Stratified random sampling might be better for survey studies. For this purpose, we use a "sampling frame" (includes a list of all members of the study population). We divide the population into categories with respect to the characteristics that their dispersion is important (such as gender, socio-economic level, etc). Then, according to the ratio of the total number of each stratum (e.g. when the population consists of 75% females and 25% males, your sample should conform this ratio) random sampling (simple or systematic) will be conducted.
Suppose you want to sample 400 out of 4000 people, everyone has a file number in excel and in another column their gender is defined as 1 (male) or female (2). If the population consists of 75% females and 25% males, then you have to sort or separate the dataset with respect to gender in order to select 100 men and 300 women in your final sample. For each one of these strata (male or female), conduct the systematic random sampling. The sampling fraction (k = 4000/400 =10) is determined. Then, randomly select the first unit (r) by random number table. Suppose that r is determined as 3, the individuals number 3, 3 + (1 × 10), and 3 + (n × 10) will be selected.
For calculation of sample size (400 in the mentioned example), you have to determine the design of the survey (longitudinal, etc), the research hypotheses, and the tested variables. All statistical packages for sample size use a similar approach. Try to determine the effect size of your research hypotheses (mean differences, correlation coefficients, odds ratios or proportions, etc) based on the previous studies or your own pilots. Next, variance or other dispersion measures are often asked. Type 1 and 2 errors (alpha, power=1-beta) are also needed to be determined.