Well, my reference is my modified CD-67, which to my ears outperforms my Rega 3 (but hey, that's not overly surprising), but I'm not one for purely subjective assessment as it'is prone to all sorts of psychoacoustic, and psychological inaccuracy. The old 'if you expect something to be better, it is'.
Mostly, my performance assessment is based on what small measurement I can do at home (using ARTA and Visual Analyser software) and what kit I have access to in work (from Distortion Analysers, the most sensitive spectrum analysers and EMC receivers you'll find, vector impedance meters etc). Some of the ideas I applied to my CD-67 have been applied to this, and I did more significant measurement on the CD-67, so I have an idea of what works and what doesn't work. The rest is based simply on sensible and proven engineering. Splitting the digital and analogue power supplies, reducing the possibility of unwanted resonance in the clock crystal (a small step towards jitter reduction), reducing the amount of radiated RF noise from the processor etc. Those are things I've found that do make measurable difference, and therefore have more potential for positive impact on audible performance. I didn't want to go too nuts on this one so I just picked the bits from the CD-67 which seemed to have most effect.
As far as changing parts goes, the LME is probably overkill, but as the CD48 only uses one, and I have some in stock, I thought it was worth doing. Change of caps is mainly to rule out leaky old ones, as properly specced ones shouldn't have any effect on sound quality (they certainly don't on anything measurable).
I've owned so called 'better' equipment in the past, and had access to so called 'high end' stuff. Some good, some crap, some snake oil. The longer I play with this sort of thing, the more I realise NEVER listen to anything the hifi press say, and subjectivism is as reliable as a chocolate fireguard.
"Two very small men cutting steps in the roof of the world"