Half bridge smps voltage drop problem

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,
I'd like to share some of my experiments about smps.

Here is my configuration:

half bridge topolgy
with 1uF 400V in series with primary winding
other point connected to the middle potential created by electrolytics capacitors (2*660uF 200VDC).

SG3525 + IR2113 + 2*IRF740
frequency : 62500Hz
Mains are 230VAC rectified is 320VDC


Core is 3C85 36mm toroid (philips) section is about 1cm2

primary winding : 32 turns (Bmax is 250mT)

secondary winding : 2*8 turns center tapped (secondary ground isolated from primary)

secondary rectifier is 4 TO220 fast diodes (10A 400V)
filtering is 2*1000uF 63V

with no load: secondary voltage is 2*41VDC
With only one secondary loaded by 4.7 oms : 32V (207W)

MY QUESTION IS:
is it a "normal" voltage drop ?
is my core saturating ?
Do i have to take this voltage drop into account and make a voltage regulation (opto...)?

Thanks for your help !
 
Looking at your turns ratio, it looks as if you are running the SMPS UN-regulated now at 50% fixed duty cycle. If you want to fix the output voltage as load varies, you will need to have opto-feedback and design the transformer turns ratio differently, otherwise your supply will sag just like a linear supply.

How much power are you trying to get? With 660uF primary capacitance, I'm not sure you will get much more power. Check to see how much voltage ripple you have on the primary caps. The less capacitance on front end, the more 100Hz ripple you will have passing to the output.

You need to check the primary current waveform to see if the core is saturating. Use a current probe if you have one or measure voltage across a small value, high wattage series resistor (0.1 or 1 ohm 5 or 10 watt) inserted into the primary somewhere. The current should be a nice ramp, like a trapezoid. If the ramp suddenly curves up with a large di/dt at the end of the ramp, you have saturation.

One more thing- a toroid in not the best type of core to use for an offline converter because it becomes trouble when you need to have the required creepage and isolation required from primary to secondary windings. To be completely safe, either triple insulated wire needs to be used on either primary or secondary, or several tape layers need to be used to completely isolate the pri. and sec. I'm not completly sure as I have not personally used a toroid for offline SMPS.

Hope this helps.
 
Well, this voltage drop is quite excessive for such a converter while outputting less than 220 watts (32^2/4.7). On the other hand, the voltage is being dropped somewhere in your circuit, so scientifically speaking, the best thing that you can do now is to take an oscilloscope and find out where. Check also primary current waveforms as it has been suggested, and try to post pictures of the results so that we can analyse them.
 
First of all, thanks for your fast reply.

As soon as possible i'll post pictures of the signals (primary current, voltage, secondary voltage...)

may the problem come from the secondary capacitors which are not low esr ones ?
Voltage ripple may occurs there.
I'm gonna check this point first.
 
Secondary Loading

Hi Alex,

You and I must've been dreaming of the exact same layout, parts list, topology, and component choice! :D

I did a half-bridge just like this about 10 years ago. Unfortunately, I cannibalized the parts from it to make other projects. I did use opto feedback and got good regulation. Output was +/- 30V at 4A (240W) and my switching freq was ~35kHz.

I used IRF740s, SG3525, MPIC2113 (Motorola's version of the IR2113), and MOC8102 Optocoupler. The hi-voltage caps I used were Panasonic TSH-series 680mF (105C and low ESR), and the core I used was an Amidon FT-140-77, (a 1.40" toroid made of a very similar material to the 3C85), and MBR10100s for the rectifiers. I chose Schottkies for their lower forward voltage drop. The filter caps on the secondary side were Panasonic HFU 1000mF/50V. I used three layers of Scotch #93 Teflon Tape to insulate the primary and secondaries. It powered 2 50W Amps into 8W.

No noticeable noise at the outputs, and no magic smoke got out, either. In retrospect, I really wish I hadn't taken it part! :bawling:

I have a couple of questions:

1) Why load only one secondary? Does anything change when you load both secondaries?

2) Is 62.5kHz your oscillator freq. or your switching freq? If it is the oscillator freq, then 32T on the primary is about right for a 31.25kHz switching frequency. But for 62.5kHz, this would be a bit much, and this could be one cause of core saturation. Amen to checking the primary current waveforms- this will tell alot.

