RifTUP! FullHD Oculus DK1 upgrade kit

[Hornet]

Very good ! I look forward for it !

[ripcurl123]

Cant wait to see a diy screen upgrade ,its been a long time in the making
who ever gets there rift upgrade first if you can think about posting maybe video or pics would be great
we need lots of feedback so as to create a good vr experience..

dave

[rozsnyo]

So the design of the base is done. It will re-use the existing breakout box, keeping all of its functions (switchable HDMI or DVI) and adding these:
- RGB status diode
- jumper to connect USB - DC jack's 5V rails (not yet designed in)

The cable ground/power/signal pins are wired the same way as in original, except that I use 5V on both power rails, so combining the old/new pieces is highly discouraged! Maybe I redo a bit the pinout to keep the power in short when somebody combines accidentally the pieces.

It is so simple that we could offer DIY kits for people with soldering skills

[mcdohl]

Could you say how much it will cost and the characteristics of the screen (Hertz and resolution)?
I know several people from different forums that also be very interested

[mr.uu]

Hi Daniel,
As far as i understood you use two converter chips for realising the HDMI-2-MIPI board. First HDMI-2-VGA and then VGA-2-MIPI.
If this is still the case, is there any possibility to use an DVI-I connector instead if the DVI-D and reroute the VGA pins through the old LVDS cable directly to the VGA-2-MIPI chip not using the HDMI-2-VGA chip? That way the "RiftUP" would even be ready to use VGA as input, which would be a great feature which even the RiftDK didn't have...
Just a thought and of course only to consider IF
- the extra routing and firmware programming is possible and not a huge extra work,
- it is even possible to use the old cabling for VGA signals,
- there is enough demand for such a feature.

What do others think?

[mr.uu]

Could you say how much it will cost and the characteristics of the screen (Hertz and resolution)?
I know several people from different forums that also be very interested

Most of the information you ask for is easily found in the last three pages. But i do summarize it for you:
- so far there are 5",5.5" and 5.9" panels to choose from,
- all these panels have FHD 1920x1080 resolution
- all these panels do support at least 60Hz, as the RiftDK did.
If there is a possibility to drive them at 72Hz (the max. of the DK) or more,
- is very likely different per panel type
- will be checked either from Daniel or latest from the community...
I.m.h.o. cost will be very likely decided after the final design and the cost for production is clear to him. But Daniel did ask everybody who is interested to send him an email with a suggested price in a post on the previous page (22).

[rozsnyo]

Hi Daniel,
As far as i understood you use two converter chips for realising the HDMI-2-MIPI board. First HDMI-2-VGA and then VGA-2-MIPI.
If this is still the case, is there any possibility to use an DVI-I connector instead if the DVI-D and reroute the VGA pins through the old LVDS cable directly to the VGA-2-MIPI chip not using the HDMI-2-VGA chip?

I am using HDMI -> 24bit parallel -> MIPI. There is no VGA/analog, where have you get that information? My circuitry is all digital. I think pushing 1080p over analog path is a little odd these days. The chip in the base is not only a passive switch but also a signal amplifier, so users wont run into trouble with having long cable + the "not for hdmi" specified rift cable. DP would make some sense (1080p can run on 1+2 pairs, not like on 1+4 pairs as in case of hdmi), but its advantage is minimal, as nearly all DP source ports are actually DP++ and can output HDMI as well.

[mr.uu]

I am using HDMI -> 24bit parallel -> MIPI. There is no VGA/analog, where have you get that information? My circuitry is all digital. I think pushing 1080p over analog path is a little odd these days. The chip in the base is not only a passive switch but also a signal amplifier, so users wont run into trouble with having long cable + the "not for hdmi" specified rift cable. DP would make some sense (1080p can run on 1+2 pairs, not like on 1+4 pairs as in case of hdmi), but its advantage is minimal, as nearly all DP source ports are actually DP++ and can output HDMI as well.

Thank you for clearing that up. It was definately me mixing up information from possibly earliear this thread or from the few other threads about an upgrade path for the rift. There did somebody state that he would realize it with two converter chips. Sorry for that.
I do drive my Xbox360 via VGA at 1080p and i am happy with the quality (and the little bit "softer" picture). Not to bad at all but maybe it is another story if you look through a magnifying lens 5cm close at analog jitter .

