UC3842 basics

[Th3_uN1Qu3]

Re: UC3842 basics

Here's a preliminary schematic of what i'm playing with right now. Does 3A "in spec". Some tweaks left to do on the compensation and the snubber across the diode, but other than that it's pretty solid. More on that further down.

I won't post the PCB i made for this schematic because there are a number of things that need to be fixed in it... Trust me, it's sitting on my bench, and it's been trouble. But it is a working PCB with good layout for the most part, not a messy protoboard. Also, as you can see, there's no input filtering yet.

Notes:
Needs a minimum load for the aux voltage to rise high enough for the 3842 to operate properly. This can be lowered by reducing R4, but expect higher dissipation in it and the zener if that's the case. Btw, like it is now, the zener is OPTIONAL... The 3842 has a built-in zener. It's not a good idea to abuse of it, but so far, it works.
It is not necessary for the secondary ground to be connected to mains earth. I made it that way in the schematic to make a clear difference between primary and secondary grounds.

The current configuration on my PCB is as follows:

• 1 ohm resistor for the fuse... i learnt from the Chinese. Temporary, don't be foolish. Will install a proper fuse once i fix the PCB... not enough room.
• 2A bridge rectifier
• Samxon 47uF 400v primary cap. I have a 10uF around somewhere, will try with that as well. Don't have any 22uF tho.
• 2SK2545 primary switcher (600V 6A 0.9 ohm TO-220F).
• UF1002F rectifier diode on the secondary. (200V 10A TO-220F).
• UCC KZE 1500u 6.3v after the diode.
• 4.7uH inductor for the pi filter.
• UCC KZG 820u 6.3v after the pi filter. I thought it can at least do that.

There are no heatsinks on any of the parts.

Test results:
At 3A load, everything runs in the low 70s at most, with one exception. The UF1002F rectifier reached 112C with 3.5A load... at which point i went back to 2.5A where it doesn't exceed 100C. Will need to use a Schottky, and/or a small heatsink.
Output ripple is very low at all loads beyond the minimum, switching noise is low as well. It was well worth spending a few hours measuring and calculating the proper values for the RCD clamp on the primary switch.

Looks like a 5vsb transformer from a cheapo ATX PSU can actually do 3A continuous output. Most cheapies fall way short of this number however, due to poor design of the switching circuit.

That's it for now, stay tuned for more.

[ben7]

Re: UC3842 basics

Post pics!

You should teach us how to calculate all that stuff xD

[Th3_uN1Qu3]

Re: UC3842 basics

Pics... once i get a few minutes on the DV5. I've lost the cable for my camera long ago, and i don't have an SD reader on anything else.

I will publish the math when it's all done, right now the schematic is preliminary and subject to change. For example, i'm changing the series resistor of the optocoupler now, with a value that is quite far from what's on the schematic. Reason being that the EL817 opto i'm using has a higher minimum CTR (current transfer ratio) than the 4N27 in the Astec appnote i've based my calculations on.

The series resistor marked as Ropto in the schematic is now 680 ohms instead of 100 ohms. The power supply is now unconditionally stable at all loads above the minimum, with just one 100nF capacitor across the 431 for compensation. Aka how most people do it. I plan to take it one step further and improve the reaction time with a proper compensation network, but that needs more math.

You will discover that component values in critical places like feedback loops and snubbers are actually quite flexible, if you keep certain ratios the same. There is no "optimal configuration", hence why every designer seems to do it differently.

[Th3_uN1Qu3]

Re: UC3842 basics

Had some more tinkering time today, updated some component values. New probes arrived for my scope and together with this excellent appnote from TI (attached), i was able to measure some aspects of the control loop of the actual physical power supply.

The current schematic is stable from 0.25A to 3A with resistive load and has a crossover point of ~1.5kHz with 60 deg phase margin. More control loop optimization to follow with pulsed load testing... Fun.

I also found a really nice document on SMPS control loops, attached that one as well.

