What It All Means: line-level preamplifier measurements Page 2

The frequency-response measurement on the Audio Precision is quite straightforward. Like most measurements on this device, the Audio Precision sweeps across the frequency band automatically, one channel at a time, to provide the graphs seen in the charts. The crosstalk measurement is a little more interesting. Each channel is measured separately. The left-to-right crosstalk is a measure of the output from the right channel when a known signal is fed into the left channel and the latter's input is shorted to minimize noise in the reading. L-R crosstalk is measured in dB below the level of the output from the driven channel. The crosstalk is also a swept measurement, the Audio Precision again sweeping across the range automatically once the test is initiated. Practically all of our crosstalk measurements show a poorer separation (or increased crosstalk---two sides of the same coin) at higher frequencies. This is largely because normal capacitive coupling between channels is stronger at high frequencies where the two channels share a common chassis and their circuits are physically close together.

The measurement of total harmonic distortion plus noise (THD+noise) is also straightforward on the Audio Precision once the proper setup is established. The primary consideration of the latter is arranging the grounds between the DUT and the Audio Precision to minimize the contribution of noise and hum. Again, the Audio Precision performs a sweep across the frequency band, automatically plotting out the result.

Polarity is checked by feeding a positive pulse---actually a positive raised-cosine signal from a test CD---into the DUT and observing the output on an oscilloscope. DC offset is measured at the output of the DUT with its inputs shorted. Gain is a simple calculator computation based on a set input and the resulting output. And the input and output levels for 1% distortion are easily read from the output display of the Audio Precision using any of its THD+noise test modes.

Given reasonable frequency response, distortion, and crosstalk results, the major item of measurement interest to most users is probably the DUT's input and output impedances. In the case of a preamp or other similar line-level device, conventional wisdom calls for the highest possible input impedance and the lowest possible output impedance. It is common engineering practice to have the input impedance of the next device in the chain be considerably greater than the output impedance of the preceding device. There is no magic ratio at which everything becomes "right," but the rule of thumb is an input impedance of at least ten times the source impedance (footnote 2). Less than that, and adverse sonic effects grow more likely. A low output impedance will also generally minimize frequency-response aberrations caused by the characteristics of the cable feeding the next device in the chain---be that a power amp or another line-level component.

Footnote 2: The only high-end manufacturer espousing a different theory is the Jeff Rowland Design Group. They prefer matching input and output impedances for maximum power transfer. This will work only if the first device in the chain can furnish adequate current; matched source and load impedances result in considerably more current flow between the source and the load than is usually the case with low-output/high-input impedance hookups---as well as a 6dB loss of effective gain. The Rowland preamps are clearly designed to operate in this manner; most preamps are not. I haven't been convinced of the need for maximum power transfer between line inputs and outputs. What we're trying to provide is an accurate voltage to the input of the second device in the chain. But Rowland believes in the technique enough to provide optional output and input impedance settings on their preamps and power amps, which practice allows for such matching.