12v 250w Car SMPS based off SG3525

[wally7856]

To understand what is going on here you have to know what skin effect is.

http://en.wikipedia.org/wiki/Skin_effect

Litz wire.

http://en.wikipedia.org/wiki/Litz_wire

and understand proximity effect.

http://en.wikipedia.org/wiki/Proximity_effect_(electromagnetism)

The main reason your transformer does not work is because of proximity effect. If you look at the proximity effect article. In the middle it talks about an RAC RDC ratio. As a general rule you should not exceed an RAC RDC of 1.5. You have a whopping ratio of 30.

The litz wire article tells you why home made litz wire will not work and the skin effect is something that interacts with the litz wire and proximity effect. All in all this is a very complex interaction and it takes a long time to understand.

The trick is to design out all of these bad variables and the answer is given every day, but not many people really know it comes from those 3 articles or how important it is to follow the advice. The ideal transformer is a split primary with the secondary in the middle. Both the primary and the secondary should be a single layer if possible and all winding’s should fill the entire winding width of the core, except for safety insulation on the ends of the bobbin, something you do not have to worry about at the voltages you are working with.

I think from here i should go in to transformer design. The above is standalone information so i will post this to break up this long post.

 

[codex653]

No worries! I hadn't thought you forgotten me! That sucks about your computer...quite annoying when those things happen.

Thanks for helping me with all of this! Can't wait to see the upcoming post!

 

[wally7856]

There are many methods to calculate transformer winding’s. Some work from primary to secondary and some work from secondary to primary. There are metric formulas and English formulas. I use the old school English primary to secondary method, but when you go to collage you will be taught the secondary to primary KGE method. ironically, both methods were invented by the same guy, Colonel William T. Mclyman.

The formula for calculating primary turns is as follows. You may want to copy this for future reference. As far as i know i am the only one who has ever gathered up all of the constants used in this formula and has written them up in one place.

Np = (Vmax * 10 ^ 8) / (X * F * B * Ae)

Np = Number of primary turns

The constant X used in the primary turns formula can be any of the following.

1.11 - compensates for the sine wave ratio of RMS to average voltage. Used for standard sine wave transformer design.

1 - Used for half wave square wave operation of a transformer. A flyback converter uses this constant.

2 - This is the square wave constant above multiplied by 2. there are two semi cycles in each full cycle (Hertz refers to full cycles) use for forward converters.

4 - This is the constant 2 above multiplied by two for topographies that have a flux swing in all 4 quadrants, like a full bridge circuit, half bridge or push pull.

These are the only constants there are for this formula.


UNITS IN FORMULA
Ae = Area of center leg of transformer in CM squared
B = Flux in Gauss, common values of 1200 to 1600
F = Frequency the transformer is switched at in Hz
Vmax = Voltage in volts the transformer sees


GENERAL RULES

For a half bridge driven off of a split capacitor supply the transformer only sees half of the bus voltage.

Vpri maxV gives you the maximum turns to use on the primary, this limits the flux to a maximum of your original number used in the formula. The flux will go down when the voltage goes to Vpri minV. This will make a safe design.


Increasing the Gauss decreases the turns.

As the turns go down the Gauss goes up




Designing a transformer is an irritative process of taking your best guess calculating it out and trying again. What we are trying to achieve is the following.

1) Make single primary and secondary layers.
2) Use the full width of the bobbin.
3) Have enough total wire diameter in circular mils to carry the current we need.

You want to build a 12vdc push pull smps and get around 250W with a +-35vdc output. So lets start gathering some data. First we will assume a poor 60% efficency because of no output inductors.

250W x 1.4 = 350W input power.

Assume 14.7 volts max input.

350W / 14.7vdc = 23.8A, max primary current.

BTW, I am not perfect and i make mistakes. If anyone see’s a mistake speak up!

For the secondary each 35 volt supply has to provide half of the power so 250W / 2 = 125W
125W / 35vdc = 3.57A

We also know that with no output inductor we will need 1.2 times the output current.
So 1.2 x 3.57A = 4.28A, secondary current.

WHAT WIRE SIZE IS NEEDED

Wire size for current rating is spec-ed in circular mils or cm. CM is wire dia in mils squared. mils is wire dia x 1000. A wire .02“ in dia is 400 cm. There are many wire charts on the Internet with wire dia and cm listed.

This is a guideline on how to use cm for transformer design. This gives you a cm per amp frame of reference.

