If your signals come from standard databases (see System Requirements), signal calibration is usually not a concern. Briefly, the header file for each database record contains calibration information for each signal. This information includes (for each signal) the gain (the number of adus per physical unit), the baseline (the sample value that corresponds to zero physical units), the type of physical unit, and the type of signal. (Not all header files contain all of this information, but WAVE can make reasonable guesses about any of it that may be missing in most cases.)
If you have digitized your own signals, however, you should generally calibrate them before processing them further. WAVE makes it easy to do so, provided that you have recorded signals with known amplitudes. The procedure is:
Figure 5.1: Signal calibration using WAVE.
Signal 3 (ABP) has been selected (note that all of the other signal
numbers have been removed from the Signal list). The `<'
and `>' markers bracket a step in the recorded calibration
pulse near the center of the signal window. In the Analyze
window, we have clicked on 
, and the result of doing
so appears in the Analysis Commands window.
in the Analyze window, and
answer any questions that appear in the Analysis Commands
window. These questions will always appear if the header file for the
record does not contain signal descriptions or physical unit
specifications; some or all of the questions may not appear if the
signal type and calibration pulse type are already known.
Figure 5.1 illustrates signal calibration using a record from the MIMIC Database. In this case, the type of signal (ABP) is known from an entry in the DBCAL file (not shown here), so that calibrate is able to determine the physical units of the signal (mmHg). Since a variety of calibration pulses are used for ABP signals, the DBCAL file does not specify the pulse levels, which calibrate has asked us to enter (in this case, we have entered 60 and 120). Based on this information, calibrate determines the offset and gain needed to convert raw sample values for signal 3 into ABP measurements in mmHg. calibrate then makes the appropriate changes to the header file for the current record. When WAVE or another DB application next opens this record, the ABP signal will be properly calibrated.
By default, calibrate generates an amplitude histogram of the samples between the `<' and `>' markers. It then identifies the low and high amplitude portions of the calibration pulse by searching for the two largest distinct modes in this amplitude histogram. For this reason, calibrate works best if the segment bounded by the `<' and `>' markers includes at least a few samples of both the high and low amplitude phases. Avoid placing either marker immediately next to the transition point between the phases if possible. If calibrate fails to find two distinct peaks in the amplitude histogram, it will produce an error message; if this happens, adjust the positions of the markers and try again.
In some cases (for example, if the calibration pulse is a sawtooth, as in the RESP signal at the bottom of figure 5.1), this strategy may fail, no matter where the markers are placed. In such cases, try again with calibrate's -q or -Q options to use one of its alternate algorithms. (These are less robust since they depend on differential rather than integrative measurements, but they can be used in a pinch.)
For further information on signal calibration, see calibrate(1), in the ECG Database Applications Guide.