Building your own power supply: An actual example

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Fixed. The transistors are fine. I found that the secondary of the transformer was ringing like crazy at 250kHz! Obviously most of that made its way to the primary as well and disturbed the switchers. The culprit was the output inductor, i was again testing one with a lot of turns on a pair of T106-26 iron powder cores. Replaced it with a ferrite core one and all the ringing has gone away. Transistors are now cold.

Of course i could've tried using a snubber. But instead of wasting power why not actually fix the offending part?

Morale: When in doubt, scope around.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Thanks. I've sped up the fall time a bit (still not using bipolar driver supply tho), and the result is.... the same. I'm puzzled.

The losses are far too great to only be rise/fall time related. It's like they are constantly hitting breakdown voltage. And they are (supposedly) 400 volt devices, while i got by with 250 volt ones in insulated TO-220 packages - how's that for heat dissipation - and yet they ran COLD at this load level, and could go for about a minute at 400W without a fan till things got really crispy. The IRFP360s heat up about the same like when i tried IRF640s (200 volt) just for kicks... Those ran for a few minutes but got really really hot and then one of them died. I don't want the same to happen with the IRFP360s because they ain't cheap.

They're either not what it says on the tin, or there's something else that i'm missing. Regardless, i'm going to hook up the two pairs of 740s in parallel, and if that works fine, then that's what i'm using. The differences in capabilities are marginal, and the two pairs of 740s still have roughly half the gate capacitance of 360s, so rise/fall times are likely to be faster too.

[b700029]

Re: Building your own power supply: An actual example

That's the Siliconix/Vishay logo, but the one on the left looks like it was stamped on upside-down and the labeling is also misaligned. All the pictures of IRFP360s I could find look like the one on the right.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Thanks, but i already had that datasheet. Anyway i think i found the issue. Due to my driving method fall time is slow. With the big capacitances of those fets, it's even slower. I have to get that bipolar driver supply going.

[Khron]

Re: Building your own power supply: An actual example

If they're from the pre-Vishay-acquisition, then they should look like it says on the last page of the datasheet (2004). The Vishay version doesn't specify how the markings should look...

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Here's how my IRFP360s look like. They both look dodgy IMO. What do you think?

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Almost completed assembly of the controller PCB. I'll post details, pics, and the layout tomorrow. I have to say that i made a little oopsie - somewhere along the way the scale got bugged up on the vertical axis. Which means that my trances ended up a little taller, and the hole pitch no longer is what it was supposed to be.

I found out only after i started drilling the board, so i let it be. I did have to persuade the TL494 a fair bit... but it all fit together eventually.

Now, i have a little problem. I got my IRFP360s. For some reason, they run really hot with not much load at all. I mean, i could go on forever at 60W load without a fan on the 2SK2341s or IRF740s, and the heatsink wasn't even getting warm. Now, it gets hot. Quickly. I have added 15 ohm gate resistors because the ringing was getting out of hand on those big FETs - it went all the way down to turning the thing off and then on again during a switching cycle - however, they didn't fix anything. I used to run with no gate resistors at all on the smaller ones. While this wasn't a good idea as ringing develops, the gate voltage remained high enough at all times for it to not matter. Even on the IRFP360s Vgs has a spike to 12v and then settles at 10v. And the datasheet specs are given at Vgs=10v. So they should work great. I have no idea why they don't.

Now, one little detail. I've never got from this shop two transistors that have identical writing. Every time i buy a pair, for some reason, the two MUST have a different font. It's like the clerk picks them up that way. And they usually have an just an "i" instead of the familiar International Rectifier logo with a diode symbol in the O. So far they have all worked fine but now i think i got fakes. I'll post some pics tomorrow.

Anyway, i also bought two pairs of IRF740s. Looks like four weak cheap FETs are better than two big and more expensive ones.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Okay, i can say i've figured out the source. Due to the messy wiring the mains electric field is making its way inside the feedback loop. For some reason it's worse with a ferrite core output inductor at low loads, but at max load it's pretty much the same whatever i use. Looks like some PCB design is in order before i can start troubleshooting.

I have fixed the main PCB dimensions to 100x200mm. This should be enough room to enable me to leave some space and decide on the output inductor after everything else is installed. The only limitation is the max height of 70mm, so the daughterboards and heatsink will need to reflect this. However, the bottom of the case is plastic, i can cut a little bit so 80mm fans will fit perfectly.

I have also decided upon the power transistors i'll be using for the final version. Might as well be doing it right so i'm gonna be running a pair of those.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

The same 400W load i've used till now.

