12 to 35-0-35V

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
,
 

Attachments

  • pict0003.jpg
    pict0003.jpg
    34.6 KB · Views: 694
blmn said:
dkemppai,

There is another possibility for 12V smps, the full bridge configuration. There is an article in Elektor Electronics (I don't remember the number) about an amplifier using this configuration. Half bridge might work too, BUT, these two alternatives are very complex (if you compare them with the push-pull type) and the push-pull works very well in this case.

There are some new mosfets with a current "sensor" terminal, but, I don't know very much about them.

I'm using the formulae at the McLyman's book called Design Magnetic Components for High Frequency dc-dc Converters - ISBN 1-883107-00-8 - and it works very well.

blmn,

Pics look good! It's always nice to see other peoples work!


Yes, it is true, that a full or half bridge will work, but is not used for efficiency reasons (usually). In full bridge, you pass the high current through two transistors in series, this leads to double the losses with no real advantage.

Push-pull works well at low voltages because the transformer is usually never filled with windings, so there is room for a second primary. For a higher voltage transformer, the number of turns needed makes it difficult to wind two primary windings. An H-bridge configuration is used to reverse the polarity on the primary and bring the transformer through it's entire B-H loop.

Push-pull would be difficult at higher voltages because the transistors need to withstand double the input voltage because of autotransformer action on the primary windings. (Imagine a 300 volt input supply... ...at leaset 600 volt switches! :)

In half bridge, a capacitor or two are used as a current carrying device on one side of the bridge (again, doing this from memory), and that capacitor would need to pass the high currents also. This works well for high voltage designs where current is low and ESR capacitor losses are minimal. Basically you get the advantages of H-Bridge with less components. :)

I have been doing amplifier repair for a local auto sound company. I've only seen Push-Pull designs. Only one supply differed from the rest, and it ran a 50% duty cycle all the time. The transformer ratio set the output voltage (Very unusual).


The 5 lead fets you talk about are neat, I've worked with them a little, but they are much more expensive than standard fets. I would think that they would be too expensive for mass production in terms of profit margin.

-Dan

P.S. Is the McLyman's book any good? I mean, overall is it a good book on power magnetics? (If so, I may just order it)
 
I did point out that MOSFET's don't share the problem -- my experience relates to an smps in an HP "portable" analyzer -- portable being an engineering joke at HP -- kept blowing switching transistors until I put in some very beefy SGS Thomson horizontal deflection transistors which handled the voltage and current stress. Perhaps I got lucky with the matching.

The current sense amp in the SG3524 is an adjunct to the voltage output sense amp.

I also note the gapped core in Pressman's book with its plusses and minuses.
 
jackinnj,

In the same book, starting at page 49, Pressman lists some ways to combat the flux imbalance, wich includes the current mode topology, stated as the best solution. I think the use of mosfets is an affordable solution for battery powered audio amps or other variable current devices, but for sure it can't solve flux imbalance problems in all design types.

dkemppai,

McLyman's book is a very good practical design book, lots of tables, charts and examples. The theory of magnetic components is there too, but it's very strong in its pratical matters. A very good acquisition I think.

regards
 
dkemppai - '3524 isn't a current mode controller

it has "overcurrent sense" on the output, but it does not do a pulse by pulse sense of the current on the push-pull transistors

I do like the chip and it is extremely cheap. I use it in motor control in conjunction with the LMD18201 H-Bridge. You can configure it for a tach-less speed control.
 
Hello all!
As the designer of Rod Elliot's 12V SMPS, I can tell you that you can considerably increase its power capabilities by paralleling more MOSFETs, using thicker wires and a bigger core. This is the most difficult part, but as someone has pointed out here, you can use standard E cores (ETD49, for example).
Have luck!
 
ssanmor
I've been building this smps (on and off ---- haha), for a while now, I get a good(ish) looking square waveform on the scope, with and without the BD139/140's (needed a 100pf bypass cap on the 3525 supply..).
I need better details on the transformer winding though..
I've purchased a
EE42/F44 (Al = 3500)
ferrite core kit from rs-components
Could you advise me on any winding tips I may need..?
I've also got a couple of toroids (as per Rod's sketchy size spec..)
Are these any easier / better to wind?

I've also tried IRF540 MOSFETS....
I think I may need to adjust the dead time, I've still got a blister from touching one of those mosfets..(best way to learn a lesson, re-inforce with pain..)

Could you please help a SMPS newbie with some advice?

Thanks for the help

Regards
Jan Holtzhausen
 
Hello, Jan.
First of all, could you provide more details about the core you have? I don't know how it is by the reference, sorry.
I did all my tests with toroidal cores. To give you an indication, you can easily get 200W from a 5cm external diameter core and 3 cm internal diameter, about 1cm height (approx.)
For me it was easier to wind toroidal transformers. Just follow the instructions I wrote in Rod's page. I couldn't explain it better in few words.

And don't use IRF540 Mosfet's if you can, they have a relatively high Rds(on) (77mOhms). Better use MTP60N06 from On-Semi (they offer free samples). They will run muuuuch cooler.

Once you have wound the transformer, test the system with a light load (100mA, for example) and see the waveforms in the primary. It should look quite square, with peaks when switching, at most.

Have luck!
 
Hiya ssanmor
WOW! Thats what I call a fast response!

