UC3842 basics

[ben7]

Re: UC3842 basics

??? any ideas?

[ben7]

Re: UC3842 basics

Hm, I just realized that the output isn't regulated, so I looked and I had the opto hooked up wrong xD So I fixed it, but now the power supply wont start up and it just makes a ticking noise. I set the TL431s voltage to 4.7v. Any ideas? :S

I am using a 5vsb transformer.

[ben7]

Re: UC3842 basics

I found a UC3844 in another PSU and I salvaged it, and built the power supply (with snubber). It seems to work well

[ben7]

Re: UC3842 basics

So, my UC3842 wasn't quite fried, until today xD

I added an RCD snubber to the primary just in case.

Of course I went to fiddle around with it more today, and first, I shorted a mosfet. Later, I tried a new mosfet, only to have the 18V zener and 1N4148 and the UC3842 go shorted on me

The second mosfet is still fine, and the 150 ohm resistor is still good too, so idk what the heck happened
(Ugh the components in the schematic aren't numbered xD)

[budm]

Re: UC3842 basics

Example diagram left out a lot of circuits that should have been included.

[Shocker]

Re: UC3842 basics

What do you think is wrong? I got the UC3842 from a working psu, but do you think I could have fried it?

With static, possibly. Same for the MOSFET.

[ben7]

Re: UC3842 basics

Hm, I made the circuit in the first post, but for some reason the UC3842s power voltage doesnt go above 12v! I used a 20v zener, 10uF cap, and 100k startup resistor. It only seems to be slighty working, with only about 2vac on the transformer primary xD Tried switching the feedback windings around with no luck, and tried both the windings on the primary of the transformer, again no luck

What do you think is wrong? I got the UC3842 from a working psu, but do you think I could have fried it?

[tom66]

Re: UC3842 basics

You need a current clamp capable of at least several MHz bandwidth.

Ghetto solution: 0.1 ohm noninductive resistor and two scope probes, use A-B feature.

Single channel solution (and isolated to some degree): Use a pulse transformer across the resistor (primary) and use one scope probe to measure across the secondary. Disadvantage is you'll probably not capture the lower frequencies...

When you get that sorted out, the most important harmonics I believe are the 3rd and 5th... don't quote me on that though... You could build a band pass filter of some kind so you only see these, then measure the amplitude...

Here's some reading:

Or you could just use a massive common mode coil, X cap (330nF or bigger), differential mode coil and X cap (>=330nF.) Those usually take care of most noise... Also don't forget some Y caps, safety rated of course... and the EMI discharge resistor, must have a certain time constant... EMI/RFI is a nightmare!

[Th3_uN1Qu3]

Re: UC3842 basics

Okay, progress. I was getting oscillations under some degrees of load regardless of compensation. So i figured there must be some parasitics getting in somewhere.

Scoping the gate drive of the mosfet revealed that even when the transformer was quiet, every other cycle would terminate prematurely. I fixed that by increasing the capacitor on the current sense pin of the 3842 - due to poor layout the spikes at the mosfet source are quite big. But it still hissed at some loads.

So, i took my soldering iron and started messing around the board. When i put my iron on the pin where the RC network that sets the oscillation frequency connects to the 3842... it shut up. Silence at last. Sure took me a long time to find that a cold solder joint was to blame.

Now i can consider the design finished, and will be starting work on a PCB layout. In the meantime, i'd like if someone could help me in regards to EMI filtering - i don't exactly know how much i need and what the rules are.

[Th3_uN1Qu3]

Re: UC3842 basics

In the UK, college is basically the same. Electronics, although best practiced in a practical manner, is not taught this way.
[...]

Hey, at least they're trying. That still looks lots better than what we do over here. I won't elaborate on that, it's getting late. Maybe some other time.

My load tester project is currently on the back burner. But it has 30 high power MOSFETs lined up for it. (Bought some surplus IRFZ44N's, good for 60V or so.) I have plans to make it dissipate up to 900W peak, 600W continuous - but I'm currently too busy on other projects. I may think of a way to integrate it into a university project, although it looks a bit ugly (made from an old stereo amplifier case.)

I have two dozen IRF530s and a heatsink that can dissipate 1kW+ on a continuous basis. I have the whole control circuitry thought out as well (nothing too fancy, just a dual precision opamp taking input from a pot, sensing current from a shunt, amplifying it and keeping it constant; one for each rail). I just don't have a case to strap it to.