500cm/A, for every amp you need 500cm. This number is for power supplies that have to put out a continues power and is considered a conservative design. For further reference 500cm = 23awg.

450cm/A, not as much copper used as above but when you can not quit fit enough wire this is acceptical most of the time.

350cm/A, for a power transformer this is considered poor quality and it will run hot.

275cm/A, considered a junk power transformer.

150cm/A, last and certainly least, is the “audio quality” transformer. You can get buy with such a low cm/A because the average music level is so low compared to rated output.

From here you figure out how many cm you need for the primary and secondary winding’s.

Primary current 23.8A x a minimum of 150cm/A = 3,570 cm total minimum size.
Secondary current 4.28A x a minimum of 150cm/A = 642 cm total minimum size.

I can kind of see how this is going to turn out already. Your transformer is oversized for this application and you will have plenty of room and more options. You will be able to run a higher frequency with smaller wires or a lower frequency with thicker wires. But lets see were the numbers take us.

Usually we are hurting for room on the transformer because an oversized transformer costs more and takes up precious space.

As we lower the frequency we can use thicker wires due to skin effect. But as we lower the frequency we need more winding’s, so you must put less strands on. When the transformer size is just big enough, there will be a small range of frequency’s and wire sizes that will work.

A a general rule for a high current smps like this one, you go down to 20 to 25khz to get the wire size up. Lets start at 25khz.

TRANSFORMER CALCULATIONS

I went to the Amidon site and found the specs for your transformer, minus the curves. The curves show the flux verses core losses as well as a few other things. But even without those curves we can make some good guesses what the numbers are. The only numbers we can really play with is the flux density and frequency, the other numbers are fixed. As the frequency goes down you can run higher flux, but as a general rule do not exceed 1600G unless you have a very good reason to. You run the risk of saturating the core and or running in a non linear region. You can almost always run 1200G even up to 100khz if the ferrite was designed for a power transformer. The only differance these two numbers make in our design is the number of primary turns.

Ae =1.84 CM squared from Amidon site
B = 1200 to 1600Gauss
The constant 4 as defined above
Frequency 25,000 Hz
Vmax, 14.7vdc

Np = (Vmax * 10 ^ 8) / (X * F * B * Ae)

Np = 14.7Vdc * 100,000,000 / 4 * 25,000Hz * 1600G * 1.84CM squared

Np = 1,470,000,000 / 294,400,000

Np = 4.9932 turns, call it 5

Now before you climb thru the computer and clobber me yelling that is the number of turns i had, remember this. You were running 52khz. Lets do quick calc for 52khz.

Np = 14.7Vdc * 100,000,000 / 4 * 52,000Hz * 1600G * 1.84CM squared = 2.4 turns.

Now lets do this again with 1200G and 25khz.

Np = 14.7Vdc * 100,000,000 / 4 * 25,000Hz * 1200G * 1.84CM squared = 6.657 = 7 turns

So what we learned is at 25khz we can use 5 to 7 primary turns by choosing different flux points. We can still vary the frequency for more choices in turns but we do not know what we need yet.

Now to calculate our skin depth at 25khz. The equation is for copper at 70 deg C.

Skin depth in mils = 2837 / squar root of frequency in hz.

2837 / sqr root of 25,000 hz = 17.943 mils

What we need to do now is pick some wire and see if it fits on the bobbin. And now i will have to make an ejamikated guess what size the bobbin is.

Winding width = 1.096“
winding height = .33“ on each side

The ideal transformer will have the primary winding’s 1 skin depth and the secondary 1 to 2 skin depths.

Primary = 17.943 mils = .017943“ = 25awg = 320 cm
Secondary17.943 mils x 2 = 35.886 mils = .035886“ = 19awg = 1,288 cm

Previously we determined that.

Primary current 23.8A x a minimum of 150cm/A = 3,570 cm total minimum size.
Secondary current 4.28A x a minimum of 150cm/A = 642 cm total minimum size.

Primary 3,570 cm total minimum size / 320cm of 25awg = 11.156 strands of 25awg
Secondary 642 cm total minimum size / 1,288 cm = .498 strand of 19awg

As you can see we are backwards of what we really want. We are hurting for primary wire size and we have much more secondary wire size than we need. What this tells us is we have to prioritize this design for fitting the primary winding. Making the secondary fit should be easy.