[ben7]

Re: UC3842 basics

Good job

[Th3_uN1Qu3]

Re: UC3842 basics

I've been chasing this ~5kHz oscillation for a while, which manifested itself at low loads. No amount of compensation would fix it, even drastically lowering the crossover frequency merely hid the issue rather than eliminate it completely.

It turns out this wasn't a phase shift issue but a DC operating point one, and the culprit turned out to be the optocoupler. The clue was that when injecting signal at low load, the waveform distorted quickly, at a much lower level than it did at 1A or higher. Something was going out of its linear region, and the opto was that something.

It may not be immediately obvious, from the designation "switching power supply", but while the power devices are operating in switching mode (saturated/cutoff), the control loop is a linear system. When any of the devices in the control loop goes out of its linear operating range, the whole compensation math goes out the window, and the loop is said to be "open" instead of "closed".

The fix was to increase the optocoupler operating current on both sides. On the primary, this was done by adding a resistor from Vref to Comp on the 3842, and on the secondary, by decreasing the series resistor of the optocoupler.

This also elegantly solved the minimum load problem: previously, the current drawn with no load was not enough to keep the UC3842 operating, it would kick in and out of UVLO (tick-tick-tick). Now the power supply works fine without any load - this avoids the use of a more wasteful minimum load resistor.

I'll post an updated schematic after i run all the numbers again and determine the optimum values.

The cookbook i attached earlier was missing its appendixes with examples. I attached a more complete one. The appendix on boost-derived topologies (such as the flyback) is still missing tho, i'll see if i can dig it up.

[ben7]

Re: UC3842 basics

Wow, well explained! Thanks!

[kaju_badcaps]

Re: UC3842 basics

I have a similar problem mentioned with a UC3842A based flyback converter.
On no load condition the converter is oscillating, but with 10% or more load it is dead silent and performing pretty well, regulation is also upto the calculation. But , as I am trying to couple it with a audio amplifier, it should not oscilate on no load condition also. The problem is the DC operating point problem. I have read the solution of adding a resister from Vref to Comp, but I need to know that if there is a calculation for measure the resister between the Vref to Comp, means what should be the value of the resister? Already I have fried one Mosfet, Two Caps, two main rectifying Diodes by trying trial and error method, I don't want to add a load resistor, it will be a efficiency loss.

It will be very nice if any of you comment on that which will help me to fix the problem.

Thanks in advance.

@Th3_uN1Qu3: I have read in another discussion that you have fixed this kind of problem by adding a 1k register from Vref to Comp, How did you calculate this value?

Converter Specification: input 100 - 265V AC, Output - 14V 5A and the circuit is same as fig 6 of "AS3842 SWITCHING POWER SUPPLY CONTROL
LOOP DESIGN.pdf" at page 5(164)(Uploaded by you on page 2 of this discussion).
Transformer Spec: Pri ~275uH(40Turns), Sec ~21uH(5 Turns), Aux ~30uH(7Turns), Core EI28 with a air gap of 0.47 mm

Please Help

[daverave56]

Re: UC3842 basics

i have a problem with a UC 3842 which is used in my oscilloscope power supply,the chip will not power up,the supply voltage is around 12 volts and the reference pin is about 2 volts,someone told me it maybe a soft start problem,if so what pin of the chip and components are involved in soft start,there is also opto coupler nearby,can anyone please advise.

[ben7]

Re: UC3842 basics

i have a problem with a UC 3842 which is used in my oscilloscope power supply,the chip will not power up,the supply voltage is around 12 volts and the reference pin is about 2 volts,someone told me it maybe a soft start problem,if so what pin of the chip and components are involved in soft start,there is also opto coupler nearby,can anyone please advise.

Check to make sure the none of the electrolytic capacitors are dried out/faulty.

Also, next time, you should just start a new thread. A moderator will likely come and move this to a new thread for you.

[daverave56]

Re: UC3842 basics

Check to make sure the none of the electrolytic capacitors are dried out/faulty.

