I am currently working on the design of a 450V powersupply. I need 2.5A at 450V as the output. The input is 400V from a PFC circuit.
Is an ETD44 core large enough for this application? I calculated that a 30 turn primary at 70kHz gives <0.2 Tesla flux. I have rarely seen people push 1100W through an ETD44 though so I am questioning if there is some other problem I will run into. I can fit enough wire on the core for 500 circular mils per amp so perhaps all will be ok.
I originally wanted to do an LLC topology with the resonant inductors as part of the main transformer. If I go this route I want to use a UCC25600 controller and possibly a gate drive transformer. Switching frequency would be in the range of 80kHz to 120kHz. I found an online LLC calculator and it gave Ls = 25uH, Lp = 126uH, and Cs = 100nF. I also did some calculations by hand and found that the results were similar. I have no idea if it will be practical to reach these values in a real transformer.
One possible complication is that I intend to use a 20uF film cap for the output filter. This cap will have very low ESR on the order of a few milliohms. For this reason the ESR zero that the cap causes will be at a very high frequency. I am unsure if this could cause stability problems.
I also considered using a full bridge but I like the fact that the LLC only needs two transistors.
Does an LLC with ETD44 core sound reasonable for 1100W? With such a high output voltage, will the capacitance of the output rectifiers be a problem?
Is an ETD44 core large enough for this application? I calculated that a 30 turn primary at 70kHz gives <0.2 Tesla flux. I have rarely seen people push 1100W through an ETD44 though so I am questioning if there is some other problem I will run into. I can fit enough wire on the core for 500 circular mils per amp so perhaps all will be ok.
I originally wanted to do an LLC topology with the resonant inductors as part of the main transformer. If I go this route I want to use a UCC25600 controller and possibly a gate drive transformer. Switching frequency would be in the range of 80kHz to 120kHz. I found an online LLC calculator and it gave Ls = 25uH, Lp = 126uH, and Cs = 100nF. I also did some calculations by hand and found that the results were similar. I have no idea if it will be practical to reach these values in a real transformer.
One possible complication is that I intend to use a 20uF film cap for the output filter. This cap will have very low ESR on the order of a few milliohms. For this reason the ESR zero that the cap causes will be at a very high frequency. I am unsure if this could cause stability problems.
I also considered using a full bridge but I like the fact that the LLC only needs two transistors.
Does an LLC with ETD44 core sound reasonable for 1100W? With such a high output voltage, will the capacitance of the output rectifiers be a problem?
From what I read on the data sheets you should keep your frequency down towards 25KHz.
At 100KHz the core losses are 8 times higher.
If you can fit the copper in then it should work.
I don't know about your copper losses though.
http://www.ferroxcube.com/prod/assets/etd44.pdf
http://www.ferroxcube.com/prod/assets/3c94.pdf
http://www.ferroxcube.com/prod/assets/etd_frnt.pdf
Good ap note
http://www.ferroxcube.com/prod/assets/sfappl.pdf
my $.02 worth.
At 100KHz the core losses are 8 times higher.
If you can fit the copper in then it should work.
I don't know about your copper losses though.
http://www.ferroxcube.com/prod/assets/etd44.pdf
http://www.ferroxcube.com/prod/assets/3c94.pdf
http://www.ferroxcube.com/prod/assets/etd_frnt.pdf
Good ap note
http://www.ferroxcube.com/prod/assets/sfappl.pdf
my $.02 worth.
Last edited:
An ETD-44 at 50khz will give you just enough power for 1.1KW watts. 70khz will give you approximately 1.6KW.
At 70khz you will have 1600G with 26 turns
I played around with transformer margins a little and wire size and 2mm margins on each side will let you have a primary of 5 strands of 25awg, 320 Circ mills each. At 500 Cir Mills / A you would be good for 3.2 pri amps.
Sec turns = 62, and oops, it looks like i made a mistake after this. I mixed up my numbers and gave myself 2x as much room as there really is. Hmm best case i can see is 62 turns of 26awg at 254 Circ Mils for about 1/2 amp at 450vdc.
