Do not forget that the quantities you list are in fact the inputs to the models that are used to derive the variables you really need in practice: irradiance and illuminance on the facets of the building (windows especially). These facets are usually NOT horizontal. Measuring all the components for all tilts and azimuths can be a costly proposition, and that's why it is rarely done (hence the need for models), but that's what should be done in serious experiments to remove the (large) uncertainties in modeled radiation.
Illuminance is measured with photometers (from, e.g., Licor), which resemble silicon-based pyranometers. Both are less costly than thermopile radiometers, which are normally the best in terms of accuracy. Measurements obtained with silicon-based pyranometers need various corrections to account for their limited spectral range. No correction is needed for photometers, though. So you have this issue of accuracy vs cost to consider.
Direct irradiance is measured with a pyrheliometer, which tracks the sun and is therefore costly, but also the most accurate of all radiometers. Obtaining direct irradiance by subtracting diffuse from global is convenient, but not accurate, as I showed in recent publications.
Chris
--
Chris A. Gueymard, PhD
Solar Consulting Services
P.O. Box 392
Colebrook, NH 03576, USA
Tel. (603) 237-8550
Fax (603) 237-5314
http://www.SolarConsultingServices.com