I thought that was standard practice? It's certainly how I was taught..
Same here. Carb heat on as part of the downwind checks, off on short final.
As the carb heat begins to have an effect, consider flying at a reduced power setting (and leaning more if need be), as less volume of air going through the carb at the lower power setting needs less total heat to warm it and melt ice.
I wonder if this is actually true. A reduced power setting also means less power from the engine (duh) and thus less warmth produced. Which in turn reduces the amount of warm air available for carb heat - assuming carb heat is generated in the traditional way, with a cuff around the exhaust.
But there might be another reason not to do this. I'm not an expert, but here's my reasoning. Carb ice is formed due to a temperature drop in the carburetor. This temperature drop can be caused by two things: The pressure drop across the butterfly valve that regulates the MAP, and the evaporation effect of the fuel in the venturi.
The evaporation of fuel, and the resulting drop of temperature, scales more or less evenly with the MAP and thus the power applied. A higher MAP leads to more fuel evaporation, leads to more power from the engine, leads to more heat produced, makes more heat available for carb heat. So assuming you have carb heat applied, it would not matter all that much how much power you apply as these effects cancel each other out.
But the ice formed by the butterfly valve has a completely different, and in fact negative relationship. If you reduce power you do so by closing the butterfly valve. This leads to a lower MAP and thus a higher pressure drop across the valve. This may produce more carb ice, right at the time when the engine is producing less power and thus less heat.
Because of this last effect my gut feeling would be to open the throttle completely in case of carb ice, instead of restricting it. And in fact this seems to be confirmed in POHs and such, which call for full throttle in case of persistent carb ice.
Anyone can confirm I'm right?