I read a bit of this thread and also the magnum opus thread owned by austinpop (“A novel way to massively improve SQ…” on computeraudiophile) where this came up. Quoting that thread, the focus of the participants is
… on direct listening impressions. We were lucky that, from the outset, the early participants set a tone of experimentation with the proposed ideas, and reporting their results. Over time, this … extended the body of knowledge with their experiments and reports. … Most of what is discussed here does not have a readily available analytical explanation. We would not have gotten this far if we had gotten mired in the “how could this possibly matter” debate.
This is fair enough if you’re engaging a group of intelligent, well-meaning folks without particular depth in the relevant science and engineering. We can observe, but we can’t explain.
But this is not good enough to motivate investigations of particular speculations about the causes. Why? Because the suggested causes of the observed differences (memory playback in this case), while apparently reasonable on the surface, make no sense to the experts in the field who do understand what’s going on under the covers. And what’s under the covers with computing technologies is never one iota as simple as the naive lay person’s mental model. While we should remain skeptical of experts, the suggestions of intelligent lay experimenters are even less likely to hold up.
So, folks, you may posit that memory playback explains the perceived performance difference between squeezelite with a memory buffer and LMS vs. Roon, and it may appear that adding the memory buffer is the reason for roon’s lesser performance, but as been hinted at, it’s probably a performance limitation of the platform itself. Who knows if it might be drivers for the disk device, or inadequate resources to run roon and LMS simultaneously, or a zillion possible other hardware limitations (IRQ conflicts, thread priority, yada yada yada.)
If we determine that the streamer hardware meets or exceeds roonlabs’ minimum hardware spec to run roon successfully on that type of device (an embedded linux system, I presume), then the only evidence that I think should motivate roonlabs to investigate is for someone to first observe a difference using ABX testing between two identical hardware stacks (streamer + DAC), one running roon and the other with LMS+sqeezelite+memory_playback. Put identical copies of some music files on each, give each it’s own controller (ipad or whatever), hook the DAC outputs to a blind ABX switch, level match them, start them both playing identical playlists, and compare. If the differences are as pronounced as announced, then one might make a convincing case using this setup without doing strict ABX testing. But you have to give each software system the minimum hardware resources it requires.
Thanks,
- Eric