| Built a 250W SMPS, possible to modify? - Click HERE for Original Thread |
| simingx |
I built this power supply and it's been working fine at 250W output power (build log is here).
I would like to scale it up to supply a pair of UcD400's (about +/-60V @ 8A). If I understand correctly, I need to:
1) Use a bigger core at a higher frequency (Currently it's an ETD39 @ 50kHz).
2) Obviously, change the turns ratio on the transformer. Referring to this site, for a minimum of 270V and maximum of 340V input and 100kHz, the transformer should be a ETD59 core with 29 (15+15) primary windings and 27 (14+14) secondary windings for a total of 120V output volts?
3) To save some trouble, I might not use the semi-synchronous rectification described in the original article. Instead, I will probably implement a standard fullwave bridge rectifier with a center tapped winding.
But then, adding a few more windings for the synchronous rectifiers probably isn't too difficult either... :)
4) The FETs will have to be upgraded also, I see the maximum current through each FET is something like 9 amps.
Will the existing drive circuitry be able to properly drive bigger FETs?
Will any other (subtle) modifications (besides the obvious changing of OVP and regulation voltage dividers) be required?
Can the existing snubbers still be used successfully?
Thanks! |
|
|
| luka |
1.) Maybe you need bigger core, but not as big as ETD59. It depends on how much wire will you try to put in, as for power, ETD44 can put over 1kw at freq. 50~100kHz.
2.)At input of 270v freq 50kHz, turn ratio would be PRI = 1.4xSEC
I put 90v for sec votage since this page calculates for unregulated supply.This way you can have it regulated.
3.)Best and simple
4.)9A x Vin/2 = 1.4kW+!!. Try to get IRF450, since they are the best for their price, at least for me :)
If you are using original gate trafo that was made to drive NPN tran., it will NOT work with fets.:att'n:
Snubbers will still be useful if you use the same freq., I think |
|
|
| simingx |
Thanks for the guide..
| quote: | 4.)9A x Vin/2 = 1.4kW+!!. Try to get IRF450, since they are the best for their price, at least for me
If you are using original gate trafo that was made to drive NPN tran., it will NOT work with fets. |
I have a pair of IRFP264's lying around (250V / 38A / Rds = 0.075ohm), these should do as the half-bridge voltage is only 170V (with 340V DC side voltage)?
I am using a re-wound gate driver transformer with 26 turns on primary (to the SG3525) and 16+16 turns on the secondary (to the FETs). This should give me slightly more than 10 volts to drive the MOSFET gates...
I will probably be rewinding another gate transformer using an RM core... these are much less hassle than trying to strip the windings from the old transformers :P |
|
|
| megajocke |
| You will have full DC bus voltage over the transistor when the other one is conducting. So you need FETs with a higher voltage rating. |
|
|
| simingx |
| quote: | Originally posted by luka
Snubbers will still be useful if you use the same freq., I think |
I'm probably going to go up to 100kHz... the snubbers on the primary side are 100 ohms + 2.2nF and those on the secondary side are 10 ohm + 10nF. Will these still be usable?
| quote: | Originally posted by megajocke
You will have full DC bus voltage over the transistor when the other one is conducting. So you need FETs with a higher voltage rating. |
So I might have to get IRFP460's or something... |
|
|
| luka |
hi
I don't know much about snubbers, becaude I don't use them so I can't tell you for sure.
Yes, or some like IRFP460. |
|
|
| N-Channel |
I have seen this project and would like to build it. If you're going to go for 120V (+/-60V) @ 8A (close to 1kW), I would give the Full-Bridge topology some serious consideration. Yes, the half-bridge is good for 200-1000W, and, yes, it is simpler to implement, but you will have less drain currents through the MOSFETs and lower primary-side current in the main Xfmr with the fullbridge. Not to mention, capacity to expand the power levels.
Steve |
|
|
| simingx |
N-channel,
Go for it! Winding new transformers isn't difficult and it's guaranteed that they will be at least the correct physical size (some of those AT psu's have such a small main transformer that I'm quite suspicious of them!).
Just one point, I would use (at least) 200V rated polypropylene capacitors in the snubbers (C5 and C8). This point isn't really clear in the text.
For the fullbridge design, am I correct in saying that I just need to put 4 windings on the driver transformer instead of just 2?
Then each pair of the driver windings will go to opposite legs of the transistors? |
|
|
| luka |
hi
Yes you are corrent, but you must pay attencion to how you connect them to fets.You don't want to open 2 mosfets that are connected between +supply and ground, or do you :D ? |
|
|
| N-Channel |
simingx,
Don't tempt me! I need another unfinished project like a hole in the head! :xeye: Seriously though, you are correct in your understanding that four separate windings will be needed for the secondary side of the driver Xfmr. I agree with you on the 200V rating for the snubbers. Probably not necessary, but better to be safe.
