Using the rift with an RCA signal

[android78]

I see what you're saying, but wouldn't the increase in parallax ( shrinking of the world) also make it feel like you're going really slow?

[geekmaster]

I see what you're saying, but wouldn't the increase in parallax ( shrinking of the world) also make it feel like you're going really slow?

It would take the same amount of time to go between point A and point B, but they appear to be close, so you feel like you are travelling slower. If you perceive a slower speed, you have more time to react, so that could be a safety device, right?

[android78]

Lol. Nice idea, but I think I'll still be top ranking in lawn darts!

[geekmaster]

Awesome little experiment: I just took two of my fresnel magnifiers, and held them vertically side-by-side so the lens centers were about 2-inches apart. I held up a sheet of paper so that my CFL ceiling light was focused on it as a stereoscopic pair. Then I did a wide-eye merge (by looking through the paper while focusing on the paper). I saw a clear 3D image of my ceiling light, that appeared behind the paper. Because the lenses were closer together than my eyes, the virtual image appeared smaller than the real ceiling light.

Anyway, putting the lens centers close together would require a divider sticking up from the paper to keep the images from overlapping. The same would need to be done for a camera image sensor with two lenses.

My experiment proved that SBS lenses (with divider) can indeed create an SBS-Half stereoscopic image pair.

Now I want to try this hack with a pair of fisheye lenses, and a video camera (without lens) in place of the paper. The fisheye SBS-Half image should be Rift-ready.

EDIT: Same experiment, but this time with a pair of 6x magnifying reading glasses. Merged image pair was again perceived as 3D. This is a fun new waste of time.

[Grix]

So I've ordered the two video processors myself now, plus everything else required for FPV with OSD. So as soon as everything arrives I'll be able to report on how well this setup works in practice. I'm a post-kickstarter backer though, so it will probably be at least a month until the rift arrives.

[geekmaster]

So I've ordered the two video processors myself now, plus everything else required for FPV with OSD. So as soon as everything arrives I'll be able to report on how well this setup works in practice. I'm a post-kickstarter backer though, so it will probably be at least a month until the rift arrives.

I got my RCA-to-HDMI adapter a long time back, but Chinese New Year delayed the RCA-to-3D adapter, which just shipped last week (on a slow boat from China, via Singapore post). That 3D adapter takes in composite video from an RCA jack, and outputs 3D video in various formats, including SBS-Half. A fisheye backup camera feeding this should create Rift-compatible video, which is then upscaled to 720P HDMI with the RCA-to_HDMI adapter.

The 3D video will even be "synthetic 3D", which for photos from an FPV aircraft may actually look pretty good. No computer needed (other than any embedded processors in the adapters).

I still need to decide which OSD to get. Any ideas on that?

[nnamd]

Oh yeah! When I first saw the Rift i immediately had to think on those guys who are flying their RC Planes with a HMD. And I imagined how it would be too look around with the Rift from above.
(I can't find the video I was looking for but this is very similar:)
[youtube]http://www.youtube.com/watch?v=pNmRXss23R8[/youtube]

Unfortunately a nice a plane is way to expensive for me. (Budget for me for such a Project would me < 200$, excluding the Rift off course).

Reading this Thread I had to think on that nano Copter "Crazy Fly":
[youtube]http://www.youtube.com/watch?v=3WBUVYZkODI[/youtube]
@2.17 someone added a small Camera to it. But I don't think his hack supports live streaming(?)
You can control it by PC via a PS3 Controller but there will be an SDK

Do I see Dog fights in the future?

[Grix]

So I've ordered the two video processors myself now, plus everything else required for FPV with OSD. So as soon as everything arrives I'll be able to report on how well this setup works in practice. I'm a post-kickstarter backer though, so it will probably be at least a month until the rift arrives.

I got my RCA-to-HDMI adapter a long time back, but Chinese New Year delayed the RCA-to-3D adapter, which just shipped last week (on a slow boat from China, via Singapore post). That 3D adapter takes in composite video from an RCA jack, and outputs 3D video in various formats, including SBS-Half. A fisheye backup camera feeding this should create Rift-compatible video, which is then upscaled to 720P HDMI with the RCA-to_HDMI adapter.

