Building VR Hardware Using A PSC1000a JavaChip? 24
malachid69 asks: "As a Java programmer [pls, no spam about that], I have been looking at the Patriot Scientific PSC1000a JavaChip for quite some time now. Last week, I finally ordered one (yes, just one, until I figure out how to use it). Some of the specs are here. In college, I took some Analog/Digital classes, but that was 11 years ago -- and I have forgotten a lot of it. I have been into VR for quite some time, and am looking for ways to make some really nice (yet cheap) VR hardware with this chip. I am investigating EInk/OLEDs (as replacements for LEDs/shutter) for the display, etc. Does anyone have any idea how to build VR hardware out of this chip?"
Re:This chip vs. MAJC? (Score:1)
Re:What are you asking, really? (Score:2)
Even with those specs, the PSC1000 is still just a small CPU - it sounds like it would make a great system for robotics, though. I imagine you could build a dedicated VR system, but it will be a TON of development work. You state it would be tailored to the task, but in the end, it is a VR system running on the CPU. You say it would be better than VR->Windows->DOS->i386, but the fat of the matter is I wouldn't build a VR system that way anyhow. An ideal VR system will have multiple boxes: One handling input, one handling video, one handling audio, a controller box, and a backend DB server. Emphasis would be placed on the graphics, controller, and DB server. I would put an RTOS on the graphics, audio, and input servers, the rest could be normal OS machines. Have them all communicate via 100BaseT ethernet over a switch. Essentially a distributed processing system.
High-end HMDs are expensive mainly due to the resolution and optics used - were talking quality stuff here. I consider an HMD high-end when it starts to cost over $1000.00 - maybe a bit higher (some of the mid-range HMDs have come down in price). Consumer HMDs can't compare.
LCDs aren't outdated for homebrew HMDs. A couple of nice TFT LCDs from a couple of hand-held TV's can actually give great results, much better than what was affordable in 1991. It is also possible to get small full color CRT displays for a resonable price. Even shutter glasses can be good, once you start homebrewing your own CAVE system (not that I have ever done such, but it is possible).
I can see using the device as part of a system - maybe as one of those "boxes" I mentioned, since you could actually make it real-time or near real-time with custom coding. With enough of them, your final box might very well be the backend DB server, and the graphics display box (I am thinking something like a 1 GHz machine with a very nice 3D accelerator on it). Doing calcs for some of the hard stuff would be nice too, to get the data as a serial stream in regular ASCII format. But for triangulation, you obviously don't need to use something like this - the Powerglove gave the information in units properly, and it didn't need such a processor. I would say a PIC would suffice for that stuff.
If you manage to create anything interesting, let me know - I would love to learn more about it...
Worldcom [worldcom.com] - Generation Duh!
Re:What are you asking, really? (Score:2)
I am not sure if you can distribute the database this way or not (via an RMI model). Sure sounds like it would be worth pursuing. Really what you are talking about is building a distributed database model, and miniturising the interface hardware (by this I mean the hardware and software that works on the world model).
The pods you speak of (Waldern-style? Never heard of it, but sounds suspiciously like W Industries former design - is that where the W comes from?) are not designed that way so much for tracking, but for keeping the participants from hurting themselves. It is actually a nice design. But the tracking doesn't rely on the pods. You see, there are sensors placed on the HMD, the glove/handheld wand/joystick, and perhaps even the fanny pack (though this is unlikely, as the player doesn't really need to know which way their body is oriented, only the head and hand(s)). These sensors contain coils that sense electromagnetic pulses, and based on the orientation and distance of the sensor, and some complex math, determine the yaw/pitch/roll and x/y/z of the sensor. It doesn't matter if you do the calculation on the body or off the body, though if you could put everything on the body (like an ultimate wearable), then you would only need one link to the outside - wired or wireless.
I don't think you will be able to build what you are asking for what you are asking. A lot of the professional HMDs use an optical system known as LEEP - which, from what I understand, manages to give a bright and wide FOV, with low distortion. It is a very high-tech set of optics, and they don't come cheap. Those that don't use this system, still use fairly expensive optics. If you want to know what professional optics cost, take a look sometime in an Edmond Scientific Optics catalog, and be amazed. A high quality prism can set you back $20-50 dollars easily!
BTW - how were you going to use OLEDs? Do they make these in panels? Or only single "lights"? If it is the latter, then you will be facing the challenge of either building your own panel (good luck), or making a scanning system. Either way will be too expensive.
Powergloves are hard to come by - I actually think the glove paradigm is only good for a few special cases (mainly where you need to manipulate things), though such devices hold promise of acting as a good interface (ie, hand/finger signals). You could build your own glove, and even build your own sensing system - perhaps using IR sensors, or using a camera and digitising software, or other means. The camera system would be good from the standpoint that it is the most lightweight - you would probably not have to wear anything to use it - or at worst, a colored lycra glove.
I really don't think the main costs are in the designing of the hardware for homebrew VR. While the upfront costs of these components can be expensive, what really will cost is the software. VR is very much a multi-displined art right now - to participate, one must, almost without exception, be a deft with a soldering iron as one is with code. This is only because there is so little off-the-shelf hardware and parts to incorporate easily in the design, and what little there is, may not work with your platform of software (ie, the Powerglove is meant for the Nintendo - but it has to be hacked for the PC, and only there will it work up to a 486 - because the drivers have hard-coded timing loops, due to the bit-banger interface - one would have to recode for this, or build a hardware interface, like the Menelli box).
