Hi guys!
Great forum you have here, already found a lot of usefull things!
But for my actuall problem i found not so much.
I need some more information about dc/dc-converters, especially 12V with power round about 1kW output voltage +/-25V. Some catchwords will be usefull for me.
My idea was to design a push-pull-concept with switching frequency round about 500kHz, RM pot core and active rectifying.What do you think about this, is it practicable?
Thanx
Wolfgang
Great forum you have here, already found a lot of usefull things!
But for my actuall problem i found not so much.
I need some more information about dc/dc-converters, especially 12V with power round about 1kW output voltage +/-25V. Some catchwords will be usefull for me.
My idea was to design a push-pull-concept with switching frequency round about 500kHz, RM pot core and active rectifying.What do you think about this, is it practicable?
Thanx
Wolfgang
Pushpull is the only way to go if you want 1000 watts, but this will draw huge amps fron your 12v source.
Iam building a smps myself and i can tell you its no game. I have blown rectifier diodes (on the output of the thing), overvolted the test amp with 200+v, chasing problems causing oscillation and noise for days.
You could salvage the smps from a busted 1kw car amp, yeah that wont be as fun as building it but i knocked out tv reception here with my smps sometimes so building from scratch is not reccomended, especially not in your power ranges.
Iam building a smps myself and i can tell you its no game. I have blown rectifier diodes (on the output of the thing), overvolted the test amp with 200+v, chasing problems causing oscillation and noise for days.
You could salvage the smps from a busted 1kw car amp, yeah that wont be as fun as building it but i knocked out tv reception here with my smps sometimes so building from scratch is not reccomended, especially not in your power ranges.
My question is now, how to dimension the smps. Is it usefull to take the formulas for the ideal converter, like it is the way for lower power?
Or where can i find a mathematic model which is for the reality?
In my opinion i think it will be better to dimension the devices
with a model which is equivalent to the reality, in such power regions. What is you opinion?
Are there any links or do you know any usefull books?
Thnax for your help!
Or where can i find a mathematic model which is for the reality?
In my opinion i think it will be better to dimension the devices
with a model which is equivalent to the reality, in such power regions. What is you opinion?
Are there any links or do you know any usefull books?
Thnax for your help!
for that power u should use 2 or more transformers
and i think your freq range is too high , up to 100khz sounds more practical .
and i think your freq range is too high , up to 100khz sounds more practical .
Try to develop your own model. For example : Assume a continuous output current value. Find out rectifier dissipation according to datasheets. Select a suitable model and amount of devices in paralell to suit your heatsinks (both mechanically and thermally).
Build or figure out a transformer prototype, measure the required magnet wire length and find out primary and secondary DC resistance and theoretical wire dissipation. If it's too high then build a new prototype with more magnet wire (the best way to know that is to measure actual temperature increase in a working prototype). Don't forget to consider the effective cross-sectional area reduction in the wires due to skin effect if your switching frequency is well over 30Khz.
Find out primary winding current. Assume a certain crossover time, for example 100nS, and calculate switching losses. Then calculate conduction losses for several switch models and amounts of devices until something that suits your heatsinks is found...
If you know some programming language, you may find useful to write small programs in order to provide some degree of automation thus making trial and error calculations much easier (It may sound a bit weird, but I use the good old GWBASIC for that purpose).
Finally, if you have doubts concerning basic component behaviour and forumulas, try to search that forum and try also google (sometimes component datasheets and application notes include precious information).
Build or figure out a transformer prototype, measure the required magnet wire length and find out primary and secondary DC resistance and theoretical wire dissipation. If it's too high then build a new prototype with more magnet wire (the best way to know that is to measure actual temperature increase in a working prototype). Don't forget to consider the effective cross-sectional area reduction in the wires due to skin effect if your switching frequency is well over 30Khz.
Find out primary winding current. Assume a certain crossover time, for example 100nS, and calculate switching losses. Then calculate conduction losses for several switch models and amounts of devices until something that suits your heatsinks is found...
