Push-Pull Forwarding Converter for Solar

[Madscientist267]

I'm starting two new threads out of one old one:

http://fieldlines.com/board/index.php/topic,144628.0.html

Which left off with two rather intriguing concepts for pseudo-MPPT for PV.

For the new true buck converter thread, see:

http://fieldlines.com/board/index.php/topic,144675.0.html

More to come...

Steve
  

[Madscientist267]

Ok, so I left off with having the basic design built and on the table, but made another somewhat careless oversight.

The unit works mysteriously well at 12V nominal, but while I adjusted one component for the inevitable increase in input voltage (~39V under no load), I managed to not give proper due to the bias/drive resistors for the much higher voltage. I'm pretty sure that this is what caused the next failure I caught...

Smoked the KA3525 after trying to tune the PWM feedback pot. I think the increase in duty cycle translated into too much heat from the excessive current flowing through the bias/drive resitors. At lower duty cycle, it was able to keep cool enough and avoid triggering of the smoke release valve.

That's ok, I have about 6 or 7 more of these boards that I can cut up and it's a cinch to wire it up and have it running, although the next one will mean 6 additional components get replaced - 4 drive and 2 bias resistors.

I also need to tweak the feedback mechanism to allow more precision control of the threshold.

Steve

[Opera House]

Are you doing anything to limit the voltage to the chip and fet to under 20V?  Hard to comment when I can't figure out a thing you are doing.  Taking a picture of a schematic isn't that hard.

[commanda]

This is an SG3525 forward converter I built using a re-purposed pc psu transformer for powering cfl's.
It may give you some ideas.

http://fieldlines.com/board/index.php/topic,129295.0.html

Amanda

[Madscientist267]

LOL @ opera -

I usually design in my head, but this time I haven't been paying as close attention as I should; been rushing things a bit much. Haven't really given it much thought; the boards that I am scavenging from are a dime a dozen... 

Usually, after it's over, I draw the schematic up so that I can replicate the ckt later after the original design has been tested and potted. Not always, but usually. Sometimes I have to design from paper to board, but these are relatively simple and it's really just been oversight that has been a problem. Would have done it on paper as well for the mistakes I have made.

This becomes especially true when I'm prototyping on an 'already-made' board like these are. I'll scribble one out soon so that you guys have a good idea of what it really looks like.

I'm convinced at this point that the next version of this I create will work perfectly.

When I saw the 7812 in Amanda's MC34063 design, I realized that I'd have to be stupid to not do that to the KA7525 as well... Solves a few problems actually. As long as the regulators can handle the panel OTV (I remember reading something about Vsmoke=37V), I will leave it in on both designs.

Amanda -

That looks very close to what I'm working with from the inverter units. The main difference is that I'm not using the error amp at pins 1/2, and am feeding the input voltage back into pin 9 instead (PWM comparator).

More as it comes...

Steve

[commanda]

Amanda -

That looks very close to what I'm working with from the inverter units. The main difference is that I'm not using the error amp at pins 1/2, and am feeding the input voltage back into pin 9 instead (PWM comparator).

More as it comes...

Steve

Pin 9 is the output of the error amp. Not a good thing to use as an input.

Can I suggest you connect pin 16 to pin 1, and use pin 2 as your input.  Your voltage divider from the input voltage should make 5.1 volts at pin 2.

Amanda

[Madscientist267]

Problem there is that it inverts. In versions 1-3, I left the inputs to the 1/2 error amp floating (well, pin 1 floats, 2 is connected to 16). Considered grounding pin 1 to make sure it didn't affect pin 9.

When pin 9 goes higher than Vref (internally), the PWM drives harder.

When pin 1 goes higher than pin 2, pin 9 goes low, shutting down the PWM. So I'm probably going to use pin 1 as my output sense, to cut off the PWM when the battery is full.

I think the output from the first error amp is clamp-to-ground rather than pull high/low, so as long as pin 1 floats lower than pin 2, or is grounded, it should 'stay out of the way' of the input from pin 9.

I'm basing this on the fact that I observed the output from the first error amp as seemingly having no effect on the input given at pin 9. It may be a tad unorthodox, but clamp-to-ground to shut down is what made it viable to use as my feedback from the input voltage.

Make sense?

