SMPS Help / General Science and Electronics / Forums

Forums

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

« Previous topic | Next topic »

SMPS Help

1 2 next

Move Thread

LAN_403

Jrz126

Wed Aug 01 2007, 11:53AM

Registered Member #242 Joined: Thu Feb 23 2006, 11:37PM
Location: Erie PA
Posts: 210

I'm working on a DC-DC buck converter for work so I cant post a schematic.
Its a 20-80V input and an isolated 12V,5A output (60W). It will typically be running at 74VDC in. When starting the engine, the voltage will drop to 25-35V while cranking.
When running at 80V in, 5A out, I am putting in 133W and getting 65W out. I'm losing 68W somewhere, and I'd like to figure out why. (Most of the power is being disipated by the fets and rectifier).

Its a push-pull type supply, using a Uc1525 IC

I have posted a shot of a gate pulse below, and I think it may be the cause. The trace is at 5V/div and 2uS/div. The lip thingy at the end is roughly 3.5 volts and lasts for 800nS. The Vgs threshold is typically 4V +/- a volt. Is the Fet in linear mode during this time? Anyone know what might be causing that and how to fix it?

Steve Conner

Wed Aug 01 2007, 11:58AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

The lip thingy is your Miller plateau. If you want to make it shorter, you need to sink more current from the gate at turn-off, or use a smaller FET with lower capacitances. You probably do want to make it shorter, since the FET is in linear mode like you say. (Scope the drain voltage too and look at the two traces together, to get a feel for what's happening.)

Steve Ward

Wed Aug 01 2007, 05:37PM

Registered Member #146 Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055

50% efficiency is very poor! Surely you can post a schematic of the power section (or find a website that already has). You must be doing *something* wrong... The gate waveform you show is indeed a good way to burn off lots of power in turn off losses. Your rectifiers should be schottkey, and not dissipate much power at all (especially at only 5A). Also, perhaps your transformer is not designed properly and is causing problems?

Jrz126

Thu Aug 02 2007, 02:08PM

Registered Member #242 Joined: Thu Feb 23 2006, 11:37PM
Location: Erie PA
Posts: 210

Hmm...Miller Plateau sounds a lot smarter than lip thingy. Its been about 3 years since I had microelectronics, and we didnt cover mosfets that much.
I did notice the drain and gate voltage relationship, but I wasnt sure if it was the fet or something else, like an inductor or something.
We copied this psu from another device and it works just fine, (very little miller plateau). I did notice that the gate drive traces are rather long on this new design, would this add any significant amount of capacitance? I tried a smaller fet and it seems to help.
How can I tell if my driver chip is current limited? Should the pin of the driver chip be at 0V during the plateau? (it currently tracks the gate voltage and is at about half the amplitude of the gate). Or does the miller plateau already indicate that the chip is current limited?

Here's a basic diagram of the power section:

The fets and the rectifier are disipating the most heat. I'm just using a UF diode right now. I looked into the schottkey diodes, but I couldnt find any with decent specs.

Marko

Thu Aug 02 2007, 02:31PM

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

This is odd for SG1525. You should not have any special problems at 80V even with larger mosfets. What mosfet's are you using?

Are you sure you are really dissipating 68W just so, no measurement error? Your mosfet's would last very short unless you used really huge heatsink.

Can you provide us with schematic? Or briefing?

Steve Conner

Fri Aug 03 2007, 10:52AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

If the pin on the driver chip also shows the miller plateau, then that implies the current limiting is coming from the chip.

The rectifiers in this circuit are important, like in any SMPS. Just picking any old diode is unlikely to be optimal. In particular, if the reverse voltage rating is too high, the forward voltage will be unnecessarily high too. (taking this to an extreme, some HV fast diodes are made of two lower voltage diodes in series internally)

When you take account of the peak voltage that comes out your transformer, you probably need a high enough PIV that you'll struggle to find Schottky diodes to suit. The required PIV is equal to the transformer turns ratio times the highest possible DC input voltage, and another 12V on top of that, and then a little more for safety.

You might want to try a device similar to the one that supplies the +12V rail in an ATX PS. It's a double fast rectifier in a TO-220 package.

These small SMPS's usually end up about 75% efficient, AFAIK. 50% is a bit low :(

Marko

Fri Aug 03 2007, 11:57AM

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

These small SMPS's usually end up about 75% efficient, AFAIK. 50% is a bit low :(

Hey, why such pessimism over there? 75% efficiency, WTF? I remember building breadboard boost converters like 95% efficient at this power level, with SG3525.

Used <50% duty cycle (almost always discontinuous mode), single IRFZ44 at about 100kHz and 100uH yellow toroid inductor.

In this example I was opposite from you; mosfet seemed to experience nil switching loses, at 5-6 amps it gotjust little warm without heatsink.
Fast recovery diode I was using got very hot, though, just from forward conduction. Still loss was not more than few watts and efficiency looked really high.

I don't have much expeirence wit these push-pull bridges but it appears that lots of people are using them happily, with SG3525 directly connected to mosfet gates.

In this case voltage difference isn't actually too big for a simple buck converter.
I always try to avoid use of ferrite if I can, high side gate drive is much smaller problem.
I wouldn't want to speculate about current design until jrz126 posts a schematic.

PS. I missed out what he said about input voltage falling to 20V. Minimum voltage must be considered when building a transformer, and in this case it would be of little use. You would need to use like 1:1 or 1:1.5 ratio at most wich makes little sense.
So unless you specifically need isolation go with a normal buck converter.

Jrz126

Fri Aug 03 2007, 12:54PM

Registered Member #242 Joined: Thu Feb 23 2006, 11:37PM
Location: Erie PA
Posts: 210

I am using a diode similar to the ones in a powersupply. We originally had some regular axial diodes in there, which quickly melted. (the boss picked them :P )

The transformer is a 1:1, and isolation is definitly needed. I would love to post a schematic, but this is something for work, and I dont think they would appreciate it too much.

I dont understand why this supply doesnt work as well as the one that we copied it from. I tried changing everything I could to make them the same, except the output voltage. This supply is 12V out and the one we copied is 28V. Would this have an effect? I cant change it because the boss used 16V caps.

Only other thing that changed would be the layout. Is there a decent app note or something that has some guidelines to follow?

Steve Conner

Fri Aug 03 2007, 01:35PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Do they both use the same transformer ratio? The 12V one should really have half as many secondary turns. (unless the 28V one used a 4 diode bridge rectifier)

If the transformer ratio is 1:1 then it will run at a very low duty cycle normally, so the RMS-to-average ratio of the currents will be poor, and it'll be inefficient.

Marko

Fri Aug 03 2007, 02:54PM

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

Hi steve

so the RMS-to-average ratio of the currents will be poor, and it'll be inefficient.

His setup needs to regulate down to 20V in so 1.5:1 is biggest ratio he can afford himself. 1:1 is safety clearance.

I don't think it will be so big problem, though. 1:8 max voltage ratio may sound like a lot; but things like motherboard buck converters operate with even larger ratios and insane current outputs.

Mosfets under 100 volts can be found with quite low Rds ON, and paralleled if needed to improve efficiency.
Vishay has things like SUP80N15-20, 150V mosfet with 20 miliohms Rds on.

Something needs to go especially bad for you to have 50% efficiency.. I don't know how much can people help you.

1 2 next

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