If you know some programming language, you may find useful to write small programs in order to provide some degree of automation thus making trial and error calculations much easier (It may sound a bit weird, but I use the good old GWBASIC for that purpose).
Finally, if you have doubts concerning basic component behaviour and forumulas, try to search that forum and try also google (sometimes component datasheets and application notes include precious information).
For example when i want to find out the dissipation for my active rectifier i must use the equivalent circuit diagram.
What is usefull for me, i think i have to decide about two models.
First the Rdson and the capacity loss like Pv=Pon+Pswitch
Second only the capacity loss in the switching moment like Pv=Pswitch
And then i will use the biggeset one, i think that is right...
But i'am afraid of the transformer
...
What is important therefore? I think the transformer is the device with the biggest loss or do i get it wrong?
What is usefull for me, i think i have to decide about two models.
First the Rdson and the capacity loss like Pv=Pon+Pswitch
Second only the capacity loss in the switching moment like Pv=Pswitch
And then i will use the biggeset one, i think that is right...
But i'am afraid of the transformer
...What is important therefore? I think the transformer is the device with the biggest loss or do i get it wrong?
Wolfgang said:For example when i want to find out the dissipation for my active rectifier i must use the equivalent circuit diagram.
What is usefull for me, i think i have to decide about two models.
First the Rdson and the capacity loss like Pv=Pon+Pswitch
Second only the capacity loss in the switching moment like Pv=Pswitch
And then i will use the biggeset one, i think that is right...
But i'am afraid of the transformer
What is important therefore? I think the transformer is the device with the biggest loss or do i get it wrong?
You are definitely going to wind up stacking several E-Cores for this job -- this necessitates making your own bobbin. Not that difficult -- if you can get a copy of the ARRL Handbook at your local ham radio club in Germany (they are plentiful) the technique is well described. The XQ2FOD desinged 1/2 kW off line switcher used 5 Amidon EE-77-625 cores (equivalent TDK EE42/42/15) -- and you will need more!
When you calculate the transformer parameters, you need to know both the minimum and maximum current that the powered device will consume (SMPS don't take kindly to devices that idle with low current consumption). You also have to know the range of input voltages that the switching devices will see -- for a car it will be something like 7 to 15 volts.
Just a rough guess -- primary 3.5 turns, secondary 15 turns for 100kHz SMPS -- something between 660uF and 1,000 uF as the output filtration capacitor -- something like 20uH as the output inductor. Should be good from 50W to 1kW.
Attachments
jackinnj said:
1kW from single EE-core should be no problem at all. EE65 should do more than 3kW at around 100khz and EE55 should fit this job wery well, as suggested by Eva.
Single EE42 would be pushing your luck @1kW , but should be do-able if operated around 200-400khz in ZVS bridge.
ARRL article used very low switching frequency because bipolar transistors.
I agree with mzzj, furthermore, I've obtained up to 1.8Kw output from two E42/21/20 transformers (a pair of cores in each) at 33Khz with some air cooling. However, this was an off-line application and low voltage windings tend to require more space and plenty of coil former pins or soldering the magnet wires directly to the PCB.
Another alternative is toroid cores, as the ones employed in most car-audio amplifiers, see the picture :
This one has 6+6:12+12 turns and produces +-23V output. It was operated around 40Khz, I think, and approx. 400W output may be expected from something of that size. Note that output voltage will always sag under load in unregulated push-pull converters due to leakage inductance, but paralelling several small transformers solves the problem.
Another alternative is toroid cores, as the ones employed in most car-audio amplifiers, see the picture :
An externally hosted image should be here but it was not working when we last tested it.
This one has 6+6:12+12 turns and produces +-23V output. It was operated around 40Khz, I think, and approx. 400W output may be expected from something of that size. Note that output voltage will always sag under load in unregulated push-pull converters due to leakage inductance, but paralelling several small transformers solves the problem.
...would also recommend switching frequencies below 50kHz.
In this power and current range high frequencies typically cause more trouble than fun, - except you are a real freak !