I would say the secondary may end up driving the transformer core choice.
At 70khz you will have 1600G with 26 turns
I played around with transformer margins a little and wire size and 2mm margins on each side will let you have a primary of 5 strands of 25awg, 320 Circ mills each. At 500 Cir Mills / A you would be good for 3.2 pri amps.
Sec turns = 62, and oops, it looks like i made a mistake after this. I mixed up my numbers and gave myself 2x as much room as there really is. Hmm best case i can see is 62 turns of 26awg at 254 Circ Mils for about 1/2 amp at 450vdc.
I would say the secondary may end up driving the transformer core choice.
Tick tock - 5 hours of calculating later. Like i said the numbers all work out!
Well as i also said we need 2X more room than we have. It came to me that either we need a huge core or just use 2 transformers. I tried the ETD-44 and -49 again with primary’s in parallel and secondary’s in series. The ETD-49 had a relatively elegant solution at your original 70khz frequency.
etd44
page 545 2009 Fcube catalog.
winding width min on bobbin 29.5mm = 1.161" - .158" for 2mm margins = 1.003" actual winding width.
Ae = 1.732
etd49
page 548 2009 Fcube catalog.
min bobbin wind width = 32.7mm = 1.287" - .158" for 2mm margins = 1.129" actual winding
width.
Ae 2.114
1100W / 2 transformers = 550W each
550W / 200vdc on primary = 2.75A min primary current on each transformer.
Secondary has a given 450vdc at 2.5A
70khz
1600G
Primary = 22 turns of 5 strands 25awg. 320 Cmills x 5 = 1600Cmills = 3.2A
Full width of bobbin filled up nicely.
Secondary = 1 strand of 19awg = 1,288Cmills = 2.58A
Need 52 turns but 54 turns fills up full width of bobbin. 54 turns / 2 transformers = 27 turns on each transformer.
Each secondary is 1/2 the 450vdc needed so a FWB rectifier is used, not a center tap.
The transformers should be wound with a split primary.
11 turn primary.
27 turn secondary.
11 turn primary.
I am feeling pretty good about this design, lets see if anyone else can come up with a better solution.
Well as i also said we need 2X more room than we have. It came to me that either we need a huge core or just use 2 transformers. I tried the ETD-44 and -49 again with primary’s in parallel and secondary’s in series. The ETD-49 had a relatively elegant solution at your original 70khz frequency.
etd44
page 545 2009 Fcube catalog.
winding width min on bobbin 29.5mm = 1.161" - .158" for 2mm margins = 1.003" actual winding width.
Ae = 1.732
etd49
page 548 2009 Fcube catalog.
min bobbin wind width = 32.7mm = 1.287" - .158" for 2mm margins = 1.129" actual winding
width.
Ae 2.114
1100W / 2 transformers = 550W each
550W / 200vdc on primary = 2.75A min primary current on each transformer.
Secondary has a given 450vdc at 2.5A
70khz
1600G
Primary = 22 turns of 5 strands 25awg. 320 Cmills x 5 = 1600Cmills = 3.2A
Full width of bobbin filled up nicely.
Secondary = 1 strand of 19awg = 1,288Cmills = 2.58A
Need 52 turns but 54 turns fills up full width of bobbin. 54 turns / 2 transformers = 27 turns on each transformer.
Each secondary is 1/2 the 450vdc needed so a FWB rectifier is used, not a center tap.
The transformers should be wound with a split primary.
11 turn primary.
27 turn secondary.
11 turn primary.
I am feeling pretty good about this design, lets see if anyone else can come up with a better solution.
Thanks for all of your work.
Let me show my work here and see if I have any mistakes in my thinking.
I have that B(Tesla)=E(Volts)/[4.44*N(Turns)*A(m^2)*F(Hz)]. Lets consider an ETD44 core.
A = 173 mm^2 = 1.73*10^-4.
Put E = 300 V (due to resonant rise)
Put F = 80*10^3 Hz
Solve for N => N = 24.4 => Choose N = 25.
So if my math is correct, 25 Primary turns on an ETD44 gives a maximum flux of 0.2 Tesla.
