This was the final project born in my workshop in our house in Norfolk, completed just as our house sale went through. This means that I wasn't able to audition the amplifier through my reference system, which had been in storage for quite some time previously, so I was relying on the modest system in the workshop when making the initial design choices.
It spend some months in the office at work, allowing me the opportunity to finalise the firmware during the lunch-hours. And in the evenings, my poor Audax mini-monitors were somewhat over-driven! I was rather impressed with the overall system performance, especially given the size of the room (approximately 25 foot square).
Now that we've finally moved in, I'm still getting used to the new listening room, and also having difficulty making time to listen to music because there is simply so much that needs doing elsewhere. But initial impressions via the ATCs are very good indeed. The eventual acid-test will be to compare this amplifier to my optimised prototype LM3875 gainclones, which perhaps might sound better because they are dual-mono?
Despite the project being largely finished, there's plenty of scope for experiments and improvements. It's still a work in progress, and one day I might get around to producing PCBs once I've set myself up for DIY production. But for the time being I'm pretty pleased with how it turned out, and I can't wait for an excuse to do another metalwork project on that scale again!
2024 Update
Some 20 years later, this amplifier is still working well, and hasn't been modified since the initial build. Only two minor faults have occured, and they both happened within the first few months of operation: firstly the 100Ω surge-limiting resistor in the capacitor-dropper standby supply needed to be replaced with a carbon composition type as film resistors don't enjoy a long life in that role, and secondly the PIC developed amnesia and stopped working. This was because I programmed it at work with the educational kits that were designed in-house by a former staff member. They've worked really well for their intended purpose, but the programming voltage doesn't meet the Microchip standard, meaning the PIC might not be reliable long-term, as was the case here. I re-programmed it with my JDM clone and all is well. It's worth saying that the PIC in my pre-amp was also last programmed in 2004 and is still fine, so although my JDM clone is far from production grade, it does a good job of producing lasting PICs.
In use, I've found no reason to want to replace this amplifier. I have used it with a large range of speakers over the years, and I have never noticed it having any character or shortcomings - it just does the job. I experimented with bi-amping my ATCs, but couldn't hear a significant difference. However, they do sound better when bridged because of the increased headroom. A quick check with a 'scope shows that this is a good demonstration of a long-established fact: mild clipping might not be perceived as such, but the character of the sound will likely change a little. This is why I'm a big believer in clipping indicators - which are included by default in professional equipment, of course. This is the only change I'd make to this design if I was doing it again.
When I'm reviewing loudspeakers, I use it in 4 channel mode, along with a switchbox I made that allows instant comparison between 2 pairs of speakers. It includes level controls to match the volume of the two pairs - well, to some extent, at least, given the inevitable tonal differences between them in most cases.
Over the last two decades, the amplifier has never had a temperature or DC fault shutdown, and the PIC has only ever reset during a genuine power outage. It has always been totally civilised in operation, with no pops or clicks as it turns on and off.
My biggest worry is the unfortunate discontinuation of the LM4780. Should one fail, I will have to re-engineer it to use the LM3886, which is still in production (and I have lots in stock!). While that should be easy electrically, given that the LM4780 is a dual LM3886, it would be a lot of work - the whole of each heat sink assembly will need rebuilding almost from scratch. But if the worst was to happen, I would definitely make the effort as it's such a nice amplifier in a really convenient, compact package.