Eagerly awaiting waveform pics. :)


Steve
 
Re: Secondary Loading

N-Channel said:
Hi Alex,

2) Is 62.5kHz your oscillator freq. or your switching freq? If it is the oscillator freq, then 32T on the primary is about right for a 31.25kHz switching frequency. But for 62.5kHz, this would be a bit much, and this could be one cause of core saturation. Amen to checking the primary current waveforms- this will tell alot.

Steve
Umm, did you swap your numbers by accident or..?
 
Hi N-channel,thanks for the interest you're bringing to my project !

1) I've loaded only one secondary because with one 4.7 ohms resitor on each secondary the core produces a HF noise !
kind of overload or great saturation ?

2) 62.5kHz is the transformer operating frequency.
(SG3525 : RT=10k;CT=1nF;Rdisch=100;Css=1uF).
Do you think it is a bit much ?
Around 30kHz should be better with 32 turns on the primary side ?

Could you tell me a little more about your regulation:

1) How many extra primary turns ?
2) Did you implement output inductors ?
3) SG3525 connection and compensation (pin 9) ?
4) opto-coupler wiring (with zener maybe) ?

I post pics as soon as possible.

Thanks for your reply.
 
if i make some maths :

URMS=4.N.f.S.Bmax

where:
4 : for square wave
URMS=160V
N=32
f=62500 Hz
S=1cm2 (cross section area)
that gives Bmax=200mT (2000Gauss)

Do you really think it is too much for a 3C85 core ?

with f=35000 Hz
it gives Bmax=357mT (3570Gauss)

so increassing frequency lowers Bmax

isnt it ?
 
alexclaire said:


2) 62.5kHz is the transformer operating frequency.
(SG3525 : RT=10k;CT=1nF;Rdisch=100;Css=1uF).
Do you think it is a bit much ?
Around 30kHz should be better with 32 turns on the primary side ?

Better not use 30khz with 32 turns primary, I am affraid that core loss is starting to get too high. 250mW/cm^3 is bit high

Also, problem must be something else than core saturation, half-bridge core doesnt saturate with increased loading.
 
alexclaire said:
if i make some maths :

URMS=4.N.f.S.Bmax

where:
4 : for square wave
URMS=160V
N=32
f=62500 Hz
S=1cm2 (cross section area)
that gives Bmax=200mT (2000Gauss)

Do you really think it is too much for a 3C85 core ?

with f=35000 Hz
it gives Bmax=357mT (3570Gauss)

so increassing frequency lowers Bmax

isnt it ?
Thats correct. I wouldnt recommend 32 turns primary with 35khz. Something like 50 turns at 35khz sounds better.
 
Mzzj,

50 turns with 35kHz gives Bmax=230mT (2300Gauss)

Bmax previous value was 200mT (62.5kHz 32turns)
so what's the difference ?

So you think it's not the reason for the secondary voltage drop ?

My problem may come from primary voltage ripple:
660uF*2 maybe isn't enough for 2A (primary current)
I'd like to get 300W.

Thanks
 
alexclaire said:
Mzzj,

50 turns with 35kHz gives Bmax=230mT (2300Gauss)

Bmax previous value was 200mT (62.5kHz 32turns)
so what's the difference ?

So you think it's not the reason for the secondary voltage drop ?

My problem may come from primary voltage ripple:
660uF*2 maybe isn't enough for 2A (primary current)
I'd like to get 300W.

Thanks
Saturation limit is something like 300-400mT

Limiting factor is core loss, ie how much core is going to heat.

Yeah, reason for secondary voltage drop is definetely somewhere else.

How much ripple you have on primary caps? Dont tell us that you dont have a scope and safety isolation transformer? :devilr: :whazzat:
 
You can calculate expected voltage ripple quite easily:

For any capacitor, dV = (average)current * time / capacitance.

Take time=7ms (approx. dead time of a rectified mains sinewave), current=half the primary current (it's only drawn from each capacitor bank during half of the time, so average value is halved), and capacitance=the value for a single bank.

For 660uF, 32:8 turns and 32V output into 4.7 ohms that gives approx 9V p-p (in each capacitor bank). Measured value should be quite close.
 
Back again,

I have an oscilloscope but my digital camera is out of order so i have to find one to share screenshots.

I was thinking about one thing:

my power supply (12V) onboard has a current capability of 200mA, maybe it's not enough to feed both SG3525 and IR2213 driving 2*IRF740.

What do you think about this ?

It could explain the voltage drop when the transformer is loaded.

