[noggin]

The trouble with getting all technical and ticking every box with HDR is you can be as accurate as you want .ie NVIDIA Shield converting everything perfectly from rec.709 to bt.2020, but when such data gets to the HDR display you are then at the whim of the TV manufacturer. As NVIDIA found out with a bunch of SDR / 4K HDR TV users.

You then have LG OLED's TV for example having issues with near black flashing currently and other LG workarounds have also had to be implemented for Chroma Subsampling (see Vero 4K+ most recent Dec. update)

---

You really have to look visually - for picture color banding, too dark a picture, washed out colors, etc. because even not every bit of technical metadata is even measured 100% correctly downstream, with Vertex HD Fury hardware for example.

Looks like Apple have gone down the conservative Metadata route, after looking at user feedback. I personally have seen very few complaints.
Same with OSMC (Vero 4K), & AMLogic HDR hardware - running recent community versions of LE or CE.

NVIDIA is finally changing to colorspace switching on the Shield like everyone else has been doing, so no more washed out rec.709 colors for affected 4K HDR TV owners in 2019.

Projector owners belong in a special category and are more demeaning due to their equiptment not being bright enough to display the full brightness levels present in source HDR content and tone mapping then being required. Seems they definitely do need MaxFALL, MaxCLL HDR metadata.         

  
None of this is any reason for not passing HDR metadata is it? After all every UHD Blu-ray player does it when you play a UHD Blu-ray doesn't it?  Passing the wrong metadata isn't that helpful is it?

The HDR metadata is there for tone mapping in all displays that can't fully display the source picture's full PQ range. (Which is almost all consumer displays).  The effect will be subtle on subtle HDR content that keeps most content in the SDR portion (<100 nits) of the PQ curve and only pushes a few speculars into the HDR range, but stuff mastered to push HDR harder, and put more content into the HDR range (>100nits) will be more noticeable.   That said I still don't think PQ is the right standard for domestic viewing - and it's clear manufacturers agree as they are introducing systems to alter PQ light levels based on ambient light levels, moving away from a proscriptive absolute bit to light relationship.  Pity that HLG wasn't picked up - ambient light level control is built into that standard as it was designed for real-world TV viewing...

https://www.lightillusion.com/uhdtv.html Is really good at explaining HDR systems and the role of metadata.

All DV content being played back on a DV device (Apple TV, Amazon Fire TV Stick 4K, Chromecast Ultra etc.) will surely be passing dynamic HDR metadata (you wouldn't get DV certification otherwise)?

Is the Apple TV not passing through the correct static HDR10 metadata in Netflix and Amazon Prime?  (Looking at MrMC forums there seem to be quite a lot of issues with HDR replay?)

Also what aspect of Metadata is the HD Fury Vertex getting wrong? (The areas I know it has issues are clock rate - and that's not a function of metadata?) The 4:2:2 bit-depth it correctly reports - there is only 12-bit 4:2:2 supported at 2160p in the HDMI standard. What the signal contains beyond this (padded 8-bit or padded 10-bit) is a content issue, not a signal issue?

*** EDIT - just checked the Shield TV.  In Netflix and Amazon with HDR content both the mastering display Min and Max luminance and the Max and Frame Average light level are output on a show-by-show basis (and are different).  In Kodi on the Shield TV it seems that if this data is present it is correctly passed through on a quick check where I compare Media Info and HD Fury Vertex reported values.  I have some files where Media Info confirms that the Min/Max luminance of the mastering display is flagged, but Max and Frame Average light level isn't, and in this case the mastering display metadata is passed through (I have files with 1000, 4000 and 10000 nit peak light level mastering flagged and these all pass through) but Max/Average light level is left at 0/0 - as reported by my HD Fury Vertex ***