The load that i built for testing this little power supply was just two resistors in series, that used to be minimum load resistors in an ATX PSU (underrated for this task, every 10 minutes or so they melt the solder that keeps them connected ), and a random TO-247 FET. An IRFP460 i think it is. The gate of the FET is driven from my laptop via my headphone amplifier. It doesn't get anymore ghetto than this. I use Sony Sound Forge for generating the signals, as it's the only piece of software that will do 192kHz on my laptop's onboard soundcard, other software that includes a waveform generator (such as Visual Analyzer) seems to only detect the soundcard as 96k capable. At 192k sampling rate i get reasonable squarewaves up to 7kHz.

[tom66]

Re: UC3842 basics

If they use 650v ones for 5vsb, it's good measure to not use anything smaller when you try and get higher voltage out of the same transformer. I have to get the compensation calculations done, now i put the "right" resistor for the optocoupler's emitter (for maximum bandwidth of the opto), and before, it was reasonably stable, now it's unstable at almost all loads, and does all kinds of nasty things at low frequency pulsed loads.

I only wish we'd do this kind of thing at uni, as somehow i have no problem spending hours on calculating a power supply control loop, but i cringe whenever i have to do calculations for lame projects at uni... I suck at math, but for some reason, whenever it has to do with electronics, it comes out okay.

School in this country is a major letdown. Lots of damn near useless theory, and next to no practical applications of the theory studied. If you ask the teachers for help in a practical issue, they are at a loss for words.

Btw, a pulsed load (a bunch of power resistors and a big FET driven by a signal generator) is a great tool to test stability of a power supply. Measuring the power supply loop response typically involves the use of a network analyzer, which is out of the reach of most hobbyists. There is another method that only uses an oscilloscope, a signal generator and a cheap transformer (can even be a regular mains transformer), but it works at low signal levels and i'm not sure i'll see anything useful on my scope. With a x10 probe i only get 100mV/div minimum. Gotta build an amp for this thing sometime soon. For now i'll do the math and see if i come up with a stable result.

In the UK, college is basically the same. Electronics, although best practiced in a practical manner, is not taught this way. For example, the laughable switch mode power supply example (and this is from the head examiner's book, I kid you not):



The bridge recitifier is unnecessary, and I'd go as far to say it wouldn't work - with a flyback supply, this would conduct in both "quadrants" (primary current ramp up then secondary current ramp down) - at best, it would be some kind of hodge-podge of a push-pull or forward converter with four times unnecessary diode cost. No winding indications on the transformer mean I can't tell if it is a flyback or something else.

Honestly if they are going to teach switch mode power supplies by far the easiest topology to understand is the boost converter. An inductor, switch, diode and capacitor. You can even make one with a mechanical switch if you've got a large inductor and capacitor. It's relatively easy to understand that interrupting the current through an inductor induces a voltage across it; put a capacitor on the output to store this energy, and add a control circuit: viola, an easy to understand boost converter. Once you introduce a layer of isolation, you really need to consider compensation loops, frequency response, transformer design and layout, feedback topologies, opto current...

This is not the only mistake in the specification but it was the most irritating for me.

There was a recent question on SMPS's which tripped me up - see if you can spot the pitfall.



I drew the output of the comparator ramping up and then switching on and off (exaggerated as very slow and of course ignoring the fact that it wouldn't be a stable oscillation.) But the exam paper showed the output starting immediately at +12V as if powered by some mysterious extra power supply. Also, if the supply were to start up normally, it would have to have the comparator reversed, or a startup timer would be needed where it would ignore the opto for the first t milliseconds.

...

My load tester project is currently on the back burner. But it has 30 high power MOSFETs lined up for it. (Bought some surplus IRFZ44N's, good for 60V or so.) I have plans to make it dissipate up to 900W peak, 600W continuous - but I'm currently too busy on other projects. I may think of a way to integrate it into a university project, although it looks a bit ugly (made from an old stereo amplifier case.)

[Th3_uN1Qu3]

Re: UC3842 basics

If they use 650v ones for 5vsb, it's good measure to not use anything smaller when you try and get higher voltage out of the same transformer. I have to get the compensation calculations done, now i put the "right" resistor for the optocoupler's emitter (for maximum bandwidth of the opto), and before, it was reasonably stable, now it's unstable at almost all loads, and does all kinds of nasty things at low frequency pulsed loads.

