I have an isolated DC-DC converter application that is well suited to the forward converter topology. However, instead of the traditional voltage source design where the output voltage is regulated with an overarching maximum current limit, I'd like to be able to regulate the output current with an overarching maximum voltage limit. Despite a great deal of searching and reading, I'm struggling to find any app note, eval board or reference design that describes such an implementation.

Forward converters are typically current mode, but the feedback into the current mode comparator still results in voltage control. Forward converters also typically have a maximum current limit by way of a primary current sense. This picture is a nice simple representation of the major elements. The output voltage, appropriately scaled, provides the error signal to the current mode PWM. The primary current sense typically provides an upper threshold.



After much furrowing of brow, I see two possible solutions:

1. Dial the primary current sense input up so the converter mostly runs in max current limit mode. Therefore the max current limit becomes the operating setpoint, while the voltage feedback only provides overvoltage protection.
2. Change the voltage feedback to a load current feedback with a sense resistor or current transformer. That way the converter will work to minimise the current error, with a now redundant overarching max current limit.

Either method has its pros and cons, but it makes me nervous that I can't find evidence that either has been documented. What do all those current controlled battery chargers and high power LED drivers and bench DC power supplies with a current control knob do?

Is anyone aware of a forward converter design with a current setpoint instead of a voltage setpoint? Failing that, is there another suitable topology worth considered for the 50-150W range?

 

Of your two ideas I like #2. Most likely you will get better regulation using the voltage feedback input since that is the regulation channel in which performance is more critical in typical applications of the device.

 

Thanks DickCappels. I agree it ought to give the finest accuracy. Though accuracy is not so important as stability in our application as we have a slow outside loop that can perform correction. I think we'll end up trying both and measuring the results!