There's some useful information about this in the manual, I think at the front end under the "Algorithmic Trading Theory" section. But the gist of it is that the Zorro team found that H4 and above are useful time frames for trading. H1 is harder, but possible. Below that, things get more difficult.
At lower time frames, trading costs tend to eat up more profit, since the amplitude of price swings is roughly proportional to the period of the cycle (I think I read that in one of Ehlers' books, and you can see that it approximately holds true just by looking at price charts). Trading costs however are constant. Thus they make a bigger impact on a strategy that trades a small time frame.
Secondly, the lower time frames are generally more noisy. That is, the impact of random fluctuations is more significant. Thus it is more challenging to extract predictive information. Another way of saying the same thing is that the signal to noise ratio increases as the sampling frequency increases.
So you have two challenges to overcome at the small time frames. It probably can be done (my first serious attempt traded 5 minute bars and did very well for about 3 months before nose diving into oblivion over the next few months), but it would certainly be simpler to start with a higher time frame. Especially if you're just starting out with algorithmic trading (not sure of your background, not making assumptions!).
One thing to note is that the sharpe ratio of a profitable strategy will generally increase with its trading frequency. Therefore, if sharpe ratio is your performance objective it would make sense to trade it at the greatest frequency (lowest time frame) at which it remains resistant to the effects of the increased noise and trading costs.