However it's not that - the ferrite core of the output inductor is saturating (needs more turns) and effectively turns it into a piece of wire, which isn't too effective at filtering ripple... I was able to get much better results by messing around with the air gap, under 100mV at 50W load. So, need to take the core apart and put some more turns on it.

[goodpsusearch]

Re: Building your own power supply: An actual example

At high loads the output inductor started to buzz. I thought the regulation loop was oscillating but hooking up the scope revealed what was really going on - 2.1vpp of 50Hz ripple! My caps are too small, both primary and secondary. I knew i wasn't gonna get away with just 2x 330uF secondary caps, but the 2x 680uF primaries i thought will be enough. Even at a moderate load there's like 400mvpp.

You don't specify the load that made the 2x 680uF primaries insufficient.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Running stable from maximum output down to nothing connected. At the moment i'm testing an EI33 core for the output inductor too, i'm so lazy i'm just using the primary that was already on it (40 turns). It works great.

At high loads the output inductor started to buzz. I thought the regulation loop was oscillating but hooking up the scope revealed what was really going on - 2.1vpp of 50Hz ripple! My caps are too small, both primary and secondary. I knew i wasn't gonna get away with just 2x 330uF secondary caps, but the 2x 680uF primaries i thought will be enough. Even at a moderate load there's like 400mvpp. Though it's not a big deal for a 80 volt supply, this is going to power an audio amplifier after all so i'd like to keep it quiet.

I'm gonna have to parallel some more caps, let's see what i can do while keeping the same board size. Current schematic is attached. Pin 16 is of course also hooked to ground atm, but it'll do the current limiting so that's why i left it free in the schem.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

I'm going to use single ended drive for the GDT like i did till now (one n-channel mosfet and a 100 ohm 3w resistor across the GDT), so i won't need +/- driver supplies anymore, and may go for a standby supply to avoid any safety risk.

I'm proud to announce i have a TL494 successfully implemented on my protoboard. Now the fun part - stabilizing the loop.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Thanks. Now the mess has been moved onto the bench and has become a little, um, less messy. I don't step on random parts on the floor anymore.

I now have a good idea of the overall design of the power supply. The whole thing must fit on a 100x150mm board. To save space and simplify PCB design, the controller and protection circuits will be mounted on their own daughterboards.

The GDT driving circuitry must use split supplies for best efficiency. Since i already have provisions for +/-16v rails in the transformers (which will be regulated by LM317/337 to whatever value i find optimal for the driver circuitry of the amplifier), i figured it'd be a good idea to drive the GDT from there. So i don't need a standby supply - but then how would i start the 494?

I thought about it a little then i figured out something that should work - a low current non-isolated power supply (derived from a small capacitor in series with AC line) will start up the 494, then once the main supplies are up a relay will break this connection, making the supply fully isolated. I'm making a protoboard with a 494 atm.

[Pyr0Beast]

Re: Building your own power supply: An actual example

What a nice mess you have there !

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Ok, i think i've proven my point. Here's what my latest transformer does. 85 turns primary, 59 turns secondary, all with 3x 0.35mm wire. I dare anyone to try and shove more wire in that EI33. The DT838 on the left is measuring winding temperature btw. Do the math on the second picture.

This one doesn't get hot to the point you can cook eggs on it, but still enough for the tape to start melting, so i stopped after a few minutes. If i had proper high temp insulation tape it would probably be able to run continuously at this level, but again, it's meant for an audio amp so it'll do okay.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

You'll *love* this: http://schmidt-walter.eit.h-da.de/smps_e/smps_e.html

That calculator has been proved to be too conservative by several people. I have my own spreadsheet which i use to calculate things, i believe i linked to it in here as well.

How do you like that Uni-T UT60E, Unique?

It's good. There are three things i don't like about it however:

1. Slow autorange.
2. The frequency meter sometimes gets fooled by switching waveforms (it displays double the actual frequency, however, it does get the duty cycle right all the time).
3. The backlight doesn't use a regulator so it only works when the battery is brand new.

It is 4000 count. Btw, the frequency meter is spec'd for 10MHz but it works up to 50 within the same 0.1% tolerance.

You might try taking two of those yellow/white ferrites (type 26), stacking them and winding your output inductor through both cores at the same time.

http://micrometals.com/software.html I've done my math already. And no matter what i do to them T106-26, even stacked, run either too hot or need too much wire. I have a larger toroid (i think it's a T130) made with -52 material from an Antec Truepower, i'll be using that, and winding both + and - rail on it so that the DC bias basically cancels itself out. Should drastically reduce core loss.

Don't worry about the mess- it needs _more_ wires.