The 3D video will even be "synthetic 3D", which for photos from an FPV aircraft may actually look pretty good. No computer needed (other than any embedded processors in the adapters).

I still need to decide which OSD to get. Any ideas on that?

I saw your post about receiving the components, I thought you had received both and verified that they worked all right. Bummer if it turns out there's a problem with the 3d adapter Anyway, are they different from the ones I linked earlier? I thought the 3d converter worked with a hdmi signal, not rca?

Regarding OSD: I have ordered a minimOSD board for my crius AIOP FC myself. I've heard good stuff about it. It's very flexible, everything can be configured. So if for example the OSD data is hard to read on the peripheral vision in the rift, you can move it closer to the center of the screen using the firmware config tool.

[geekmaster]

I saw your post about receiving the components, I thought you had received both and verified that they worked all right. Bummer if it turns out there's a problem with the 3d adapter Anyway, are they different from the ones I linked earlier? I thought the 3d converter worked with a hdmi signal, not rca?

You may be right on which box plugs into which, but no matter, the end results should be the same. I will have to go look it up again. I think mine were from your links. The 3D adapter does the same thing as what is already built into my 3D HDTV. It converts monoscopic video and converts it to various 3D formats (anaglyph, over/under, SBS-Half). The synthetic 3D in the TV works great for most stuff, but I do not use it other than just to test it now and then. The converter box should work as well. For the Rift, we need to set the RCA/HDMI converter to 720p (a switch on the side). For the 3D converter, we need to select SBS-Half. Unfortunately, although these would have worked great for the original 5.6-inch displays, I suspect the optical centers will not be close enough together for the new 7-inch Rifts. But it should work well for the Fov2Go or Hasbro My3D viewers that have smaller displays and can handle full overlap. Most 3D video sources (including 3D movies, these 3D adapters, and the Doom 3 BFG originally planned for the Rifts) are in fully-overlapped FoV. Everything changed when the number of Rift orders exceeded the number of 5.6-inch panels available on the planet, pushing us into new territory regarding video formats.

My TV lets me adjust the 3D disparity for partial overlap, so perhaps these adapter boxes will too. If not, perhaps "somebody" can add it by tweaking the 3D adapter firmware (one of my long-neglected skills).

Worst case, you can just cover one of the lenses in your Rift, until we learn how to adjust the 3D offsets.

[geekmaster]

I did some searching and I found these two thingies. If I run the signal through both of these, with the 3d converter set to side by side and lowest possible depth, would that work? Would there be any notable latency? Does anyone have experience with these sort of converters?

http://www.ebay.com/itm/280993783685
http://www.ebay.com/itm/170976138135

What I have so far (from your links above):

New Mini RCA AV CVBS to HDMI 720P /1080P Vdeo Converter Adapter
Price: US $32.99
Shipping: FREE Economy Shipping from outside US

The MINI AV to HDMI converter is a universal converter for analog composite input to HDMI 1080p (60HZ) output. The analog to digital conversion in this module employs 10 bits maximal 162MSPS sampling, black/white level expansion, color transition improvement, dynamic range expansion, blue stretch , auto-detect and auto-convert the composite signal to 1080p(60HZ) output. Making video come alive, delivering the sharpest, most realistic HD visuals available.

Features:
1. No need to install drivers, portable, flexible, plug and play.
2. Provide advanced signal processing with great precision, colors, resolutions, and details;
3. Support PAL, NTSC3.58, NTSC4.43, SECAM, PAL/M, PAL/N standard TV formats input.
4. Support HDMI 1080p or 720p output.

Specification:
Input ports: 1xRCA (Yellow, White, Red).
Output ports: 1xHDMI.
Composite input: PAL, NTSC3.58, NTSC4.43, SECAM, PAL/M, PAL/N.
HDMI output: 1080p/60Hz, 720p/60Hz.
Dimensions (mm): 66(D) x55(W) x20(H)
Wight (g): 40
Power Adapter
Input: AC100V-240V, 50/60Hz, 0.15A; Output: DC5V, 1A; Plug Type: US

Contains:
1x Mini CVBS to HDMI converter
1x Mini USB Cable
1x Power Adapter
User Manual

What I am waiting for (delayed by Chinese New Year):

Portable 2D to 3D Converter HDMI 1080P Switcher With Glasses For TV DVD PSP WII
Price: US $37.99
Shipping: $5.99 ePacket delivery from China
Overview: Portable and compact, this 3D converter makes the best 3D effect on your normal TV or normal projector. Just connect to your DVD player, set top box, PS3, etc, free your vision and enjoy the wonderful 3D world with the 3D converter and glasses.