Your design is not usable for a dual setup, right? Or would it be possible to use the HDMI and the DVI as seperate inputs for driving a dual setup? Then it would require a different "inrift" setup to connect to the two driver boards. But as i realized in my own Frankenrift fun, there are some not so trivial modifications on the housing necessary (not even to mention the portrait or landscape options) to fit two panels, which is the way only few users might go. So i think that a dual setup will be a later target?

EDIT for clearity and typos...

[3dvison]

Hi rozsnyo,
Do you have a new updated timeline for when you can start taking and shipping orders ?

[mr.uu]

Btw, Daniel, did you think about

- contact Oculus about your upgrade kit. Maybe they would even feature it on their website? Will boost your sales hundredfold. Why they would do that? Well, first your kit is not a threat for them because the Rift will feature both positional tracking and low persistence, which they know is what we all want, too. And very likely an even higher resolution. And it would probably give them a boost on the RiftDK sales and help them to clear their stock for the Rift. Hey, they sold 50.000+ RiftDKs. They know, that it could be a small nice extra to offer an upgrade, but they do know, that they have to look forward only and focus all their resources for the Rift development. Which brings me to the rumored second DK. It looks like, as if they will release the Rift in Q4 this year (Cristmas). So for them it makes no sense to ship a second DK, if the only new features are the one we know. For me it would make absolutely sense to integrate stereo cameras in the housing, too. But they have now big inverstors. And they want revenue. So they will drop the DK2 which is necessary only for more groundbreaking features than what Crystal Cove has. So they will stick with their RiftDK till the end of this year. They even say, that everything what you design now for the RiftDK will work perfectly fine and even better with the Crystal Cove prototype. So maybe they use your RiftUp and add a new front housing with leds and a camera and sell it as an upgrade for the RiftDK? At least the one more important, different feature for the programmer would be offered (positional tracking). So they would have a DK1.5 and would not need to sacrifice loads of resources for it.
Maybe it is all a big brainfart of mine - but maybe not. The worst case for you would be that they do not react, which means, that you can only win .

- regarding dual landscape setup. Contact the person behind the InfinitEye, because what you are doing is the same what he might already is working on (heard that he founded a company) and he is probably interested in your work. Just a thought...

[rozsnyo]

I have just noticed that the new LG G Pro 2 has also a 5.9" FHD screen. Any replacement screens for it ?

I have sent the breakout board to a prototype service. Now its time to design the board for the headset - with 5.9" screen. If everything goes as planned, by end of next week we will have a prototype - will be tested by forum user Kernel32, who is also in Prague.

[stevetb]

Email sent, very interested in becoming one of your testers / early adopters.

-Stevetb

[remosito]

I have just noticed that the new LG G Pro 2 has also a 5.9" FHD screen. Any replacement screens for it ?

I have sent the breakout board to a prototype service. Now its time to design the board for the headset - with 5.9" screen. If everything goes as planned, by end of next week we will have a prototype - will be tested by forum user Kernel32, who is also in Prague.

I guess the chips you use can't go beyond FHD?

Oppo just announced a new smartphone. One version is supposed to sport a 1440p screen: http://www.gsmarena.com/oppo_find_7_wil ... s-7800.php
(I don't think worth waiting/delaying for, just wondering if your board would have the oompf to drive such a display once it's out (neglecting all the other unknowns like connector,signalling,...)

[rozsnyo]

We can always make a new board/modification depeding on real availability of screens and documentation. The >FHD panels are usually DUAL MIPI DSI, so it is not possible to drive them with the current hardware. Think of having a dual-link dvi monitor and you just got a single link or hdmi I have solutions prepared for such screens - it is based on the hdmi 1.4 splitter for the dual screen setup.

[kavanagao]

i just found this: VIVO XPLAY 3S PHONE'S 2560X1440 ,6 INCH screen ON MARKET:http://item.taobao.com/item.htm?spm=a21 ... =&sku=&ug=
screen price is chinese¥299=50$USD ;
, but I dont know if it is real


screen of this phone:http://list.taobao.com/itemlist/default ... .1000187.1

sorry for my english

[cgp44]

fhd panels are dual dsi mipi???
what do you mean? 4 lanes, 8 lanes??

[rozsnyo]

fhd panels are dual dsi mipi???
what do you mean? 4 lanes, 8 lanes??