Well, I would not dare my first steps with 100A above 50kHz.
Your switches will have to be quite fat types, with high gate charge. And this will cause increasing driver losses with higher frequencies.
Also you will probably fight with parasitic ringing effects, and
with higher frequencies, the time relation of ringing vs. clean signal will become worse.
Large E-core windings are suffering from proximity effect losses, so you would need thin multi stranded wires. For optimized low losses the strands must be much thinner than you would expect just from skin effect.
In this regard torroids have a more fortunate geometry and simple skin effect calculation is closer to reality.
Losses of the core:
The larger the core is becoming, the unfortunate the relation between core volume and surface will result.
Volume increases by size^3 while the surface just increases by size^2.
Unfortunately the losses (for a given f,B,material) are proportional to the volume, while the heat dissipation ability just increases by the surface.
So you probably cannot run a E65 with 100kHz push pull and +/-300mT, ...I guess, but never tried, and never caculated up to now..
When you are going for thermal tests: Don't let you trick by the time constant. Measurements after 15-20min full power are meaningless, such transformers have much longer time constants ....
But from thermal view we hifi guys have an advantage vs most
industrial applications.
Music does not draw full power constantly. Typically, when you drive an amp up to clipping the consumed average power is just around 10%-20% of the clipping peaks, due to the nature normal music program.
Have fun
Markus
In this power and current range high frequencies typically cause more trouble than fun, - except you are a real freak !
Well, I would not dare my first steps with 100A above 50kHz.
Your switches will have to be quite fat types, with high gate charge. And this will cause increasing driver losses with higher frequencies.
Also you will probably fight with parasitic ringing effects, and
with higher frequencies, the time relation of ringing vs. clean signal will become worse.
Large E-core windings are suffering from proximity effect losses, so you would need thin multi stranded wires. For optimized low losses the strands must be much thinner than you would expect just from skin effect.
In this regard torroids have a more fortunate geometry and simple skin effect calculation is closer to reality.
Losses of the core:
The larger the core is becoming, the unfortunate the relation between core volume and surface will result.
Volume increases by size^3 while the surface just increases by size^2.
Unfortunately the losses (for a given f,B,material) are proportional to the volume, while the heat dissipation ability just increases by the surface.
So you probably cannot run a E65 with 100kHz push pull and +/-300mT, ...I guess, but never tried, and never caculated up to now..
When you are going for thermal tests: Don't let you trick by the time constant. Measurements after 15-20min full power are meaningless, such transformers have much longer time constants ....
But from thermal view we hifi guys have an advantage vs most
industrial applications.
Music does not draw full power constantly. Typically, when you drive an amp up to clipping the consumed average power is just around 10%-20% of the clipping peaks, due to the nature normal music program.
Have fun
Markus
Chocoholic!! Nice to see you again! 😀
There is not so much proximity effect at 30Khz but there is a whole lot above 100Khz. It appears to be not linearly related to the frequency at all, but I can assure that 1mm diameter magnet wires may be employed at low frequencies without worrying about drastic effective cross-sectional area reductions (ie: lower switching frequency requires a bigger core but winding is easier).
Concerning temperature rise and time constants, I've done several tests that lasted for a few hours in which E42/21/20 ferrite transformers were involved, and unlike what happens with bigger 50hz transformers that take hours to stabilise, these smaller high frequency ones appeared to reach steady state in 15 minutes or so. Bigger transformers may take longer, though.
There is not so much proximity effect at 30Khz but there is a whole lot above 100Khz. It appears to be not linearly related to the frequency at all, but I can assure that 1mm diameter magnet wires may be employed at low frequencies without worrying about drastic effective cross-sectional area reductions (ie: lower switching frequency requires a bigger core but winding is easier).
Concerning temperature rise and time constants, I've done several tests that lasted for a few hours in which E42/21/20 ferrite transformers were involved, and unlike what happens with bigger 50hz transformers that take hours to stabilise, these smaller high frequency ones appeared to reach steady state in 15 minutes or so. Bigger transformers may take longer, though.