Features: Convert SBS-H 3D format contents to amber-blue mode. Convert 2D contents to amber blue mode, with 6 levels fine tune. Convert 2D contents to SBS-H 3D format to work with 3D TV/projector, with 6 levels fine tune. HDMI 1.4a input and output, compatible with HDMI 1.3/1.2/1.1. Maximum resolution of 1920 x 1080P @60Hz. Controls by buttons and remoter. Portable and compact, easy and simply to use.
input device : TV Box ,Xbox 360 ,PS3,WII ,Projector ,tablet PC ,DVD which support HDMI output.
Output device : 2D HD TV ,3D HD TV which support HDMI Input.
Model: VT1I10-HD
Power supply: DC 5V±5% @ 1.5A
Power consumption: Max 4.5W
Input/Output port: HDMI 1.4a
Size: 124 x 78 x 13mm
Weight: 104g
Color: blue
Operating temperature: -10 to 80℃
Storage temperature: -20 to 85℃
Package size: 177 x 143 x 80mm

I see that amazon wants a bit more for this VT1I10-HD 3D converter box:
http://www.amazon.com/Converter-Glasses ... B00973IQZW

I am interested to see what the "6 levels of fine tune" is. I hope it is the same 3D disparity adjustment my 3D TV has, in which case it should work fine on the 7-inch Rift. We will know more when one of us gets our 3D converter and our Rift.

[Grix]

Looks like the 3d converter indeed has a setting for the 3D depth (I assume that's what the "6 level fine tune" is).

I guess the only two possible problems are 1) optical warping and 2) latency. Either way, we'll see when everything arrives.

[geekmaster]

Looks like the 3d converter indeed has a setting for the 3D depth (I assume that's what the "6 level fine tune" is).

I guess the only two possible problems are 1) optical warping and 2) latency. Either way, we'll see when everything arrives.

Actually, a fisheye camera lens should provide a reasonable amount of barrel distortion, which is exactly what the Rift needs. There are plenty of FPV videos with fisheye lenses and watching them does not seem to be a problem, so having the Rift partially compensate for that should be even better.

If the fine tune is 3D depth, what they does on my 3D TV is to shift one image to the right or left by an adjustable amount. The Rift really wants both images shifted toward the center though, so we will have to see how effective that adjustment is. We could possibly use custom firmware in the 3D convertor, if somebody wants to invest the time. I do not have spare time like I did when I was younger, so hopefully we have some other firmware hackers willing to tackle that if needed.

Another option may be to shift the image optically with offset lenses (perhaps from a fresnel lens stack). I will not know until I get my new toys and have time to play with them...

[geekmaster]

My 3D adapter came today. The "6 levels of adjustment" are not exactly what I would like. Instead of 6 different shifts inward or outward, some of the changes are almost imperceptible. It does have a setting that shifts inward noticeably (but not what the 7-inch display needs), and another that shifts outward noticeably. It also adjusts the bottom of the display only (pulling the bottom inward more that the top to make the bottom look closer in 3D). I suppose that makes the ground look more flat. It also has some nonlinear shifting where it squeezes the inner edges more than the outer edges, probably to simulate tilting the display edges.

Again, this should work fine with a 5.6-inch Rift clone, or smaller (like Fov2Go). But 7-inch really does not like fully-overlapped SBS-Half (typical 3D movie content, and the Doom 3 BFG warp from the early Rift demos). It may also work on a 7-inch Rift with offset lenses (which I plan to experiment with), to shift the images inward optically.

This adapter can also convert SBS-Half to mono (left only) or to anaglyph (it comes with Blue/Amber glasses, not Red/Blue).

It supports 1920x1200 (beyond TV 1080p). But it changes the black levels (probably using the common 16-235 brightness range instead of 0-255).

