Recently this project has lead me to a sad realisation. I came to London some years ago eager to find an endless pool of disposable minions, ready to play their tiny but collectively vital parts in whatever thrilling scheme of world domination one might wish to embark upon. And so it went for the first dozen or so crews that were liquidated in unfortunate accidents once they began asking too many questions. Eventually however it did become harder to find fresh bodies to press-gang into service. I had to reduce the turnover rate, and the workforce has taken this as a sign of weakness and started asking for such things as living wages and break periods. One can only hope the project completes before they ask for... (shudder) ...benefits. And for this reason I am opposed to Brexit: clearly it is shallowing the disposable minion pool. On the plus side, they still show absolutely no interest in basic health and safety, so no impact on budget or schedule from that direction.
Anyway, I digress, the external structure is broadly complete with the railings and planter on the patio level and the strip of roof at the front.
The original plan had insubstantial wooden railings, I replaced that with 16 stainless steel pipes, embedded in thick brick-clad R/C piers, fully plumbed to the bottom level for use as water cooling radiators. This is sufficient in total for about 16 servers, to run the full cluster this way I'd need to build a pergola of similar construction.
My younger son has also been attempting to help with construction.
Drainage in the planter, it is now filled with soil and a hedge at the front, need to install the smoked glass screen down the middle (posts anchor into the blockwork piers) before the last few plants can go in.
Image from last month in the final stages of tiling: the front face is now complete except for the channel drain across the threshold. The interior has also been progressing but with one significant setback:
I spent a week building the secondary rear projection mirror starting with the support frame.
Which was a 5m x 2.4m plane inclined at a tricky angle over a concrete pit (which holds the primary mirror and the subwoofers). Getting the acrylic mirror to stick to it required trying three separate adhesives.
Unfortunately testing revealed this to be infeasible. The geometry worked correctly, and the main problem that web forums had warned me about; getting the flexible arcylic material perfectly flat; wasn't a big issue. The real problem with that first surface arcylic mirror seems to have non-uniform reflectivity, not enough to be visible to the naked eye when looking at it as a mirror, but enough to give mirrored projections a pronounced cloudy texture. Presumably this is due to variance in the thickness of the aluminium layer; I tested a couple of samples from other suppliers after the issue became apparent, and they had the same problem. Obviously I should have tested this before building the mirror system, as it was I wrote off 900 GBP in materials and a week of work (fortunately the structural materials are reusable elsewhere).
The obvious option to fix this was to switch to glass, but this would cost about 10K GBP for reputable optical glass, assuming I didn't break any mounting it, and would require a lot more segments. Segmentation isn't a dealbreaker because the mirror plane is out of focus, so the unavoidable slight cracks don't show up as black lines so much as slightly darker strips, but going to 20 segments on the secondary mirror instead of 5 plus having to segment the primary mirror was asking for trouble.
Fortunately, advances in projector technology and the extended project timeline have bailed me out. When I designed the building the only 4K HDR solid state projectors available were the Sony VW5000-ES at ~50K GBP (5000 lumen) and the JVC DLA-Z1 at ~30K GBP (3000 lumen). With a DLA-Z1 I'd need to zoom to fill the full 8:3 screen or use a couple of cheap 1080p aux projectors in data mode. By the time I had to replace my old home theatre projector early last year, the Optoma UHZ65 was out at a far more reasonable 4K GBP for 3000 lumens. It doesn't have power zoom but at that pricing it was feasible to pair two of them, edge blend across the middle and create a 6K ultra-widescreen display. However the UHZ achieved that price by using a (laser-pumped) colour wheel, instead of 3 separate laser sources on the high-end projectors.
Since then the Optoma UHL55 has been released and impressively managed to deliver 4K HDR and a surprisingly good gamut from a genuinely solid state light source: LED. Unit price is only 1300 GBP, so it's completely feasible to use six of them in a 3 x 2 array to create a 12K display. The UHL55 is fixed throw, no zoom at all but that is actually preferred for rear project video wall application, as it eliminates a source of discrepancy between tiles and image drift. The 1.2x throw and tiling does unfortunately bring back the issue I was hoping to avoid with the very long light path in the original design; brightness variation (hotspot) due to incidence angle changes, but looking at the geometry I think it's managable. If I'd designed for this unit from the start I could've actually managed direct project, as it is with 2m usable depth I will have to use mirror bounce, but six small glass mirrors co-located on the projector mounts is far cheaper and easier to build than a 12m2 perfectly flat optical glass secondary mirror. As such I've ebayed he UHZ65, picked up six UHL55s and tested them in the living room: individual picture quality is a little below the UHZ65 due to the smaller chip and cheaper lens, but combined quality in 12K configuration is substantially higher. Plus combined brightness 12000 lumens, well above even a blended dual VW5000-ES config (7500 effective at 8:3). Setup will be a bit tedious since I'm not blending and the pixel size is now 0.4mm, but should be much easier than trying to do this with a long throw front projection array or bulky dual mirror bounce RP rig.
The only remaining issue is driving 6 4K60 HDR HDMI2 displays as a single screen. For the initial test I just used two graphics cards and some DP to HDMI dongles, which works ok for Windows desktop but not gaming. AMD Eyefinity can present a 6 screen matrix as a single display of any chosen resolution, but they all have to be driven from the same card (even with Crossfire on). Which means dual-stream Displayport splitter/hubs, the vast majority of which are only DP 1.2 i.e. only support 4K30. I've tracked down some DP1.4->HDMI2 splitters which should do the job, when they arrive from the US...
In the mean time at least the audio configuration is working well. The full-planar Atmos array of 4 x electrostatic + 8 x magnetoplanar probably doesn't image as accurately as the equivalent spend in conventional cone speakers, but the ambience is unique and frankly pretty amazing (as in, visitors are consistently amazed by it). The quad 15" subs are not up to AVS forum hardcore standards of output level, but they are all servo-controlled and thus give more than enough bass & subsonics for me while not suffering from any noticeable distortion.