OK, lets do our first calc for wire fit. We will use a split primary and have a choice of 5 or 7 total primary turns. You can not have half turns on a transformer (you can, but not for this discussion, it is tricky business) so we need an even amount of turns to divide by 2, or live with an uneven primary to secondary coupling and have an increase in RAC RDC ratio. Which in this case would not be so good because we have few turns and many strands. So there would be a large area of uneven flux. If it was just one turn of 1 strand it would not be a big deal.

It is always safer to add turns than take them off, so lets try adding first. 6 turns or 8 turns divided by 2 is 3 turns or 4 turns. The best case for current is 3 turns because we can fit more strands.

3 primary turns x 11 strands needed = 33 wraps of wire
33 wraps of wire x 25awg .017943“ = .592“, WOW, that surprised me. We have a Winding width of 1.096“ to work with. This core is way over sized.

Let’s re-calc with 4 primary turns.
4 primary turns x 11 strands needed = 44 wraps of wire
44 wraps of wire x 25awg .017943“ = .7895“
Winding width of 1.096“ - .7895“ = .3065“ left over, not a very good fit.

Next thing to do is recalculate everything at a lower frequency. This will increase the number of turns but we already added one turn. I need to run the numbers to know if this will help us.

This post is already long and we have a long way to go. This might be a good place to break it up again.

 

[wally7856]

We can not go much lower than 20khz or we will be in the audible range, so lets go there and see were we are.

Np = 14.7Vdc * 100,000,000 / 4 * 20,000Hz * 1200G * 1.84CM squared = 8.3 turns, use 8
Np = 14.7Vdc * 100,000,000 / 4 * 20,000Hz * 1600G * 1.84CM squared = 6.24 turns, use 6

So primary divided by 2 = 4 and 3 turns, exactly what we had before. But the fact is no transformer can beat me. On the surface we made no progress but i see a perfect transformer in our future. What we gained by going to 20khz is we will be running the gauss we wanted. Adding a turn at 25khz lowered our gauss we will now be using our chosen value. More importantly, the lowering in frequency increased our skin depth. We can now go to the next size wire or so and completely fill the bobbin.

First we need to know our new skin depth and find our new wire size limit.

Skin depth in mils = 2837 / squar root of frequency in hz.

2837 / sqr root of 20,000 hz = 20.06 mils skin depth

20.06 mils skin depth = .0206“ = 24awg = 404 cm

We were at 25awg so we only gained 1 wire size. Lets see how it comes out.

4 primary turns x 11 strands needed = 44 wraps of wire
44 wraps of wire x 24awg .0206“ = .9064“
Winding width of 1.096“ - .9064“ = .189“ left over, still not a very good fit.

We still have one more trick we can use, more strands of wire.
4 primary turns x 12 strands needed = 44 wraps of wire
48 wraps of wire x 24awg .0206“ = .9888“
Winding width of 1.096“ - .9888“ = .1072“ left over, getting better.

I think 1 more strand will be to much but lets check.
4 primary turns x 13 strands needed = 44 wraps of wire
52 wraps of wire x 24awg .0206“ = 1.0712“
Winding width of 1.096“ - 1.0712“ = .0248“ left over, not bad at all.

We could also go back to the 25awg and add strands also. But the fact is since i guessed on the bobbin size you will have a different answer than i have.

Next calc is what is our final circular mil area of the primary winding.

13 strands x of 24awg of 404 cm = 5,252 cm

Lets calculate our final cm/A

5,252 cm / 23.8 pri amps = 220 cm/a, well above our 150 cm/A minimum.

The good news is we made the primary fit, and reflecting on this a bit, we actually are not to bad on over sizing the transformer. We only had to add 2 strands and 220cm/A is not overkill. So i guess the lesson to learn here is, it ain’t over till it’s over.

This looks like a good place to break up the post again. Next we will calculate the secondary winding’s.

 

[wally7856]

Lets bring some of the secondary specs here.

Winding width of 1.096“
+- 35vdc
4.28A

Calculate secondary turns

35vdc + 1 volt for diode drop = 36vdc min output

N35 = 4 pri turns x 36vdc / 12vdc min pri voltage (key off)

N35 = 144 / 12vdc = 12 turns

Our secondary is center tapped so we need two 12 turn winding’s the full width of the bobbin.

At 20khz our skin depth is.

2837 / sqr root of 20,000 hz = 20.06 mils skin depth

20.06 mils skin depth = .0206“ = 24awg = 404 cm

and we can go up to 2 x skin depth for.
2 x .0206“ = .0412“ = 18awg = 1624cm

We needed 4.28A at 150 cm/A minimum = 642 cm minimum for the secondary winding. Two strands of 24awg will easily handle the current. As 404 cm x 2 = 808 cm.