Steve

[commanda]

Problem there is that it inverts. In versions 1-3, I left the inputs to the 1/2 error amp floating (well, pin 1 floats, 2 is connected to 16). Considered grounding pin 1 to make sure it didn't affect pin 9.

When pin 9 goes higher than Vref (internally), the PWM drives harder.

When pin 1 goes higher than pin 2, pin 9 goes low, shutting down the PWM. So I'm probably going to use pin 1 as my output sense, to cut off the PWM when the battery is full.

I think the output from the first error amp is clamp-to-ground rather than pull high/low, so as long as pin 1 floats lower than pin 2, or is grounded, it should 'stay out of the way' of the input from pin 9.

I'm basing this on the fact that I observed the output from the first error amp as seemingly having no effect on the input given at pin 9. It may be a tad unorthodox, but clamp-to-ground to shut down is what made it viable to use as my feedback from the input voltage.

Make sense?

Steve

It does make sense actually.

[Madscientist267]

I got the basic circuit done... It works fairly well, didn't take any measurements yet, but it appears to be passing a fair amount of power.

To get around the Vmax issue with the 78xx regulators, I'm using a 7812 with a ground lift from a 5.1V zener, providing ~17V to the chip, and allowing the regulator to operate on as much as '40V'. I wanted to use a 7805 with 2 zeners, but I am out of them at the moment. Would give me a little more clearance on the input, as well as lower the chip supply to ~15V.

The MOSFETs run a little on the warm side, the bias resistors (gate to source) are probably a little bit high (10K), with the drive resistors being 150R.

I think the result is some crossover causing both sides to be on momentarily during switching. I'm going to try dropping them somewhat, maybe down to 1K and see where that leads me.

Another problem is I'm also out of my 10K cermet trimmers at the moment, so I'm having to use regular pots and this thing is touchy.

I'll work up a schematic as soon as I get the finishing touches applied.

Steve

[Opera House]

Put that next zener in series with power to 7805.  You don't want gate drive going over 17V.

[Madscientist267]

Hadn't thought about it that way... think thats could be partially responsible for the excess heat?

But I wonder one thing... Small zeners aren't generally able to pass much current, and I got a little bit of heat out of the 7812 last night during the smoke tests...

They are 1N4733A.

Think they can handle it? I mean, they're only rated for 49mA, albeit an 890mA surge (8.3ms).

Steve

[Opera House]

39-17 leaves 22V dissipating in the 7812.  IF the current is  50ma, that will be over a watt and that is hot.  A 5V zener would take a quarter watt off that.  A 180-200 ohm 1W is probably ideal in place of the zener .

[commanda]

The MOSFETs run a little on the warm side, the bias resistors (gate to source) are probably a little bit high (10K), with the drive resistors being 150R.


Steve

Steve,

The 3525 has a totem pole output. No need for a gate-source resistor. The gate drive resistor needs to be a lot smaller. You also need a decent sized cap on the supply line to provide the high gate switch-on current.  That gate capacitance on the fet needs to be charged as quickly as possible to get clean switching.  Dedicated driver chips are often able to supply turn-on peaks of several amps.

As for the high input voltage to the 78xx regulator.   SOP is to use an emitter follower pre-regulator.

Amanda

[Madscientist267]

Hmm, the pull-downs were already there in the OEM implement, so I just left them... ?

Define 'decent sized cap' - 100uF, 470uF, 1000uF? And I'm assuming right at the Vcc/Gnd pins on the chip?

The one that is there is a 47uF/25V...

As it sits right now, there is a 4700uF/50V right at the input from the panels, which leads across to a set of 6 22uF/50V electrolytics tied on a 'reinforced' buss that sits as close to the toroid/MOSFETs as possible.

From this, there is a trace that leads off to the 3525 headed for Vcc, which I broke in the middle to put the 7812 in. There is another 22uF/50V directly across the input of the regulator, and a 47uF/35V directly across the output, both negatives tied to pin 2 of the 7812, not to battery ground. 5.1V zener ties pin 2 of the 7812 to battery ground.

The 47uF/25V is OEM and is the last cap in the chain.

As far as the driver resistors, can they just be completely eliminated, along with the bias resistors, or will that put too much stress on the 3525?

Steve

[commanda]

Your cap situation seems pretty well covered.

As far as the driver resistors, can they just be completely eliminated, along with the bias resistors, or will that put too much stress on the 3525?

