The performance of a free-space optical (FSO) communication system is strongly affected by optical scintillation. Scintillation fades can cause errors when the power on a detector falls below its noise floor while surges can overload a detector. The very long time scale of scintillation compared to a typical bit in an FSO link means that error-correcting protocols designed for fiber optic links are inappropriate for FSO links. Comparing the performance effects of different components, such as photodetectors or protocols, such as forward error correction, in the field is difficult because conditions are constantly changing. On the other hand, laboratory-based turbulence simulators may not really simulate the effects of long-range propagation through the atmosphere. We have investigated a different approach. Scintillation has been measured during field tests using FSO terminals by sending a continuous wave beam through the atmosphere. A high dynamic range photodetector was digitized at a 5-KHz rate and files of the intensity variations were saved. Many hours of scintillation data under different environmental conditions and at different sites have been combined into a library of data. A fiber-optic-based scintillation playback system was then used in the laboratory to test modems and protocols with the recorded irradiance files. This enabled comparisons using the same atmospheric conditions allowing optimization of such parameters as detector dynamic range. It also allowed comparison and optimization of different error correcting protocols.