Our laboratory uses a bank of 5 FS20 lamps that emit a peak spectra of 313nm, since we are interested mainly in the effects of UVB. We use a Kodacel filter to remove any traces of UVC.

One important question is the size of your sample. How big a vessel do you need to irradiate? and do you want to use an irradiance similar to that in summer noon sunlight? or would a lower irradiance do?

Simulating sunlight (including its spectrum) with lamps is very difficult. If this is your aim then the solar simulator is the best option. If you know the action spectrum of the reaction, then you could choose a source that emits radiation at this wavelength. Fluorescent lamps that emit UV radiation are available (Q-Panel UVB-313, UVA-340; Philips TL12, TL01). The Philips TL01 are available as compact fluorescent and as tubes. UVA LEDs are cheap only for longer wavelengths (365 nm), those emitting at shorter wavelengths are expensive.

With any UV source you should be very careful with eye and skin protection, especially if the ambient level of visible light is low. UV-absorbing goggles are a must, even if UV levels on your sample are similar to those in sunlight, as the irradiance near the lamp, or UV exposure in the eye if looking into the lamp could cause eye damage in a similar way as looking straight into the sun does.

Hope this helps. (An open-access handbook on UV research methods is available for download from my profile and at http://uv4growth.wordpress.com . Focus is on plants, but a lot of its contents should also apply to photochemistry.)

Book Beyond the Visible, A Handbook of Best Practice in Plant UV ...

Consider a use of  OTS floods with unfiltered Xenon "solar flash" lamps (1-10kW) with a nearly perfect match of the solar spectrum across the range of 250nm – 1000nm (CCT ~6500K. (e.g. Heraeus). Xe lamps are used in many solar simulators.