Anything above FHD is Dual DSI. According to RENESAS documentation and roadmap, who makes the DSI receiver / Tcon chips for these screens.

[cgp44]

yes I recently saw 8 lanes mentioned

a quick google on RENESAS dual dsi


http://www.rsp.renesas.com/en/news/2013 ... 131018.htm

[cgp44]

mipi cells.JPG

the above picture is somewhat misrepresentative ...

Device utilization summary:

PIO (prelim) 18+4(JTAG)/336 7% used
18+4(JTAG)/115 19% bonded
IOLOGIC 5/336 1% used <=====

SLICE 110/3432 3% used <=====

CLKDIV 1/4 25% used
PLL 1/2 50% used
ECLKSYNC 2/4 50% used

this is the placement of the mipi 4lane reference IP
placed on a lattice MachXO2-7000HE device. As seen
there is plenty of room.

I have ordered the machXO2 breakout board, which has
good straight leads to 110 io pins.

If you are interested to order this board I suggest you get
a group of local fellows together and order more than one @26usd!!!
yet a 60usd handling and postal wacko (for a 3x3 inch board).

I have just done a quick learn of both Xilinx and lattice fpga development
environments. They are not too bad to get familiar with, disregarding
knowing only a bit of the technogy behind these fast signals, clocks, plls, dcms
etc.

The lattice IP is open source for their devices and with most of the IP black-boxed.

It is the first open IP ready available for DIYers.

links:
http://www.latticesemi.com/en/Products/ ... Board.aspx
http://www.latticesemi.com/en/Products/ ... DSItx.aspx
http://eewiki.net/display/LOGIC/Lattice ... d+Tutorial

[cgp44]

It really looks like we can get dual head VR
with a $26us fpga board and two x $50 fhd 5.5 inchers.

[ExZero16]

I'm on my phone so didn't do too much checking but wanted to post this before I lost the link
http://m.alibaba.com/product/1283576404 ... _inch.html
7" 1080p lvds

[cegli]

I'm on my phone so didn't do too much checking but wanted to post this before I lost the link
http://m.alibaba.com/product/1283576404 ... _inch.html
7" 1080p lvds

That's the same screen that everyone claims to have, but no one actually has. I looked around for it quite a bit in December, and every seller I contacted said they were not available. Seems like everyone is moving on to MIPI now.

[tbrotte]

So the design of the base is done. It will re-use the existing breakout box, keeping all of its functions (switchable HDMI or DVI) and adding these:
- RGB status diode
- jumper to connect USB - DC jack's 5V rails (not yet designed in)

The cable ground/power/signal pins are wired the same way as in original, except that I use 5V on both power rails, so combining the old/new pieces is highly discouraged! Maybe I redo a bit the pinout to keep the power in short when somebody combines accidentally the pieces.

It is so simple that we could offer DIY kits for people with soldering skills

Missing ESD protection and current limiters for usb etc. looks like a really basic amateur layout. Still need someone with definite fpga/cpld experience ?

[3dvison]

Missing ESD protection and current limiters for usb etc. looks like a really basic amateur layout. Still need someone with definite fpga/cpld experience ?

Wow..Thats nice of you, and that's only your first post. Can't wait for more ideas..

Everyone talks, but rozsnyo is the only one that does anything.
We needed this six months ago, not six months from now.
The Dev.kit-2 and CV-1 are on the way.
We need someone like rozsnyo that will do the work not talk about it.

PS..This is not rozsnyo first Rodeo.

[cgp44]

hey don't scare him/her off. we need people with fpga skills.
there is a good chance that my fpga fumblings will fail
due to some novice error.

[OzOnE2k10]

@3dvision - I tend to agree with you - saying "looks like a really basic amateur layout" is rather rude...

@tbrotte - You may have some valuable input into this project though, so let's hear what you can do first.

Also, I'm sure Daniel is well versed in this stuff, and looks to me like there are ESD diodes on the HDMI / DVI inputs. :p
I can't see anything wrong with that design at all. All the differential pairs are in place, and it's designed to fit into the original box.


I have been trying to help Daniel with this effort as much as I can with limited time and budget.
But, I've kind of screwed up my finances atm, so won't be able to buy any more panels or parts for a few weeks.