Now lets see if the winding fits on the core. I have an ETD44 bobbin and a pair of calipers here so I am taking physical measurements to get my numbers. The window on this bobbin is 1.167" wide and 0.27" tall. Lets assume I have a spacer between the primary and secondary that is 0.125" wide. That leaves 1.042" for windings. Lets put the spacer in the middle. So the primary winding window is 0.502" x 0.27". In mm this is 12.75mm * 6.85mm.
Lets say I need the primary to carry 5.5A. If I want 500 circular mills per amp then I need 2750 cmils. Circular mils are calculated by squaring the diameter of the wire in mils, so lets take the square root of 2750 cmils to find the required diameter. sqrt(2750) = 52.4 mils diameter. Now lets find the cross sectional area that this corresponds to.
From a wire table, a 52 mil diameter wire has an area of about 1.31 mm^2 (Yes, I am mixing metric and imperial). I am using Chinese litz wire that is many strands of 0.1 mm diameter wire. 0.1mm diameter wire has an area of 0.00797 mm^2 per strand. 1.31/0.00797 = 164. So I need 164 strand wire or greater for 500 cmils per amp or better.
Lets say I pick litz wire that is 1.3mm diameter for a bundle of 100 strands of 0.1mm wire. To get above 164 strands I will use two pieces of litz in parallel on the primary and one strand on the secondary. If I use a 1:2 ratio then this means I need 50 turns on each side.
The window is 12.75mm * 6.85mm. Lets assume a poor packing density to start, to do this we assume the wire is square. The window area for each winding is 87.3mm^2. Now for the square wire: (1.3mm)^2 = 1.69mm^2. Now lets see if it fits!
87.3mm^2 / 1.69mm^2 = 51 Turns!
So unless I have made a mistake, I can just fit the required winding on an ETD44 core while keeping the flux density below 0.2T. The ETD44 datasheet says the core should dissipate less than 9.4 W at 0.2T at 100kHz. This is close enough to my 80kHz to 120Khz operating range to be a decent estimate.
It may sound crazy but my math shows I can push 1100W through an ETD44 core.
Let me show my work here and see if I have any mistakes in my thinking.
I have that B(Tesla)=E(Volts)/[4.44*N(Turns)*A(m^2)*F(Hz)]. Lets consider an ETD44 core.
A = 173 mm^2 = 1.73*10^-4.
Put E = 300 V (due to resonant rise)
Put F = 80*10^3 Hz
Solve for N => N = 24.4 => Choose N = 25.
So if my math is correct, 25 Primary turns on an ETD44 gives a maximum flux of 0.2 Tesla.
Now lets see if the winding fits on the core. I have an ETD44 bobbin and a pair of calipers here so I am taking physical measurements to get my numbers. The window on this bobbin is 1.167" wide and 0.27" tall. Lets assume I have a spacer between the primary and secondary that is 0.125" wide. That leaves 1.042" for windings. Lets put the spacer in the middle. So the primary winding window is 0.502" x 0.27". In mm this is 12.75mm * 6.85mm.
Lets say I need the primary to carry 5.5A. If I want 500 circular mills per amp then I need 2750 cmils. Circular mils are calculated by squaring the diameter of the wire in mils, so lets take the square root of 2750 cmils to find the required diameter. sqrt(2750) = 52.4 mils diameter. Now lets find the cross sectional area that this corresponds to.
From a wire table, a 52 mil diameter wire has an area of about 1.31 mm^2 (Yes, I am mixing metric and imperial). I am using Chinese litz wire that is many strands of 0.1 mm diameter wire. 0.1mm diameter wire has an area of 0.00797 mm^2 per strand. 1.31/0.00797 = 164. So I need 164 strand wire or greater for 500 cmils per amp or better.
Lets say I pick litz wire that is 1.3mm diameter for a bundle of 100 strands of 0.1mm wire. To get above 164 strands I will use two pieces of litz in parallel on the primary and one strand on the secondary. If I use a 1:2 ratio then this means I need 50 turns on each side.