Also, next time, you should just start a new thread. A moderator will likely come and move this to a new thread for you.

sorry im new to this site,i did try and put this in the right section but i was blocked from doing so,not sure why i cannot post in correct section,my capacitors ive checked with esr meter and no bowed tops,im very frustrated trying to get this power supply to work.

[ben7]

Re: UC3842 basics

sorry im new to this site,i did try and put this in the right section but i was blocked from doing so,not sure why i cannot post in correct section,my capacitors ive checked with esr meter and no bowed tops,im very frustrated trying to get this power supply to work.

Do you know if the 3842's supply voltage is pulsing, or is it just a steady 12V?
The 3842 needs a minimum voltage of 16V to start up, and a minimum voltage of 10V for regular operation (after start up). The supply voltage really should be more than 12V.....

There will be a resistor (sometimes a few in series) that is connected to the rectified mains filter cap, which should be around 320V or so. Check to make sure that resistor isn't out of tolerance. (I know, it's hard to read the bands on those tiny SMD resistors LOL)

Oh, and, why are there tantalum capacitors (one is yellow and the other is blue) near the 3842? It looks like someone might have repaired (or tried to) this before. Looks like they were hand soldered. What voltage do you get on those tantalum capacitors?

Also, what oscilloscope is this? (Model number?)

[daverave56]

Re: UC3842 basics

the scope is a metrix ox 8620...the 12v supply is steady but the ref pin of the UC 3842 only measures about 2v...the tantalum capacitors in the photo are ones that i replaced some months ago but made no difference the scope being still dead.

[ben7]

Re: UC3842 basics

the scope is a metrix ox 8620...the 12v supply is steady but the ref pin of the UC 3842 only measures about 2v...the tantalum capacitors in the photo are ones that i replaced some months ago but made no difference the scope being still dead.

Have you tried replacing the chip? Only other thing I can think of is that the chip's VREF output is having too much current drawn from it. Hard to tell what that could be without a schematic.

[daverave56]

Re: UC3842 basics

i replaced the uc 3842 but no difference,i find the surface mounted resistors and capacitors hard to read.

[daverave56]

Re: UC3842 basics

checked all the resistors around the chip and all are ok,i wonder if the opto coupler circuit may be causing this ??? its so frustrating trying to fix this switch mode power supply

[rievax_60]

Re: UC3842 basics

the scope is a metrix ox 8620...the 12v supply is steady but the ref pin of the UC 3842 only measures about 2v...the tantalum capacitors in the photo are ones that i replaced some months ago but made no difference the scope being still dead.

The VCC pin needs to get to 16V for the UC3842 to start. After it starts, it can settle down to a lower voltage but needs to stay above about 10V to stay working.
Some faults can cause the VCC pin to rapidly cycle between 16V and 10V causing an average reading of abut 12V. You should change the VCC bypass capacitor again with an ordinary electrolytic.
Are the output rails absolutely zero volts?

[stj]

Re: UC3842 basics

vcc bypass should be atleast 47uf 50v

[ben7]

Re: UC3842 basics

The VCC pin needs to get to 16V for the UC3842 to start. After it starts, it can settle down to a lower voltage but needs to stay above about 10V to stay working.
Some faults can cause the VCC pin to rapidly cycle between 16V and 10V causing an average reading of abut 12V. You should change the VCC bypass capacitor again with an ordinary electrolytic.
Are the output rails absolutely zero volts?

Yeah, I'm still wondering if the meter is reading a steady 12V even though the voltage could actually be pulsing, that would be an indicator that there might be an output short.

Check the output rectifier diodes and output rails of the power supply for any short circuits. Want to make sure we weren't chasing a dead end.

[daverave56]

Re: UC3842 basics

i replaced the two tant capacitors with electrolytic 63v 47uf and did some voltage tests on the UC 3842 chip.

Pin 1 - 0v
Pin 2 - 0v
Pin 3 - 0v
Pin 4 - 0.7v
Pin 5 - ground
Pin 6 - 0v
Pin 7 - 13v
Pin 8 - 2v