I've worked on the massively parallel DAC for a few months now and got to the point where I really need transformers made by a factory to create some kits for sale.. Winding them by hand is only my idea of fun when its a research project, not a commercial undertaking. Seeing as nCoV has hugely slowed commercial activity here in China I figured it would be quite some time before I'd get any significant volume of transformers (or even PCBs) made I therefore decided to re-visit the original PhiDAC (2019) once again and have a play with that design.
The two major attractions of the sound of 'grossDAC' over PhiDAC are considerably increased ambience and cleaner high frequencies. I began to wonder how much of the excellent ambience of grossDAC was due to the paralleling of chips and how much might be down to some effect of the transformers. Transformers I've used before on the output of my mobile phone gave a certain 'spaciousness' to the SQ which was alluring. Over on DIYA a contributor to my lingDAC thread had asked if it was possible to modify PhiDAC to accept 8 paralleled DAC chips, instead of the usual 1. The answer was a quite firm 'no' to that but I was curious how many extra chips might be OK to use.
4 chips turns out not to be too difficult. Just a few tweaks to gain and current sources (one of which controls the DC offset) with a cap change in the filter. What was surprising on listening was how much closer the SQ was to grossDAC in terms of the ambience. I'm not sure I'd be able to tell them apart on that aspect. My hypothesis is that the DAC generates some of its own LF noise which gets reduced 6dB when 4 chips are stacked up. Pretty damn encouraging for such a simple, cheap circuit.
So then, what about the HF 'haze'? Still working on that but adding more caps helped quite a bit with that on PhiDAC SE. For what I'm calling 'PhiDAC Quad' I don't want to go to such complexity as an additional PCB for caps so I'm experimenting to see if there are any other ways around this.
Another thing I've discovered is that under certain conditions the MLF2012 inductors used in the passive filter generate some measurable distortion. This might be because I'm using them close to their maximum rating (5mA) - 4 chips running at 6V put out 4.8mA max. So I'm exploring different inductors and slight variations on the basic filter configuration.