I'm designing a SMPS for 2x 70V @ 300W (yes i only need 8 ohm capability in my amp). My main reference is "Switching Power Supplies, Third Edition", and i'm a bit confused with the transformer calculations. More specifically the wire diameter which i need. I'll be using two-transistor forward topology, and either EI33 or EC35 core (more likely EC35, as one can find a proper datasheet for it, the infos i could find for the EI33 are kinda dodgy). Switching frequency is 96kHz.
In the book it is recommended to use 500 circular mils per RMS ampere. Trouble is, at ~3A that adds up to an awful load of wire! The results of the calculations are 59 turns for the primary, and 35 + 35 turns for the secondaries. It's kinda obvious that it won't fit in the tiny EI33 at that thickness.
But, in cheap computer PSUs you often see the EI33, and they claim to get 300-400W out of it. It is invariably wound with a single 22 AWG wire in the primary, and six of them for the fattest secondary which is the 5v one. There is one notable difference which is that those power supplies are running halfbridge which if i remember right can squeeze twice the power from a given core size. Or was that just about the magnetic aspect, not the copper one... And of course, those power supplies are of the kind that explode when you try to draw any more than half their rated load.
However i have also seen people around here claiming they can get 200W or more out of that tiny core. My question is, of course, HOW??? I've seen the number 200 circular mils per ampere for short wire runs, and that does seem to indicate that a single 22 awg (or two 28 awg in my case, skin effect at play), would be appropriate for the task. That would just about fit in the tiny EI33 core.
But what would be the copper loss in this case? Anyone who is knowledgeable in SMPS transformers please chime in. I'm interested in the minimum number of circular mils per ampere needed to keep the core temperature at 30C above ambient or lower, without taking core loss in consideration for now, as i can figure that out by myself.
In the book it is recommended to use 500 circular mils per RMS ampere. Trouble is, at ~3A that adds up to an awful load of wire! The results of the calculations are 59 turns for the primary, and 35 + 35 turns for the secondaries. It's kinda obvious that it won't fit in the tiny EI33 at that thickness.
But, in cheap computer PSUs you often see the EI33, and they claim to get 300-400W out of it. It is invariably wound with a single 22 AWG wire in the primary, and six of them for the fattest secondary which is the 5v one. There is one notable difference which is that those power supplies are running halfbridge which if i remember right can squeeze twice the power from a given core size. Or was that just about the magnetic aspect, not the copper one... And of course, those power supplies are of the kind that explode when you try to draw any more than half their rated load.
However i have also seen people around here claiming they can get 200W or more out of that tiny core. My question is, of course, HOW??? I've seen the number 200 circular mils per ampere for short wire runs, and that does seem to indicate that a single 22 awg (or two 28 awg in my case, skin effect at play), would be appropriate for the task. That would just about fit in the tiny EI33 core.
But what would be the copper loss in this case? Anyone who is knowledgeable in SMPS transformers please chime in. I'm interested in the minimum number of circular mils per ampere needed to keep the core temperature at 30C above ambient or lower, without taking core loss in consideration for now, as i can figure that out by myself.
Last edited:
Inches? Bah, i have to dig out the tables again. Well, with that formula, it looks like the 22AWG will carry exactly one amp, not even remotely close to three. With the math in the book i ended up with ten strands of 28AWG, and that's pretty much what your formula says too. So that's settled, i need a bigger core. 
I use J_Cu = 7.75 A/mm^2 for most calculations. Smaller transformers can go higher, larger transformers have to go lower. 2000 A/in^2 is typical of large iron transformers where lots of turns are buried inside.
SMPS transformers get away with a lot of current density because of direct airflow and consumer grade product lifetimes.
Tim
SMPS transformers get away with a lot of current density because of direct airflow and consumer grade product lifetimes.
Tim
Mm, so that's how they get away with a single 22AWG at 200-something watts.