We will first try to fit 24awg since we used that on the primary and i think you have that wire anyway.

12 secondary turns x 2 strands of 24awg at .0206“ = .4944“
Winding width of 1.096“ - .4944“ = .6016“ poor fit.

Now you have to make a decision. Do you want to use many turns of 24awg or fewer turns of 18awg and have to buy more wire, assuming you have enough 24awg left over.

Small quantities of any size magnet wire can be bought on Ebay fairly cheaply. Search for “18awg magnet wire” and many choices will come up. Magnet wire just means high temperature coated single strand copper wire, preferably double insulated.

OK i made a mistake. I forgot that magnet wire has an insulation about .0015“ bigger in diameter than the bare copper wire dimension i was using for all calculations. This error adds up fast. Take the primary of 52 wraps of wire x .0015“ = .078“. Little mistakes like this are easy to do, sometimes i forget to divide the primary by two also.

Well we did not have the correct bobbin dimensions anyway so it does not make any differance.

lets try the secondary again with 18awg double heavy wire. The nice thing about the Ebay ads is most of them give the diameter of the wire with double heavy insulation.

http://www.ebay.com/itm/18-AWG-Gaug...530?pt=LH_DefaultDomain_0&hash=item3cc6721af2

Notice you can get the wire in red or green insulation.

18awg heavy = .0431“ in x 12 = .5172“ almost as much room as the 2 strands of 14 awg above.

We have to get to a Winding width of 1.096“ it looks like .5172“ x 2 would almost fit.

.5172“ x 2 = 1.0344“ width of 12 turn winding with 4 strands of 18awg
1.096“ - 1.0344“ = .0616“ left over, not to bad but we could do better if we wanted. I find if you want to you can dial it in to .02“ with a little work. In this case to get closer you have to go up 1 awg at a time and add a strand or 2 in to make up the differance.

18awg is slightly small.

With 18awg we had 12 x 4 strands = 48 wraps and we were a little short.
Try one more strand.
12 x 5 strands = 60 wraps
1.096“ winding area / 60 = .01826“, looking at bare wire chart i see 25awg is close.
So check Ebay for 25awg magnet wire heavy = .0199“ this is bigger than .01826 so it will not fit.

26awg heavy = .0178“ x 60 = 1.068“
1.096“ winding area - 1.068“ = .028“ , left over This is pretty good and close to my .02“ prediction.

So now you could wind each 12 turn secondary with 5 strands of 26awg and have a perfect match to the primary.

One thing that should be apparent at this point is why magnet wire comes in every gauge size. To get your winding’s to fit on the bobbin.

So to wrap it up we have 5 strands of 26awg = 254 cm x 5 = 1,270 cm

1,270 cm / 4.28A = 296 cm /A, compared to the 220 cm / A we had on the primary. Our copper I squared R losses are fairly equal.

This will make a first class smps transformer. But it will be a big pain to wind. In fact anything more than 4 strands is very hard. If anyone has any winding suggestions for the 13 strand primary please leave a post.

 

[codex653]

HOLY CRAP.....I'm gonna be going through this later tonight when i have the time, but I think this is gonna need to end up being turned into a sticky!!!

 

[codex653]

WOW! I love how in depth you were with all of that! I don't think I have ever seen anything like this before compiled all in one place...quite amazing! Are you an engineer or something or are you just very knowledgeable? I'm gonna have to go through all this again with the right bobbin specs. Guess I'm gonna need to go and get those calipers!

Ah this is so fantastic! I can see now just how completely screwed up my transformer is for this smps....no wonder this thing hasn't been working right. I wonder how Rod Elliot got his to work?

 

[wally7856]

Codex653, I am someone who has studied smps and mostly transformer design for several years. The lack of available information makes learning very difficult. For example the transformer constants i mentioned earlier and what they are used for i have never seen in one place. Also my chart of cm from 150 to 500 cm with a description of each level will not be found anyware. I have tried to develop methodology for transformer design as i have posted here to make it easy to design a successful smps transformer. If this method is followed the transformer has to work and will work well. It is also a fairly forgiving design, meaning it will tolerate some degree of sloppy assembly.

With that said I have been thinking over the results of this current design, info as follows.

This is a winding height check for a total of 6 layers of winding’s.

Pri 4 layers of 24awg heavy = .0223“ x 4 = .0892“
Sec 2 layers of 26awg heavy = .0178“ x 2 = .0356”
Total winding height = .1248“
We have .33“ so have plenty of room for insulation.


This layer chart is a summation of the transformer design
Pri-2, 4 turns, of 13 strands, 24awg.
Pri-1, 4 turns, of 13 strands, 24awg.
Sec-2, 12 turns, of 5 strands, 26awg.
Sec-1, 12 turns, of 5 strands, 26awg.
Pri-1, 4 turns, of 13 strands, 24awg.
Pri-2, 4 turns, of 13 strands, 24awg.
THE TRANSFORMER CORE IS HERE, each one of those layers wrap around this.

The reality of this chart is a little scarry. With the two split primary’s and the 13 strands required we have to bring out wire’s for 6 separate winding’s and get everything connected up right. Winding this transformer is going to be very difficult. Your bobbin only has 6 slots, 3 on a side. If you were to attempt winding this i would say use 4 of those slots 2 on a side for the secondary and on the other side of the bobbin make 4 slots on a side for the primary’s.

The object of my post was the design of a ideal type transformer that will work the first time. Splitting the primary is a good way to lower leakage inductance, improve coupling, and reduce proximity effect. But after thinking how hard it will be to wind this particular transformer i am having second thoughts. Hand winding 13 strands of wire on 4 different layers and keeping them all straight and in order will be a monumental task. This probably explains why on push pull smps i have not seen people split the primary’s, and they are wound as pri-pri-sec-sec. This is far from ideal, but at least it is windable.

I am also rethinking using just one skin depth on the primary. This gives us the lowest RAC RDC, and keeps our winding’s cool, but 13 strands are hard to manage and i think we should sacrifice some of that coolness for thicker wire and less strands.

With that said i would still use my design methodology above for a half or full bridge as they usually have far less strands to worry about and a single primary winding.

Let me know what you think about winding this transformer. I am going to play with the numbers and see what else i can come up with.

 

[wally7856]

OOPS, found another mistake. I used the wrong number of primary turns.

“Calculate secondary turns

35vdc + 1 volt for diode drop = 36vdc min output

N35 = 4 pri turns x 36vdc / 12vdc min pri voltage (key off)

N35 = 144 / 12vdc = 12 turns

Our secondary is center tapped so we need two 12 turn winding’s the full width of the bobbin.”

For the secondary winding calculation i put the wrong number of primary turns in. The total primary is 8 turns. I used the split winding number of 4 turns. Lets re-calculate this.

N35 = 8 pri turns x 36vdc / 12vdc min pri voltage (key off)

N35 = 288 / 12vdc = 24 turns.

fortunately this should help us not hurt us. This will reduce the number of strands needed on the secondary.

 

[wally7856]

I think i did it. If i did not make any mistakes, this will be the simplest and most elegant 12v smps transformer i have ever seen.

You will only have 2 winding layers. One primary and one secondary. The primary will have a total of 4 strands and the secondary will have a total of 2 strands. This design is good enough to put in production. It has the minimum number of turns, maximum use of copper, and highest transformer utilization at 1600G and full width winding.

First i recapped all of the specs in one place.

Ae =1.84 CM squared from Amidon site
B = 1200 to 1600Gauss
Frequency 20,000 Hz = 20.06 mils skin depth
Vmax, 14.7vdc
pri 8 turns, sec 24 turns at 1200G
pri 6 turns, sec 18 turns at 1600G

Winding width = 1.096“
winding height = .33“ on each side

Primary current 23.8A x a minimum of 150cm/A = 3,570 cm total minimum size.
Secondary current 4.28A x a minimum of 150cm/A = 642 cm total minimum size.

I started playing with primary wire size, wanting to limit the number of strands for each winding to two.

15awg = .0571" = .0603" heavy = 3,256 cm x 2 = 6,512 cm
16awg = .0508" = .0538" heavy = 2,582 cm x 2 = 5,164 cm

I then did a check for the wires fitting the bobbin.

Winding width = 1.096“ / 16awg = .0508" = .0538" heavy = 20.37 wraps
8 turns x 2 strands = 16 wraps, way to much left over

I noticed something here with this calculation. 16awg has plenty of cm with 2 strands and I was close to 24 wraps. 24 wraps would hold 4 strands, enough for both primary’s on one layer if a smaller wire had enough cm to fit. 24 wraps / 4 wires = 6 pri turns. And of coarse since we took the time to calculate our min and maximum primary turns before, I knew that 6 primary turns is a valid number.

Here i started to figure out how many cm i would have with wire that might fit.

3,570 cm total minimum size. / 2 = 1,785 cm
18awg = 1624 cm = .0431" heavy x 2 = 3,248 cm
17awg = 2048 cm = .0482" heavy x 2 = 4,096 cm

Winding width = 1.096“ / .0431" heavy 18awg = 25.4 wraps, this will fit.
Winding width = 1.096“ / .0482" heavy 17awg = 22.7 wraps, this will not fit.

Check actual cm for two 18awg wires.
1624 cm x 2 = 3,248 cm
3,248 cm / Primary current 23.8A = 136.47 cm/A
This is a little lower than our desired 150cm/A but 150cm/A is not cast in stone and i figured the 23.8A at 60% efficency, so i do not feel to bad about being slightly under our desired number.

The primary's will be one layer of 6 turns with 4 strands of 18awg. Each primary winding will be two strands of 18awg.

Here the secondary turns is recalculated with the new 6 turn primary number.

recalculate secondary.

N35 = 6 pri turns x 36vdc / 12vdc min pri voltage (key off)
N35 = 216 / 12vdc = 18 turns.

Re cap what i am designing for.
Need 642 cm minimum = 22awg min wire size.

Figure out what wire diameter i need for two secondary’s.

Winding width = 1.096“ / 18 turns x 2 secondary’s = .03044"

Make a chart of double heavy wire size to look up on Ebay.

20awg heavy = .0346" to big.
21awg heavy = .031" still to big.
22awg heavy = .0276" 3rd times the charm. 22awg = 642 cm exactly what we need.
The secondary’s will be one layer of 18 turns with 2 strands of 22awg. Each strand is one of the secondary winding’s.

You will wind and connect this as you did before with A, B, A1 and B1. The only differance is that you will not twist the wires together you will lay them flat.

I have thought about the best way to lay the primary wires. There are two choices, two A wires next to two B wires or alternate, A-B-A-B. One way is probably going to give better specs like leakage inductance but i have never seen any tests run with this type of winding so i will give it my best guess. I am thinking that A-B-A-B will give the best flux distribution. So go with that one, unless another reader comes up with a good reason not to.

One more note about twisting wires together. Outside of the transformer to keep the wires together you may twist them and even over twist them, as long as you remember that the insulation may be compromised and you are not letting the wire bundle touch another part of the circuit.

All in all i am very happy with this design. The transformer ended up being the perfect size and your specs of 250W at +- 35vdc have been met. The simplicity of the winding is almost scary, I will continue to mull it over for the next day or two to see if i forgot anything or made a mistake, but i do feel good about this design.

 

[codex653]

Incredible I cannot thank you enough my friend! I'm gonna go through every calculation myself just to make sure I learn how to do this on my own. What's the fun in just taking the results without knowing how to even get there?

 

[DCPreamp]

Hi Guys,

Wow wally7856, I am very, very impressed! Outstanding work! While indeed I have build quite a number of SMPSs over the years, and went through similar formulas back (way!) in college, I couldn't have worked through what you just did. Thanks to experience, I'm more of a seat-of-the-pants SMPS designer. Funny thing is, I came up with just about the same winding's count, but I had larger wire gauge! The larger wire gauge was, naturally, seat-of-the-pants numbers, plus, for DIY purposes, the added cost for making onesy-twosy transformers is no big deal. If I was designing for production volumes and manufacturability, well, you get the idea.

Hey codex653, wally7856 beat me to the SMPS frequency-doubling problem you reported before. Your SMPS didn't really change frequencies under load. Instead, with the increased load came increased noise which caused your scope to trigger not on every-other switching cycle, but twice as often on the noise. It reported a frequency increase, but in reality, it was switching just the same as before.

Regarding using inductors on the secondary of your SMPS, again, wally7856 nailed it with his (I assume he, but my apologies if you're a her) description. However, it is very common in professional car audio not to use secondary inductors. Given the size and cost of the inductors needed to effectively lower noise and smooth ripple currents, they can quickly become prohibitive. And given the z-out total of the transformer plus rectifier diodes, the durability of decent caps (especially well-bypassed caps), and the supersonic switching frequency not bleeding into the audio stream, they really don't add much other than better measurements on the test bench. We built car audio amps ranging from 45 to 800 WRMS, none of which had secondary inductors, and they operated on a daily basis, in pretty harsh environments, for decades without failures, is a testament to them not being needed. So for in-home and commercial designs, I always include secondary inductors, but for most push-pull SMPSs to be used in cars, I just omit them.