@Daniel, apologies for not being in contact recently, or keeping up with the forum posts.
We're looking to move house very soon, so it's been a bit hectic here boxing stuff up and sorting everything out.

I'm not sure that I'll have a great deal of time nor money to help with this for a while.
Really loving the design of the new boards though.

Just to confirm though - this is still HDMI / DVI passed from the base board through to the headset, correct?
Sounds great to me. It means a longer cable can be used.


@cgp44 - I wish I could help somewhat with FPGA coding, but what you're looking into is WAY advanced of my skill level to be fair...

I just spent a fair bit of money on an HDMI output board for retro consoles, but got quite stuck now with figuring out how to scale / deinterlace things without onboard SRAM / SDRAM (even then, I'd struggle).

Anything you can do towards cracking this MIPI thing and making it more generic / accessible will be a big help though, so we'll be watching your progress with a keen eye.

btw, Is the MIPI reference IP free on the Lattice chips, or will it require some "creative coding"?

OzOnE.

[rozsnyo]

Missing ESD protection and current limiters for usb etc. looks like a really basic amateur layout. Still need someone with definite fpga/cpld experience ?

ESD is there for HDMI and DVI. Look carefully. Current limiter is always on USB HOST side, not on USB device side. One can argue if an ESD for usb is missing or not - I say it is not missing, as the USB on this board connects to a cable, on whose end the ESD is. Do we really want to protect a piece of cable from ESD??

If you still need to make a fool of yourself, you may reply

[rozsnyo]

@Daniel, apologies for not being in contact recently, or keeping up with the forum posts.
We're looking to move house very soon, so it's been a bit hectic here boxing stuff up and sorting everything out.

No problem with that. I know you do your stuff, as I do my stuff. People who do something are usually busy... but I would say we are super-busy.

Just to confirm though - this is still HDMI / DVI passed from the base board through to the headset, correct?
Sounds great to me. It means a longer cable can be used.

Yes. The switch is actually an active re-driver (receiver, equalizer, transmitter) and should be capable of processing up to 3G signals - which means that the 297MHz HDMI (for dual FHD) is still usable with that breakout board design.

[3dvison]

Just to confirm though - this is still HDMI / DVI passed from the base board through to the headset, correct?
Sounds great to me. It means a longer cable can be used.

Yes. The switch is actually an active re-driver (receiver, equalizer, transmitter) and should be capable of processing up to 3G signals - which means that the 297MHz HDMI (for dual FHD) is still usable with that breakout board design.

That does sound great.
So I could keep the rift control box near my computer with Power, USB and Video plugged into it. Then I could use a long,HDMI extension cable connected to the RIFT-UP HDMI cable running to the Rift headset. With wireless headphones that HDMI extension cable would be the only cord going to the Rift. Leaving me with plenty of cord to move around if I want to...NICE

OH, Wait. EDIT:
Your RIFT-UP HDMI cable leaving the Rift control box would be a permanent connection right ?
If so, no big deal, but make it as long as you can...(SMILE)

[rozsnyo]

@3dvision - not as you imagined. The breakout box uses the original DK1 cable from it to the headset, so unless you are willing to do mechanical modification (or fine soldering / crimping), you are stuck at length between head and box which there was originally.

I decided to use the original cable as I do not want people to drill or break the DK1 - just swap boards for mine.

[3dvison]

@3dvision - not as you imagined. The breakout box uses the original DK1 cable from it to the headset, so unless you are willing to do mechanical modification (or fine soldering / crimping), you are stuck at length between head and box which there was originally.

I decided to use the original cable as I do not want people to drill or break the DK1 - just swap boards for mine.

No that's more than good.
Have you seen the Valve Home Streaming thread ?
Sounds like thats the way to go for wireless, so no problem with your cord length.

Now get back to work you HACK...(BIG SMILE)

PS.
Hey rozsnyo, it looks like it's near that time, for your RIFT-UP to have it's own thread, don't you think ?
When you do, make sure you link it here, so we don't miss it.

