That's still pretty vague, but I'll give it a shot.
Suppose you want to calculate y = log1030. That's equivalent to solving the equation 30 = 10y.
Start with an initial guess of y0 = 1.
Check it: 101 = 10 - too small. We want the answer to be 30.
Try a larger number. Let y1 = 2.
Check it: 102 = 100 - too big.
Note: Although I haven't found the number I'm looking for, my first two estimates tell me that log 30 is somewhere between 1 and 2.
As long as I have the true value bracketed between two numbers, one of which gives an answer that's too small and the other an answer that's too large, I'm in good shape. I can refine my estimate by choosing the number halfway between the two trial numbers.
Try y2 = 1.5.
Check: 101.5 = 31.6227766 (approx.) - too big, but getting closer.
Of my three trial values, 1, 1.5, and 2, 101 = 10, 101.5 = 31.6227766, and 102 = 100. For the next value I try, I want to find the number in whichever half-interval brackets the answer. Since 30 is between 10 and 31.31.6227766, but ISN"T between 31.6227766 and 100, I'll pick the midpoint of the interval [1, 1.5], namely 1.25.
Try y3 = 1.25
Check: 101.25 = 17.7827941 (approx.) - too small.
I continue with this process (which is called bisection) until two successive trial values give results that are "close enough." You can define how close you want to get - for example, if the two trial values produce results that are within .005 of each other (i.e., agree to two decimal places).