This is a very tricky thing to put a definite figure on, not least because one farming household would rarely produce all the food it would require on a regular basis: cereals, vegetables, fruit, fish, meat of various types, eggs, milk, cheese etc, so by definition we already have to assume some level of surplus production to facilitate trade.

An oft-repeated statistic is that one "hide" of land in England was enough land to provide for a household for a year, but that's a statement that bears absolutely no relation to the figures that we see in Domesday Book (assuming, of course, that the "hide" is essentially analogous to the "ploughland" that replaces it as a unit of assessment, as jt indeed appears to be. The most numerous class of peasant farmer in Domesday Book is the villein, who typically held a virgate of land, a quarter of a ploughland and thus very roughly some 30 acres. Villeins, however, were a relatively comfortable class, and the second most numerous class was a motley association of "smallholders", bordars and "cottagers". These households held anywhere between 5 acres and 15. Given that a significant chunk of the population wasn't starving to death on an annual basis but that famine was a genuine risk in years with a poor harvest, we have to assume that, in good years at least, it was theoretically possible to support a household on as few as 5 acres, even if it didn't necessarily present a particularly full or varied diet.

Our other problem comes from the fact that documents such as Domesday and other medieval land assessments were predominantly financial documents, and only had a cursory interest in demographic information. As a result, we have the number of households in a settlement, but there's a significant debate as to how many people were actually in the average household. Estimates range as low as 3 and as high as 8, but a common consensus is that a household typically represented around 4-5 individuals.