Flyback output dipping 2V

[spinach]

Hi,

I have just built a 16V output flyback which is fed from a 36V mains transformer.

Unfortunately the output volts suffer a dip in output voltage of 2V......this happens usually about every 100ms or so.....the output volts are ok inbetween these events.

..When the output volts fall, it goes down 1 to 2V in some 5ms...it then snaps back up to 16V.

It's weird that its ok between this.....and weird that its such a "long" (100ms) time between the volt dips.

This flyback has no capacitors after the full wave diode bridge and just switches the "humpy" DC Bus.

....I use a PIC for the oscillator and vary the duty cycle throughout the mains period (there is a mains zero crossing detector) to try and get contiuous conduction mode throughtout.

PWM is done by two LM393 comparators (one taking the current sense resistor voltage and the other the output voltage divider feedback volts) which feed into a CMOS NAND flip flop along with the oscillator.....the oscillator and flip flop output then feed the mosfet driver via an AND gate.

I've had this type of controller working fine with a constant DC input voltage.

Flyback spec is

Oscillator PIC12F675 (4MHz internal oscillator)
Switch freq = 100KHz
Secondary diode = MBR1100 schottky (100V) 1A
Max Load = 75R
Lp = 487uH
Np:Ns = 2:1
Np = 42 turns.
Output caps = two NRSA 470uF 35V electrolytics
Non-isolated
Saturation current of ferrite core with these turns= 2.2Amps

(cap datasheet =

http://www.niccomp.com/Catalog/nrsa.pdf)

When i scope across the primary current sense resistor...it appears that it leaps in voltage every 100ms or so....it leaps up to about 1.5V which across 0.2R represents 7.5A. The leap seems to last some 200us which is strange because the MOSFET is always pretty stone cold after the circuits been running for 10 minutes or so. The current sense comparator has a reference voltage of 0.45V


.....i appreciate that the ripple current rating of my output caps is only 600mA at 120 Hz....However, with two that makes it 1200mA and i've got some more coming so i'll put those in as soon as i can in parallel.

...I believe caps can withstand higher than usual ripple current rating for a while if they're brand new.

Anyway, I am thinking that maybe the volts are dipping due to reverse recovery problem of the secondary diode when low primary volts are being switched...because then the referred volts to the secondary on MOSFET_OFF would be lowest and the secondary diode will not be able to snap off quick enough...then the output caps are discharging into the secondary causing the voltage dip...why it only happens every 100ms i don't know.

...I put an RCD snubber across the secondary diode and that did improve the situation slightly (R=100R metal film, C = 1.36nF ceramic, D = MBR1100). However its not good enough and as i said it dips 2V every 100ms or so.

I am wondering about getting some saturable inductors (square law) and putting them in series with the secondary diode....this will hopefully cure reverse recovery as they massively resist change in current but then saturate.

-However, i can't seem to find any of these anywhere in the RS catalog.....i guess i'll just have to get some high permeability ferrite beads and wrap them with as many turns as i can.

BTW, I can't use a standard PWM controller here as they don't have a settable Max duty cycle....whereas my PIC's duty cycle is settable (albeit in 1us chunks).....i need duty cycle to be variable as input volts are of course varying from zero to 49.5V.

Anyway, if any reader has ideas as to why my output volts are suffering these strange dips then i would be most grateful.

 

[LionelHutz]

Care to provide some type of circuit drawing, even somewhat simplified?

Overall though, there must be some event you can find with a scope that corresponds to these voltage dips. Trace it back through the circuit to the source.

 

[spinach]

Hello, thankyou for replying...here is a simplified diagram of the circuit......

The idea is eventually to do this at 240V mains.

...Also, to vary the frequency (of the PIC oscillator) as the input volts rise so as to avoid the incidence of sudden high peaks of current being drawn from the source (mains).

This power supply will eventually be used as a "daughter" power supply inside a Power factor corrected power supply and will provide current for switching relays, PFC/PWM controllers, gate drives and the auxiliary electronics. There will not be any auxiliary coils used to provide power
(ie coils sharing the core of the PFC boost inductor or full bridge converter which is downstream of the PFC)

The PIC oscillator receives a 1us pulse on every mains zero crossing...it then splits the mains (~10ms) period into 32 equal sections and varies the frequency and duty cycle for each such time section)

I could just use an 8 pin current mode controller in continuous mode flyback and let it switch away........however, say during low to high load transitions at high Vin, -there would then be some cycles where a high narrow spike of current is drawn from the mains....i want to avoid this. -Eventually when later using this power supply for different ~40W power supply.

Though i get these dips in Vout, which i am trying to remove.

On scoping across the current sense resistor, there are times when it looks like the source side of the current sense resistor does not go back to ground when the FET switches off....but stays at a raised voltage level......then i see the current pulses rising up from this raised level....this "level" is not a striaght line but kind of slopes....the current pulses look like stumps sticking out of the "slope".

When i switch the circuit off after say 10 minutes the MOSFET is cold.

(I've just realised that it couldn't have been secondary diode reverse recovery, as MBR1100 is a schottky and these don't have reverse recovery....maybe i imagined the snubber making the problem better)

All my resistors are metal film type....i am looking into their inductance. (the current sense resistor is 5 off 1R in parallel.)

 

[Skogsgurra]

Is THIS your engineering documentation!??

BTW, ever heard about 100 Hz ripple when rectifying 50 Hz? Just a hint. (If you need more information, read up on filtering after a rectifier bridge).

I think you have been kidding us all the time. This is a forum where engineers discuss engineering problems. Not a place where kids and hobbyists demonstrate their complete inability. Sorry, you will be red flagged.

You will be more than welcome back once you have done your homework.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[LionelHutz]

Yes, if you are in a 50hz part of the world then something wrong in your circuit due to the line frequency is causing the dip every 100mS.

Otherwise, I've never seen a power supply without filtering before the switching section as you have done. You do know that a PFC front end for a power supply basically goes like this?

rectifier, inductor, FET to com, diode, filter caps.

As you can tell, I'm not wasting time drawing it.

Then, the switching supply is run off the filter caps of the PFC front end.

Finally, you do know that there are a number of companies each making a number of specialized IC's specifically for SWPS circuits?

 

[cabraham]

It sounds like servo loop instability. The error amp is oscillating. Without a circuit schematic I can only guess, but it's a good possibility.

I would recheck your frequency compensation network and recompute the error amp and modulator gain and phase response. I'd suggest generating a Bode plot to examine the gain and phase margins. Excel is great for this sort of thing, and I always use it. BR.

Claude

 

[spinach]

OK, Thanks for these replies.

 

[DHambley]

If you've never desinged a PFC boost circuit before, I would suggest downloading a copy of the data sheet for either the LT1248 or UC2854. You can even buy a PCB already built and you just solder the parts on. The application notes provided with either IC will lead you through the math in a very easy-to-understand way.

DH

 

[GonzaloEE]

Maybe your PIC12F675 is running too slow to handle your line and load voltage changes. The 12F675 uses 4 clock cycles per instruction, so your program will be likely running a tight loop of some 10 instructions to make the 100KHz PWM clock. If you're using interrupts for zero crossing detection, the PIC will process it and forget momentarily your line changes.

Try switching to external 20MHz crystal osc. for the PIC, and reduce your count registers in s/w accordingly.

 

[waross]

The old switch tail ring counter configuration would give you a repeating 10 count without reset overhead every 10 counts.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[waross]

Can you program shift registers in your PIC?

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[3kids]

you need to filter the power supply , always a must,