Consider the gate - source to be a (discharged)capacitor. Now switch it high in zero nanoseconds. How much current flows in that first minute (very very small, not the 60 second kind)  brief instance while the gate cap charges.
Where does this current come from? And how much noise will it put on that supply line.  Unlikely to damage the 3525. Not impossible, but unlikely.
There is a reason why pcb's have 100nF caps across the supply lines all over the place. To filter out all the switching transients.  So why skimp on 2 x 1/2 cent resistors when good engineering practice says they should be about 10 - 15 ohms, limiting the gate transient current to an amp or so.  This should allow the fets to switch fast enough. Your 150 ohm will slow them down too much.

On another note. If you check out the datasheet on that chip, there is the possibility to adjust the dead-time.  There's also another pin specifically for shutdown.

Amanda

[Madscientist267]

Ok, that sounds good. I'll give it a shot...

In lieu of the soon to come schematic, here are the most recent shots of the board...


The parts side.





The underside. You can see the SMT bias/drive resistors along the top edge of the board.





Here is a shot of another board where the traces got mangled when I was removing the original transformer. The path, though, is still visible.



So you think that the heating is a combination of the dead time being too close, too high value for the drive resistors, and inability to throw a sharp punch at the MOSFETs?

Also, meant to ask this last post - What is causing the heat in the 7812? Just the dissipation? Or could the above have something to do with it?

Steve

[Madscientist267]

Also, what is the 10K in the 'Test Circuit' section of the datasheet (attached it for convenience) between pins 9 and 1?

Is that some form of hysteresis?

Steve

[Madscientist267]

Ok, smoked a couple more components... ok actually 4 more MOSFETs and the KA3525 at the same time... all due to a very careless mistake where I accidentally shorted a couple pins on the chip while it was running...

Got smart tho, reduced back to 2 MOSFETs instead of 4, so I don't blow so many at once when it DOES hit the fan, and socketed the KA3525 so I can swap it out easily...

I have an interesting problem though - there's some hysteresis in the feedback going to pin 9 that I can't quite figure out... Anybody have any ideas?

I know... schematic.. I'm working on it!

Steve
 

[Madscientist267]

I think I found out at least why the MOSFETs keep popping... it seems to happen when there's a sudden change of load (either way).

I thought of the idea when I kept getting zapped slightly when I was feeling the heatsinks and had my fingers across the phases. It wasn't a painful shock, but definitely above threshold for feeling it.

I threw a couple of the same diodes I used to make the bridge on the output across each of the ends of the primary to ground, and the tingling went away, but I haven't abused it enough yet to know whether I have eliminated the tranny popping frenzy.

I still have the hysteresis issue, so I'm starting to think I may have been wrong about the first error amp not having a totem pole output to pin 9. Not completely convinced just yet, but it looks like a real good possibility. The datasheet is of little help for this, unfortunately. This has been a bit of hacking from beginning to end... 

I may have to revert to something similar to what I had going for the MC34063, with a bipolar tranny to invert the input from the panels so that it's properly in phase as it moves it's way through the chip.

We will see. Timing is a bit off ATM, and guess what? It's CLEAR outside today, but I have that whole family thing to do, so... Figures... LOL

Merry Christmas everybody!

Steve

[commanda]

Steve,

For Xmas, buy yourself an oscilloscope.  I have a feeling you're going to need it, sooner rather than later.

Can I suggest you connect pin 16 to pin 1, and use pin 2 as your input.  Your voltage divider from the input voltage should make 5.1 volts at pin 2.

AND, put that resistor from pin 9 to pin 1 like it shows on the datasheet.

And merry xmas everybody.

Amanda

[commanda]

Steve,

Check out the first cct diagram in this link.

http://sound.westhost.com/project89.htmhttp://sound.westhost.com/project89.htm

Particularly R1 C1 & R2 C2. It's called a snubber network.

Also,
C17
C18
R14
R13 - rhs to Vref
remove R16
non inv ip to voltage divider from input.

Amanda

[Madscientist267]

Will have to look into it..

I have an old military surplus 4MHz scope in the attic that still works surprisingly well.

But for the moment I'm sick of popping MOSFETs, and so far the efficiencies I'm seeing are not in the amicable range, as was pointed out earlier.

With that, this particular version is on hold ATM. 

Steve