

 

[prawncracker]

i see... i see... that cleared a lot of things KX36 thank you.

i am using a TL431 standard feedback. divider to Ref, opto in series with anode. im just curious of it is possible to use a transistor to drive the opto instead?

i shall try your suggestion when i work on it, might have to rewind the transformer too. i didnt account for duty cycle.
max duty city is 80%. so i guess calculating the transformer at 43% like you said will give the converter some headroom before feedback kicks in. am i right?

i was also thinking about sensing the output current via the secondary. using the sensing R between the load[battery] and ground. might need another opto or discard isolation which is dangerous.

 

[KX36]

Yeah, you said it's a forward converter, so the maximum duty cycle is set by the transformer's requirement to reset each cycle. Exactly what that limit is depends on your reset circuit, but since it's an off-line converter I expect you're resetting with a reset winding and a primary:reset turns ratio of 1:1, which gives a maximum duty cycle of 50% to avoid saturation and a peak voltage on the switch of twice the input voltage plus a spike. IIRC, this reset method tends to have the lowest switch voltage stress and that's still a high voltage for most FETs if you have a high mains line voltage. For that you'd need an IC which limits the duty cycle to 50% and to give a little headroom for transient loads, quick charging on the output capacitor etc you'd design that in the worst case (low input voltage), the duty cycle is somewhere around 0.40-0.45.

http://www.onsemi.com/pub_link/Collateral/TND381-D.PDF Here's a reference for the use of TL431. I can't really see how or why you'd put a transistor in there. You have to know where you are with your plant gain and where you want to be with the loop gain to design the compensation and stabilise the thing.

If you have a primary side cycle by cycle current limit on the IC, that should limit the current to the output once you have the rest of the thing stable. It is possible to have a higher current briefly at the output which will discharge the output capacitor (as it can only be charged by the limited current) and lower the output voltage so that won't last long.

 

[prawncracker]

thanks for the reply.

hyep, ive got a lot to learn. i thought the compensator is just sending binary [over/undervolt] info back to the controller.
i seem to have snapped the core while splitting it for rewinding. might be better off getting a slightly bigger core and a proper UC chip to learn with.