T^2 TyMist [gd0138]

Ah yeus... something that, once again, looked fine in my mind but isn't geometrically viable, so the resultant solution looks bulkier than expected.

The issue stems from the final fresnels across the IPD range. Actually, I don't think I properly remembered how far from the front of my face the combiner actually sticks out.

Yeah this is a lot closer to what I was imagining.

Maybe a fusion of the 2 concept ideas could work out?

Yeah like increase the diameter so that there's ample amount of cusioning space at the back of the head, then add a thing at the top. Disregard the part where the 2 transparent parts collide for now...

Ok, the look doesn't outright fail. It looks like a royal guard from 2060. I probably wouldn't fit the decollimator in that front lip though. I guess if space was less of a limitation, it could look like this:

Ok now I'm getting somwhere that looks large but modern.

I actually like the tall hat more, but this does look sleeker to ride a bike with. It certainly looks more like something that would come out of a kickstarter and not a cosplay catalouge (do those exist?). 

This gives me an idea. Now that there's another window to handle the sides of my eyes, I could have a reflective panel that can fold away:

I think this could work, since it's got the same base design idea as #Tetent TestCut and #Tetent Timespy [gd0136].

Then add an LED ring...

I can already feel the crisp and fresh air just looking at this concept. The middle of the LED ring is dark because the combiner folds over it. The only shame is that I can't have that Thunderbird 5 idea where the PLDC covered cylindrical door slowly rotates around the background of the virtual monitors.

In related news, I'm computing a solution that looks past the TyMist. If the test rig works and is no more complicated to set up than levelling a 3D printer build plate, I might skip the TyMist and go straight to this idea.

Essentially, the Solynder is a name I recently thought up to call the helmet I've been thinking over for the past 24-48 hours. It'll vaguely be like #Tetent TestCut if the top face was a circle instead of a square. It's a cylinder for 3 reasons. 

1. I want the front door to be motorised open, I don't want any of my 180 degree horizontal field of view obstructed, and I don't want the door sticking out when open. 
   1. I remember seeing a 3D printer enclosure concept years ago but I can't find it now.
   2. The door is to keep as much of the same air in the helmet as possible. There would be air scrubbing and conditioning in the helmet, because like I'd rather eat on a washed plate instead of an unwashed one 30 restaurant customers before me ate on, I'd rather breathe some washed air and not the invisible mysteriums floating about. 
      1. At least with food, I only need to eat a couple times a day and the consumables go though some hydrochloric acid first before entering the bloodstream. Air? I think it's a direct absorption into the bloodstream and I have to breathe at least once every 10 seconds.
2. The helmet also includes over-ear audio. Looking at normal headsets, the front:back left:right head dimensions are roughly cylindrical already.
3. I want to be able to rest in bed with the helmet on. This means that the rear of the helmet has to follow a similar curve to the back of my head, or at least a continuous curve on the back of my head.

Now, as some future thinkers in the industry would be able to tell you, attempting to fit a lot of tech into something even remotely consumer-edition is somewhat far off. That's partially the reason of the helmet idea. I may look a bit closer to a futuristic Mojo Jojo than some 80's early adopter, but from my point of view, there's a lot of benefits and less of the drawbacks (and looking futuristic is probably better anyway).

Audio on the Solynder will remove the requirement from #Tetent [gd0090]. I also want to try (or at least research into) a 2 tweeter + subwoofer-ish configuration on each ear. The idea is that the tweeters are 120 degrees appart (with the main speaker inbetween them) and I could recreate 7.1 surround sound. 

Ideally, the door will be able to rotate a full 360 degrees. That means if it opened and went into the right side of the helmet, it could close from the left side and keep doing that. I think it'll look pretty cool having it slowly rotate like Thunderbird 5 whilst working.

I've used "y" instead of "i" in Solynder because I think it looks more obvious that the name was derived from "Cylinder". I guess I could go full Startup(TM) with the name with "Solyndr", a bit like "Savr" in SecondSavr.

Lastly, I'd look into wireless video methods. Having 0 cables, or just 1 when charging, sounds like a nice idea. My main lead is WiGig, but I'd take a 90Hz stream over WiFi too. WiGig supports DP 1.4, right?

[16:00]

I've been able to scramble up a replacement set of optics for this rather strong F-9.56 lens.

The replacement (compromised of 3 lenses) is another 2 D40 F-30's and a single D43 with a focal length of -50. Conviniently, the resultant image is about the same size, and the 3 lenses are in contact with each other so they probably could be thought of as 1 single lens assembly.

Using a F40 fresnel instead of F35 isn't an option, as the resulting beam is too large:

[16:45] Wait a second... maybe I don't have a solution, if that's the lens thickness:

I stand corrected. With the available options, I'd need an F40 fresnel lens, along with 2 F-30 lenses and an F-40:

The assembly length would be over 20mm longer. Fine for a test rig but not ideal for a final solution.