It has connectors (missing pins) for the I2C bus, and the EDID I2C bus (for display parameters). I suspect one of those is for the firmware serial ROM. If it is flashable, perhaps the firmware could be tweaked. But it could also be a cheaper OTP ROM, which cannot be changed (so no firmware updates). I do not have time to mess with that (using my bus pirates).

[Grix]

I'm not quite sure I follow 100%. Could you explain in simple terms why this would work well on a smaller display rift clone but not the 7 inch dev kit? Is there too much optical warping? Would a fish eye lens fix this?

And is it possible to turn off or set the 3d adjuster to a minimum to get two nearly identical images side by side?

Most importantly, do you mean that the adapter is not suitable at all with the 7' rift or is it just not optimal?

[geekmaster]

I'm not quite sure I follow 100%. Could you explain in simple terms why this would work well on a smaller display rift clone but not the 7 inch dev kit? Is there too much optical warping? Would a fish eye lens fix this?

And is it possible to turn off or set the 3d adjuster to a minimum to get two nearly identical images side by side?

Most importantly, do you mean that the adapter is not suitable at all with the 7' rift or is it just not optimal?

For a standard SBS-Half 3D image, the optical centers are too far apart for stereoscopic convergence. The centers need to be brought closer to your eye IPD, either optically, or electronically, or with software. Otherwise viewing the horizon will require diverging your eyes considerably beyond parallel. You could end up looking like Marty Feldman!:

[geekmaster]

duplicate post deleted.

[EdZ]

I'm not quite sure I follow 100%. Could you explain in simple terms why this would work well on a smaller display rift clone but not the 7 inch dev kit? Is there too much optical warping?

The 5.6" prototype has a symmetrical side-by-side view, which other than the warping (barrel distortion) is the same as the side-by-side view you would feed a 3DTV.
The 7" Rift, however, does not have a symmetrical view. The centre of each viewpoint will be closer to the centreline than the edge, as will the lens centres and thus the centre of the desired warp.
To use the 7" Rift with an off-the-shelf side-by-side generator, you'd need some way to pad the sides of the view with black bars ('windowboxing') after performing the SBS conversion. This might be possible if the Rift's display controller allows the image size to be modified, but this may be unlikely (for 'proper' use, anything other than pixel-perfect 1:1 is undesirable behavior).
You could also shift the camera lens position to be off-centre for each sensor, if you can adjust the mountings.

[geekmaster]

You could also shift the camera lens position to be off-centre for each sensor, if you can adjust the mountings.

Lens shift can be done in the camera, or in the HMD, or perhaps a little in both would be best. Lens offset causes asymmetrical warping, so smaller offsets are better unless you compensate for that asymmetrical warp. You get extra image stretching proportional to the distance from the lens optical center.

[EdZ]

Doing lens-shift in the Rift would be a Bad Idea. Not only would it make it incompatible with a computer (without tweaking the pre-warp shader), but it would produce a non-optimal arrangement of the optics. You want to have the lens centres aligned with the IPD and directly forward of the eyes (i.e. so when looking straight forward, you look directly through the centre of each lens). Any other arrangement means you are looking at significantly more stretched pixels in the most critical area of the image. Also, the warp would then also be asymmetric, so significantly harder to compensate for (the warp in the 7" Rift is symmetric but clipped).

[Grix]

Well poop, I missed that.. So the stock 3d adapter is useless? I doubt I have the skill to modify the firmware anytime soon.. Are there any cheap video processing chips that would be relatively easy to program an adapter on? I am using only one camera and I don't really want to risk breaking my rift so manual adjusting of the lenses aren't really possible..

[geekmaster]

Doing lens-shift in the Rift would be a Bad Idea. Not only would it make it incompatible with a computer (without tweaking the pre-warp shader), but it would produce a non-optimal arrangement of the optics. You want to have the lens centres aligned with the IPD and directly forward of the eyes (i.e. so when looking straight forward, you look directly through the centre of each lens). Any other arrangement means you are looking at significantly more stretched pixels in the most critical area of the image. Also, the warp would then also be asymmetric, so significantly harder to compensate for (the warp in the 7" Rift is symmetric but clipped).