STEAM WIRELESS THREAD:
http://www.mtbs3d.com/phpbb/viewtopic.p ... 33#p148333

[cgp44]

the lattice mipi IP (for intellectual property: read desperate electronic engineers seeking a business plan)
is about 200 lines of Verilog code, which is a simple wrapping of the black boxed guts
is available free of charge to do want you want, as long as you maintain the usual
copyright. Why they blackbox the guts is probably due to the original IP holder and the mipi
organisation. It is a pity that mipi org don't just open up a very successful technology that has
all the bigboy mobile phone players wrapped up, but is now a vital technology for second uses.
There will be heaps of activity once mipi is fully open sourced. I expect maker magazine to
feature panels for maker electronics- think about the touch screen interfaces as diverse button
boxes, not just VR. Once maker get into fpgas it will be all on just like the arduino revolution.

The lattice mipi IP is advertised as able to do 1080p. We will see.

I am currently researching the Xilinx xapp460 notes on hdmi for spartan3 no serdes,
and the lattice ecp3 serdes (serial deserialization) hard code-
alas the machXO2 chip does not. As well we have local boy hamsters fantastic fpga site with needed explanations for the many novices like me who realise the immense power of recent (2009 on) fpga devices.

I've noted some recent mention of edid data. The full handling of that is trivial and all in the
various sources. You have the rom or a address/byte case and a i2c slave, and various triggers
for plug disconnects. I'll have a script that just fills the Verilog code with the 128 byte edid data.

[cgp44]

hey rozsnyo,

you will use the dk1 cable?

You mean your video pipeline is hdmi -> rgb -> lvds (ie the occulus dk1 controller box) -> rgb -> mipi ?????

I presumed your pcb would be hdmi -> rgb -> mipi straight, no dk1 cable, just a hdmi cable.
Also a usb cable for the motion sensor.

[OzOnE2k10]

Yeah, I hate it when these companies hide all the details to a protocol, and so locking out all but the big guys from using the latest stuff easily.

It's sounding like most small high-res screens will be using MIPI from now on, so good to know the IP can be used.

Ahh, the hamsters site - guess where I got that HDMI / DVI code from for my first N64 experiments.

Such a great resource of clever code on there. Even has the SPDIF one which worked fine through a LED on my DE1 board.

Have you seen those few sites with the LVDS code for laptop panels btw?
Can't find it atm, but there are a couple of good ones available now.

With HDMI direct from the FPGA, I was using the LVDS pairs though of course, so it didn't work at all on my monitor (wrong voltage swing / current loop levels).
I'll see if I can find those few sites though, and post you the links.

I somehow managed to "port" hamster's Xilinx TMDS code to run on the Cyclone as well, but wasn't too bad.
It just involved figuring out the basic blocks, then adding the DDR IO library block and Cyclone PLL.

Yep, EDID should be straightforward.
AFAIK, most PCs will just read the EEPROM directly, so there's no fancy stuff going on there luckily.

I see how Daniel has done the new board now. It makes perfect sense...

We were worried about extending the MIPI signals via a cable as I don't think it was ever intended for driving a cable directly anyway.
And also thought about what would be the easiest method for the "average" Rift user to swap out the panel + controller, keeping the stock DK1 cable.

Daniel is now using an HDMI switch in the "controller" box, but with boost amp / EQ for driving long cables.

So, AFAIK, it's...

PC DVI / HDMI -> control box (switcher / booster) -> HDMI + power + USB to headset -> HDMI-to-MIPI board -> Panel.

The MIPI board will still sit inside the headset behind the screen, and I'm sure Daniel is busy designing a perfect fitting board to go next to the tracker.

(HDMI -> 24-bit RGB -> MIPI is just the signal path on the small board behind the screen too IIRC?)

With the stock DK1 cable, it will also pass the USB for the tracker as before.

But, not to worry - it would still allow us enthusiasts to replace the cable with HDMI sockets if need be.
That will pass USB and power as well, but with a few caveats...

* The HDMI cable must have half decent shielding and quality (most our nowadays), but the amp chip should allow longer runs too.

* The wires for power / USB in the "HDMI" cable need to be fairly sturdy too, which brings us back to the above point.

* The user / modder needs to be aware that the "HDMI" output of the control box should never be plugged into a standard HDMI input / output!
(any minor risks can be minimized though, by re-using the +5V pins for the panel / tracker power, like in the DIY thread / HDMI guide.)


In other words, Daniel will be using the stock DK1 cable at first (only one cable to headset), but don't worry about it - it will be simple enough to extend.

(@Daniel, please chime in if I'm off-base here.)


