Coda Technologies S5 power amplifier Measurements

Sidebar 3: Measurements

When I first attempted to measure the Coda S5, the internal rail fuses blew after a few minutes of high-power running into 8 ohms. I returned the review sample to Coda, who investigated what had happened. Apparently, they had had a bad batch of output devices, and two of these had found their way into our review sample. They had survived Wes's music, but gave up the ghost when hit with sustained sinewave signals. The amplifier was repaired, then sent back to me so that I could complete my measurements.

The Coda S5's output stage is biased heavily into class-A (footnote 1), which means it runs hottest when not delivering power into a load. Even so, at the end of the one-hour preconditioning period, running at one-third power into 8 ohms, its heatsinks were around 55ºC; ie, not so hot that I couldn't keep my hand on them. The THD+noise percentage at this level was 0.006% when the amplifier was cold, rising to 0.032% at the end of the preconditioning. All the measurements were performed with the amplifier hot.

The voltage gain into 8 ohms was 28.45dB from the S5's unbalanced input, but 1dB lower from the balanced jack. This discrepancy was explained when I looked at the amplifier's input impedance: a high 47k ohms unbalanced, it was just 1085 ohms balanced. This will tax many preamplifiers, which like to see an input impedance of at least 10k ohms. And if you intend to use a capacitor-coupled tube preamp with the Coda, make sure you use an unbalanced connection or the bass will be prematurely rolled off.

Both balanced and unbalanced inputs preserved absolute polarity (ie, were noninverting), and the XLR jacks were wired with pin 2 hot, the AES standard. The output impedance was a moderately low 0.15 ohm across most of the audioband, rising slightly to 0.17 ohm at 20kHz. As a result, any modification of the amplifier's frequency response due to the usual Ohm's Law interaction with the speaker's impedance will be mild. With my standard simulated loudspeaker, the response variation remained within ±0.15dB limits (fig.1, top trace at 2kHz).

Fig.1 Coda S5, balanced frequency response at 2.83V into (from top to bottom at 2kHz): simulated loudspeaker load, 8 ohms, 4 ohms, 2 ohms (0.5dB/vertical div., right channel dashed).

The S5's response was flat up to more than an octave above the audioband, with then both a higher-order ultrasonic rolloff than I usually see and a bit of peaking, particularly in the right channel. This behavior results in very good 10kHz squarewave reproduction, but with a light degree of overshoot (fig.2). No ringing can be seen, however. Channel separation (measured with the undriven input shorted) was good, at 90dB or better in the midband. However, capacitive coupling between the channels increased the level of crosstalk slightly at the top of the audioband (fig.3). DC offset was low, at 10mV right channel, –5mV left.

Fig.2 Coda S5, small-signal 10kHz squarewave into 8 ohms.

Fig.3 Coda S5, channel separation (R–L dashed, 10dB/vertical div.).

The S5's linearity varied more than usual with changes in load impedance, which can be seen in fig.4, a plot of the THD+N percentage present in the amplifier's output while the power level varied from 100mV up to clipping (defined as 1% THD+N). Only at levels above a few watts into 2 ohms did the distortion rise above 0.1%, however. At clipping, the S5 exceeded its rated power by quite a large margin. With both channels driven, it clipped at 65W into 8 ohms (18.1dBW), a full dB higher than specified, and managed 94Wpc into 4 ohms (16.7dBW). With one channel driven into 2 ohms, the S5 clipped at no less than 182W (16.5dBW).

Fig.4 Coda S5, distortion (%)vs 1kHz continuous output power into (from bottom to top): 8 ohms, 4 ohms, 2 ohms.

The Coda's signal/noise ratios were excellent, the A-weighted figure (ref. 1W into 8 ohms) measuring 99dB. Even the wideband, unweighted figure was a still excellent 87.6dB. This is a quiet amplifier. The upward curve of the traces above 500mW or so in fig.4 means that the distortion is starting to emerge from the noise floor at this level. I therefore measured how the THD changes with frequency at 2.83V, equivalent to 1W into 8 ohms (fig.5). While the increase in THD with output current can again be seen, the THD is both low in level and doesn't change with frequency, and the usual rise above the audioband is not evident, suggesting that the S5 circuit has a wider-than-normal open-loop bandwidth.

Fig.5 Coda S5, THD+N (%) vs frequency at 2.83V into (from bottom to top): 8 ohms, 4 ohms, 2 ohms (right channel dashed).

Figs.6 and 7 show that what distortion is present is predominantly the subjectively innocuous second harmonic. These measurements were performed using the balanced input. Repeating the 50Hz spectral analysis using the unbalanced input (fig.8) reveals that the third and fifth harmonics rise under this condition, but that the distortion harmonics smoothly drop with increasing order, something that correlates with good sound quality. Driving the S5 almost to visible clipping on the oscilloscope screen into 4 ohms with an equal mix of 19kHz and 20kHz tones gave the spectrum shown in fig.9. The 1kHz difference component at 1kHz is the highest in level, at –65dB (0.06%), which is good performance.

Fig.6 Coda S5, 1kHz waveform at 12W into 4 ohms (top), 0.078% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).

Fig.7 Coda S5, balanced, spectrum of 50Hz sinewave, DC–1kHz, at 40W into 8 ohms (linear frequency scale).

Fig.8 Coda S5, unbalanced, spectrum of 50Hz sinewave, DC–1kHz, at 40W into 8 ohms (linear frequency scale).

Fig.9 Coda S5, HF intermodulation spectrum, DC–24kHz, 19+20kHz at 82W peak into 4 ohms (linear frequency scale).

The Coda S5's low input impedance in balanced mode mandates that care be taken in choosing a partnering preamplifier. Otherwise, this is a well-engineered design that performs well. Impressive.—John Atkinson

Footnote 1: Each of the 30 output devices for each channel stands on a 0.22 ohm resistor. With the input shorted, I measured voltage drops ranging from 15mV to 22mV across these resistors, which adds up to a standing bias current of 1.26A spread across the 15 complementary pairs. This in turn is equivalent to around 25W of class-A operation into an 8 ohm load. Above that power level, the output stage operates conventionally in class-B.—John Atkinson
The Continuum Group
8274 Mediterranean Avenue
Sacramento, CA 95826
(916) 383-3653