Series resonant SMPS question

Hi:

I am working on a Half bridge SMPS as a power supply for 2 B&O Icepower 500a audio amplifier modules, but it could be useful for another similar amps. So I use a pair of IRFP450B MOSFETS driven by an IR2156 ballast control & driver IC. Fixed oscillating frec at 126kHz with an 80% of duty cycle(20% dead time) no PWM.
The output transformer is build using an E42 Epcos core and it is connected from the common point of the two mosfets to a common point between + and - legs through a pair of capacitors calculated to obtain a resonant frec. (1/2*PI*(Leakage L*C)^0.5)
I tested the design and the prototype works fine(2 amps plugged simultaneusly) and it looks very stable.The transformer waveforms are clean and I do not have thermal problems by now. My question is the way to check if the real resonant frecuency is close to my calculations.(I measured the transformer primary Leakage L needed for my formula conecting a LC Meter on the primary with the secondaries shorted).
I,m thinking about to put a sense resistor with a very small value (maybe 0.1 or 0.2 ohms-10 watts) in series with the transformer primary to "see"" the primary current on my scope comparing to the drain-source voltage on the other probe of the scope to check if it is in phase and to check how it looks.

Someone has another method to check this?
 
I can calculate the winding leakage pretty accurately if I see the winding sheet.....Also, when using the short circuit method of measuring leakage, you should limit the primary test current to the same as it would be durring normal operation....

Also..If you post the schematic I can better have idea of how to help you measure...I am familiar with the IR parts you mention, since I have used them before....especially the ballast controller...Have you considered the IR 1856 for a driver????

Chris
 
Hi,

safer method of current measurement would be current transformer. Just wind around 100 turns on 10mm OD or larger ferrite toroid. Preferably use single layer winding, so that is usually 0.15 or 0.2 mm magnet wire. Terminate the winding with low ohm (several Ohms) noninductive resistor. Your current probe is ready. You just stick toroid on wire where you want to measure current (high frequency AC only of course).

Best regards,

Jaka Racman
 
Hi:

I have some current sense samples from Coilcraft so I can put one of these in series with the output trafo for current measurement. Anyway, I would like to know a better way to measure or calculate the Leakage inductance needed.
I attach the basic schematic and winding sheet.

Regards.
 

Attachments

  • smps1.zip
    56.2 KB · Views: 613
I think that you have made several mistakes.

You should connect one end of the primary, through a several-microfarads DC filtering capacitor, to a low-impedance node centered half way between Vcc and ground, for example the junction between R2 and R3.

The other end, the one that is being switched, is the one that may be allowed to resonate by placing suitable capacitors. But it's hardly required because the voltage will always inmediately spring towards the opposite rail after turn-off. So you will get zero-voltage zero-current turn-on by just operating the converter near 100% duty cycle.

Also, your output diodes are improperly connected. A2 of D8 should be connected to S2- of TR3, and K2 of D3 should be connected to S1+ of TR3. This is requied in order to guarantee symmetric operation, otherwise the capacitively coupled end of the transformer may slew towards one of the rails and +-50V outputs would also lose balance.

Concerning current measurement, the most practical way to do it on floating nodes is by means of current transformers, otherwise you would need very expensive differential oscilloscope probes. However, current transformers have its limitations, they must be either excited in a fully symmetric fashon or allowed to reset its flux, otherwise they will saturate and stop providing a precise measurement.
 
Eva said:

Concerning current measurement, the most practical way to do it on floating nodes is by means of current transformers, otherwise you would need very expensive differential oscilloscope probes. However, current transformers have its limitations, they must be either excited in a fully symmetric fashon or allowed to reset its flux, otherwise they will saturate and stop providing a precise measurement.

Would a Hall effect based current sensor be suitable? E.g. http://www.allegromicro.com/sf/0704/

Regards, JZ
 

mzzj

Member
2005-07-02 8:17 pm
65N 25E
Eva said:

The other end, the one that is being switched, is the one that may be allowed to resonate by placing suitable capacitors. But it's hardly required because the voltage will always inmediately spring towards the opposite rail after turn-off. So you will get zero-voltage zero-current turn-on by just operating the converter near 100% duty cycle.


I think in many cases some extra capasitance is beneficial, without extra caps you rely entirely on mosfet internal output capacitance, it might be too much variable or not enough for high leakage/series inductance. (even more since this is relatively low freq case)
Running very close to 50% pulse width with high dV/dt requires beefy gate drivers if mosfet dissipation is to be kept reasonable and is not so good from emc-point. So you lose all the benefits of ZVS/ZCS. I guess something like 40-45% duty cycle migh be good compromise(depends on freq), peak currents are still reasonable and resonant switching has plenty of time to happen.
 

AP2

Banned
2009-12-30 7:52 am
Italy
You can see the behavior of signals on the primary probe on the secondary connecting the AC-GND on one side. with no load at output.
DT and some defects are clearly visible.
schema to make some changes. (my opinion)
C20 10uF add electr.
C22 10uF add electr.
eliminated C25-26 and put in series at P1 + 330nF.
p2-connect at center snap-in (R2-R5)
if you can turn on primary side winding 1.5 + one diode, it solve +14 V at TP3.
try to see these changes (differences) in SEC-AC probe when the output load.