Funny you say that. Just a couple days ago i had a 19" LG monitor in for repair. It would not power on, power supply in protect mode. Reason? Caps gone bad because they were too close to the SMPS transformer which ran really really hot. The monitor was exactly 3 years old (made in July 2007), and i had a good laugh when i read the MTBF in the service manual - 50k hours. Yeah sure.
Anyway, i was bored enough that i wound that EI33 with some 22AWG i had around, which isn't too appropriate at my frequency due to skin effect and blah blah, but oh well. I'll try it out tomorrow and see if anything blows up.
This one is kinda "in between"... Anyway i finished winding the EI33 with some sort of secondary (two pieces of wire both too thick and too short), and put it to work. I went for 200mT max flux, thus 47 primary turns, and i didn't count the secondary ones as they're wrong anyway.
With all this, initial impression is positive. I'm currently running open-loop, and i have 10.6 volts output, while heating the heck out of a resistor that used to be the minimum load in the 5v rail of an ATX power supply. The 100W bulb i put in series with the mains lights up at about 40%. The core gets just a bit warm, the switching transistors get slightly hot but not worryingly so (i'm guessing 50-60C, i'll get some exact temperatures later), considering that they were supposed to be cooled by a fan and i don't have any right now, i think it's fine.
There's one interesting thing about this power supply though. The optimal duty cycle seems to be WAY under 40%, i think that right now it's set at 20% or so. At 40% duty cycle i would get 8.5v and the bulb would light up at like 80% brightness, indicating higher consumption. Even at this low duty cycle the bulb lights up at full brightness (and of course output voltage drops along with it) if i try to do dumb things such as shorting the output, so i'm guessing this thing will be able to deliver a fair amount of power once i get the secondaries wound properly.
I'm starting to get the hang of this.
Edit: I noticed there were actually TWO resistors which i was heating up. With the tougher one removed, i'm getting 20 volts and zero light from the bulb. Looks good.
Last edited:
I threw the feedback loop in the mix too, and everything looked good, i got a nice stable 12.5 volts. I gave it 30 watts load and the lightbulb was already limiting me a lot, so i replaced it with a fuse and NTC, just how it's supposed to be. Result? Working perfectly.
But if you hear anything hissing, that's gotta be my transformer. I use pulse skipping as method of regulation, and it's making A LOTTA NOISE... But i think some varnish will help cure this issue. The upside to it is that during testing i can tell the load % by the noise that the transformer is making. And it had a lot more to go from 30W, but i don't really have suitable load resistors for 12v so i have to get the secondaries done first. The transformer doesn't get hot at all, but the transistors kinda do. I'll try a larger heatsink, after all, they came from a cheap chinese supply with a very loud fan, so the fan probably had its purpose.
Anyway, i built something that didn't blow up, i'm really happy.
But if you hear anything hissing, that's gotta be my transformer. I use pulse skipping as method of regulation, and it's making A LOTTA NOISE... But i think some varnish will help cure this issue. The upside to it is that during testing i can tell the load % by the noise that the transformer is making.
And it had a lot more to go from 30W, but i don't really have suitable load resistors for 12v so i have to get the secondaries done first. The transformer doesn't get hot at all, but the transistors kinda do. I'll try a larger heatsink, after all, they came from a cheap chinese supply with a very loud fan, so the fan probably had its purpose.Anyway, i built something that didn't blow up, i'm really happy.
Last edited:
Hissing means your control loop sucks. Specifically, you have a chaotic system where nonlinearities and time delays contribute to make a "never quite the same each cycle" system. Often, a forward converter under short circuit will hiss because the control loop is on the threshold of the lowest PWM possible; PWM generators seem to have hysteresis, so at one point it's 0%, then next it's 0.5% and up. It hisses as the control loop attempts to maintain this boundary condition, obviously doing a poor job because it's highly nonlinear and sensitive.
