The laboratory evaluation of the Iris was carried out in two stages. First, although it might seem excessive to use a component of this kind as a "bare" bridge, we evaluated its interface with the home network and server through the digital output. We then assessed the D/A converter and analogue stage, using the coaxial digital input and the single-ended analogue output. For this part of the evaluation, the unit was set to low gain, with the attenuator engaged. Unless otherwise stated, all signals were 24-bit/96kHz.
As far as interfacing with the home network is concerned, the Iris performed flawlessly, with very low jitter at both the data and clock levels. With a 16/44.1 signal, we measured 0.014UI (2.35ns p-p) in both categories. These figures increased to 0.048UI for data and 0.045UI for clock at a 24/96 sample rate.
The eye diagram is clean, with low noise levels, and remains comfortably clear of the minimum acceptable limits at both 16/44.1 and 24/96 sample rates. The latter shows slightly slower edges and somewhat greater uncertainty at the zero crossing. This is also confirmed by the corresponding eye-narrowing figures: 0.012UI/0.035UI (2.7/3.0ns) at sample rates of 44.1/96kHz, respectively.

The coaxial digital input interface proved effective at suppressing incoming jitter, with the corresponding curve hovering around -40dB and showing a mild tendency towards reduced effectiveness at very low and very high frequencies.
Inferred jitter, generated by the unit itself, is also quite low, with the lowest calculated value in the region of 55ps p-p and the highest at 380ps.

Output-level linearity proved very good, with the curve staying very close to the ideal down to -100dBFS, where noise begins to make its presence felt.
The Iris's distortion performance appears to be influenced almost exclusively by the tube output stage. With the signal level very close to the DAC's limit, the corresponding curves are in the region of -50dB. The fact that the THD and THD+N measurements almost coincide indicates that the dominant factor is the behavior of the tubes as stage approaches overload.

This becomes clearer in the distortion-versus-level graph. Here, THD alone remains very low at levels below -20dBFS and stays low even at extremely low signal levels. Its performance at -60dBFS is particularly good, at -45dB. Noise, which becomes increasingly significant as the signal level drops, places the THD+N curve considerably higher (about 25dB at the lower end of the measurement range). Above -20dBFS, however, the dominant factor is the tube's behavior as it approaches overload, so the two curves almost coincide. A figure of -45dB at maximum output level (0dBFS) is a good result for a tube-based design.
The distortion-versus-frequency-and-level graph supports all of the above. One particularly interesting finding here is the uniform behavior, with stable figures even at high frequencies, showing that the absence of negative feedback does not have a detrimental effect.

Ayon's choice of digital filter is interesting. Although we cannot say with certainty which chip is used, it is clear that a “slow” filter has been selected, with the corresponding frequency-domain response.
However, the "sacrifice" made by the slow filter in the stop-band does not appear to translate into a corresponding improvement in the time domain, where pre-ringing is clearly visible. Unfortunately, Ayon provides no further information on this particular aspect, so we cannot tell exactly what the thinking behind this choice was.

The Iris is a fairly quiet component. The noise graph for the idle channel consistently remains below -90dBFS. What's more, the contribution from power-supply-related frequencies is very small, which vindicates the company's choices in the design of the power supply.
The noise and interference spectrum confirms the quiet operation of the power supply, with all the measured components below -90dBFS. There is, however, an interference component of unknown origin that affects the region below 1000Hz. Higher up, interference components are visible, but remain below -100dBFS.

The graph showing signal modulation caused by very low-frequency noise is very clean around 50Hz, where power-supply-related effects traditionally show up, and contains no components that could be attributed to jitter. It does, however, show indications of interference from noise around 18Hz. The origin of this finding cannot be established from the measurements alone, although theory suggests a number of possible causes, which are beyond the scope of this review.
Finally, channel separation was good, with the corresponding curves around -80dB and indications of mild capacitive coupling above approximately 5kHz.

Previous | Next | More Reviews