That's what i've been thinking too...
- Andrew

[kaboom]

Re: Building your own power supply: An actual example

You'll *love* this: http://schmidt-walter.eit.h-da.de/smps_e/smps_e.html

I wound a new transformer for what was a Hipro 200W AT PS. Turned it into a 45A 13.8 -14.2V supply. New/bigger caps, two SBL3045CT LV rectifiers, MJE13009s instead of 2SC4242s, bigger line rectifier, line wires, heavied up/paralleled traces.

And the most important part- completely changed the feedback paths. Many times, you'll see 'conversions' that consist merely of a big load on 5 to keep things happy. That doesn't do anything about the outputs being dependant upon each other, the too-small 12v rectifiers or the smaller 12v winding in the transformer.

I'll start another thread for that, don't want to hijack.

How do you like that Uni-T UT60E, Unique? I've got the next lower model, without RS-232 nor TRMS, labeled as MCM Electronics' Tenma 72-7740. It has a yellow grip instead of orange. Does yours have a 4000 or 6000 count display?

You might try taking two of those yellow/white ferrites (type 26), stacking them and winding your output inductor through both cores at the same time. If too little inductance or they saturate (softly- 26 is iron powder), you'll get high dI/dT charging the output caps, as you know. Reflected into the primary, your switches will 'splode. Don't worry about the mess- it needs _more_ wires.

Another link: http://www.bcae1.com/trnsfrmr.htm

Stacked toroid example is shown in the antenna tuner.

-Paul

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

Ghetto PSU Company Inc here. The previous EI33 still blew my MOSFETs up, so i gave it one more shot. Third time lucky they say.

I used a 100mT flux density and ended up with the following configuration. 100 turns primary, 68 turns secondary. Wound as follows: 50 turns primary with 2x 0.35mm wire, 13 turns secondary with 3x 0.35mm wire (for the tap for feeding the 12v linear regs), the remaining 55 turns secondary with 2x 0.35mm wire, and finally the last 50 turns primary with 2x 0.35mm wire.

I knew the copper loss was going to kill me on this one. But a hot transformer is still better than blown MOSFETs. And... results? See for yourself.

And yes, talk about a messy floor. At the moment the whole thing is being driven from an adjustable generator using a 555, a couple trimpots and a MOSFET for driving the GDT, no TL494, no feedback, no nothing, all is controlled by hand. The fan is a 24v 92mm fan being run at 12v so it doesn't move as much air as you could imagine. The heatsink holding the MOSFETs stays cool, the transformer is at about 50C estimated, i'll probe it with the thermocouple later.

You can see the primary caps at the bottom (two 680uF 200v caps in series), the tiny GDT, the crap diodes which i took from the worst of ATX supplies (200v 3A, and actually they're quite well suited for this job), output coil (which btw is taken straight from an ATX supply, i'm using the 5v windings in series, and it runs extremely hot, actually the reason it's sitting on another coil is because it got so hot that it started to leave hot glue on my floor! ). The tin which didn't make it in the pic in whole, is filled with water and is holding some of the (undersized) load resistors to keep them from burning up. Romanian load testing at its finest. The secondary caps are currently 2x 330uF Viva... anyone remember them?

And yet, with all of this, i'm getting 350 watts. I expect at least 20W more from a proper output coil, some 5W from better rectifiers, again a few more watts from beefier caps... and looky, there's my 400W. And all ATX PSUs which use this core blow up at 250W. Romania 1, China 0. And btw, the switching devices are a pair of those - based on their voltage rating (220v mains here), they should have blown in the second the supply was plugged in - but apparently nobody told them. The primary diodes are some 5A ultrafasts from Central, don't remember their exact model.

I'll be dealing with the output coil now, then winding the second transformer, then i'll get to designing a proper control circuit. AND YES it is gonna be regulated.

[Th3_uN1Qu3]

Re: Building your own power supply: An actual example

I calculated for 310VDC which is precisely what i am getting at the primary caps. Voltage does go down as expected with high load but the variation is too much.

I had to tackle another problem - the transformer was saturating. Fortunately our good old friend the air gap saved the day. Does it do 400W now? Just barely, but at least it doesn't blow up. The core gets hot but the MOSFETs are just warm now (i blew up two pairs before), sign that i am now far from saturation. Since it's for an audio amp... it'll do. You know all cheap PSUs that use the EI33 core blow up at 250W. I have 400W. I think that's enough of an achievement.

Pics tomorrow. I'll be starting work on the second transformer and on the TL494 and protection circuitry.

[Pyr0Beast]

Re: Building your own power supply: An actual example

Did you calculate for 230VAC or for 320VDC ?