The window is 12.75mm * 6.85mm. Lets assume a poor packing density to start, to do this we assume the wire is square. The window area for each winding is 87.3mm^2. Now for the square wire: (1.3mm)^2 = 1.69mm^2. Now lets see if it fits!
87.3mm^2 / 1.69mm^2 = 51 Turns!
So unless I have made a mistake, I can just fit the required winding on an ETD44 core while keeping the flux density below 0.2T. The ETD44 datasheet says the core should dissipate less than 9.4 W at 0.2T at 100kHz. This is close enough to my 80kHz to 120Khz operating range to be a decent estimate.
It may sound crazy but my math shows I can push 1100W through an ETD44 core.
First of all, i hope you understood i was designing for a standard half bridge. Second of all i know very little about resonant transformer design so all i can do is look at your numbers and check the math.
“So if my math is correct, 25 Primary turns on an ETD44 gives a maximum flux of 0.2 Tesla.”
Close enough i get .1952 Tesla.
“Now lets see if the winding fits on the core. I have an ETD44 bobbin and a pair of calipers here so I am taking physical measurements to get my numbers. The window on this bobbin is 1.167" wide and 0.27" tall. Lets assume I have a spacer between the primary and secondary that is 0.125" wide. That leaves 1.042" for winding’s. Lets put the spacer in the middle. So the primary winding window is 0.502" x 0.27". In mm this is 12.75mm * 6.85mm.”
I think there is a mistake here. measured bobbin width = 1.167“ - .125“ for spacer = 1.042“ for winding’s. Then you say “put the spacer in the middle” and the primary window is .502“ wide. If you meant loosely in the middle and you wanted to bias the winding’s window then you are OK. If you literally meant in the middle then there is a slight mistake. 1.042“ / 2 = .521“ not .502“. This would actually give you more room to work with. So your design should be OK.
At the end i do not agree with your rosy scenario of .2T being OK to run. I do not pretend to be an expert, just like you I have been trying to learn switching power supplies on my own. But I think 9.4W is high, and after you add in copper losses ( P = I sqrd R) I would bet that core is running over 100 deg C.
“It may sound crazy but my math shows I can push 1100W through an ETD44 core.”
ETD-44 at 80khz = 1,300W, according to pressman in his switching power supply book (page 311) for a full or half bridge. He has no charts for a resonant supply.
“1100W through an ETD44 core.” The secret here is that you have litz wire. This is why you can fit 500 Cmils of wire in that core at 80khz. Try it again with regular magnet wire, no can do!
I like your use of metric wire calculations i think i will save this post for a reference.
“So if my math is correct, 25 Primary turns on an ETD44 gives a maximum flux of 0.2 Tesla.”
Close enough i get .1952 Tesla.
“Now lets see if the winding fits on the core. I have an ETD44 bobbin and a pair of calipers here so I am taking physical measurements to get my numbers. The window on this bobbin is 1.167" wide and 0.27" tall. Lets assume I have a spacer between the primary and secondary that is 0.125" wide. That leaves 1.042" for winding’s. Lets put the spacer in the middle. So the primary winding window is 0.502" x 0.27". In mm this is 12.75mm * 6.85mm.”
I think there is a mistake here. measured bobbin width = 1.167“ - .125“ for spacer = 1.042“ for winding’s. Then you say “put the spacer in the middle” and the primary window is .502“ wide. If you meant loosely in the middle and you wanted to bias the winding’s window then you are OK. If you literally meant in the middle then there is a slight mistake. 1.042“ / 2 = .521“ not .502“. This would actually give you more room to work with. So your design should be OK.
At the end i do not agree with your rosy scenario of .2T being OK to run. I do not pretend to be an expert, just like you I have been trying to learn switching power supplies on my own. But I think 9.4W is high, and after you add in copper losses ( P = I sqrd R) I would bet that core is running over 100 deg C.
“It may sound crazy but my math shows I can push 1100W through an ETD44 core.”
ETD-44 at 80khz = 1,300W, according to pressman in his switching power supply book (page 311) for a full or half bridge. He has no charts for a resonant supply.