EDIT: Oh, to answer your second question more clearly - Daniel and I did discuss about having a standard HDMI cable direct to the headset, but that left us with the issue of getting USB and power to it easily.

Using the original scaler box housing solves many issues at once, as it means everyone with a dev kit will already have a suitable power supply and all cables.
It also means not having to remove and replace the stock cable for the panel and tracker (unless you want to extend it at a later date).


OzOnE

[rozsnyo]

I have just sourced two 4.7" screens - and for my surprise they have significant design differences (different FPC, different COG position).
What shall I think of this? Can we trust the marking to be okay? (It should be Sharp LS047T1SC01K, from HTC One)

I am little bit afraid to buy a 120pcs box of 5.9" after such experience - just to find out they are not the ones I have tested. The wayengineer.com site (which seems to be the only source for 5.9") is keeping price of these screens at $90 and does not want to go lower + 100% pay in advance is quite risky.

[cgp44]

Don't extend mipi wires more than 30cm. That's for FHD resolution at Gpbs data rates,
This is the official spec. Perhaps going over this just makes more mobile power rather than
Signal jitter.

[cgp44]

I wonder if putting hdmi differential signals through the dk1 cable will work?
The five differential tmds pairs in a hdmi cable are twisted wires. The dk1 cable is a bunch of
Lvds wires not twisted together and they are more wires at 10 times less the
Serialised tmds data per wire. I suppose you guys have just done it .

I have cut up hdmi for harvesting bare thin wire for wire cradles for soldering to 0.4 mm
lug arrays. The wires in the hdmi (15 and 16 used for the aux and a power wire. And maybe
a ground wire), should be fine. There are a few non connected pins too.

[rozsnyo]

Don't extend mipi wires more than 30cm. That's for FHD resolution at Gpbs data rates,

I wont. MIPI is kept inside head, few cm max.

I wonder if putting hdmi differential signals through the dk1 cable will work?

I wonder too. We just try and if it does not work, we waste the prototype and some time. But given the fact, that LVDS has no special care, while TMDS got at least equalizers in the receiver, I think it can work. The TMDS frequency is about 1.5-1.6G for FHD ( TMDS is 10x base clock which is 148.5MHz for standard FHD and slightly lower for portrait mode and/or reduced blanking). LVDS in original DK1 is around 85MHz x 7 = 600 Mbps.

I am changing the design on the fly, so now it looks like I will have both of USB+HDMI plugs and DK1 cable plugs. But until I post some layout, it can be changed if somebody make a strong point. The decision how to connect the screen is so far open - not sure if it shall be a 5.9" specific (cheap as no cable or extra board is required) or universal to some extent (but here the rift plugs would not be required).

[mr.uu]

I wonder too. We just try and if it does not work, we waste the prototype and some time. But given the fact, that LVDS has no special care, while TMDS got at least equalizers in the receiver, I think it can work. The TMDS frequency is about 1.5-1.6G for FHD ( TMDS is 10x base clock which is 148.5MHz for standard FHD and slightly lower for portrait mode and/or reduced blanking). LVDS in original DK1 is around 85MHz x 7 = 600 Mbps.

I am changing the design on the fly, so now it looks like I will have both of USB+HDMI plugs and DK1 cable plugs. But until I post some layout, it can be changed if somebody make a strong point. The decision how to connect the screen is so far open - not sure if it shall be a 5.9" specific (cheap as no cable or extra board is required) or universal to some extent (but here the rift plugs would not be required).

Well, if you could add a "internal" HDMI plug in the free area next to the "DK1-old cable-tracker connector", we could use either the old cabling (if it works), or desolder it and use a new (even longer) HDMI cable instead of the old DK1 cable. Of course you would need the same on the driver board in the Rift, too. Parallel wired and we could choose which one to use (if the signals are not influenced too much by the open ends of the not used connections). This way we have minumum change of the original parts (the HDMI cable would use the original cable holes) and certainly one way will work...

[cgp44]

I would cut the cable connector to the mask right out and repurpose the whole
dk1 controller box, lvds cable, and 1280x200 panel as something else. Either a
low res vr mask with say the use of the other two lenses in dk1 or as a tiny
pc screen. I've done both. A tiny screen is quite neat when you are reclining,
not wanting to sit up to a monitor.