I only wish we'd do this kind of thing at uni, as somehow i have no problem spending hours on calculating a power supply control loop, but i cringe whenever i have to do calculations for lame projects at uni... I suck at math, but for some reason, whenever it has to do with electronics, it comes out okay.

School in this country is a major letdown. Lots of damn near useless theory, and next to no practical applications of the theory studied. If you ask the teachers for help in a practical issue, they are at a loss for words.

Btw, a pulsed load (a bunch of power resistors and a big FET driven by a signal generator) is a great tool to test stability of a power supply. Measuring the power supply loop response typically involves the use of a network analyzer, which is out of the reach of most hobbyists. There is another method that only uses an oscilloscope, a signal generator and a cheap transformer (can even be a regular mains transformer), but it works at low signal levels and i'm not sure i'll see anything useful on my scope. With a x10 probe i only get 100mV/div minimum. Gotta build an amp for this thing sometime soon. For now i'll do the math and see if i come up with a stable result.

[tom66]

Re: UC3842 basics

The FET won't avalanche unless you exceed its breakdown voltage, which in my case does not happen.

Yeah, just saying a 450V FET would work for low power. Probably not 12W though. The only way to find out involves letting out magic smoke from expensive transistors, so best to design in a good margin.

[Th3_uN1Qu3]

Re: UC3842 basics

The FET won't avalanche unless you exceed its breakdown voltage, which in my case does not happen.

[tom66]

Re: UC3842 basics

You can get away without a snubber if and only if:

• dV/dt won't kill your switcher or turn your FET on for too long
• Avalanche energy is kept under device maximum and power dissipation is within limits

Both are very hard to achieve, especially the second one for outputs more than a couple watts... plus an RCD network is much cheaper than a bigger heatsink.

[Th3_uN1Qu3]

Re: UC3842 basics

I don't see any RCD snubber on the primary, that FET won't last long under high load.

Also has anyone spotted that the fet is labeled as 450v? That won't work... I used a 650v fet, and scoping the drain reveals that the spike tops at 550v, thus no need for a snubber. But a 450v most certainly won't work. I'll be leaving room for the snubber when i make the PCB.

Using a 350V rated input cap isn't a great idea either, as it's not too hard to exceed it. And what about EMI filtering???

At UK maximum 253V that puts 351V across cap (accounting for diode bridge drops.) 400V caps are cheap and give a much safer margin.

I used a 47uF 450v Samxon as that's what i had lying around. As for EMI filtering, i'm working on it. Also note that in the original schematic there is no fuse either...

[tom66]

Re: UC3842 basics

Using a 350V rated input cap isn't a great idea either, as it's not too hard to exceed it. And what about EMI filtering???

At UK maximum 253V that puts 351V across cap (accounting for diode bridge drops.) 400V caps are cheap and give a much safer margin.

[Shocker]

Re: UC3842 basics

Using a 350V rated input cap isn't a great idea either, as it's not too hard to exceed it. And what about EMI filtering???

[tom66]

Re: UC3842 basics

I don't see any RCD snubber on the primary, that FET won't last long under high load.

[Th3_uN1Qu3]

Another episode in the power supply saga courtesy of yours truly.

This time, we will be focusing on the UC3842 current-mode controller. We will be making a small, cheap, low parts count, low power flyback supply using a 5vsb transformer from an ATX PSU.

A current mode controller has two feedback loops. One for voltage, one for current. Along with sensing the output voltage, it directly senses the current in the power switch, enabling better line regulation, and overcurrent/short circuit protection comes free. Done properly, a current-mode SMPS cannot be damaged by a secondary side overload or component failure. Trust me, i tried.

It also makes compensation easier, but that's something i'll get into a while later, gotta figure it out myself first.

I will be loosely basing my work on this design:


The general idea is good, but there are a few things that could use improvement. I've built my version on a perfboard. It looks like a disaster waiting to happen, but it works just fine. It'll drive anything all day long, and gets away without any heatsinks. Outputs 12W or so before going into current limiting, and i made it adjustable up to 15v. Flyback topology is nice in that regard. I'll be posting an updated schematic soon, and i'll also be making a PCB i will share here. Time to put those scrap 5vsb transformers to good use. I see this as an all-round wallwart replacement.

Right now i'm using it to charge the SLA i drained during its development. I don't have an isolation transformer so i'm using an inverter and battery when i want to scope stuff in the primary of a SMPS.