Small flyback SMPS's / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Small flyback SMPS's

previous 1 2 3

Move Thread

LAN_403

uzzors2k

Tue May 01 2007, 09:56AM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

I just bread-boarded it up. Don't rely on my schematic being correct Joe, the only successful runs I've gotten were without a snubber, and the output voltage rose constantly until the IRFP450 couldn't take avalanching anymore. Its fairly unstable atm, so consider it untested.

Well it works, and it doesn't work. Everything relies on the clamp voltage, and with a 500V device running from 320V you're left with very little headroom. If I understand the whole reflected voltage thing right, since my turns ratio is 16:1 the reflected voltage is 320V for 20V out. So in my case the mosfet must be rated for 2*320V + at least 100V for clamping + 20% for mosfet derating! I find the reflected voltage deal really confusing, how does the turns ratio really affect output voltage in flyback converters? For example, could I use a 1:1 ratio for a reflected voltage of 20V for 20V out? RCD set for (0.8*500) - (320+20) = 60V max. That would make it possible to use a 500V device, but with such a low clamp voltage wouldn't there be alot of power loss? And what would the output be? 20V or 320v? Bleh.

Other than that it seems to be working, I'll dig out a large HOT or something so I can do things properly.

Marko

Tue May 01 2007, 01:06PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

hi... this is how I imagined a typical UC3842 SMPS to look like.

This one uses an optocoupler and TL431 shunt regulator for feedback.
Mosfet I currently use is 1N50.
I use DCR snubber, but it just wastes power and doesn't seem to do anything but cosmetically affect the waveform.

I heavily used SMD to get densest possible layout, but it was tedious.

Running at 10W, the mosfet gets only slightly warm, snubber somewhat more and in overall there is very little dissipaton in the circuit.

At more than that current limit trips in... I blame the much lower operating frequency than before.
I use 1,6nF timing cap instead of 1n, and current operating frequency is about 60kHz.
I was safe at 100 but when getting over 200kHz mosfet started to get hotter witohut performance benefits.

I have few things more to refine and then I hope to draw a schematic.

I actually got around to building it today, and woe and behold it worked! For a little while. Snubbing is a major problem and the snubber resistor was smoking in seconds. I ran it without the snubber just for a test run, and within a few seconds the IRFP450 became dangerously hot, so it definitely needs a snubber. Relying on avalanching isn't smart for long term use...

If anything overheats in seconds with no load, something is terribly wrong with the circuit.
I haven't tried something as big as IRFP450, but still.

Your snubber is definitely bad, I used 10k resistor and 10n cap in mine and it still got too hot in some cases.

joe wrote ...

I looked into the UC3842 again, did you know that it can be sampled from TI?

I already considered using it about a year ago when I discovered it in almost all of the old CRT monitors I had taken apart. There it is used to control the main (200W +) supply, so it sure is a good chip to use.

I can buy the IC in local store for $0,9...
Compared to blocking oscillators, I think benefits are really worth of price.

Dr. Shark

Wed May 02 2007, 09:29AM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

Doh, you guys are p0wning me so hard! I also had a go at prototyping the circuit from that App Note on a breadboard

but all it does is blowing IGBTs. IF I get it working, I'll try making a PCB for it.

uzzors2k

Wed May 02 2007, 10:11AM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

Firkragg is pwning you mean, I'm still trying to get it to run properly too.

I've found that I get alot of heating no matter what, without a snubber the mosfet takes the beating, with a snubber the snubber resistor takes the beating. I think I found the problem though, I suspect it has to do with my flyback inductor. For one thing I put too few primary turns on it, so it might be saturating, for the other the step down ratio is so high that it has to run full power and still won't give enough voltage. Therefore the PWM is relying on the current sense entirley, so around 50% duty, even when powering a LED. I read up on flyback converters and found out how the output voltage is controlled, it might have been smart to do that first...

