Digital benchtop power supply design / General Science and Electronics / Forums

Forums

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

« Previous topic | Next topic »

Digital benchtop power supply design

previous 1 2

Move Thread

LAN_403

Sulaiman

Tue Dec 22 2009, 11:32AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3141

Personally I would BUY a LINEAR psu - either new or used.

I prefer LINEAR psu designs because it's quite a challenge to remove all of the switcing noise,
and if you ever work with rf your psu will probably be too 'noisy'

I would BUY because once you factor in the cost of hardware (case, connectors, metering etc.)
it is usually cheaper to buy.

50v @ 5A is quite a 'big' psu, more like part of a specific project.

I would go for a dual supply e.g. 2x 0-30V @ 2A for op-amp supplies, or +5V and +12V etc.
In my experience, for most psu applications you need;
3V3 or 5V fixed for logic
+/- 12V or 15V fixed for analogue
24V fixed for industrial
and a variable 'testing' voltage.
So a few 78xx/79xx is a good choice.

But if you want to learn a lot, building your own is a good project.

For home construction my personal favourite is Variac - transformer - rectifiers - capacitors.
With switch/fuses/indicators etc.
It's very reliable and repairable and re-configurable (change transformer etc.)
Not 'sexy' but you will probably still be using it decades from now!

GeordieBoy

Tue Dec 22 2009, 11:49AM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Hi cjk,

> First, Is there an easy way to figure out how many volts per turn my core can take without saturating? I have a function generator, inductance meter etc. If it turns out to be hard to get a real number, ill probably just use 5v per turn or so.

Ferrite transformers in power applications are core-loss limited, not by saturation. What I mean by this is that if you design the transformer to run at the flux density where it just doesn't saturate, then the core will ultimately overheat and melt the bobbin during continuous operation. So forget about the saturation flux level, and look at the core-loss data on the datasheet.

Basically you have a trade-off between core-loss and copper loss. If you use too few turns on the primary of the ferrite transformer the operating flux-density will be too high and the core will overheat, but copper losses would be low because there's not many turns so you can use lots of strands in your litz wire for the primary. Conversely, if you use too many turns on the primary, the operating flux density will be low, but you require so many turns on the primary and secondary that the wire must be thin and therefore copper losses are higher.

I recommend you do a bit of reading about the transformer design in some power electronics books, otherwise you might be in for a long and tedious learning process.

If this is your first experience with designing a SMPS and it's transformer I have a recommendation for you which you would be advised to follow. Don't try to design the most compact, efficient, high-frequency power supply to match what you see on the market - you won't get there without a team of people with magnetics skills, thermal management skills, PCB design etc and many many man-hours of effort. Over engineer everything slightly and accept that your first venture into SMPS design might be a bit bigger than what Aztec offer, not quite as efficient, and a bit clunky, but at least it works. Your next attempt can then refine the design and get closer to ideal. So start with a reasonable operating frequency like 50kHz, a large ferrite core set with plenty of winding room, and a conservative design like voltage-mode control. Also don't aim for kilowatts, aim for a couple of hundred watts. That way you are guaranteed to arrive at something that works in the end.

I hope this helps,

-Richie,

PS. Gate drive waveforms look good

Steve Conner

Tue Dec 22 2009, 03:03PM

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

As usual, I agree with Richie... Except for one thing...

I've found that current mode control is actually easier to get working than plain voltage mode. (Though on the other hand, it is harder to understand.) Current mode gives a better phase margin, so the compensation and output filter design is more forgiving, and the transient response is better. It does lead to a more complex circuit, but you can get all-in-one SMPS ICs with it built in, such as the LM2586 and LT1170. These are to SMPS what the LM317 is to linear supplies. They're limited in power, but might be enough for a benchtop supply.

If you want to design your own current-mode switcher, then the UC384x series of controller chips are what you'd use instead of the TL494.

The basic concept behind a digitally programmable power supply, as you put it, is to first make one that's programmable by analog voltages, and then strap some DACs onto it.

I have two Xantrex 1kW DC supplies in the lab, which are remotely programmable by analog 0-10V inputs. You get a GPIB option board, but neither of them has it. I imagine it has DACs that connect to the 0-10V inputs of the PSU section.

PS: if you're having trouble finding ETD39 cores, it's maybe because they're made in Europe. I'm sure there's some similar US-made thing, check the Fair-Rite catalog.

GeordieBoy

Tue Dec 22 2009, 05:33PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Well, I also agree with Steve!

Current-mode control is actually better in a lot of situations, and often results in a supply with better transient response too. However...

The only reason why I would recommend voltage-mode control to start with is that you can operate the power electronics part of the supply open-loop to start with. Sure, line regulation and load regulation will be crap but you can get it running and do some tests on it without having to get to grips with the whole closed-loop feedback and control systems stability stuff or slope-compensation. You can load it up, check device temperatures, output ripple, control range etc, and it should be unconditionally stable when open-loop. Then you can move on to looking at closed-loop compensation.

As Steve said some flavours of current-mode control like peak-current mode control can be easier to implement than trying to stabilise a voltage-mode SMPSU with its under-damped second order LC filter at the output! You also get voltage feedforward compensation and the short-circuit protection features for free with current-mode control, whereas you have to work at desinging both of these in with voltage-mode PWM control.

The hardest thing about peak current mode control is often sensing small currents in the presence of large switching spikes when devices turn-on.

If you're going to buy a core set for this project you might want to look at going up a size or two from ETD-39. A larger core will provide you with more margin of error for a given power rating before it overheats. For DIY use it's better to spend a few cents more on a larger core, than waste hours trying to cram in wire and insulation that simply won't fit! It also allows you to rewind the bobbin at a later date to increase the power level when your a SMPS transformer guru

-Richie,

previous 1 2

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