I can't make any measure right now,I've blown the power stage (IR2113 and IRF740)...
 
alexclaire said:
Back again,

I have an oscilloscope but my digital camera is out of order so i have to find one to share screenshots.

I was thinking about one thing:

my power supply (12V) onboard has a current capability of 200mA, maybe it's not enough to feed both SG3525 and IR2213 driving 2*IRF740.

What do you think about this ?
(IR2113 and IRF740)...
200mA should be enough if guess IRF740 specs right. For big mosfets like IRF450 200mA would be possible if switching way over 100khz.
Or calculate it from gate capacitance and freq. P=f*C*U^2
 
Back Again

Hi Alex et al.

First off, mzzj: I don't t hink I swapped my #s, but I may have worded my question to Alex a little poorly. Nothing gets past this crowd! :D I referred to F(osc) and F(sw) because, as we all know, for dual-channel PWM chips, F(osc) =2F(sw). That's where the confusion may have come in. :xeye:

Alex,

My primary was 38T of #22 (tri-filar), V(pri) was 160V (Voltage doubler for 120VAC on our side of the Pond). The primary turns equation I used came from Chryssis' book, chapter 4 "The High Frequency Power Transformer". I will have to dig out the equation, but after a fashion, I had reduced the transformer design process down to 8-12 very concrete cut-n-dried steps. The variation in the number of steps was determined by the complexity of the transformer (# of secondaries, etc.). When i find the equation, I will also type out my design process. For the timing and compensation components:
R(t) = 2.21KW
C(t) = 10nF
R(d) = 0W,

yielding an oscillator frequency of = 64.6kHz, and a switching frequency of 32.32kHz.

Compensation at pin 9 is 49.9kW paralleled with 1000pF, going to ground. Opto current was set a 1mA over 60V (so 60kW. No Zener, but in a previous version, I did use a TL431 for a reference, compensated at the TL431. Have to go for now (child is stirring from nap), but I will be back soon.

Steve
 
Start-up

Alex,

How do you derive the 12V for the PWM & Driver chips? I gather it is an auxiliary winding on the main core. Do you use a Zener-pass regulator for the start-up ckt?

This is how I did mine. TIP50 (400Vce) with a 16kW power resistor in series from the high-voltage bus (~320-330VDC) to the collector of the TIP. 1N5242 (12V 500mW) and 440kW provide the reference for the transistor. My auxiliary winding put out ~16V, so when it started producing voltage, this automatically shut down the start-up ckt. The SG3525 needs a start-up ckt capable of providing atleast 50mA, beacuse it's start-up current is the same as it's operating current, ~20mA. As mzzj stated, 200mA should be enough to power the '3525 and the '2113.

I have decided to rebuild this power supply, as I still have most of the parts lying around. I think I will replace the IRF740s (400Vdss) with IRF840s (500Vdss), because I might want to PFC the front end. Everything else, except the transformer's primary turns, I think I will keep the same.

Oh yeah, one other thing- I found that stacking two toroid cores helped me in the power capacity department. I've done this using the same FT-140-77 cores for some 12V push-pull converters I did in the past. Haven't tried it for off-line yet. This might be worth looking into. Sure, it will change some of your numbers around, but you will have more headroom and avoid core saturation.

still looking for those equations.....

Ciao for now,

Steve
 
Magic Smoke

Eva,

If it's because of what I suspect, probably his '2113 caused one or both of the '740s to blow. Perhaps the floating drive for the upper MOSFET crapped out. This happened to me. After I took apart my 3525-2113 based PSU, I tried to re-do it on a pc board. Powered it up, and the 2113 and one of the IRFP350s (TO-247 version of the '740) promptly let the Magic Smoke out. :hot: :dead: This led to the failure of the lower MOSFET about a second and a half later. :bawling: I did not have a can of magic smoke to put back in.....

I believe in one of the other threads, eva, you stated your reluctance to use these floating gate-driver ICs because they tend to fail more often than not. I bleieve you said that driver transformers were still better from a reliability standpoint. Please correct me if I'm wrong. :whazzat:
 
I have a few IR2112 in an antistatic bag somewhere, they look fancy but I have never used them. I feel more attracted towards transformers as they allow to provide both power and signal to your own custom gate drive circuit, that may be as simple as a PNP transistor to speed up turn-off and a couple of resistors and diodes, while keeping galvanic isolation and allowing for secondary-side control. Also, transformers doesn't blow everytime a power device dies, and they protect the control circuit.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.