The nice thing is that the unique BOM count decreases:

[17:45] Unreal engine. If I swap the collimating lens so that it goes F-30, F-40, F-30, I can get the lenses back-to-back and the output looks to be a perfect 2:1 image!

Actually, it alternates if you think of the absolute value of the focal length: 40, -30, -40, -30, -40 -30, 40.

Shame I'd need a bigger lens though:

Below is what the default 4:3 monitor looks like in the visible area, along with the full 1:1 (dotted lines) and dual QHD (16:18 / 8:9) looks like.

The 4:3 is the only one I truly care about, and it fits so I may be able to use this optical layout. The dual 1440p was just a bonus extra idea, though I'd probably go lower anyway just so that I have some padding between monitors and space for icon pointers to other virtual monitors.

I found an actual DP multiplexer hub thing (that can extend screens not just mirror or redirect them) for the low price of <£30 and it uses a female port for display in, meaning that it could be possible to have it integrated into the system such that only 1 DP is needed from the perspective of the user. There's also a silver one with a cable pre-attached. 

Now, I've also been window shopping for mini PCs, laptops and the emerging x86 handhelds because I like the idea of SimulaVR but don't like the look of the price or the massive head turning angles between monitors, as well as the 35PPD.

Considering that I can easily see the text on the virtual in this image that only takes up a fraction of my entire screen, I could probably fit more content on 1 TyMist virtual monitor than 10 SimulaVR ones. Back in my day, I managed quite a bit with 1366x768 screens, so I could have 6 1410x940 (3:2) screens visible (with padding, 1440x960 without) visible with the 2880x2880 visible area.

Anyway, back to what I was saying. Finding something with dual HDMI is straightforward, but 1 or more DP is much harder to come by. I'm thinking of limiting my search to TB4 devices, because I just realised that the bandwidth exceeds that of an Nvme drive (and later that TB can route a direct PCIe connection) so it should be possible to connect a Windows drive and boot assuming the BIOS has the relevant features implemented. This would mean that I could have multiple PC's without any of the synchronisation issues. It might be a good thing I specifically looked for TB ports on #Teti [gd0022]'s motherboard. They're not as popular on desktops as I'd thought they'd be. If there are 2 TB ports and no DP, I could use that £11 adapter I also found to get me the video out I need.

I also want to keep tabs on Pimax's recently announced MiniStation, which is a WiGig enabled mini pc with battery:

I don't really know if I should put this log under #Teti [gd0022] or TyMist, but this discovery was made because I was looking at the TyMist screens so I've decided here. 

I was on a tangent and found https://www.displaymodule.com/collections/ar-vr where they had things like the datasheet to the 1440px (300nits) screen I plan to use. I was on this tangent because I was wondering why the screens cost as low as £15 each (see below for the 150nit version) but the driver board added another £100+ to the BOM. 

I do wonder if I should prioritise higher brightness+contrast or lower blur. I'm leaning onto the latter since the effective contrast on a HUD isn't that great anyway, the PLDC is supposed to diffuse bright sources of light behind the display and I might want to configure a game such that I can turn my head to look around (I hear that motion sickness is a thing that exists in VR and the low duty backlight is used to mitigate it). Additionally, I think a transparent virtual monitor gets a free pass regarding the amount of tolerable imperfections. Just think of how fuzzy, low resolution and severly blue tinted holograms are in Star Wars.

Anyway, those DisplayModule guys sell this interesting thing:

I was looking up and down the internet in 2020 for something like this, which takes power+usb and a display interface as an input and outputs an Alt Mode USB-C. It's what the Gigabyte Titan Ridge and Gigabyte Z590I Vision D does with the 3060Ti's DisplayPorts installed in Teti. If I search "HDMI to Type-C", I'd get nothing except "Type-C to HDMI" adapters.

A tad more searching revealed a chip that could be powering it: lontium LT6711A.

And that lead me to Delock. I haven't done a full search, but I have a feeling that if an adapter could ever exist, these guys have made it.

Just look at some exciting things I didn't even know were possible:

Something to power a 10 metre DisplayPort cable! I'm not sure if there's a USB-C one that also conviniently exists, but I think a >=5m cable for TyMist would be nice..

I've used USB3 to VGA or HDMI offerings, but they only went to FHD 60Hz.

This sounds really cool too. I wonder if they've got a DisplayPort version. Or perhaps I could get a USB version for #Tetent [gd0090].

Now I'm sure all these things are out of my league budget, but it's nice to know what's actually possible since I can't Google what I don't know exists. This excitement reminds me of the days looking though an Argos catalogue.I'm hoping that I can find out the specific IC's that run these kinds of things, because I'd ideally like to take a single DP cable and send the video feed across all 4 displays. Practically speaking, it's probably better to spend the £200 just so that I have something that exists. I'm already finding it tedious to route the PCB for #Tetrinsic [gd0041] and I've then got things like #Tetent [gd0090] and #Tetent Timespy [gd0136] PCB's to design.