Marko

Wed May 02 2007, 01:30PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Uzzors, if you know the number of primary turns and core length, considering it isn't gapped, you can calculate your maximum allowable current for induction of about 0.2T.
B = (uNI)/L

So a core wich is 5cm long and owns 50 turns will stand maximum I = (B*l)/(mi*N) =
(0.2*0.05)/(3.14*4*10^-7*2000*50) = 0,0796 A.

You then set your current limit resistor for that current.

If you are using a generic transformer, just increase value of your sense resistor until you get out of problems. You can't actually set it too big; when you run out of power, just increase the frequency.

Note that saturation current has nothing to do withh core cross-section, and that it is inversely proportional to number of turns!

Larger number of turns and cross sectional area will only increase inductance, store more energy and make dI/dt slower.

Gap will lower the inductance but increase the saturation B many times and allow for more stored energy.

Due to difficulties of flyback transformer design I prefer much more to use ready made transformers wich I have in dozens laying around.
Practically all transformers wich worked before with mosfets are fine.

They work OK for output voltages they were used for before, and you can even get 100% or so margin if you know what you are doing, but you simply can't push a 3V cell phone charger transformer to 20V out and live with 500V switch, ratio is just too high.

5:1..10:1 ratio is around fine for off-line converters wich require ~10V of feedback; mosfet will in that case see only about 80V over the supply voltage.

IIRC, the appnote schematic uses specified transformer with 1:10 ratio for auxiliary supply, so uzzors's transformer is indeed a bit in edgy configuration.

I think you could halve the number of primary turns and be fine.

Other things would be removing the 11V zener from the gate, it does nothing but overloads the controller.

Regarding snubber, it is advisable to use a softer recovery diode, anything with specified recovery from 500ns-2us. Reverse current during ringing is good because it transfers some of snubbed energy to the output.

uzzors2k

Wed May 02 2007, 07:49PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

W00t! I got it working right. Sure, its only rated for driving it's own standby LED, but it should just be a matter of using a smaller current sense resistor. The current 33 ohms sure beats 0.75 I was using earlier. No wonder things were heating, running at 1.3A peak and an estimated 50% duty. Currently nothing gets warm save the bleed resistor.

Firkragg, so to get more power I should use less turns to allow for more induction current on a given core? And use a corresponding current sense resistor of course. Right now I'm using a 50 turns primary with a 2 cm long core. 10 turns for aux and secondary. That equals 0.03A, and my guess is 0.5% duty. The output power is low, and even a 560 ohm resistor loads the power supply down.

1178135365 95 FT24184 Flyback Converter First Light

Marko

Wed May 02 2007, 08:31PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Generally experimenting with walue of sense resistor will make you right.

The supply will act as a voltage source but when current limit is overreached it effectively becomes a current source and output voltage sags. When it is set right, output can even be indefinitely short-circuit immune.

So the max current per cycle (energy) and frequency are what determines your output power.

In my featured supply, I used 4ohm sense resistor but after lowering it down to 2 it continued to work well.
My transformer is although many times bigger than those little PC power supply transformers and seems to have a small default airgap.

uzzors2k

Sat May 05 2007, 01:30PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

Well, I finished the design, but its not quite as good as I wanted it to be. The output power is only around 10W. I was in the process of tweaking it when the UC3842 just failed spontaneously, and it was my last one too.

1178371853 95 FT24184 Uc3842 Flyback Converter

Marko

Sat May 05 2007, 02:50PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Ok, here's final schematic for mine:

Optocoupler isn't linear enough and you need another ''error amplifier'' on secondary side, wich can be a zener or TL431 if you need variable voltage and good stability.

Regarding your circuit, I actually wouldn't expect for the small ATX auxiliary tansformer to output much more than 10W... you can try increasing the frequency, but mosfet will start getting hot if you go too high.

And regarding UC3842 failure, I really don't know what happened there.
I killed a dozen of mosfets by all ways I could think off but I never managed to kill the controller.

IC has robust 1A output driver, and failure of mosfet removes power from the IC instantly.
You can increase gate resistor to something >20ohms to be safer.

previous 1 2 3

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob