Confused regarding ferrite core power output

Silvio

Well-known member
Dear all, I am a bit confused regarding power output of ferrite core transformers. According to data sheets and comparison what everyone else did by reading forums over here it was found that though data sheets (TDK Epcos cores) say for example an ETD 49 can give around 700 watts at 100Khz in a forward converter mode. Thus Microsim and also Ludo got around a kilowatt and more using half bridge topology. From what I noticed the core area and also the frequency play an important part in this factor. Can anyone tell me why and how is this achieved?

I already build an smps with an EI 40 core from an old atx psu but only got a marginal 300 watts till now. I build it from scratch and also took a few ideas from previous articles found here

Here is my work so far::SD
 

Attachments

  • IMG_0212.JPG
    IMG_0212.JPG
    1.8 MB · Views: 102
  • IMG_0201.jpg
    IMG_0201.jpg
    1.4 MB · Views: 71
  • IMG_0211.jpg
    IMG_0211.jpg
    1.1 MB · Views: 80

mone

New member
I have a question at you on another topic. Is transistors are getting warm without a load? I have a similar power but my mosfet are getting warm when it is soldered snubber. Not hard, but noticeably warmer.

i have etd39 and 33R/470pF
 

Silvio

Well-known member
Not really my transistors are just a little bit warm at around 30 deg C , without load. The values of the snubber have to be found by trial and error with an oscilloscope otherwise you will not get good results. The snubber is there to suppress as much as possible the ringing of the switching square wave. Every build has minor differences in it and often what one uses in a circuit is not necessary suitable for another.
 

MicrosiM

Administrator
Staff member
Hello

The only way to know the core limit is until it saturates.

ETD49 can make 1KW easily for audio applications, however not for other applications.

Increasing the frequency will increase the power of the transformer. but running at 100KHZ and above is the start of a challenging SMPS.
 

Silvio

Well-known member
Hello

The only way to know the core limit is until it saturates.

ETD49 can make 1KW easily for audio applications, however not for other applications.

Increasing the frequency will increase the power of the transformer. but running at 100KHZ and above is the start of a challenging SMPS.

Hi MicrosiM thanks for coming back to my call, I have read a lot in this site and learnt a lot of things. This is my first try at SMPS from scratch. I have been following your build at 1KW. Although I am a hobbyist in electronics I fiddle quite a lot and try my skills in everything. I learnt a lot what an IR2153 is, the hard way as I burnt quite a few mosfets and chips, well they blow like a blink of an eye :) Well I had issues with inrush currents charging the secondary caps but that was cured with the over current protection following Borysgo 2 (capacitive coupling) I dismantled the trafo and going to re-wind it with some homemade Litz wire with 0.16 mm twisting 15 strands for primary and 25 for secondary (also confused about ampicity) I tried to calculate the cross sectional area of each strand and adding up all but compared to solid copper it is nowhere near. Looking back what Ludo used for his smps the wire seemed rather thin for all that current handling!
As a rule of thumb for current rating I use the cross sectional area of the wire in mm2 multiplied by 6 for audio purposes and by 4 for continuous rating (correct me if I'm wrong) The wire gauge in the previous winding was according to frequency used in my case I used 0.46 bare (0.5mm with enamel) frequency 70Khz. 3 wires primary and 5 wires secondary. (450 watts)
Well I am waiting for some more fets to come and until then I am to wind my trafo with Litz tomorrow, will make some more tests and come back with my results. I am ready to give all the details here for all to replicate.
 

Silvio

Well-known member
I am considering of building a 1kw smps for audio based on Microsim and Ludo3232 and have a few questions regarding litz wire and current handling.
1) How is the duty cycle considered in an audio amp? 60%, 70% etc.
2) What kind of litz wire did microsim used for his smps? what thickness and how many strands for primary and for secondary? (total area of copper strands)
3) Primary winding for 155volts for an etd49 N87 material need 2 layers 0.5 + 0.5. Secondary voltage around 110 (60 + 60) letting some for voltage drop due to load and diodes (target voltage is around 55 to 57 volts, as it is unregulated so current needed at 1000watts at 110volts is 9 amps. Did this much fit in a single layer? I know coupling is important here but I like to be inspired and have some advice. Having 2 layers for secondary brings more leakage inductance and losses?

I made some tests with some litz wire consisting of 30 strands of 0.16mm ecw (0.6 mm2 total) and around 6 foot long approximately the length needed for primary winding and winded it on a piece of plastic conduit pipe having a dia of 20mm and taped it up with some mylar tape. I loaded it with a car head lamp at exactly 5 amps and measured the temperature rise. The result was that from 28 deg C ambient the temperature rose to around 62 Deg c in around 10 min. I also covered the coil with a thick cloth to simulate other winding wrapped upon it. Is this temperature acceptable? I know perhaps I need to nearly double that for 9amps however will this become too thick? Shall I consider 75% of the total power needed instead?

Bobbin length of etd 49 is 34mm winding space is 30mm leaving 2mm space at the edges and winding height 8mm maximum.

Can Wally or MicrSim give some advice here?

Thanks.
 

wally7856

New member
1) Duty cycle can be as low as 30%, but some equipment is rated for continuous use.

2) microsim told us long ago what litz wire he used but i do not remember what thread.

3) The ideal transformer would be pri-sec-pri.
But if the secondary's do not fit on one layer then the next best is
pri-sec-pri-sec

The desired internal temperature of the transformer is typically around 70 deg C. But there are some higher temperature ferrites around at 100 deg C.

I would double the amount of strands you are using. The ferrite gets hot and the pri and sec get hot. It all adds up.

0.16mm wire is about .08 A per strand

A forward converter is not a half bridge. 100khz is to high in frequency, skin effect will be a problem. Go for something around 60khz and you can still get 1000W.

Why do you want 1000W anyway?
 

Silvio

Well-known member
Thanks Wally for coming back with a few answers((@

1) My class D amplifier is a 2 X 500w per channel into 1.5 ohms. My speaker impedance is 4 ohms so this will give me approximately 350watts per channel. the Maximum current into 1.5 ohms is 16 amps 42 volts(minimum 30 volts) single rail thus having 2 amplifier boards powered with the same supply I will be needing two rails each at around 12-13 amps at around 60% duty cycle so average current is around 8 amps per rail.

2) The power supply should be capable of supplying this current continuously and also capable of handling peaks at higher currents at least 15 amps.

3)What is the recommended input capacitance for the bulk caps? I have 550Uf X 450vdc and also 680Uf X400vdc available, got 3 of each.

4) Are 2 X 0.47uf X 400v adequate for dividing the voltage for half bridge or perhaps 2 X 1Uf?

5) My design frequency is going to be 55 to 65 depending number of turns required and how are they going to fit in the trafo.

6) Litz wire will give a better performance I guess but is more bulkier due to construction (EX. grouping strands etc and more insulation on wire involved thus making it thicker than normal solid AWG 24 ecw for the same cross sectional area) It may cause some difficulty in fitting it in the window space available.
If for example using 0.16mm ecw making groups of 5 strands twisted and again having 8 groups of 5. Now instead of twisting these together to form a single wire(40 strands in all) I will lay them flat side by side and overlap half of them on the previous turn on the bobbin. The winding will be overlapping style hence wider but flatter. What do you think about this? will it disturb the magnetic field?

7) The drive circuit is going to be with an SG3525 and IR2110 driving a couple of IRFP460 unregulated so I will not be needing output chokes and enhance dynamic range.

This is my taught about this SMPS and any advice is greatly appreciated.

regards,

Silvio
 

wally7856

New member
3) Rule of thumb for input capacitors 3uF min per watt, 4uF overkill.
1000W = 3000uF minimum, 4000uF maximum

Your voltage divider will be 3000uF-center tap-3000uf. So you need two banks of 3000uF minimum capacitor, at 200vdc minimum rating, 250vdc better.


4) the transformer series capacitor really takes a beating. It has to carry the entire primary current.

Most people tend to use 1uF but i think 5 to 6uF is the proper size of a high current film capacitor 400VDC or higher.

It will take a lot of study to pick out the best capacitor. You can use one or parallel up several to make your uF. These are the DigiKey choices i narrowed it down to.

http://www.digikey.com/product-sear...=1&stock=1&quantity=0&ptm=0&fid=0&pageSize=25

5) I tried several different frequency's and 60 Khz looks the best.

6) I think you have a good idea for laying your bundles flat.

7) SG3525 and IR2110 sounds good, the IRFP460 is old, the SIHG20N50C looks like an newer chip with lower gate drive required. The SIHP33N60E has much lower RDS on hence the smaller case.

IRFP460APBF-ND 500V 20a 270m-ohm 105 nC to-247 $3.52
SIHG20N50C-E3-ND 500v 20a 270m-ohm 76 nC to-247 $2.89
SIHP33N60E-GE3-ND 600v 33a 99 m-ohm 150 nC to-220 $5.83

You need the output chokes!!! Without them the power supply has to struggle with large current surges in the primary and voltage regulation is really bad.

Any of these inductors will work. In stock at DigiKey, you need two of these, one in each 46.4 VDC rail output, after the diode and before the output capacitors.

CTX50-7-52LP-R, Best
CTX50-7-52-R, Best
AIRD-03-390K, OK

Output capacitor

Any of these will work, in stock at DigiKey, i am mostly looking at the ESR and ripple current.

Personally i would use two of these per rail, so total of four pieces.

MAL205158472E3 best
MAL215868332E3 OK
ALC10C103EC063 worst

I will add transformer info in next post.
 

wally7856

New member
Half bridge

155 VDC on primary.

etd49
Ae 2.114

Wire = .0063"

30mm winding space with 2mm margins.

Primary

turns = Vpri 10^8 / 4 x freq x Flux x Ae cm^2
15,500,000,000 / 4 x 60,000 hz x 1600 G, 2.11 Ae cm^2 = 19.13 turns = 20 turns.

Primary will be two winding's of 10 turns each.

Two bundles laid flat with 45 strands each.
(I think this will fit)


Secondary

155 volts/20 turns = 7.75 v/t

6 turns x 7.75v = 46.5v, wants 42vdc

8 amps .08A per strand = 100 strands minimum secondary

100 strands / 4 bundles laid flat = 25 strands per bundle.
(I think this will fit)

Wind, pri-sec-pri-sec
 

Silvio

Well-known member
Hi Wally, thanks for the info but still a bit confused what is MicrosiM was suggesting to others regarding wire thickness and overall CSA of litz wire. It just doesn't make sense for example he said that for primary winding in Thread: 1kW smps project (based on MicrosiM design) page 41 responding to TAJ 1941 reply #405. So for powering an amplifier of 900w he suggested that he should use 10 X 0.1mm ecw for Primary and secondary turns, the csa of all this is 0.47 amp calculated csa x6. Is there a kind of magic in litz wire regarding current? ;ww:

1) Well coming to your design makes more sense to me. Well MicrosiM suggested a B Max of 1200 gauss for N87 material, will this help the core not to saturate? Your calculation is based with a B max of 1600 gauss, any particular reason? I want to learn.

2) I am posting a few pics of what I have in hand regarding bulk caps these are 400volt rated so capacitance will not be divided. I have also drawn a very basic schematic showing my intentions regarding dividing voltage for half bridge topology. Are two of these (1360 Uf) enough for my requirements? Also take a look at the divider caps. I got all these from Inverter air condition PCB which have failed in other areas and could not be repaired. I tested the capacitance and are still within tolerance.

3) I measured my homemade litz wire and had another idea regarding winding. if for example I use groups of 5X3 (15 strands 0.16) these having a diameter of app. 1mm. then I can wind six of these (90 strands) side by side and put two layers for half Primary as only 5 turns will fit in bobin but the height will be a bit over 2mm considering tape etc. This will also bring the center tap near the starting point.

4) Have you a better idea then burning and scraping strands with steel wool to remove the enamel for soldering, my wire happened to be the one that it doesn't go off with a hot soldering iron. Such a pain in the ass cleaning all those strands LOL.

IMG_0249[1].JPGIMG_0257[1].JPGIMG_0252[1].JPGIMG_0261[1].JPG
 

MicrosiM

Administrator
Staff member
Hi Wally, thanks for the info but still a bit confused what is MicrosiM was suggesting to others regarding wire thickness and overall CSA of litz wire. It just doesn't make sense for example he said that for primary winding in Thread: 1kW smps project (based on MicrosiM design) page 41 responding to TAJ 1941 reply #405. So for powering an amplifier of 900w he suggested that he should use 10 X 0.1mm ecw for Primary and secondary turns, the csa of all this is 0.47 amp calculated csa x6. Is there a kind of magic in litz wire regarding current? ;ww:

1) Well coming to your design makes more sense to me. Well MicrosiM suggested a B Max of 1200 gauss for N87 material, will this help the core not to saturate? Your calculation is based with a B max of 1600 gauss, any particular reason? I want to learn.

2) I am posting a few pics of what I have in hand regarding bulk caps these are 400volt rated so capacitance will not be divided. I have also drawn a very basic schematic showing my intentions regarding dividing voltage for half bridge topology. Are two of these (1360 Uf) enough for my requirements? Also take a look at the divider caps. I got all these from Inverter air condition PCB which have failed in other areas and could not be repaired. I tested the capacitance and are still within tolerance.

3) I measured my homemade litz wire and had another idea regarding winding. if for example I use groups of 5X3 (15 strands 0.16) these having a diameter of app. 1mm. then I can wind six of these (90 strands) side by side and put two layers for half Primary as only 5 turns will fit in bobin but the height will be a bit over 2mm considering tape etc. This will also bring the center tap near the starting point.

4) Have you a better idea then burning and scraping strands with steel wool to remove the enamel for soldering, my wire happened to be the one that it doesn't go off with a hot soldering iron. Such a pain in the ass cleaning all those strands LOL.

View attachment 5932View attachment 5933View attachment 5931View attachment 5934



Selection of the wire gauge is based mainly on the current. using Litz wire or any other type of wires dosent mean a direct success in the SMPS.

My suggestion there was for only reference, I cannot give direct answers as you can see, due the fact that people use my hard work for there own use.

Litz wire will reflect one direct thing when using is, much LESS HEAT,and some Lower losses for sure.

Using any value for BMAX will affect your calculations, and will affect the whole design.

Stay at Bmax of 1200 and remain there. unless you use higher frequency 100KHz and above, playing at 100KHZ is not going to be a good start for you.

The value of your capacitors can be any where from 1000uf to 2200uf. and it will mainly affect the ripple.

Hope that helps
 

Silvio

Well-known member
Hi Microsim,Thanks for your reply
I see what you mean regarding hard work with success and failing, I am new to smps and trying hard to learn only for the reason because I like electronics though I never learnt it at school. I became a radio armature around 30 years ago and that was my start. I do it purely for a hobby and share my work with others. You can take a look here https://www.youtube.com/channel/UCsfTz-ktxOgs3-7be4YYHlg. I am also doing a lot of reserch and following a lot of treads here. I will publish my work here when I finish this project. I have been burning quite a lot of fets and chips and not finished yet with my first smps. My first oscilloscope is on the way and will be here in a week or so. Without it you are blind folded.

Selection of the wire gauge is based mainly on the current. using Litz wire or any other type of wires doesn't mean a direct success in the SMPS.

I done a bit of reserch here and tried a 6 foot length of home made litz wire app the length needed for primary winding and loaded it with DC current app 5 amp and found out that it got hot (30 strands 0.16mm) at around 70 degrees C in 10 minutes. I measured also the voltage drop and discovered that loss must not be more than 1.5-2watts otherwise heat will be exessive . That is why I taught things where not making sense, but as you are saying it has lower losses. well I still have to try that as I will be testing the first smps shown at beginning of tread. Now is wound with litz. I believe coupling between windings is very important in smps.

Using any value for BMAX will affect your calculations, and will affect the whole design.


Yes I know and should be in the region of 1200 to 2000 gauss ( reading Tahmid blog ferrite transformer calculations web site). How you came up with 1200g for your smps for N87 was followed here on this site and learnt how it should be done, as long as you know what core material you are using.

The value of your capacitors can be any where from 1000uf to 2200uf. and it will mainly affect the ripple.

Yes I know I believe 1uf per watt is more or less ok 2 is better 3uf is an overkill i guess (reading some Russian web site using google translate) and also as Wally suggested earlier I will be using a couple of output chokes to help out with ripple.
 

wally7856

New member
“Hi Wally, thanks for the info but still a bit confused what is MicrosiM was suggesting to others regarding wire thickness and overall CSA of litz wire. It just doesn't make sense for example he said that for primary winding in Thread: 1kW smps project (based on MicrosiM design) page 41 responding to TAJ 1941 reply #405. So for powering an amplifier of 900w he suggested that he should use 10 X 0.1mm ecw for Primary and secondary turns, the csa of all this is 0.47 amp calculated csa x6. Is there a kind of magic in litz wire regarding current?”

As a general rule you start with 500 circular mils/A
.1mm = 15.5 Circular mils
.031A per strand x 10 strands = .31A

I looked at the post I am sure he made a typo.

“1) Well coming to your design makes more sense to me. Well MicrosiM suggested a B Max of 1200 gauss for N87 material, will this help the core not to saturate? Your calculation is based with a B max of 1600 gauss, any particular reason? I want to learn.”

My guess is the microsim knows the transformer gets hotter with higher gauss. But you had the transformer core picked out and the wire. To make it all work together I had to use 1600 gauss. With 1200 gauss you would not be able to fit the windings on.

“2) I am posting a few pics of what I have in hand regarding bulk caps these are 400volt rated so capacitance will not be divided. I have also drawn a very basic schematic showing my intentions regarding dividing voltage for half bridge topology. Are two of these (1360 Uf) enough for my requirements? Also take a look at the divider caps. I got all these from Inverter air condition PCB which have failed in other areas and could not be repaired. I tested the capacitance and are still within tolerance.”

3uF per watt minimum, for continuous use.

“3) I measured my homemade litz wire and had another idea regarding winding. if for example I use groups of 5X3 (15 strands 0.16) these having a diameter of app. 1mm. then I can wind six of these (90 strands) side by side and put two layers for half Primary as only 5 turns will fit in bobin but the height will be a bit over 2mm considering tape etc. This will also bring the center tap near the starting point.”

I spent about 8 hours going over all of the calculations trying to get things to fit. I posted what I think is the best solution.

“4) Have you a better idea then burning and scraping strands with steel wool to remove the enamel for soldering, my wire happened to be the one that it doesn't go off with a hot soldering iron. Such a pain in the ass cleaning all those strands LOL. “

I searched for “chemical stripping of enamel wire” and found very nasty chemicals to remove the insulation. I think you should find new wire.
 

tibenko

New member
Hi all!

I made my first SMPS and it's work. But, after 1 hour at work ferite core will be at 50¤C without load. Duty cycle is 50% (I can not control duty cycle). Is the problem with heating in saturation of ferite core? If the answer is yes: how can I avoid saturation: did I need to increase number of turns per volt or decrease? (this is shematic by Aco).

Thank you in advance.

Tihomir Benko
 

Silvio

Well-known member
@ Tibenko

Please post schematic and what you did wind on the transformer, what core are you using and also what driver.

cores tend to get a little hot with when flux density is a bit high, But let us first see what you did and we can talk from there

regards Silvio
 

tibenko

New member
Thanks Silvio for very fast answer. I use schematic from Aco. Use standard ferrite transformer from ATX power supply (without any changes), only I add 3 wounds as addition for oscillator loop. I will found in 5 minutes schematic on forum.

TB
 
Top