Well poop, I missed that.. So the stock 3d adapter is useless? I doubt I have the skill to modify the firmware anytime soon.. Are there any cheap video processing chips that would be relatively easy to program an adapter on? I am using only one camera and I don't really want to risk breaking my rift so manual adjusting of the lenses aren't really possible..

Rift lenses can be adjusted by replacing the lenses in a set of lens cups. They come with three sets, so I many try modifying one of my sets when I get my rift. You can make an offset lens by grinding down a larger lens. For example, start with a lens twice the size you need, then grind of one half of it, and round off the corners to fit the lens cups. I need to see if the asymmetrical distortion would be acceptable for FPV use. You are correct that pulling the images inward would stretch the inner side unevenly, which is in the stereoptic region where it is least desired. This can be corrected by software, but not for this hardware-only approach (unless firmware can be patched).

I will come up with some sort of use for this device, but it is not as easy as I had hoped because the "6 levels of adjustment" are not as useful as the adjustments in my 3D TV synthetic 3D...

[Grix]

By the way, there are also other cheap 3d converters out there that might work better or be easier to hack, in case someone else wants to join the experimentation:

http://dx.com/p/mini-1080p-hdmi-2d-to-3 ... ver-152723
http://www.buyincoins.com/new_en/detail ... 12458.html
http://www.buyincoins.com/new_en/detail ... 12457.html

Btw, geekmaster, how soon do you think you will get your rift? Are you a kickstarter backer living in the US?

[android78]

@geekmaster. If the goal is to use this for fpv then I would think it easier to do the shift at the source. If you're good with dremel and hot melt glue you should be able to shift the sensor relative to lens there. This way you keep the distortion correct so it will better match the distortion in the rift.

[geekmaster]

@geekmaster. If the goal is to use this for fpv then I would think it easier to do the shift at the source. If you're good with dremel and hot melt glue you should be able to shift the sensor relative to lens there. This way you keep the distortion correct so it will better match the distortion in the rift.

The point is to take a single camera and make it display two images. The only optics that can do that are a stereoscopic lens, which may contain mirror or prisms to get two lenses focused side-by-side on the same image sensor. We only have one radio channel to send this image. Even monoscopic would be fine, converting it for the Rift later.

Ideally, a consumer Rift should be able to view monoscopic video sources, like other HMDs. But for these Dev Kits, we need to hack it somehow...

[geekmaster]

... Btw, geekmaster, how soon do you think you will get your rift? Are you a kickstarter backer living in the US?

Hopefully soon. Kickstarter backer in the USA.

All this impatience and anxiety flourishing in this forum will be long forgotten after we get out Rift face-plants.

[Grix]

When looking at demos of the 7", I don't quite see this problem.. It looks like the center of each image is indeed the same spot on both images, at least if what's displayed on the monitor is exactly what is displayed in the rift.. https://www.youtube.com/watch?v=m1785afLNFw for example.

If the centers are offset, it must only be by a tiny amount. Perhaps we would be able to focus decently even if the two images are identical? *crosses fingers*.

[geekmaster]

When looking at demos of the 7", I don't quite see this problem.. It looks like the center of each image is indeed the same spot on both images, at least if what's displayed on the monitor is exactly what is displayed in the rift.. https://www.youtube.com/watch?v=m1785afLNFw for example.

If the centers are offset, it must only be by a tiny amount. Perhaps we would be able to focus decently even if the two images are identical? *crosses fingers*.

That video shows the "gaffer taped" 5-inch display model that the user puts on his face...

I plan to make some offset lenses to replace in one of my lens cup sets, that should (hopefully) shift the images inward enough for FPV use without a computer.

[Grix]

You're right.. Why would they use the old prototype for demo when they've got a new dev kit just lying there though? Makes me suspect the prototype is superior somehow.

How do you plan to make offset lenses? Doesn't that require expensive equipment?

Also, I just realized that if nothing works out, although awkward, it should work to just use it with one eye closed, right?

[geekmaster]

How do you plan to make offset lenses? Doesn't that require expensive equipment?

Also, I just realized that if nothing works out, although awkward, it should work to just use it with one eye closed, right?

I mention in a previous post the "one eye closed" suggestion. Yes, either close an eye or cover one lens.

