DC-DC converter with multiple outputs 3

[Henneshoe]

I am designing a DC-DC converter that will have a 24V input and outputs of +/- 15V, 5V and 3.3V (~1A for 15V, .5A for 5V and 3.3V). I was thinking of using something like a flyback converter (I think) where a MC33063 would be used to drive a transformer primary with two secondarys. The first would be a split secondary (turns ratio 1:2) to provide the +/- 15V. The second would have a turns ratio of 3:1. Therefore, 15V at the primary would provide 30V across the split secondary and 5V at the other. A linear would be used for the 3.3V from the 5V. I will tie the grounds together because isolation is not an issue and Feedback would not be isolated and would come from one of the outputs. I have spent about a month researching this on my freetime and have 2 questions. 1) Anyone know a good spot to get a custom transformer like this for prototyping? 2) Has anyone tried this topology? Any huge pitfalls?

 

[itsmoked]

I would first ask... Why not just buy this? Building PS's is rarely a good use of one's time.

1) You need to hunt around and find the core size you will need then find a junker and disassemble it. Use the core to wrap your own as you will find your suggested ratio's are only the ideal, NOT, what you will actually need. Once you have the whole mess working you can then send your transformer to whoever and they will then be able to quickly ascertain what they need to provide you with - without a bunch of trial and error.

You have not said what your power levels are. Amps? It may well be that other methods would be better. Like switched capacitors, or drop in mini DC/DCs.

2) I have used that methodology but it was a while ago. There were lots of pitfalls, yes. The end result is even if you have one running, what is the reliability? If you buy a PS every test, under-the-sun, has been done and documented. The whole thing is certified to not burn down the immediate vicinity.

The worst pitfalls are ringing problems that result in damaged switching elements, noises emanating from the circuit, and radiated electrical noise.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[BobM3]

Power Integrations (

http://www.powerint.com)

can provide custom transformers. They also have links to other transformer vendors.

Their free software (PI Expert) can produce a nice spec sheet for a transformer that the vendor can build from.

 

[Renovator1]

Henneshoe - You're on the right track: the flyback topology is the best choice when you need multiple outputs at a modest power level. Magnetics, Inc., is my favorite core vendor and they will sell you small quantities.

The biggest pitfalls with the flyback:

* do not exceed a 50% duty cycle.

* give the switch plenty of voltage margin (e.g. use a 100V MOSFET, minimum, here).

* Ferrite cores for the flyback need an air gap and grinding your own isn't practical; MPP cores have the air gap "built-in", which makes them a whole lot more convenient here.

For this input voltage and power level you definitely want to check out National Semiconductor's "Simple Switcher" series - these combine a PWM controller and switch in one package (a 5 pin TO-220 is typical).

This is an excellent power supply toget your feet wet with switchers - I say go for it.

 

[gingell]

another problem to watch is the reflected voltage from the secondary back to the primary (ie referring to discontinuous mode, when the switch is off)....your secondary spans 24+15 = 39 volts.......when the switch is off this will be reflected back to the primary.....and if you have say , turns ratio of Np=3:Ns=1, then you get 3*39 = 117 volts reflected back to primary, which ADDS to your 24v input to give a switch off-state voltage of 141 volts.

 

[Renovator1]

gingell - The voltage reflected from the secondary that adds to the input voltage during flyback is calculated by multiplying the output voltage of any given secondary by the turns ratio of the primary to that particular secondary. In this case, the input voltage is 24V, the output voltage is 15V and an initial turns ratio of 3:1 is selected, giving a reflected voltage of 45V and a peak voltage across the switch of 24V + 45V = 69V.

One of the nice things about the flyback topology is that the turns ratio between the primary and any particular secondary isn't important (within reason). In this case, a turns ratio of 2:1:0.67 might be better (e.g. - a 12t primary, 24t and 8t secondaries for the 15V and 5V outputs, respectively).

This would lower the maximum voltage stress across the switch to ~56 (plus the leakage spike) and, at the same time, it also lowers the peak current stress in the output rectifiers (which lowers the amount of ripple current the filter capacitors will need to tolerate).

And here's a useful website for calculating out some of the basic flyback parameters:

http://schmidt-walter.eit.h-da.de/smps_e/spw_smps_e.html

 

[sreid]

One problem with multiple output flyback circuits is that usually, only one output is regulated. If the supply loads change a lot (especially the regulated one), the other outputs can change voltage quite a bit. This is sometimes fixed by summing all the outputs approiately but can mean that none of the output are well regulated.

The Simple Switchers are a good way to go.

 

[itsmoked]

Nice tips Renovator1.

If Henneshoe is going to go for this I wonder if National's design Workbench, which will now send you proto kits, and boards for your designs, would have a flyback dev board.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[Henneshoe]

Thank you everyone for your input. I am pretty new to the game (graduated '05); sense then I have made 2 buck converters using simple switchers. National does a nice job of basically designing the circuit for you. Now I am redesigning a board that is using 4 power supply chips to get the different voltage levels I specified in my previous post. The flyback topology looked intriguing, but I could not find much information online. With your input, I think I have enough to go to the next level and prototype a few options. Thanks again, hopefully I can return the favor in the future.