“1100W through an ETD44 core.” The secret here is that you have litz wire. This is why you can fit 500 Cmils of wire in that core at 80khz. Try it again with regular magnet wire, no can do!
I like your use of metric wire calculations i think i will save this post for a reference.
I have done a bit more work on this design and I conclude that I need to limit the flux density to 0.1 Tesla max. I found a document that claims that 100mW/cm^3 produces about a 40 degree C rise in ferrite cores. An ETD44 has a volume of 17.8 cm^3 so I should limit my core loss to about 1.78W. The datasheet shows this happens around a flux density of 0.1T.
Here is that document for future reference: http://www.lodestonepacific.com/distrib/pdfs/Magnetics/Design_Application_Notes.pdf
Lets say I double the primary turns for a total of 50 turns on the secondary and 100 turns on the primary.
Lets calculate the copper loss in this case. 100*0.1 litz wire has an area of about 0.78mm^2. This has a resistance of about 0.022 Ohms/meter. The average turn length on the ETD44 bobbin is 77mm from the datasheet. So for 50 turns it will take 3.85m of wire. This is a total of 0.0847 Ohms. Lets push 5.5 A through this resistance.
P = I^2*R so P = 2.56W for the primary.
Lets do the secondary now. I need to fit 100 turns so lets use wire with half the cross sectional area. This piece of wire will have four times the amount of resistance because it is twice as long and half the area.
P = 2.75^2*0.0847*4 = 2.56W, the same as the secondary.
So is 5W too much copper loss? I suppose I should read more about this. This design gives about 7W total loss in the transformer. This seems low to me but perhaps the winding will get too hot.
An easy solution is to pick an ETD49 core and fit larger wire on it. I still think that an ETD44 could work with a little more optimization.
Here is that document for future reference: http://www.lodestonepacific.com/distrib/pdfs/Magnetics/Design_Application_Notes.pdf
Lets say I double the primary turns for a total of 50 turns on the secondary and 100 turns on the primary.
Lets calculate the copper loss in this case. 100*0.1 litz wire has an area of about 0.78mm^2. This has a resistance of about 0.022 Ohms/meter. The average turn length on the ETD44 bobbin is 77mm from the datasheet. So for 50 turns it will take 3.85m of wire. This is a total of 0.0847 Ohms. Lets push 5.5 A through this resistance.
P = I^2*R so P = 2.56W for the primary.
Lets do the secondary now. I need to fit 100 turns so lets use wire with half the cross sectional area. This piece of wire will have four times the amount of resistance because it is twice as long and half the area.
P = 2.75^2*0.0847*4 = 2.56W, the same as the secondary.
So is 5W too much copper loss? I suppose I should read more about this. This design gives about 7W total loss in the transformer. This seems low to me but perhaps the winding will get too hot.
An easy solution is to pick an ETD49 core and fit larger wire on it. I still think that an ETD44 could work with a little more optimization.
I have just done some more work on this. I calculate about 9.5 in^2 area for my transformer if I use an ETD44. There is a nice table of temperature rise vs power for various surface areas in the Pressman book on p. 319. I estimate I will see an 80 degree C rise above ambient with no active cooling. However, In my design, there is a 120mm fan about half an inch away from the transformer blowing straight down onto it and through it. From what I have read in the past, I can expect the thermal resistance of a heatsink to reduced by a factor of 2 or 3 times under these forced air cooling conditions.
I suspect I could pull off the design on an ETD44 if I really wanted to.
I suspect I could pull off the design on an ETD44 if I really wanted to.
“I have done a bit more work on this design and I conclude that I need to limit the flux density to 0.1 Tesla max. I found a document that claims that 100mW/cm^3 produces about a 40 degree C rise in ferrite cores. An ETD44 has a volume of 17.8 cm^3 so I should limit my core loss to about 1.78W. The datasheet shows this happens around a flux density of 0.1T.”
