Liquid blood phantoms are a tool to calibrate, test and compare near-infrared spectroscopy (NIRS) oximeters. They comprise a mixture of saline, blood and Intralipid, which is subsequently oxygenated and deoxygenated to assess the entire range of tissue oxygen saturation (StO) from 0% to 100%. The aim was to investigate two different deoxygenation methods: yeast versus nitrogen (N) bubbling. The phantom was oxygenated with pure O in both experiments, but deoxygenated by bubbling N in the first and by addition of yeast and glucose in the second experiment. A frequency domain NIRS instrument (OxiplexTS) was used as reference and to monitor changes in the reduced scattering coefficient (μ') of the phantom. Both deoxygenation methods yielded comparable StO values. The deoxygenation was slower by a factor 2.8 and μ' decreased faster when bubbling N. The constant bubbling of N mechanically stresses the Intralipid emulsion and causes a decrease in μ', probably due to aggregation of lipid droplets. Deoxygenation by N requires a more complex, air tight phantom. The gas flow cools the liquid and temperature needs to be monitored more closely. Consequently, we recommend using yeast for phantom deoxygenation.