Self-oscillating SMPS with saturable drive transformer

Dem

Member
2005-05-07 12:32 pm
Israel
Does someone neve an experiense with such topology?
I have designed such supplies for up to 200W , but now I need 500W to 1kW... It should not be stabilized, but simple and cheap...
I'm especcially interesting in current-proportional-drive in this SMPS (I didn't use it in 100-200W designs ) and reliability in this mode.
Does someone have proved (or not so-proved:) ) designs of this topogy? Thank You all.
 

Attachments

  • self-osc.gif
    self-osc.gif
    36.3 KB · Views: 2,039
Hi there......A well designed high power switch-mode psu should exhibit s/c ; o/v protection amongst a host of other things.

On low power designs the self oscillatory mode or (gain limited switching) is capable of higher o/p currents than a flyback.

Drawbacks (serious).....Core driven to saturation, for high frequency operation, a square loop low loss ferrite is used...however may not always start with extreme squareness ...critical material selection; good Fe material selection can result in poor switching performance.... high loss and low effic.
Primary side switching stress is high........semi's rated for 2x bulk Vin DC.......
Result.....limited to low Pout applications.

Personally (sorry) I would put cold water on this idea!

richj
 

Dem

Member
2005-05-07 12:32 pm
Israel
richwalters - thank You for the reply,
Usually I love to put cold water on ideas too...
But this is not just an idea - I have built such supplies and successfully used them for powering 2x90W audio amplifiers, it worked just excellent. I was too young to measure all it paramerers as I plan now :), but as I remember at 180W (+/-35v) it had about 10% rippel @100hz and about 4% @30kHz, output impedance was ~0.6-0.8Ohm - all parameters are comparable with "classic" 50/60Hz PS designs. I didn't measured efficiency, but no parts were hot above 60C.
Now I need simple and cheap PS and I think "to restore" my old design, but I'm not sure that it can be simple forced from 200 to more than 500W by transformers enlarging... I think proportional-drive should be used, but I have never dealed with it...

About "a host of other things" in good PS - this simple design has a lot of hidden things - it hasn't overvoltage problem, it does have overcurrent protection and magnetic loop autobalance etc.

I'm thinking about more modern designs too (like IR2153 + IRFxxx), but this design is also is an option, and not worst option to my opinion. It's very old topolgy , of course, so what ?
 
I've used proportional drive for more tan 1Kw output with bipolar transistors in a full bridge configuration. It works reliably and outperforms MOSFETs in efficiency at low frequencies ~30Khz, but it requires a control IC in order to provide a stable oscillator and active turn-off

In proportional drive circuits the Volts*Second product applied to the pulse transformer is usually proportional to the load current so saturable reactors are not practical
 

Dem

Member
2005-05-07 12:32 pm
Israel
Eva - it's not exactly: Volts*Second product applied to the pulse transformer depends only of Volts applied to transformer and Seconds that it's applied. The voltage applied to drive transformer is always = Base-Emitter voltage of saturated transistor, and it doesn't depend of how we obtain it - from voltage feedback or proportional current feedback. It more or less stable (about 1V at deep saturation). Off course, with proportional drive base current is not constant, but Vbe changes will be not more than 10%.
Only the problem I see in this mode - with proportional drive we can lose "built-in" current limit - with voltage only FB, base current is limited, so collector current is limited too, when overcurrent occures - power transistors just exits from saturation to active mode, it's very hard mode for transistors, but no problem to stand in this mode couple of milliseconds until protection fuse will blow. With proportional drive power transistors will kept opened and saturated in overcurrent mode, so no current limit in this mode. A possible solution is to minimize proportional drive and prevent keeping power tr-rs be saturated (if for example hfe of tr-rs is about 10, so proportional drive current ratio should be less than 1/20 - only a half of required base current will obtained from the collector current), so it will not enough to keep tr-r saturated due to high output current, but on-off behavior will improved. But it's ony my theoretical ideas and should be tested.
 
Eva said:
but it requires a control IC in order to provide a stable oscillator and active turn-off

Hi there......that's where the component count goes up....although I've been <designing> smps for many years, I've weaned myself towards standard topologies with pfc included....the cost of exotic ferrite that isn't a standard product makes the self oscillating confiuration esp 500W upwards more non viable......Here in the Alps I get billed for poor cos factor ...