Fantastic teamwork guys! I have a few projects I need to hurry up and post for all to review and criticize. And hopefully learn from and maybe even copy too!

 

[codex653]

I agree with you dcpreamp! Man I'm blown away with how little time I have right now! I just moved in to my dorm this past weekend and have been going non stop since! I thought I would have a lot of time to get things done with this since orientation didn't pertain to me (i live in the same town as the college i am attending) but my wing and my sister wing have been doing A LOT of stuff together and my days are jam packed from morning to night! It sure is fun, but really tiring!

 

[AMSA]

Hello there guys.

Dear wally7856, your approach is valid too for forward converter?

Regards.

 

[wally7856]

Hello there guys.

Dear wally7856, your approach is valid too for forward converter?

Regards.

Yes you can use this method to design a forward converter transformer. However there are many types of forward converters. Depending on the type of converter the flux may use 1 or more quadrants and you need to pick the correct constant in the primary turns formula of 2 or 4. If you post a schematic of the one you want to use maybe someone will know what constant you need to use.

Np = (Vmax * 10 ^ 8) / (X * F * B * Ae)

Np = Number of primary turns

The constant X used in the primary turns formula can be any of the following.

1.11 - compensates for the sine wave ratio of RMS to average voltage. Used for standard sine wave transformer design.

1 - Used for half wave square wave operation of a transformer. A flyback converter uses this constant.

2 - This is the square wave constant above multiplied by 2. there are two semi cycles in each full cycle (Hertz refers to full cycles) use for forward converters.

4 - This is the constant 2 above multiplied by two for topographies that have a flux swing in all 4 quadrants, like a full bridge circuit, half bridge or push pull.

These are the only constants there are for this formula.

 

[AMSA]

Hi Wally and thanks for your quick reply.

About the schematic, I am looking for a nice schematic in order to build the forward converter. The goal is to build a transformer. Some time ago, I started to read the McLyman and Pressman book but, because I didn't had much time, I didn't read that much. One thing that I realize is that there several methods to build a transformer and I think that they are more complicated than that of yours.

From what I have read, he uses the Kg approach. We calculate the Kf (that is that X that you mentioned, right?), Ke and then finally the Kg. That Kg is intended to choose the appropriated core, right? Then, after that, when we choose the core we have all the data related to the core (window area, area product, core geometry, surface area, AL, etc).

Let me ask you a couple of question now. The formula that McLyman use has 10 to the power 4 and you use 10 to the power 8. Why?
From the example that you gave, how did you choose the appropriated core? I haven’t seen you calculating that Kg, in order to select the core. How did you managed to choose the core?
McLyman uses too another method to calculate the current density J, based on that Kf, Ku, Ap, etc.

McLyman has 3 chapters that refer to the design of transformer, more specifically, the first one entitled as “Power transformer design”, then he has the other two chapter related to the forward and flyback and here the method changes. And we must that into account a demagnetizing winding right? But if we don’t want to use that demagnetizing winding we can use a RC network. Am I saying right?

From what I have read, the types of converters that use transformers are: Fly-back converter, forward with one transistor, forward bridge, Forward half-bridge, Forward push-pull. Beside those, I wonder if the fly-back converter can be used using push-pull or half-bridge. There are such configurations?

The forward converter that I'd like to build is that one that uses only one transistor as a switch, since the goal is to learn how to do a transformer.

I’m sorry for my long post!

Regards.

BTW, what you mean with mils? Millimeters?

 

[wally7856]

AMSA,

“About the schematic, I am looking for a nice schematic in order to build the forward converter. The goal is to build a transformer.”

After reading the last part of your post i think you should look for an isolated DC to DC converter.

“ Some time ago, I started to read the McLyman and Pressman book but, because I didn't had much time, I didn't read that much. One thing that I realize is that there several methods to build a transformer and I think that they are more complicated than that of yours.

From what I have read, he uses the Kg approach. We calculate the Kf (that is that X that you mentioned, right?), Ke and then finally the Kg. That Kg is intended to choose the appropriated core, right? Then, after that, when we choose the core we have all the data related to the core (window area, area product, core geometry, surface area, AL, etc).”