You may find a constant-on-time circuit more appealing than "pulse skipping". This is one example:
http://webpages.charter.net/dawill/tmoranwms/Circuits_2010/Fast_DCDC.png
The MOSFET stays off until C3 is charged to threshold, where it fires again. On time is limited by R4/Q3, which turns it off when the inductor is charged to a nominal current. As the load increases from zero, frequency rises, then as oscillation becomes continuous, frequency begins to fall again, because on-time is forced to increase. At full load, Q1/R1 saturates and additional current will not be delivered.
Tim
You may find a constant-on-time circuit more appealing than "pulse skipping". This is one example:
http://webpages.charter.net/dawill/tmoranwms/Circuits_2010/Fast_DCDC.png
The MOSFET stays off until C3 is charged to threshold, where it fires again. On time is limited by R4/Q3, which turns it off when the inductor is charged to a nominal current. As the load increases from zero, frequency rises, then as oscillation becomes continuous, frequency begins to fall again, because on-time is forced to increase. At full load, Q1/R1 saturates and additional current will not be delivered.
Tim
Foe a start; how thick are the ac powercable in the wall that you`re you`re hookin up to? The primary wire should be +/- the same gauge if you`re goal is to build a strong amp.
The rest is well known; compensate for lower voltage on secondarys by doubling up secondarys diameter as many times as the voltage is reduced.
The poweramps trafos are the heart of a system, weakness here can`t be regained anywhere else.
The rest is well known; compensate for lower voltage on secondarys by doubling up secondarys diameter as many times as the voltage is reduced.
The poweramps trafos are the heart of a system, weakness here can`t be regained anywhere else.
Hmm, self-oscillating. I could never get that kinda stuff to work properly for some reason. Isn't that the same type of circuit you find in CFLs, but with isolation?
It's true that pulse skipping might not be the best way to go about this. But i lifted the controller board from another converter that i built, a 5v -> 12v boost. The boost inductor did not hiss at all in that converter, so in this one it's probably just the crappy windings, as i used recovered wire. It isn't even the right size, so i probably have some skin effect too... I'll buy a new spool of 28AWG and use as many strands of that as required. I found a couple more EI33 cores in my junk box, so if needed i'll use one core per rail.
Closer to a VCR. For some reason, all VCRs I've taken apart in the last while use a self oscillating circuit. Other consumer equipment on the same power level doesn't do this. There must be something about VCR-design-culture which uses these things.
Self oscillating circuits can be designed and built safely. The above circuit is current mode, so it won't short out the supply if it stops oscillating.
Tim
Self oscillating circuits can be designed and built safely. The above circuit is current mode, so it won't short out the supply if it stops oscillating.
Tim
Thanks for the explanations, but for now i prefer to stay with my current circuit with the crappy BJTs, and get it fully working. I kept on blowing the switching transistors whenever i attempted to run the power supply open-loop or under heavy load at duty cycles greater than 0.2 or so, and now i know the reason why. I need to prevent Ic/Vce overlap as the transistors turn off, and i will try to accomplish that by means of a non-dissipative snubber. That should make heat go down, efficiency go up, SOA curve respected and happy BJTs.
Today i bought new MJE13007s and a spool of 28awg (0.3mm here in Europe) wire. Before i try the non-dissipative snubbers i added regular RCD snubbers, and things worked out well - no more bang and pop! The snubber resistors get quite hot even though they're rated for 5W, now i understand why the poor BJTs kept blowing.
Right now the power supply is delivering 40W at 14 volts out and the transistors are just warm (~60C) with their only cooling being a heatsink from an ATX power supply, no fan. I'll wind a new transformer tomorrow and see what i'll get with that.
Then of course, the really fun part will follow... designing the amplifier.
The snubber resistors get quite hot even though they're rated for 5W, now i understand why the poor BJTs kept blowing.Right now the power supply is delivering 40W at 14 volts out and the transistors are just warm (~60C) with their only cooling being a heatsink from an ATX power supply, no fan. I'll wind a new transformer tomorrow and see what i'll get with that.
Then of course, the really fun part will follow... designing the amplifier.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.