Sidebar 3: Measurements
As might well be expected, harmonic distortion levels were negligible—better than –83dB over the measured 20Hz–20kHz range at a nominal 0.5V IHF output level. With class-A circuitry, fully complementary harmonic orders were low, with quite negligible upper-range harmonics. The two-tone high-frequency intermodulation test generated a very low –89dB level of in-band difference tone. Little variation was observed at higher or lower signal levels. Virtually a little power amplifier in its own right, the KSL could deliver a maximum voltage swing of 8.7V RMS from a very low DC-coupled source impedance of 13 ohms. With a simple adapter, it would make an exquisite amplifier for high-quality dynamic headphones for personal listening—sacrilegious in this context. Electrical noise levels were comfortably low, –90dB CCIR-weighted, this using the more demanding 1kHz reference frequency. "A"-weighted, these results would be 5 or 6dB better. Note that these figures are to the IHF 0.5V input standard, and if referenced to the 1.5V input sensitivity of the KST-100 a practical "A"-weighted result for a CD input level would reach 110dB or so. No electrical hum was present.
While the tape output showed no offset (direct-coupled to the source), the main output showed a larger than expected 21mV worst-case offset, single-ended, which, given a typical 26dB of power amplifier gain, translates into 0.4V at the amplifier output in the absence of servo correction (in practice, the KST-100's servo does correct this offset after a short interval). In balanced mode, the residual offsets balanced out to some degree.
As a matter of record, the frequency response is shown in fig.1, perfectly flat from below 5Hz to 50kHz. The –0.5dB limits were 2Hz and 100kHz, –3dB at 0.88Hz and 200kHz. I could well support more rolloff in the ultrasonic band to provide sources, especially digital ones, with some RFI filtering. The input overload figures were simply the maximum available from the signal generator; ie, 9V or so.
Fig.1 Krell KSL, balanced frequency response at 1V into 100k ohms (5dB/vertical div.).
Very high channel-separation figures were not expected, bearing in mind the KSL's shared internal power supply, but the measured figure of 71dB midband was fine, falling to 46dB by 20kHz—more than satisfactory according to the psychoacoustic references. An overall single-ended gain of 7dB was recorded, 13dB for a balanced input to a single-ended output. The values for line gain may be on the low side for low-output tape decks and tuners used with low-sensitivity amplifiers and speakers.
Volume tracking for the two channels was fine at high settings but deteriorated at low settings. At a –60dB "whisper" setting, the imbalance reached 5.4dB, unacceptable, I feel, in a product of this class. Krell should attend to the volume-control tolerancing. No mechanical hum was produced by the KSL. As regards input impedance, the parallel capacitance was negligible while the resistance value was 48k via the buffered balance input, typically 30k via the direct single-ended input.—Martin Colloms