The important thing about the magnetics choice is the Br/Bs ratio less than 80%.....with steep squared amorphous ferrites the loop core oscillation may not start as the flyback action on start up is too weak. However for high frequency appls.....square loop ferrites are ideal. As EVA points out......using a control ic solves this problem.

richj
 
Self-oscillating supply

Dem: Hi! I tend to agree with Rich on this one, at the power levels you require. And if you have to employ 'add-on' control electronics to get it to work properly, you might as well go for a more conventional topology suited to the 500W-1000W range: If fed from 300-400VDC, half bridge or full bridge is typical. If fed from battery supply for car stereo applications, push-pull is popular.

Whatever you choose: Design in some cycle-by cycle current limit mechanism, preferably with latching shutdown if the overload is sustained. When 1000W goes BANG :eek: right under your nose it's not nice.

Rich: I stand corrected, but is the design posted by Dem not what is called a Royer oscillator? I remember similar circuits being popular in 1970's capacitor discharge ignition system designs, like the Practical Electronics Scorpio. Also in inverters for fluorescent lights and other stuff 100-200W.

Cheers

John Hope
 

Dem

Member
2005-05-07 12:32 pm
Israel
Rich:
Billing for poor cos factor ... just nightmare... Did You buy PFC power supply Your PC? :)
Could You decribe more detailed about "squared" B-H ferrites in this application? I used simple ferrites (like 3C85, 3F3) for drive transformer without problem, I paid on this only by not very optimal design of output transformer (lower than need B - less than 0.15T) to be sure that drive tr-r will saturated before even during asymmetric start-up.

John Hope:
You are right - this is so-called Jensen variation of Royer oscillator, and it really great for up to 100-200W.
 
Royer Oscillator

DEM-

What you're looking for is the classic Royer-driven Oscillator. And, the good news here is........they can be found in every TL494-based computer PSU. By removing the PWM section and keeping the control-driver transformer, you will retain the basic oscillator.

Upon start-up, the Royer Oscillator gets things going, then as voltage builds up on the secondaries, and becomes available to power the '494, the '494 takes over for the proportional-drive operation.

There is a great article from QEX magazine that you can reference for this. It is about 3-4 pages long, and is quite thorough. It centers around a 300-400W power supply, for modding it from the original output voltages (+3.3, 5, 12, -5, and -12V) to +13.8V at something like 25-30A out for powering Ham radios and such.

Anyway, this is a good source of info for doing just what you're looking for.

Best of luck with this project. :cool:

Steve
 

Dem

Member
2005-05-07 12:32 pm
Israel
N-Channel: Thank You, I know about this feature of PC power supplies, I plan to use it as my developement platform. But I can't use it "as-is" because it has only proportional-drive feedback that unstable on light loads when magnetizing current is less than reflected output current. I'll add or voltage FB, or minimal load...

But before I will start, I want to collect all available info, and the article that You referenced on, can be very helpful for me. Can You post link to it? (Or to send by e-mail a soft copy, if You have it of course) - I don't know about QEX, and Google didn't help me with this. Do You know exact name of the article? Thank You.
 
Dem said:
Rich:
Billing for poor cos factor ... just nightmare... Did You buy PFC power supply Your PC? :)
Could You decribe more detailed about "squared" B-H ferrites in this application? I used simple ferrites (like 3C85, 3F3) for drive transformer without problem, I paid on this only by not very optimal design of output transformer (lower than need B - less than 0.15T) to be sure that drive tr-r will saturated before even during asymmetric start-up.

Hi there.....my pc runs off a few VA mains (don't bother)....my tube power amp is more greedy....250VA min quies (8x KT88's (pout=2x250W) but used pfc with ZVT as my mains fluctation is so wild.....For real simplicity at the time of design I had thought of PFC flyback for min 200 ACV operation but from previous experience with another commercial design, interference problems mainfested. Perhaps I should revisit.

A few years back Allied signal did some exotic amorphous core materials (metglas)....extremely low loss and mighty square...were perfect tempter for smps designs....The bog standard ferrites as you mentioned are more suitable...

Somehow I was thinking of the Royer design.....that was short circuit proof......however the smps mode isn't so forgiving. It's turning the page back to Baker clamps and when trannies were in smps before mos took over the show......

richj
 

Kenshin

Member
2005-01-09 9:51 am
China
I'm also planning to do a SMPS of 50~100W power level use some scheme like this / electronic ballast / halogen lamp electronic transformer. I'm keeping wondering if somebody has done it before. In 1990s, many people modify electronic ballast for SMPS use, but seldom any now. Also, commercial SMPS seldom use these "electronic ballast" schemes.

