Continuous observation of carbon dioxide (CO₂) flux is an important issue to understand the global carbon cycle. Although the eddy covariance (EC) method is a promising means of measuring the fluxes, its high cost is often constraint. We therefore developed a technique for a high accuracy measurement system using a relaxed eddy accumulation (REA) method with an inexpensive gas analyzer, and tested the measurement system compared with the EC measurement at a cypress forest. Since spectral analysis revealed that the gas analyzer had both high and low frequency noise, we eliminated the former by using a moving average, and then calculated mean concentration with a shorter averaging period to avoid the low frequency noise, whereby the accuracy of the gas analyzer was improved at least 10-fold beyond the specification. The measured diurnal variation and the magnitude of CO₂ fluxes by the EC and REA methods were almost identical, suggesting that our REA system was highly accurate as well as the EC system. The estimated root mean square error (RMSE) of the CO₂ flux between the EC and the REA methods was 0.17 mg CO₂ m^-2 s^-1. Although the open-path EC method could not be used to measure the flux during the rain event, spectral analysis suggested that the measured flux by the REA method was reliable during rainfall. The proposed measurement system will be applicable to the backup of the EC measurement as well as other trace gas flux measurements.