...thanks Eva. 😉
Also missed your voice a little bit.
But I guess, I will have more chances to read & post from now onwards again. Even if I will by DIY-disabled throughout about February..
Yes, at 30kHz I would also feel comfortable.
You use 1mm strands in E42/21/20 ?
Wow! Hm,...would have guessed about 0.5mm strands as an optimum in such a design.
Are you sure about your time constant?
In my 1kW boost, I used this double EVD36... and I had to wait
at least 1h to get stabilized temperatures... OK, this was measured at build in situation ...there it showed that slow heating up (did not measure in a domesticated separation), - but build in is the real situation, isn't it?
Also missed your voice a little bit.
But I guess, I will have more chances to read & post from now onwards again. Even if I will by DIY-disabled throughout about February..
Yes, at 30kHz I would also feel comfortable.
You use 1mm strands in E42/21/20 ?
Wow! Hm,...would have guessed about 0.5mm strands as an optimum in such a design.
Are you sure about your time constant?
In my 1kW boost, I used this double EVD36... and I had to wait
at least 1h to get stabilized temperatures... OK, this was measured at build in situation ...there it showed that slow heating up (did not measure in a domesticated separation), - but build in is the real situation, isn't it?
I think that the contrast between your 0.1mm litz wire and my 1mm solid wire has a lot to do with that difference in heat exchange. Also, my transformers had something like 20% or 30% of free window area and were subject to some amount of forced air cooling (test conditions were not the same at all).
Attached are the basic formulas for calculating skin depth through a round copper wire. Hopefully this can help you out in your quest.
This last summer, with the help of diyaudio.com I built an SMPS that could ouput over 1100Watts RMS. Here's a basic summary:
SG3525 Controller
No Feedback
Push-Pull Topology
6-Mosfets per leg 12 mosfets total IRFP052N
2.25" OD Ferrite Core from Mag-INC [P-Type Material I believe]
10,000uF Output Filter Capacitance
400V 30A Ultrafast Bridge Rectifier
Eva is absolutely right you MUST set up models and go through all pertinent calculations otherwise you will be in for a very bumpy ride! I found the application note: "SWITCHING POWER SUPPLY DESIGN: LM5030 PUSH-PULL CONVERTER" by Michele Sclocchi [freely available pdf download] by National Semiconductor to be quite comprehensive]
Good Luck!
-=Randy Knutson
Mankato MN
This last summer, with the help of diyaudio.com I built an SMPS that could ouput over 1100Watts RMS. Here's a basic summary:
SG3525 Controller
No Feedback
Push-Pull Topology
6-Mosfets per leg 12 mosfets total IRFP052N
2.25" OD Ferrite Core from Mag-INC [P-Type Material I believe]
10,000uF Output Filter Capacitance
400V 30A Ultrafast Bridge Rectifier
Eva is absolutely right you MUST set up models and go through all pertinent calculations otherwise you will be in for a very bumpy ride! I found the application note: "SWITCHING POWER SUPPLY DESIGN: LM5030 PUSH-PULL CONVERTER" by Michele Sclocchi [freely available pdf download] by National Semiconductor to be quite comprehensive]
Good Luck!
-=Randy Knutson
Mankato MN
ChocoHolic said:
strands can be up to skin depth thick but then its 1 layer max. 😉 Gonna need primary and secondary split to 10 pieces and interleaved to accommondate entire window area in big EE-core. Huge pri-sec capasitance and not recommended..
EE65 with deltaB=600mT 😱 Absolutely no way. Or you mean +-150mT? that may be do-able with better core materials like PC44 or 3C92
-------
easy availlability of nanocrystalline strip-wound cores would make high-power/high current supplies sooooo easy... deltaB of 1.6T (+-800mT) at 20khz makes these impossible to beat.
mzzj said:
That's exactly what I would guess, too.
If we estimate that losses are going around f^2 and B^2.8 for many standard ferrites, then you have to go down with the frequency overproportional to allow higher flux density.