As for transformer core size from all those AT & ATX boxes, here is a dirtly little secret every manufacturer of all these standard (and fancy 500W & up) boxes don't tell ya': for a 200-250W rated PSU, the core is good for only about 60-70W continuous and 200-250W on a time-limited basis. This notion is predicated on the idea that, for powering loads with a wide dynamic range (such as disk drive motors and gain-card amplifiers) the power supply will see full power levels for only a second or two, and not continuously.
If you crack open some older AT boxes, I mean like first-generation ones, you will find that the cores are at least twice the size as those found in contemporary boxes. Couple o' reasons for this: 1) Efficiencies got better, eliminating the need for oversized components, and 2) smaller parts were cheaper, and, for the most part, could do the job as well.
Steve |
|
|
| simingx |
Of course not :D
If it's actually THAT simple then I don't see why I couldn't do a fullbridge converter...
But the transformer turns ratio will be different, correct?
At least according to the turns ratio calculator here... |
|
|
| N-Channel |
| Yup. Since the primary side of the main Xfmr will now see the full voltage (320-360V) you will double the number of primary turns. |
|
|
| simingx |
| OK... I managed to pop the 2 IRF730's by (accidentally) shorting the output. Is there any way to protect against short circuits? |
|
|
| luka |
| Yes there is. Short circuit protection :D. For that you will need to place current trafo somewere in primary and sence primary current.If you search this forum you will find a lot of shemas of what I mean. |
|
|
| simingx |
I did do that... but it seems not fast enough to protect the FETs against short circuit (it works for overload by sending the circuit into some sort of "current limit" mode, I'm not sure whether it's supposed to work like that?).
And I'm not very excited about trying that little "experiment" again.. :D |
|
|
| luka |
Is not fast enough?! I think it is faster then fuse blowing up... I seen a lot of supply that have let say 105% current limiting, but if current goes to 130% then it shuts down.
Where did you put your current trafo in your design?
If it did work for overload, it must have been working for short circuit, since in both cases is intended to use more current then wanted.
There must have been any other type of error to blow up your mosfets or your current limiting didn't work. |
|
|
| ifrythings |
I've been doing a lot of learning for something like this too, I have a schematic for a full bridge but not sure if it will work.
I was tring to desgin this so it can run off of the TL494, the SG3525 should work too, I have to premaid drive transformers for fets that I robed out of a cooked server PSU. |
|
|
| ifrythings |
May be a stupid question, but I'm still learning here, What have I put in?
I was tring to find some pictures Eva posted for different turn on and off circuit for Bipolars and fet but the search function on this fourm hates me today, do you know where they are?
Also what is more of a problem slow turn on or slow turn off? (or both)?
Thanks Dave.
BTW the schematic I posted is for 120V mains getting x2 by caps, remove link from bridge to center of caps for 240V operation. |
|
|
| luka |
Well as I understand that is both.Some what slower turn on is good for core saturation, since fast turn off is good all the time I think.What you have put is: For turn off mosfet gate is discharged through resistor and trafo.If there is leakage inductance in trafo, mosfet will not discharged fully, what could result in destruction of fets or high heat.But if you use that shema the fet is turnd on slowly(resistor), but turnd off fast with transistor.
As for pictures: Eva have deleted them from her server, and I have only that shema of what there was. |
|
|
| ifrythings |
I know this isn't quite the same thing but the horizontal circuit of a TV requires a very fast turn on but the turn off isn't as important. That why I'm a bit unsure of what is required for a fet.
I've been reading SWITCHMODE™ Power Supply Reference Manual (SMPSRM-D.pdf) by on-semi, hence why I have implemted that design into my schematic. I will put the circuit you posted earlier in my schematic so when build I don't have :bawling: from toasted fets.
Is it required to have a cap in series in the secondary of the drive transfomer?
Agin, I'm still learning here so all your paciance and advice is extremly appriciated.
P.S: Hey Eva :innocent: :innocent: could you post or re-serve those pictures?
Thanks
Dave. |
|
|
| luka |
Don't mix tv with supply, coz they have nothing in common :).
I don't think the cap is needed there because trafo can't produce DC in case of in primary side of gate trafo failure. |
|
|
| simingx |
Let's not go off-topic... :D
I put the current sense transformer in series with the center point of the capacitors and the main transformer. On overload it limits the current (but doesn't shutdown? I haven't really tried overloading it THAT much).
However, if I short the output and THEN power it on, it (correctly) does not start... what it doesn't survive is a short AFTER power on... |
|
|
| luka |
| That is very strange. Some thing must be happening that shouldn't be.Did you look for any error with scope during increasing overload? |
|
|
| simingx |
Here're some shots of the gate drive waveform... directly on the primary (HV) side of the driver transformer. Note! Don't do this unless your scope has isolated inputs!
The spike looks worrying... what could be causing it? The circuit seems to be limiting the maximum duty cycle (the voltage starts to drop on overload) but it doesn't shut off.
1) No load |
|
|
| simingx |
| 2) Medium load (3x 12V 50W halogen bulbs) |
|
|
| simingx |
3) Overload (6x 12V 50W halogen bulbs)
Of course I could just make sure I don't short circuit the output... :D
Also what could be causing the little "ripple" during the off time? It's not causing any stray turn-ons of the FETs though...
