Maker Pro
Maker Pro

Problems of transformer, snubber and freewheeling design of push pull DC-DC converter using SG3525

Hi everyone, as part of inverter project I need to build a step up
converter that converts 12VDC to 360VDC using push pull topology.
Switching frequency is at 55kHz and using SG3525 as PWM chip. I have
a few question:

1. Datasheet of SG3525 from manufacture is too brief. It does not
explain clearly how exactly to size Rd for dead time nor how to
calculate value of slow start capacitor. Does anyone know where I
might find a reference design or app note of at least a buck converter
using this chip, as I could find none from ST website?

2. What is the exact use of snubber circuit with MOSFET? Is it to
absorb voltage spike due to leakage or magnetising inductance of
primary winding when MOSFET is turning off? Or to damp ringing at
rising edge due to to fast dV/dt? Where should it be placed? As in one
example http://sound.westhost.com/project89.htm the snubber is placed
across primary winding, but I also saw some circuits where it is put
across D and S of MOSFET.

3. MOSFET use gate resistor to slow down turn on transition which
could otherwise cause ringing. Isn't that overlap the purpose of
snubber? Sorry I've read too much app notes I messed up the concept.

4. When switching high frequency transformer, the ideal transformer
output waveform should look like:

------- --------
| | |
| | |
-- --- ---
| |
| |
-------
it's necessary to discharge energy stored by leakage inductance at
crossover. Does snubber do this as well? How about the energy stored
by magnetizing inductance and magnetizing current? does it also need
to be discharged? And if so, should I use a freewheeling diode (fast
recovery type) across primary winding? My first thought was to connect
two diode this way http://www.geocities.com/w2kwong/Untitled.gif which
proved failed in simulation as D1 will short the upper winding when
voltage is induced by lower winding as lower MOSFET is on. So instead
I added two zener diodes http://www.geocities.com/w2kwong/Untitled2.gif
to form zener clamp. Is it appropriate? How about using RC Clamp ( as
shown in http://www.geocities.com/w2kwong/Untitled3.gif )?

5. When winding transformer, should the direction be "back and fro"
as in http://www.geocities.com/w2kwong/Untitled4.gif, or always follow
one direction http://www.geocities.com/w2kwong/Untitled5.gif?

Thankyou for your kind helping ; )
 
S

Salmon Egg

Hi everyone, as part of inverter project I need to build a step up
Parts of speech called articles were invented to aid written and spoken
communication. USE THEM!

Bill
-- Fermez le Bush--about two years to go.
 
S

Salmon Egg

I'm not native speaker of English and sorry for my poor expression

I apologize. I am sure your English is better than my capability in your
language whatever it is.

Bill
-- Fermez le Bush--about two years to go.
 
L

legg

Hi everyone, as part of inverter project I need to build a step up
converter that converts 12VDC to 360VDC using push pull topology.
Switching frequency is at 55kHz and using SG3525 as PWM chip. I have
a few question:

1. Datasheet of SG3525 from manufacture is too brief. It does not
explain clearly how exactly to size Rd for dead time nor how to
calculate value of slow start capacitor. Does anyone know where I
might find a reference design or app note of at least a buck converter
using this chip, as I could find none from ST website?

In figure 3 the deadtime is illustrated as being dependant on the
timing capacitor size, modified slightly by the optional resistor in
series with pin7.

Softstart: using the 50uA present on pin8. the rise in softstart
capacitor voltage will be linear - i = C dv/dt. How this voltage rise
will affect your circuit, you'll have to determine yourself.

SG3525 is originally a Silicon General device - now Microsemi
http://www.microsemi.com/datasheets/sg1525a.pdf

Application notes for jellybean two-phase pwm drivers may all be
applied to a degree. There's not much special about the IC itself.
2. What is the exact use of snubber circuit with MOSFET? Is it to
absorb voltage spike due to leakage or magnetising inductance of
primary winding when MOSFET is turning off? Or to damp ringing at
rising edge due to to fast dV/dt? Where should it be placed? As in one
example http://sound.westhost.com/project89.htm the snubber is placed
across primary winding, but I also saw some circuits where it is put
across D and S of MOSFET.

In your push-pull topology, the snubber should act somehow to limit
voltage peaks and reduce radiation when the mosfets turn off. A
current snubber will limit dv/dt. At the low voltage you're operating
at (12V), a voltage clamp might not be needed to protect parts.

You do not mention a power level.
3. MOSFET use gate resistor to slow down turn on transition which
could otherwise cause ringing. Isn't that overlap the purpose of
snubber? Sorry I've read too much app notes I messed up the concept.