By the way, does anyone know of any good books that cover SMPS in general? I looked through my old text books but found next to nothing.

 

[Renovator1]

If you think you will be designing switchers on a regular basis then there are two books you must own:

Switching Power Supply Design - Abraham Pressman
Switchmode Power Supply Handbook - Keith Billings\

If you can only afford one, though, I'd give the nod, ever so slightly, to the Pressman book. There is not much difference between the earlier and later additions, by the way, so don't place too high a premium on that (IIRC, there is only one additional chapter (on power factor correction) for the most current edition of the Pressman book.

Additionally, the single best supply of truly useful applications information was from the former Unitrode, Corp, (now part of Texas Instruments) in their "Power Supply Design Seminar" books. They can definitely be downloaded as PDFs but if you can get them in print it's worth it.

Let us know how it goes.

 

[ScottyUK]

This one sits on my bookshelf and is pretty good.

http://www.amazon.co.uk/Switchmode-...cs-Handbooks/dp/0070067198/ref=pd_sim_b_img_1

The magnetics of an SMPS are frequently the hardest to design. Somewhere on Eng-Tips I've posted links to a couple of books on ferrites before. One was the old Philips MA04 databook - now Ferroxcube IIRC - and the other is 'Soft Ferrites' by Edgar Snelling. The latter is long out of print but is an excellent reference so second-hand ones are expensive. The cheapest on Amazon is £95 for a book published nearly 40 years ago!


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[gingell]

Hi
I second Renovator1's comment about getting a higher rated voltage MOSFET than you appear to need.....I once built a 24V 2A offline flyback (discontinuous mode) and the DC bus was 320V. On MOSFET turn-off, the voltage across the MOSFET spiked up to 1400V. (Even with the secondary referred voltage it shouldnt have gotten that high but it did).

I guess my secondary diodes were cheap and weren't turning on quick enough or something.
I found that no amount of snubbering could get rid of this 1400V spike.
I tried an RC series across the MOSFET and that didnt solve it. Then i tried a diode into a Resistor_and_Cap_in_parallel (-all of this across the primary) and still it didnt solve the problem.

I am now going to build a 24V 2A offline flyback using the TEA1507 offline flyback controller -this looks like being able to do it with very low component count (and is quasi-resonant) -though i hope my MOSFET doesnt again suffer the dreaded turn-off voltage spike.

Another thing -from what i hear, stay away from continuous-mode flybacks....i once went for an interview in a SMPS company. They asked me to draw flyback waveforms -so i drew them continuous mode flyback waveforms. -They told me that they never use continuous mode and made me draw the discontinuous waveforms.

 

[sreid]

The reason that discontinous flybacks designs are often avoided is that is that they have the dreaded "Right Hand Zero" in their response. This limits the transcient response or causes insability.

 

[Renovator1]

gingell - sounds like you had some vicious amounts of leakage inductance in your transformer. Were you, by any chance, trying to meet VDE isolation and creepage specs? If you multifilar wind the windings (i.e. - twist all of the wires for each winding together), use a MPP toroid core and don't try to hard switch faster than 20nS or so you can reduce the leakage spike to near non-existence (especially with multifilar winding). Unfortunately, not even the UL will let you multifilar wind the primary and secondaries for an offline supply and the more air space between the windings the more leakage you end up with.

At any rate, I think you will find the design and optimization of a quasi-resonant converter (whether ZVS or ZCS) a LOT easier than the corresponding hard-switched version. It seems hard to believe because of the extra switch(es) needed to create the zero current or zero voltage condition for the main switch(es) but when you no longer have to worry about leakage spikes letting the magic smoke out of your switches, rampant EMI failing you at approval, noise/PARD wreaking havoc with downstream loads (especially DACs!?! Oh how I hate those evil mongrels), well, the soft-switched topology doesn't seem so complicated after all.


sreid - er, that would be *continuous* mode

But, indeed, the right half plane (or hand) zero is a difficult nut to crack. Gingell - it is avoided like the plague because you must accept either drastically reduced transient response or your converter turning into an oscillator. That said, drastically reduced transient response is *often acceptable*, but it still is considered bad manners to let your flyback go continuous before your current limit kicks in.

 

[Warpspeed]

Flyback supplies have some very good points, but quite often the peak voltages on the switching device can become inconveniently high.

One topology that can be quite useful is the half bridge diagonal. This hard clamps the drain flyback voltages to the incoming dc supply rail, and returns any unused stored magnetic energy to the main input filter capacitor.

This often enables much lower voltage mosfets to be safely specified, with a resulting total reduction in Rds on, even though there will now be two switching devices in series.

 

[sreid]

Renovator1; er, yes that would be continuous, duh!

 

[itsmoked]

Now I'm getting confused!

[red]discontinuous[/red] flybacks designs are often avoided

Should have been [red]continuous[/red]?

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[GonzaloEE]

Hi.

You may want to check some PS controllers from

http://www.national.com/powerwise

I guess Texas and Altera also have related devices.

I've made a few PS before, with high switching freq. between 500-800KHz, very useful to reduce coil sizes and better cope with EMI/noise.

Good luck with your project.

Regards,
Gonzalo