I might use a proper RM core for the next project... |
|
|
| luka |
With no load pic look great. But as for othe two :whazzat: .You will need some snubber for those big spikes and that 2v ripple to minimize it as much as possible.Did you put that gate circuit to use yet?
The ripple can come from badly winded trafo
If you have to have over current, but this won't work there are allway overrated fuses that you can use. |
|
|
| simingx |
Gate circuit? Please elaborate... :)
I might incorporate the snubbers from here...
The 'scope (Tektronix THS720) is going back to the office tomorrow however... :( it's isolated inputs are making me want to buy one :D
(product plug) Everyone who is even remotely serious about doing such projects should have one of these... :D (end product plug) |
|
|
| luka |
Gate circuit from post#18
You will have to see if thoose will work for you, but try gate circuit first if you have components available.
Scopes like that are great but if they wouldn't cost as much :bawling: |
|
|
| simingx |
Thanks I might try that.
From casual inspection I suppose the diodes can be any small signal type such as 1N4148 and the transistors any small PNP type like a BC557/8?
I'm trying very hard not to buy one of those scopes... buying a pre assembled power supply is cheaper :D |
|
|
| luka |
HEHE
Yes diods are small signal, but it can be a fast one too, transistor is that type.
What is your freq.? |
|
|
| simingx |
| Right now it's about 50kHz, but I hope to move to at least 100kHz for future projects.. :) |
|
|
| N-Channel |
| quote: | Originally posted by simingx
Right now it's about 50kHz, but I hope to move to at least 100kHz for future projects.. :) |
Is 50kHz the oscillator or switching frequency? Also, I went back and re-read the earlier posts: Your FETs are IRF730s, correct? Their peak I(d) is only 5A with an Rds(on) of 1W. You might consider upgrading to '740's or even IRFP450s or '460s. Just a thought. |
|
|
| luka |
| 50kHz is more then enough.I have tested my today with 100v that droped to ~96 with 22 ohm load. It is unregulated. |
|
|
| Adnancoskun |
| hi to all, we are making 2000 watt SMPS for 220 volt input and 2x70 volt Output, we will fnish to it in short time, and ý will share,, just you must wait for two days:) |
|
|
| luka |
| Really, that is great news. Can't wait... |
|
|
| carvinguy |
sinimgx,
I've been following this thread and have a few points to add:
1) There really are no generic snubbers- every application needs to be dealt with on an individual basis depending on the ringing frequency and what damping needs to be done. You generally cannot just use values from schematic unless the PCB and all the parts are the same. You should be able to find notes on calculating snubber components from Unitrode/TI app notes.
2) The location where you have your current sense/protection will not protect the supply from FET shoot-through. Another words, you could have cross condution between the 2 FET's in the bridge and the current sense xfmr would not see it. By the looks of your gate waveforms (the ringing) at overload, there could be a chance of both FET's turning on at the same time which would cause smoke very fast. I think you should take a really good look at your FET drive circuit and try to eliminate the ringing. It's hard to elaborate any more on this without seeing your actual supply.
Hope this helps. |
|
|
| Adnancoskun |
| [IMG]http://[/IMG] |
|
|
| N-Channel |
Huh? :confused:
Here's my pc AT supply conversion project: I have an old 235W AT supply, pretty typical: +5 @ 23A, +12V @ 9A, -5V @ -0.5A, and 1-2V @ -0.5A, nothing ground-breaking. That is, until I decided to modify it to power an as yet unbuilt LM4765T chipamp amplifier.
Now, my intention is to modify the output section to get +/-15V as follows: make +5V the -15V line, make +12V line the +15V line, eliminate the -5V & -12V lines all together, and have C-L-C output filtering.
After inspecting the main transformer, I found a separate low-current winding for -12V, so this will remain disconnected. I plan to re-wind the big yellow/white toroid bifilar with the same number turns as the existing +12V line. The two existing common-cathode rectifiers will be replaced wit hMUR1620CT (common-cathode) and MUR1620CTR (common anode) for the +/- outputs.
Since the PWM (TL494) & driver transformer section are already optimized for each other, I will leave them alone, but plan on upgrading the MJE-13007s (400V, 8A) to MJE-13009s (400V, 12A).
I polan on taking pics as I go along, and will post them here.
Steve |
|
|
| luka |
Hi Steve
I'll be waiting for pic as a hawk:D .
I have rebuild one 230w supply too, with output being 13v and 18 amps up to 25. I use it for my home build car amplifier, or for testing car supplys. |
|
|
| N-Channel |
| Perhaps this afternoon. BTW Luka, what was up with that virus warning in your last e-mail? |
|
|
| luka |
Hi Steve
There was virus, but I have discovere it after already sending you that mail with files.Sorry again.
Did you get two other files? The FlukeView 1.1 and 3.04? |
|
|
| N-Channel |
| Their text was there, but not the actual attachments. |
|
|
|
|