4. When switching high frequency transformer, the ideal transformer
output waveform should look like:

------- --------
| | |
| | |
-- --- ---
| |
| |
-------
it's necessary to discharge energy stored by leakage inductance at
crossover. Does snubber do this as well? How about the energy stored
by magnetizing inductance and magnetizing current? does it also need
to be discharged? And if so, should I use a freewheeling diode (fast
recovery type) across primary winding? My first thought was to connect
two diode this way http://www.geocities.com/w2kwong/Untitled.gif which
proved failed in simulation as D1 will short the upper winding when
voltage is induced by lower winding as lower MOSFET is on. So instead
I added two zener diodes http://www.geocities.com/w2kwong/Untitled2.gif
to form zener clamp. Is it appropriate? How about using RC Clamp ( as
shown in http://www.geocities.com/w2kwong/Untitled3.gif )?

How you clamp is your choice. Ask yourself what's going to get hot.

If a lot of energy is involved, you might consider trying to recover
it, rather than burning it off.

Energy stored at any time is ( Li^2)/2
Leakage energy is stored in the leakage inductance. See references
below.

5. When winding transformer, should the direction be "back and fro"
as in http://www.geocities.com/w2kwong/Untitled4.gif, or always follow
one direction http://www.geocities.com/w2kwong/Untitled5.gif?

Thankyou for your kind helping ; )

Check out the old unitrode seminar at TI
http://www-s.ti.com/sc/techlit/slup123.pdf
http://www-s.ti.com/sc/techlit/slup126.pdf

2001 magnetics design handbook chapters are downloadable

http://focus.ti.com/docs/training/catalog/events/event.jhtml?sku=SEM401014

RL
 
Hello legg, thankyou for your reply which has been useful. But there
are still some bits which I'm not quite sure:

1. You said deadtime is illustrated in figure 3 on datasheet, so is
deadtime = oscillator discharge time? And which application note on
jellybean two-phase pwm driver are you referring to, since I couldn't
locate on microsemi website?

2. I ran a simulation on push pull stage of the circuit
http://sound.westhost.com/project89.htm, and found that as soon as M1
and M2 has just turned off, the energy stored by MAGNETIZING flux from
Lm (not leakage inductance) will get discharged by lower section of
primary winding. Demagnetizing current will flow upwards back to 12V
source through body diode of M3 and M4, as shown in this picture
http://www.geocities.com/w2kwong/Untitled10.gif . I've heard that body
diode of MOSFET is a by product during fabrication, and is of slow
recovery time. So should I place a fast recovery diode or schottky
across each MOSFETs to facilitate the flow of this current?

P.S. I'm working on a 300W DC-DC converter which steps up 12V to
360VDC. The transformer core is ETD49, Np=6T+6T, Ns=198T. Calculated
magnetizing inductance =200uH. Total primary current at rated load =
25A.

thanks again for helping ^_^
 
L

legg

Hello legg, thankyou for your reply which has been useful. But there
are still some bits which I'm not quite sure:

1. You said deadtime is illustrated in figure 3 on datasheet, so is
deadtime = oscillator discharge time?
Mostly.

And which application note on
jellybean two-phase pwm driver are you referring to, since I couldn't
locate on microsemi website?

Look at SG3524, UC3525

Intersil an6915
http://www.nalanda.nitc.ac.in/industry/appnotes/Intersil/an6915.pdf
Nat Semi AN-292
http://www.national.com/an/AN/AN-292.pdf
Philips AN126
http://noel.feld.cvut.cz/hw/philips/acrobat/8140.pdf

The links are straight Google hits.

You will be looking for power stage info, not info on the driver or
it's interface.
2. I ran a simulation on push pull stage of the circuit
http://sound.westhost.com/project89.htm, and found that as soon as M1
and M2 has just turned off, the energy stored by MAGNETIZING flux from
Lm (not leakage inductance) will get discharged by lower section of
primary winding.

Your reference is not to a conventional push-pull converter, it is to
a DC-DC transformer, with no energy storage element in the output
rectifier filter. This is basically capactor-to-capacitor energy
transfer, with no regulation ability.

The control chip is completely wasted in this application.

As to your model, I have no comments, as there are models and then
there are models.
Demagnetizing current will flow upwards back to 12V
source through body diode of M3 and M4, as shown in this picture
http://www.geocities.com/w2kwong/Untitled10.gif . I've heard that body
diode of MOSFET is a by product during fabrication, and is of slow
recovery time. So should I place a fast recovery diode or schottky
across each MOSFETs to facilitate the flow of this current?

For example, current will not appear 'across' the mosfet switches - it
will flow 'through' them. If your GUI makes this kind of an error, it
can hardly be expected to keep track of it's own bootlaces.

Remember to apply a load to your model, as unloaded operation will
hide the relative importance of the currents involved.

With both switches off, the magnetization current will have the
opportunity to ramp to zero (in a perfectly coupled transformer),
through which-ever path presnts itself. As your model image does not
illustrate drain voltages, you can have no idea where the magnetizing
current is flowing, or why.

Once you apply a load and view the resulting waveforms, you may
understand why capacitive-capacitive energy transfer has it's
problems, without the presence of effective intermediate inductive
storage. Perhaps leakage terms will suffice. Don't forget them in your
model.

RL
 
Top