The Kg approach is the newest method and uses the metric system. I use the older style method. I can not help you with the Kg method.

“Let me ask you a couple of question now. The formula that McLyman use has 10 to the power 4 and you use 10 to the power 8. Why?"

His formula is different than mine, different units.

“From the example that you gave, how did you choose the appropriated core? I haven’t seen you calculating that Kg, in order to select the core. How did you managed to choose the core?
McLyman uses too another method to calculate the current density J, based on that Kf, Ku, Ap, etc.”

I did not choose the core. The original author of this thread had that core available and wanted to build a 250 watt +-35 vdc push pull power supply. I only calculated the turns for what transformer he had.

“McLyman has 3 chapters that refer to the design of transformer, more specifically, the first one entitled as “Power transformer design”, then he has the other two chapter related to the forward and flyback and here the method changes. And we must that into account a demagnetizing winding right? But if we don’t want to use that demagnetizing winding we can use a RC network. Am I saying right?”“

The McLyman book is only a reference, you can not design anything from that book. The demagnetizing winding is only used for certain topologies.

“From what I have read, the types of converters that use transformers are: Fly-back converter, forward with one transistor, forward bridge, Forward half-bridge, Forward push-pull. Beside those, I wonder if the fly-back converter can be used using push-pull or half-bridge. There are such configurations?”

I think the flyback is only single primary winding you use simple control of turn on and off.

“The forward converter that I'd like to build is that one that uses only one transistor as a switch, since the goal is to learn how to do a transformer.”

Well then you should pick a low voltage project. The simplest may be a DC to DC converter. 12vdc in and 5vdc out, or 12vdc in and 24vdc out. Assuming you have a 12vdc supply to start with.

“I’m sorry for my long post!

Regards.

BTW, what you mean with mils? Millimeters?”

mils = thousands of an inch.
1 mil = .001“
Aluminum foil is about 2 to 3 mils thick.

 

[AMSA]

Thanks Wally. That was very helpful.

So that's why you said that you are accustomed to design the transformers in the old way. I see.

I will study your method and try to design a transformer. But let me ask you another thing: You said that "The McLyman book is only a reference, you can not design anything from that book." What you mean with that?

To finish, since you said "I did not choose the core. The original author of this thread had that core available and wanted to build a 250 watt +-35 vdc push pull power supply. I only calculated the turns for what transformer he had." This means that, for example, I can grab a core from a faulty PSU from a computer and take advantage from that?

BTW, the project that codex is working on is a SMPS with forward configuration?

 

[wally7856]

“I will study your method and try to design a transformer. But let me ask you another thing: You said that "The McLyman book is only a reference, you can not design anything from that book." What you mean with that?”

McLyman’s book is not a tutorial on building smps or transformers.

“To finish, since you said "I did not choose the core. The original author of this thread had that core available and wanted to build a 250 watt +-35 vdc push pull power supply. I only calculated the turns for what transformer he had." This means that, for example, I can grab a core from a faulty PSU from a computer and take advantage from that?”

Yes you can, but if you do first you have to trace out the circuit to the transformer to identify what topology it is being used in, usually 2 transistor forward of half bridge. The transformer material may be different and a 2 transistor forward converter may have a gapped transformer.

Also you have to disassemble the transformer and write down how it was constructed.

1) How many primary turns.
2) how many wires used in primary.
3) What size wire was used.
4) And the same information for the secondary.
5) Plus any extra winding’s like reset winding for 2 transistor forward converter.
6) The core has to be measured to identify it. And a picture will help of both pieces of the core.

“BTW, the project that codex is working on is a SMPS with forward configuration?”

It is a push pull.

You are going to find out that buying the transformer is hard to do. Even here in the USA it is difficult to buy transformer cores. The distributors only want to sell 1000's of them to manufactures and not deal with us small guys. So what usually happens is people try to make use of what they find in broken equipment or make use of the limited selection for sale. That is how i wrote up the transformer design. To show how to use what you have or can get.

 

[AMSA]

Ok. I see. Well, I think I'll study your method. It is very straight forward.

Two question. Imagine that I get a transformer from a faulty PSU, like I said. I take all the wire off and dismount the core and the core support.

What information I need to take from the core? Dimensions? If so, Which one? (The Area of center leg of transformer? Anything else?)

About the push-pull topology, for example, as we know, we need a center tap on the primary and secondary. How that is done? In which part of your calculations you take that into account?

Regards.