EE42/F44 (horizontal)
2 half cores, W.43, H.21, D.15 3500 32-110-44
bobbin, 12 pin horizontal 59-113-66

Is what I get from RS.........

I need about 600 watt ... will this core do?

Also for the toroids..
Do I wind 4 windings (over the length of the toroid) and then the tap, or do I wind 4, tap,4 over the length (ie does each of the windings cover the entire toroid, or do both).. I'm asking because it just seems to be very little in the way of copper on core..?

Thanks a lot for your help!!

Regards
Jan
 
It is difficult to know if your core will be able to supply such amount of power, but perhaps it is a little bit small.
Why don't you try to make 2 x 300W supplies instead of the big one? For example, if you want to feed 2 channels, use one supply for each one. The idea is to share the load. But don't parallel the supply rails!!!

TOROIDAL WINDING:
---------------------------------------------------------------------------------
You have to wind 4 (or whatever, but that's a good start) turns in EACH primary. That is: 4 turns, tap and 4 turns. The same in the secondary (for example, 15 turns, tap and 15 turns more).

Wind the primary, then the secondary. Don't forget to peel the varnish before in both ends of each wire (hard task):

Let's assume that you will use 3x0.8mm wires in each primary (I wouldn't use less!). Then take six wires of enough length to wind 4 turns + 10cm more for the connections. Wind them altogether in 4 turns so you cover ALL the core and finish very near where you started. Then you must join 3+3 wires to form the tap. Now you have 3 wires (solder them together), 6 wires (solder them together, these are the center tap) and another 3 wires (solder them togeter). So you have finished the primary.

Now the secondary: Do exactly the same, with the proper diameter and number of wires, but in the opposite direction. Start (and finish) the 15 turns in the opposite side of the core. For example, if you use 3 0.6mm wires, wind 6 wires and you will end with 3 wires for one end, 6 for the tap and another 3 for the other end.

I don't know if it is clear, but read it carefully and ask me what you want.

Best regards
 
Gottit!!
The penny finally dropped!

Thanks a lot, you are now officially my (smps) hero!

Should I twist the joined strands, or coil them flat (to be as close as possible to the core?

As far as the (2x)E type core is concerned, should I wind the same number of turns?


Best Regards
Jan
 
dkemppai said:
Push-pull would be difficult at higher voltages because the transistors need to withstand double the input voltage because of autotransformer action on the primary windings. (Imagine a 300 volt input supply... ...at leaset 600 volt switches! :)
Take a look inside your pc power supply sometime! with any transformer winding, the amount of volt / seconds it gets pulled down, it has to swing up by the same amount of volt / seconds to reset the core. So if you pull it down 300v for 50% it has to swing up for 300v 50%. The switching fet has to withstand this 300v upswing *plus* the 300v dc supply feeding it. If you run less than 50% duty cycle you can reset the core at a lower voltage for a correspondingly longer time.
 
Sorry...

Circlotron said:

Take a look inside your pc power supply sometime! with any transformer winding, the amount of volt / seconds it gets pulled down, it has to swing up by the same amount of volt / seconds to reset the core. So if you pull it down 300v for 50% it has to swing up for 300v 50%. The switching fet has to withstand this 300v upswing *plus* the 300v dc supply feeding it. If you run less than 50% duty cycle you can reset the core at a lower voltage for a correspondingly longer time.

Hi!

I guess I don't follow your point. I'm looking at the guts of a PC power supply right now, and see a sort of half bridge configuration. There's only two pins on the primary, so it can't be push pull.

I'm not sure what this has to do with the origional post? Maybe you could explain in more detail what you mean.

sorry!
Dan
 
ssanmor
Finished winding the new toroidal last night (in bed, watching TV ... the wife was NOT impressed ...:))
I have one blue and extremely sore thumb, but it's as neat as I can get it (wires parrallel / flat, evenly spaced, six for primary, 5 turns, four for secondary, 15 turns....looking for 34V - 40V).
ONsemi are unable to register me at the moment...system gone pear-shaped I guess..
Can't find a supplier for those MTP60N06 (replaced with NTP60N06 according to ON), except for RS-Components..... you do NOT wnat to know what their prices are like..... guess that's what you pay for variety / availability...

What viable alternatives are there? (BTW NTP75N06 has even lower rds -9.5- ... any good?)

Regards
Jan
 
Congratulations!
Could you post a photo of the wound transformer? What are its dimensions?

MTP60N06 is what I used, but OnSemi released NTP75N06 after that, but it is better because of its lower Rds(on), as you have pointed.
If you retry several times and the web is not ok (it is strange, it seems to work for me), you can use another part easy to get, BUZ11, used in 90% of car SMPS, altough it is slighty worse.

Best regards.

(By the way, you can talk directly to me at ssm@tid.es)
 
blmn said:
Try IRFZ44.

I was just going to suggest the IRFZ44E. I've also used several IRF1010N's paralleled. I did a little cost vs. performance comparisons a while back, and in my application the IRF1010's were the best fit... ...However, I tried replacing some IRFZ44's with the IRF1010's in a mass produced amplifier, and they switched too fast. The switching transients were way too high with the IRF1010's.

I've also used IRFI3205's, low Rds on Fets... ...expensive, but nice... ...insulated package (usually for lower power applications)

Good luck transistor hunting!
-Dan
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.