Regarding offset lenses, just start with a lens with the same magnification and focal length, but a lot larger in diameter, and grind off one side. After you round it off, you mount it with the thin edge next to your nose.

You can experiment with a fresnel page magnifier, which is the equivalent of a 6-inch diameter lens. Look though one side, and you will see it shift the image. When held near an object to view.

For normal use, if you cut fresnels for the eyecups, you want the ridges away from your eye.

But for my lens stack experiments, I wanted the ridges TOWARD the eye to cause additional distorion and image wrapping into the peripheral view. For FPV, you do not want extra distortion. With a Rift, you may need to stack 2 or 3 layers of page magnifiers to get enough magnification.

For quality, grind down a solid acrylic lens, such as the 2-inch 5x lenses recommended for DIY Rifts. In fact, without cutting or grinding those lenses, to hold them very close to the eye they MUST be offset or your nose gets in the way. If you hold them with the lens frames against the sides of your nose, the lens center is outward from your eye center, and the offset pulls the images inward (exactly like we want).

While holding a pair of those lenses close to my eyes (so my eyelashes brushed them), they did allow me to easily merge OLD-STYLE image pairs from the 5-inch pre-warp while viewing SBS on my 7-inch Nexus 7 PDA display. They would work the same if the outer portion not being used would be removed from the lens, making them small enough to insert into Rift Dev Kit eye cups. Rotation of the offset lens would be important (thin side toward nose). The image gets pushed in the direction of the offset from optical center.

Offset lenses are commonly used in high-end video projectors, allowing the projected image to be pushed down, so the projector can be ceiling mounted without pointing downward. That gives a better image with more pixels, because keystone (trapezoidal) correction is not needed for an offset image.

[fireslayer26]

Geekmaster- Keep us updated on this. I am really interested in seeing how this turns out. I really want to use the Rift for FPV as well, but am worried as to how much latency it adds to the live feed. Too much and flying with it would be unsafe.

[geekmaster]

Geekmaster- Keep us updated on this. I am really interested in seeing how this turns out. I really want to use the Rift for FPV as well, but am worried as to how much latency it adds to the live feed. Too much and flying with it would be unsafe.

A long as you run the LCD in its native resolution (1280x800) [hopefully 720p runs natively], there should be very little additional latency.

[geekmaster]

Sad news: even though the Rift will display any resolution from my computer video card HDMI DVI output (with Rift DVI/HDMI adapter), it will NOT display anything output from the 3D converter HDMI output. The rift just displays a message saying the HDMI is out of range... Oh, wait! For some reason, while trying it for about the sixth time, to get the exact error message wording, it WORKS!

Hmm... Go figure! Now to try the converter settings...

But still, sad news. I see left and right images that would work on fully-overlapped displays like the old 5.6-inch Rift prototype, but for this new 7-inch partially-overlapped Rift, the 3D converter only adjusts the convergence of the two images about 10% of what it needs for the Rift.

The converter puts out standard SBS-Half content, with NOT ENOUGH adjustment to allow stereoscopic convergence on the Rift UNLESS we use custom lens cups with offset lenses. I plan to make just such offset lenses, which SHOULD work with this converter. I will post the results here.

Another slight annoyance is that the Rift lenses are held in place with snap rings, and I was told in the past that I should have no problem re-using existing lens cups for my experiments, but somebody decided to GLUE my snap-in lens holder rings, and I do not want to risk damaging my lenses trying to remove that glue. I suppose I will need to make my own lens cups...

[fireslayer26]

Well, it doesn't look like its going to be a simple solution for FPV with the Rift. Guess I'll have to get a pair of FatShark's as well.....

[Grix]

Damn.. Does it work to cover one eye though? Is it really distracting or do you think it would be doable to fly FPV like that until the consumer version comes out?

Also, I still don't quite understand how you are going to modify the lenses. Grind them you say, but with what? Doesn't that pretty much ruin the glossiness and transparency?
If you manage to make it work, do you think you would be willing to modify a pair of my lenses as well, for a price of course?

[geekmaster]

Damn.. Does it work to cover one eye though? Is it really distracting or do you think it would be doable to fly FPV like that until the consumer version comes out?