Now that I've found out how to use the lens formula correctly, I've been able to achieve this ray path. The respective optical elements are about the same lengths as the previous path, which is nice. Another thing is that all the lenses are off-the-shelf. The factory has a MOQ of 10pcs for custom fresnels.

I have the choice of using a 50x50mm or 55x55mm fresnel, and I'm leaning onto using the 50mm as the output image is 46mm. Coincidentally enough, if the size of the image is unchanged, my eye needs to be 90mm away from the output fresnel.

The current design has it 110mm away, but that's an easy issue to fix:

For the 46.8mm output image, I'd need it to be 80mm away from the eye. If it's 90mm away, 100ppd would be achieved. From trying different values and seeing different screens, I can say that

• 70ppd looks pixelated and annoying. Must avoid.
• 80ppd is on the edge of looking pixelated. I'd like to avoid because I'm always going to be squinting, thinking "can I see pixels?".
• 90ppd is a nice balance of sharpness and Windows GUI at 100% scaling. 

So it seems that 100ppd would be ultrasharp if I extrapolated it, and that increments of 10ppd make a very noticable difference, so I'd rather not imagine what Nreal Air and its 50ppd looks like.

Anyway, the simulation mostly agrees with what I expect, other than the distance between the last two lenses being 20mm instead of 23mm

This results in a square size of 42.8mm. Not great, not terrible. At least the beam size is an easier 8.7mm instead of sub 3mm.

[23:45] I just cleared up a listing conflict with the first seller. The D32mm lens listing is only for 1pcs. I've found a different seller that has the same D40.5 F-30mm lens at a lower price (which is great because I need 6pcs) and a D42 F-40mm lens.

The lenses and beam path are awfully close, but remember that the beam modelled is for the diagonal of the square image and that the main desktop area requires a smaller usable diameter (in other words, won't be cut off).

Another good thing is that the square output image is 46.5mm, which I think is an ideal size for a 50x50mm fresnel lens.

I blame this confusion on the fact that I couldn't find it mentioned anywhere that the effective focal length is from an image of the same size, and not the first lens. It was all equations without labelled diagrams. The scale is 1mm = 10px and I've got a 52mm beam, F30mm converging and F-10mm diverging lens which is 22mm away. The first lens is at (0,0).

For Fusion360, it would be easier to define the resultant top ray as an angle, which would be

arctan(0.5*screen length / effective focal length)

Expectation:

Simulation I found:

Equation I used:

Now, the simulation isn't completely accurate. I can't change the grid size so I've multiplied all measurements by 10 and the lens is only 26 scaled mm instead of 27, but the difference is quite drastic between the formula and raytracing.

At least I'm correct in thinking that if there is a F-4mm lens 26mm away from a F30 lens (in other words, 4mm away from the F30's focal point, the light will be parallel:

However, I'm getting what looks to be F33 for what should be closer to 51!

I think... I think what's gone wrong is this assumption:

In terms of a physical fresnel lens, yeah sure maybe it's thin, but the fresnel is essentially approximating a lens that looks like this:

So what if I trace it backwards from the double concave lenses?

Okay... alright...

The beam is too large for the 32mm lens but let's not worry about that for now.

Well that failed. Oh right, the distance between lenses moved. Ok, what if I calculate a few permutations?

Ok now the two values match up after around 5 iterations... and the simulation still doesn't.

Okay... the focal point is about -6mm at a distance of 6mm... hmmm.

And the equation for thick lenses isn't even that different:

So I'm quite lost as to where the issue could be.

Well if I move the lenses so that I do get a parallel beam, I get approx F40 when the first lens is at 24mm from the fresnel. And the calculation I've got is about twice that. Coincidence? Yeah probably:

Effective focal length is only 48mm and 2*48 does not equal 120.

I found this webpage that explains a few things about fresnel lenses and lenses in general. The notable thing I took from that is that the recommended F number is between 0.5 and 1.5. From my understanding, the "diameter" of the lens would be the longest edge-to-edge length of a square/rectangular lens, which for a 52mm square image, is about 73.5mm. With a required focal length of 30mm, the F number is 0.41 if I'm calculating correctly. The smaller the F number, the smaller the "airy disk" for a given wavelength, which is a good thing. From what I can understand, the issues for using F numbers smaller than 0.5 don't apply to this application. Another thing is that the limit of spherical lenses is f/0.5 allegedly, meaning that the fresnel I plan to use is aspheric by nature.

I'm thinking of buying 1 set of lenses for a test, conviniently because this off-the-shelf fresnel exists:

I'd also be getting some 5mm right angled prisms and maybe a 9 or 10mm beamsplitter cube. I'd like to validate the optical path though some specialist software first, because it's still kind of pricey for a test run:

I dunno... a single test set is almost half the cost of the full order, partially because I get free shipping if I order a certain amount.