The problems of breath-by-breath measurement of respiratory gas exchange to analyze the transient responses during complex load work are described and treated. With review for the experimental systems developed by other investigators, our signal transducers and data processing were improved. We found the computer-processed data were in good agreement with those obtained from simultaneous Douglas bag gas collection.
1) Errors in the measurement of volume arose from errors in the flow signals such as nonlinearity, base-line drift, noise, and frequency response, as well as composition and temperature of gas, water vapor condensation, quantization errors, and breathing valve leakage. ±1.74% error in the flow integration reproducibility resulted from the modification of the upstream geometry of the pneumotachometer and the pressure tubing as well as a compensation for base-line drift and filter smoothing.
2) Errors in the gas concentration signals were attributed to inaccuracy, drift, noise, and water vapor concentration. The transport delay of gas concentration signals was overestimated in order to cancel the underestimation in Vco₂and Vo₂·Other compensation methods for the response time were discussed.
3) Error magnitudes below±0.97% in the A/D amplitude quantization were found by means of signal simulation.
4) The optimal compromises between breath recognition threshold and the fluctuation in flow signal were examined to permit identification of irregular breath.
5) Since the breathing valve dead space was modified to tidal volume dependent, errors in the gas exchange variables were reduced.
6) To validate the accuracy of the equipment operation and the gas exchange algorithm, problems in signal simulation and the model lung were described.

In order to study respiratory transients during exercise, we examined breath-by-breath differences between gas exchange kinetics measured at the mouth and those estimated at the alveolar level. The gas exchange data at the mouth were obtained by measurement of expired gases only (expiratory flow method) . Correction for breath-by-breath changes in lung gas stores was applied to the total gas exchange, which was obtained by subtracting expired from inspired gas volume (alveolar gas exchange method) . Constant work loads (150, 200, 250 W) and a ramp work load (30 W/min) preceded and followed by a 50 W load were generated by a computerized cycle ergometer. Best-fit first- or second-order model values for gas exchange kinetic parameters were found by the non-linear least-squares method.
1. Regardless of work intensity and forcing function, the breath-by-breath variation in gas exchange measured at the mouth was larger than the gas exchange estimated at the alveolar level, in both a non-steady state and a steady state. The variation was caused by the invalidity of assuming zero N₂ exchange at the mouth, which was attributed to changes in lung volume.
2. Vo₂ kinetics at the alveolar level were faster than those at the mouth, while the converse held for Vco₂ at the onset of constant load work, due to the effects of fluctuations in lung gas stores on the kinetics of gas exchange at the mouth. During ramp load work, Vo₂ and Vco₂ kinetics at the alveolar level were faster than those at the mouth.
3. Steady state gas exchange values at the alveolar level and at the mouth were the same during constant load work, since the lung gas stores corrections added up to small fractions of the total gas exchange when summed over the long term.
4. Consideration of both the proper end-expiratory lung volume and ventilationperfusion inhomogeneity was required in order to estimate the true alveolar gas exchange.