"All", "always", etc. are dangerous words.
Most epidemiology studies are observational - as a field epidemiology tends to concern itself with study questions that are not amenable to randomization and controlled trials. The dominant form of studies one would encounter while doing graduate work in epidemiology or working in a public health department, or reading epidemiology journals will be observational in nature - cross-sectional prevalence studies, cohorts, case-control studies, etc.
There are however caveats to this:
"all observational and not controlled": One of these things is not like the other. A study can be observational and controlled - see caseif we just mean "people who didn't have the exposure/treatment/whatever" than the bulk of most cohort studies are controls. Case-control studies have "controls", but this means non-cases. In a study I've run, "control" meant "This hospital didn't have a particular policy in place". There isn't the deliberate assigning of people as controls as there is in a randomized trial, but depending on what one intends to convey by "control" it's possible for an observational study to have them. Equally, you can have a non-observational experimental study with no controls. That's probably not a good idea, but its certainly possible.
There are studies in epidemiology that are arguably not observational. The most obvious example of this would be mathematical modeling studies, which aren't really observational or randomized, but are definitively part of epidemiology. They also happen to be what I do.