Also, I still don't quite understand how you are going to modify the lenses. Grind them you say, but with what? Doesn't that pretty much ruin the glossiness and transparency?
If you manage to make it work, do you think you would be willing to modify a pair of my lenses as well, for a price of course?

Yes covering one eye would work fine, like using a monocular HMD. A set of mirrors or other stereoscopic lenses on the FPV camera might also work (if they can be adjusted for partial overlap), so the 3D converter box would not be needed.

I am thinking the RIGHT solution may be to use a Raspberry Pi with a USB video input device, and let it do full Rift pre-warp correction and partially overlapped images, then output to the Rift over its HDMI port. It could even read the Rift IMU to control FPV camera pan/tilt with head tracking data.

[fireslayer26]

[quote="geekmaster"I am thinking the RIGHT solution may be to use a Raspberry Pi with a USB video input device, and let it do full Rift pre-warp correction and partially overlapped images, then output to the Rift over its HDMI port. It could even read the Rift IMU to control FPV camera pan/tilt with head tracking data.[/quote]

Awesome, let me know when you finish!

[gearwolf]

Would a camera that outputted side by side video with individual cameras be better than trying to use a mono camera that is fake 3D? There is a camera already out there for 300 dollars that can do that for fpv but I think this one might be better. http://www.rcgroups.com/forums/showthread.php?t=1802511 . What do you guys think? I've got a rift on the way and also thought this would be great for a large fov hmd or 3d fpv. I was hoping that it wouldn't have to lead to pass through/post processing on an external device.

[geekmaster]

Would a camera that outputted side by side video with individual cameras be better than trying to use a mono camera that is fake 3D? There is a camera already out there for 300 dollars that can do that for fpv but I think this one might be better. http://www.rcgroups.com/forums/showthread.php?t=1802511 . What do you guys think? I've got a rift on the way and also thought this would be great for a large fov hmd or 3d fpv. I was hoping that it wouldn't have to lead to pass through/post processing on an external device.

A simple way is to use a pair of side-by-side lenses focusing into the same camera. Or use mirrors. With electronics and two cameras, they need to be genlocked (sync pulses occuring at the same time), and a simple timing controller (arduino, or even a PIC chip) that switches the analog video halfway across the frame. One half of each camera view gets combined into the video output, and you turn the cameras a little inward so the center of the VISIBLE portion is pointing forward.

To play the video fullscreen in the Rift, the camera centers would have to be closer together.

For cameras farther apart like were used here, you need to get the centers closer together to view them in the Rift. I just viewed some of those YouTube 3D videos in SBS, by playing them in a window and shrinking the size so the image pair was easy to view in 3D on my Rift. A benefit is that all of the video was visible, viewed as rectangles with corners stretched out, as if projected onto the inside of a spherical movie screen (like a theater dome).

It worked quite well. Because NTSC composite (RCA) video is only 360-pixels wide anyway (720 needs S-Video), displaying it in the central portion of the Rift does not lose any of its resolution, making it as good as it gets. PAL video would have similar limitations, but would provide slightly larger images at lower frequency.

You can use one camera with an optical 3D splitter, or two genlocked cameras with an electronic switcher that switches in the middle of each scan line (a VERY simple circuit). And for Rift-compatible video, just place the camera close enough together and pointing in the right directions (i.e. align your cameras to match the Rift). Using some degree of fisheye lenses in the cameras would AUTOMATICALLY pre-warp them just as the Rift needs (i.e. barrel distortion).

[gearwolf]

Would a camera that outputted side by side video with individual cameras be better than trying to use a mono camera that is fake 3D? There is a camera already out there for 300 dollars that can do that for fpv but I think this one might be better. http://www.rcgroups.com/forums/showthread.php?t=1802511 . What do you guys think? I've got a rift on the way and also thought this would be great for a large fov hmd or 3d fpv. I was hoping that it wouldn't have to lead to pass through/post processing on an external device.

A simple way is to use a pair of side-by-side lenses focusing into the same camera. Or use mirrors. With electronics and two cameras, they need to be genlocked (sync pulses occuring at the same time), and a simple timing controller (arduino, or even a PIC chip) that switches the analog video halfway across the frame. One half of each camera view gets combined into the video output, and you turn the cameras a little inward so the center of the VISIBLE portion is pointing forward.

