After much work I got to the point of testing the attenuation provided by the latest version of the DÆ stepped attenuator (version v4.2). The results are shown in this chart.
The chart shows the measured attenuation of right (Orange) and left (Green) channels of a v4.2 DÆ stepped attenuator compared to the target attenuation (Blue). All three curves lie on top of each other showing how closely the attenuation of the two channels match each other and the target.
The difference between the attenuation of the two channels is the central red curve with a finer scale on the secondary y-axis. From about -60 dB to 0 dB the left and right channels are within ± 0.2 dB of each other. This is exceptionally good. By way of comparison, the tolerance bands for the Alps RK271 series potentiometer are show in the dashed black lines. The Alps RK271, also called Alps Blue Velvet, is a well regarded potentiometer by the audiophile community.
The tightest tolerance for the Alps RK271 is a gang error of 2 dB max between -60 dB and 0 dB. The gang error tolerance is even greater at ± 3 dB lower volume levels with attenuation between -70 dB and -60 dB. Gang error is the difference between the attenuation of the right and left channels at a particular shaft angle. The shaft angle sets the volume level or attenuation.
The v4.2 DÆ stepped attenuator easily meets the gang error tolerance specification for the Alps potentiometer. This is not surprising considering the DÆ attenuator uses 1% precision resistors which is a much greater precision than possible with the carbon trace of the Alps or any other potentiometer.
The attenuation of the left and right channels match the target curve within ± 1 dB across the whole attenuation range and within ± 0.25 dB for attenuations from -45 dB to 0 dB. This is also a very good result.
A note on the target attenuation curve and the number of attenuator steps. The version of the v4.2 DÆ stepped attenuator tested has 24 steps. While a greater number of steps up to 96 is possible with the v4.2 attenuator, I have only built and tested 24-step attenuators. In my listening tests I have found 24 steps to be work nicely.
Part of this may be due to the target attenuator curve employed here which is - 3 dB per step from full gain down to - 66 dB. This accounts for 23 of the 24 steps. The last step is really a mute feature with attenuation of greater than 90 dB.
There is a good rational for a linear attenuation curve with three decibel steps. A three decibel change in sound level is considered to be barely perceptible and the attenuation range -66 dB to 0 dB appears to be well suited for music listening. In practice, the 24-step v4.2 DÆ stepped attenuator provides smooth and precise control of volume.