If you choose a low loss ferrite then for smaller cores like EF20 it should be possible to handle +/-300mT at 30kHz, as they have a lot of surface compared to their volume.
So for Wolfgang the operating frequency should also depend a little bit on the available magnetics. I would definitely propose a frequency range between 30kHz and 50kHz, not higher.
mzzj said:
Easily available?? Not at all !!! Do I live on another planet??
I got some U shaped cut cores of similar stuff, for some trials with flyback and boost converters, where I additionally needed an air gap.
But they were expensive like hell and definitely not easy to get!!
First experiments showed that the saturation level around 1.4T (deltaB 2.8T !) as given in the data sheet was OK.
But the losses at 50kHz were already high and I could not go continuosly over +/-80mT in a large U-shaped cut core (2 pieces with approx. 35mmx40mmx20mm and Ae around 200mm^2).
So for applications where you might have short duty cycles and low flux density during most of the time it could be OK and allow quite massive peak power. But for continuous high flux conditions it showed poor behaviour.
So even for the 1kW PFC boost converter of my sub woofer, I decided to use a stepped gap ferrite choke.
For push pull applications (or any ungapped power conversion transformer) there seem to be very good torroids from VAC in Germany. Their data sheets and application notes are promising ....
http://www.vacuumschmelze.de/dynamic/docroot/medialib/documents/broschueren/kbbrosch/Pk003.pdf
With a respectable portion of good luck I managed to rescue some single and quite old Vitroperm samples from beeing scrapped. Thanks to
I guess/hope they are still OK. 😎
Up to now, I could not test them. I plan to use them for the
second stage of my sub woofer smps. (The first stage was the 1kW PFC boost converter, which is more or less OK now.)
The second stage will be a half bridge forward converter running from the 470V DC output of the boost and delivering the supply voltage for a Class D amp. So far my current dreams....
ChocoHolic said:
Would make easy.... I meant that their availlability just sucks.
Yes, vacuumschaisse vitroperm F is the material i have been dreaming of 🙂
You lucky bastard! Vitroperm F or some other sort of vitroperm?
btw. Do you have any ide how much for example 50mm vitroperm F cores cost? I would be more than intrested to hear this...
...uhps, now I understand your previous comment..! 🙄
Yes it WOULD be great to have them available in any
electronics DIY-store...
Glad to hear that we are living on the same planet.
It's the Vitroperm 500 F.
Well the price is one of the major disadvantages....
Compared to ferrite it is like diamonds.
Please understand , I cannot tell any prices !
But for private purpose I would be willing to pay it,
even if I guess that my small quantities would end up in even
less fortunate offers.
But as far as I know, they offer it for professional use only.
If you manage to convince the VAC guys, to sell such cores in for privat persons, please let me know.
Yes it WOULD be great to have them available in any
electronics DIY-store...
Glad to hear that we are living on the same planet.
It's the Vitroperm 500 F.
Well the price is one of the major disadvantages....
Compared to ferrite it is like diamonds.
Please understand , I cannot tell any prices !
But for private purpose I would be willing to pay it,
even if I guess that my small quantities would end up in even
less fortunate offers.
But as far as I know, they offer it for professional use only.
If you manage to convince the VAC guys, to sell such cores in for privat persons, please let me know.
Wolfgang,
to get more detailed support on your dedicated smps, you
should think about a preliminary schematic and transformer design.
When you post more details of your plans, I am sure that you will get further detail support from this thread !
Don't hesitate. Nobody will kick you, just because some design flaws.
This forum is always giving great support to avoid many fatal errors during prototyping !
Bye
Markus
to get more detailed support on your dedicated smps, you
should think about a preliminary schematic and transformer design.
When you post more details of your plans, I am sure that you will get further detail support from this thread !
Don't hesitate. Nobody will kick you, just because some design flaws.
This forum is always giving great support to avoid many fatal errors during prototyping !
Bye
Markus
- Status
- Not open for further replies.