To play the video fullscreen in the Rift, the camera centers would have to be closer together.

For cameras farther apart like were used here, you need to get the centers closer together to view them in the Rift. I just viewed some of those YouTube 3D videos in SBS, by playing them in a window and shrinking the size so the image pair was easy to view in 3D on my Rift. A benefit is that all of the video was visible, viewed as rectangles with corners stretched out, as if projected onto the inside of a spherical movie screen (like a theater dome).

It worked quite well. Because NTSC composite (RCA) video is only 360-pixels wide anyway (720 needs S-Video), displaying it in the central portion of the Rift does not lose any of its resolution, making it as good as it gets. PAL video would have similar limitations, but would provide slightly larger images at lower frequency.

You can use one camera with an optical 3D splitter, or two genlocked cameras with an electronic switcher that switches in the middle of each scan line (a VERY simple circuit). And for Rift-compatible video, just place the camera close enough together and pointing in the right directions (i.e. align your cameras to match the Rift). Using some degree of fisheye lenses in the cameras would AUTOMATICALLY pre-warp them just as the Rift needs (i.e. barrel distortion).

Ah. I'm guessing the live feed would be similar unless you used s-video. If it's the same windowed effect, then it would be fairly similar to the current set of head sets for fpv as they look like watching a 40 inch tv at 6 feet away minus 3D (something like that). I guess it's back to using a rpi, which I guess isn't a bad thing considering most will use a ground station if their goggles don't have a built in receiver. geekmaster maybe you could chat with the camera developer to see if it could be a solution as I am not knowledge able on the subject of 3d (want to learn but very behind).

[geekmaster]

Ah. I'm guessing the live feed would be similar unless you used s-video. If it's the same windowed effect, then it would be fairly similar to the current set of head sets for fpv as they look like watching a 40 inch tv at 6 feet away minus 3D (something like that). I guess it's back to using a rpi, which I guess isn't a bad thing considering most will use a ground station if their goggles don't have a built in receiver. geekmaster maybe you could chat with the camera developer to see if it could be a solution as I am not knowledge able on the subject of 3d (want to learn but very behind).

Moving the cameras closer together would reduce 3D parallax, and for flying you actually want WIDER parallax so distant things also look 3D. We could slide the right camera image center to the left electronically, by inserting a delay in the genlock sync pulse driving the right camera (horizontal centering). A simple R/C delay circuit should do the job. The idea is to keep wide parallax but move the electronic image centers closer together to make them Rift compatible.

I think a camera on each wing would be great (except when near stuff, where you would get 3D window violation errors). A window violation is where a 3D image extends out of the screen, but is clipped by a screen edge. Not a problem in the Rift, except most people cannot comfortably converge on stuff that is touching (or inside) their nose...

Wide camera separation would make the world below look like a scale model that you tower over, like a toy train set. But GREAT 3D, as long as you do not get too close.

Having cameras too close together makes everything look HUGE. I played some VR Quake in my Rift with a virtual camera separation that was much too small, and the gun I was holding looked like a freaking railroad locomotive! I was told that densohax will github VR Quake when ready. That and a lot of other things are correct now, but until we get the FoV adjustments correct, it is a bit of a stomach churner in the Rift. But still AWESOME!

And fully immersive huge FoV FPV flying will be even better, if we use stereoscopic cameras with fisheye lenses, and post-processing to make the video Rift-ready. Knowing that your expensive FPV gear is at risk makes it more "real" than a video game like VR Quake can ever be (unless you have some real bitcoin at risk in a multi-player deathmatch).


Most FPV video is analog, which uses composite video (i.e. "RCA" video) with limited resolution. To transmit S-Video (or HD) you would need some kind of digital video transmission, and then you have extra FCC rules that may get in the way. In the USA, they want video in NTSC format when broadcast on the 2.4GHz amateur bands. But they do allow high-power wifi for licensed amateur radio use, on certain wifi channels, so that may support digital video. (Of course, I have an amateur radio license, and a high-power wifi transmitter, and a high power 2.4GHz analog video transmitter).