I have seen 100mW/cm^3 before and use that as a guideline for picking flux density, at least i think i do. I look at the material data sheet for say 3C94 and look at the chart for 100mW/cm^3, flux density and plotted frequency. The chart says it is plotted at 100 Deg C so i shift the freq plot over a little to the left for less temperature. When i look at the chart and allow for windage. I see that at 70khz you should run 1600G max. But with this method all cores with 3C94 should run 1600G. I do not really know if this is right or wrong.
“An ETD44 has a volume of 17.8 cm^3 so I should limit my core loss to about 1.78W.”
I did not see this calculation in the document you posted. So i am unsure what to make of it.
“So is 5W too much copper loss? I suppose I should read more about this. This design gives about 7W total loss in the transformer. This seems low to me but perhaps the winding will get too hot.”
What i have read, is that as a total power loss goes, a good design will have about half core loss and half copper loss. But the only thing that matters in the end is that the total loss is not exceeded. I am not sure were the total loss you can have comes from or what the 1.78W you calculated means for sure.
“There is a nice table of temperature rise vs power for various surface areas in the Pressman book on p. 319. I estimate I will see an 80 degree C rise above ambient with no active cooling. However, In my design, there is a 120mm fan about half an inch away from the transformer blowing straight down onto it and through it. From what I have read in the past, I can expect the thermal resistance of a heatsink to reduced by a factor of 2 or 3 times under these forced air cooling conditions.”
I agree with you, 2x easily and probably 3x, after all thats how our computer power supplies are built. Your last design would work fine even with a low CFM fan.
I have seen 100mW/cm^3 before and use that as a guideline for picking flux density, at least i think i do. I look at the material data sheet for say 3C94 and look at the chart for 100mW/cm^3, flux density and plotted frequency. The chart says it is plotted at 100 Deg C so i shift the freq plot over a little to the left for less temperature. When i look at the chart and allow for windage. I see that at 70khz you should run 1600G max. But with this method all cores with 3C94 should run 1600G. I do not really know if this is right or wrong.
“An ETD44 has a volume of 17.8 cm^3 so I should limit my core loss to about 1.78W.”
I did not see this calculation in the document you posted. So i am unsure what to make of it.
“So is 5W too much copper loss? I suppose I should read more about this. This design gives about 7W total loss in the transformer. This seems low to me but perhaps the winding will get too hot.”
What i have read, is that as a total power loss goes, a good design will have about half core loss and half copper loss. But the only thing that matters in the end is that the total loss is not exceeded. I am not sure were the total loss you can have comes from or what the 1.78W you calculated means for sure.
“There is a nice table of temperature rise vs power for various surface areas in the Pressman book on p. 319. I estimate I will see an 80 degree C rise above ambient with no active cooling. However, In my design, there is a 120mm fan about half an inch away from the transformer blowing straight down onto it and through it. From what I have read in the past, I can expect the thermal resistance of a heatsink to reduced by a factor of 2 or 3 times under these forced air cooling conditions.”
I agree with you, 2x easily and probably 3x, after all thats how our computer power supplies are built. Your last design would work fine even with a low CFM fan.
"“An ETD44 has a volume of 17.8 cm^3 so I should limit my core loss to about 1.78W.”
I did not see this calculation in the document you posted. So i am unsure what to make of it."
Well, we know that 100mw/cm^3 is what we want to aim for. My core is 17.8 cm^3 so I have 17.8 times more than 100mw to play with. 0.1*17.8 = 1.78W. That is how I arrived at that number.
I did not see this calculation in the document you posted. So i am unsure what to make of it."
Well, we know that 100mw/cm^3 is what we want to aim for. My core is 17.8 cm^3 so I have 17.8 times more than 100mw to play with. 0.1*17.8 = 1.78W. That is how I arrived at that number.
CJK2, Different subject. Now that you have had a couple of years to test you lab supply how is it working.
http://www.diyaudio.com/forums/power-supplies/166272-smps-nearly-finished-not-stable.html
Does it really go to 0vdc, is it stable, have you made any mods since this post.
http://www.diyaudio.com/forums/power-supplies/166272-smps-nearly-finished-not-stable.html
Does it really go to 0vdc, is it stable, have you made any mods since this post.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.