Some of these schemes can ZVS, and some are said to have a efficiency up to 95%.

The circuit I planning is more like the electronic ballast circuit -- positive feedback is taken from a current transformer in main circuit. The satuation magnetics device will be taken from a electronic ballast . (I have collected many dead CFL lamps.) And current limiting will be done by a inductor series with the transformer.

Electronic ballast circuit has series resonance, electronic transformer not. Whether to have a (diode clamped) series resonance or not, is undecided.
 

Dem

Member
2005-05-07 12:32 pm
Israel
Kenshin:
50~100W is not a problem for such SMPS, but typical electronic ballast / halogen electronic transformers use pure current-feedback with proportional-drive, and it has 2 disadvantages:
1. It stops to work at no-load or very light load;
2. It continues to work under shorted load.

It doesn't matter when load is more or less stable (electronic ballast / halogen ) but for audio (It's DIYaudio forum, isn't it? :) ) 50~100W I suggest You voltage-feedback, up to 200W it should be OK to my opinion.
 
Dem :

Proportional drive requires a proportional negative base current of about Ic/2 for optimum turn-off. This is achieved by inserting some components between the drive transformer and the base pin, and as a side effect, the volts*second product seen by the pulse transformer is no longer constant but proportional to Ic
 

Kenshin

Member
2005-01-09 9:51 am
China
Dem said:
Kenshin:
50~100W is not a problem for such SMPS, but typical electronic ballast / halogen electronic transformers use pure current-feedback with proportional-drive, and it has 2 disadvantages:
1. It stops to work at no-load or very light load;
2. It continues to work under shorted load.

It doesn't matter when load is more or less stable (electronic ballast / halogen ) but for audio (It's DIYaudio forum, isn't it? :) ) 50~100W I suggest You voltage-feedback, up to 200W it should be OK to my opinion.


Can it keep running if the transformer's primary turns have a relatively low inductance?
 

Dem

Member
2005-05-07 12:32 pm
Israel
Kenshin: Can it keep running if the transformer's primary turns have a relatively low inductance?

I don't have deep knowlege or experience with royer/Jensen oscillators , that's why I started this thread - to look for people who knows more...
But I know that the problem of current-feedback driving self-oscillator is when magnetizing current is greater then load-reflected current, in other words - at light load.
If You reduce the primary inductance, You increase magnetizing current, so as I understand it, it will even worse. In addition, lower primary inductance increases max flux density towards saturation point, but You want be sure that driver tr-r will saturated before the output tr-r, so I always kept high inductance and low B at primary.
But for Your power level, voltage FB is good enough to my opinion.
 

Dem

Member
2005-05-07 12:32 pm
Israel
Kenshin said:
Since voltage feedback is good, I want to build one, just as yours...
Another problem: how to simulate the satuable magnetics behavior?

I am working now on 200W prototype, I'll update You about the progress. My main goal is to make something as simple as possible, without hard to find parts, I'm attempting to use parts only from old PC power supplies, dimmers, halogen tr-rs etc.

About simulation - I do it with a piece of paper and a pencil :), sometimes Excell for recursive calculations. I don't use p-spice, because I'm sure that simple simulation can be done manually with the same accuracy...:) I built a small jig to measure saturation point of drive tr-rs.
I am building this not only for myself, so I'm thinking about building and testing methods, I'll update You too...
 

Kenshin

Member
2005-01-09 9:51 am
China
Dem said:
About simulation - I do it with a piece of paper and a pencil :), sometimes Excell for recursive calculations.

The main parameters could be calculated on a piece of paper.
But how to get the switching waveform and calculate the switching loss?

I have done some experiment on a 220V AC electronic ballast and got more than 400V peak output using 30V DC input. It's bigger than my former calculation. Maybe because the inductance of driving pulse transformer falls gradually during the satuation, so the effective satuation current is much bigger.

Apart from the difficulty of control, it's great! Low voltage debugging, very smooth switching waveform, and almost no heat from the transistor. (Though the inductor and capacitors are heated hard by the large resonance current.)

Maybe switching waveform measurements using low voltage supply and transformer output voltage / waveform measurements can replace waveform measurements powered directly by mains voltage. The latter one require either a mains isolation transformer or a isolation probe.

After get control of the resonance amplitude (with 30V DC to 220V AC input), I will connect a isolation transformer at the output and use it as a PSU.