Inaccurate primary current sense in half-bridge smps.

[grizedale]

Hello,

I am doing a half-bridge smps and the primary current sensing is via a bridge rectified current sense transformer like on figure 8, page 18 of the following…

**broken link removed**

The problem of this current sensing method is that it is difficult to get good accuracy of the current sense because of the variability of the Vf of the diodes in the current sense rectifier

In my circuit, (I am actually using the MC34025 controller) it is voltage mode and the current sense is just to give a primary current limit.

-but I really do need this current limit to be accurate, because the FET has a distinct current limit…..and the output inductor can saturate.

The diodes that I’m using in the (full-bridge) current sense rectifier are the following:-

**broken link removed**
http://www.fairchildsemi.com/ds/BA/BAV70.pdf

It is seen how widely the Vf of eg the BAW56 diodes vary with temperature.

Are there any diodes that have a much tighter tolerance on Vf than those above?

-otherwise my current sense is just too inaccurate.

 

[FvM]

The problem of this current sensing method is that it is difficult to get good accuracy of the current sense because of the variability of the Vf of the diodes in the current sense rectifier.

Unikely with reasonable part choice. Presuming the current transformer isn't saturated at the actual output voltage and has sufficient main inductance, the diode voltage drop has no influence on the measurement. Simply consider primary current transformed by (n1/n2)², creating a proportional voltage drop at R1.

 

[mtwieg]

FvM is correct (surprise!). If designed properly, the current transformer acts as a current source, and thus the Vf of the diodes are meaningless. With the diodes, you'll want to make sure they have a fast recovery, but it seems like you've selected good diodes (so long as they can handle the current!).
The main source of error in a CT is from magnetizing inductance. A good model of a CT is a current source in parallel with magnetizing inductance (just the inductance of the CT by itself). The current source delivers the primary current reduced by the turns ratios. Ideally all this current flows into the burden resistor, but in reality some of it flows into the magnetizing inductance, inducing error in the measurement. As FvM implies, the magnetizing inductance can saturate, causing the errors to increase greatly. So for a decent measurement, make sure the magnetizing inductance has a large impedance relative to the burden resistance, and that it can never saturate under any circumstance. Preventing saturation means making sure no DC flows in the core, and that Br resets itself to nearly zero every cycle.

 

[grizedale]

Hi
mtwieg, so do you mean that at the switching frequency, the impedance of the primary of the current sense transformer should be about ~20 times more than the resistance of the sense resistor at the secondary side?

By "burden resistance" i presume you meant the sense resistor at the seconday of the current sense transformer.?

 

[FvM]

If you refer to burden (secondary shunt) impedance, than secondary transformer impedance is the related quantity of interest. But the diode rectifier voltag drop has also to be taken in consideration. As mtwieg suggested, the ratio of (secondary) magnetizing current to secondary output current makes the current measurement error.

 

[mtwieg]

Also it occurs to me that it's important for the Vf of the rectifier diodes to be balanced, so that CT secondary doesn't see a DC bias voltage. So use a monolithic bridge rectifier, if possible.

 

[Orson Cart]

As long as the CT sees the same volt-seconds on each opposing half cycle of operation - the Vf of the diodes does not matter, if the CT is not properly reset every switching cycle (or very nearly not) - this will be the only problem. You can calculate the max flux swing in the CT core from the volt seconds applied and the core area and the number of turns (2.Bpk = V.sec/(N.Ae)) if the Bpk is too high, say > 50mT then you may encounter innaccuracies and core heating problems...
Regards, Orson Cart.

 

[grizedale]

ok, though i thought B could go up to 300mT in ferrite?

 

[Orson Cart]

It can - but doesn't make for the most accurate CT.