Worldcom [worldcom.com] - Generation Duh!
Re:What are you asking, really? (Score:2)
Yeah, I started looking into Waldern/Virtuality connection - they are still in business - as CyberMind (http://www.cybermind.co.uk/).
From what I understand, there are transmitting "antennas" (basically each is a loop antenna, probably several turns) in the base of the unit, and each sensor has three orthogonally aligned coils in it (very small, many hundred turn coils). Each antenna is "pulsed", and each coil in the sensor is read to get a reading. The phase angle of the pulse in each coil indicates the orientation of the sensor, while the strength (amplitude) of the pulse indicates distance from that particular coil. At minimum, three pulses are required for a full reading (where all three of the orthogonal coils are read at one time), or maximum, nine pulses are required (three from each base coil to each coil in the sensor). Then a ton of math is performed to determine the distance from the base coil to the sensor, and orientation. Someone once sent me some math for my "idea" at implementing this - I may have it up on my site (maybe)...
If you are going to work with a manufacturer to build these things (the OLED displays), are you thinking of building a whole "ton" of them - in order to get the cost down (ie, this is not homebrew VR at this point, it is commercial R&D in VR)?
The camera/digitizing system could be used for full body VR interaction, and has seen a little bit of homebrew use (look into the Mandela system - IIRC - and there was also a game controlled by movements seen by a Quickcam). Most of the expense would be in writing the software to interpret the motions, but it can be done with reasonably powerful hardware.
Force feedback at the hand level is complicated, simply because of the bulk. I am sure, homebrew wise, it would be cheap enough to try a piezo/muscle wire combo to product feedback.
Regarding needing the hardware before the software - that was my dillema - I had the hardware, but no way to get the VGA signal to the HMD (I going old-skool route here). I finally picked up an Averkey iMicro, and it works great with my HMD (currently a hacked Stuntmaster, but later will be trying the Halted specialty - ie, Cybermaxx optics and displays, all canned in a custom HMD - I am also toying with the idea of taking two Stuntmaster displays, and building one HMD). Now that I have a display, I want to put together the rest of the hardware (including an interesting sourceless head tracking device that I have planned), so I can do the coding for...
Worldcom [worldcom.com] - Generation Duh!
Re:What are you asking, really? (Score:2)
I think for the full body thing you could get by with a better QuickCam type device doing 640x480 - maybe use multiple ones on USB and custom software. Or, have a video capture card (like a WinTV or similar), and some kind of video switcher to swap between cameras. Your biggest issue is once again, the software needed to make it work (two cameras could track a point in 3D space, though - in theory).
The link you gave to the Images Company seems neat - I will have to go over that site with a fine-toothed comb, later...
My HMD is a cheapo Stuntmaster - my original one cost me $250.00, open-box item at Best Buy - I bought it in late 1993, when it first came out. You can now sometimes get them on Ebay and other places for less than $50.00 (my other two cost me less than $50.00 for both). It isn't great, but boy is it cheap! The optics and LCDs from the other system I got from Halted a couple of years back (www.halted.com). When Victormaxx went out of business, Halted bought a bunch of the parts, which were quickly snatched up. I managed to get some, enough to build a single HMD. It is also possible to build a fairly good quality HMD using hand-held TVs and such. Takes a lot of work, but is supposedly worth it in the end (PCVR ran quite a few articles on doing this).
I didn't get the email you sent me: Did you send it to the right address? Remember to take out the NOSPAM (some people forget)...
Worldcom [worldcom.com] - Generation Duh!
Re:What are you asking, really? (Score:2)
I don't think that pattern recognition would be needed for more than gesture recognition - the main thing would be using two cameras to track a point in 3D, and working out the X/Y/Z and yaw/pitch/roll of the "part" being tracked. A body suit to track using, say potentiometers or strain gauges would have to be pretty well fitting - but still, you face error issues, because of the way skin (and the suit) would stretch and move about the joints and such (now, maybe if you could mount it to the bones - ouch!). But, it could be cheaper. I figure that if you can already track a single point with cameras, you can track a whole bunch - so just glue those points onto the body. Use 3M reflective tape, and "light" the user up with IR (or near IR) light - and remove any IR filters from the cameras. The points should stand out well.
Thanks for the link and info about JG - I hadn't tried that one. I sent an email to Raytheon asking for information - if he still worked there. If I don't get a response, I will try the school as well.
My hope is to get a CD-ROM or such with all the articles on it, in some document format, that I could turn into PDFs or HTML, and put up on my site for others to use. At minimum, I would like to put up selected articles. But, I have to get permission before I can legally do this (one thing I wonder - if I can never contact him, can I legally re-publish, as long as I can prove that I tried everything I could to contact him? - Ahh, an issue for a lawyer, no doubt)...
Worldcom [worldcom.com] - Generation Duh!
Re:What are you asking, really? (Score:2)
Joseph Gradecki wrote a book called "Virtual Reality Construction Kit" (the ISBN and such are on the site), that detailed various head tracking systems, and whether they could be done at the homebrew level. He presents a couple of projects for head tracking in the book. Many of the projects in the book were nearly straight out of PCVR, the only difference being the drawings looking better.
One head tracking system described in PCVR was actually kinda funky: Imagine a bare room. Now, place the user in the room, and run strings from auto-retract mechanisms placed in the corners of the room. The strings run up the wall, over a pulley and down from the upper corners of the room to meet at a central point on the user's HMD. At this central point are orthogonally arranged potentiometers, to gather yaw/pitch/roll info about the head. So, you have these four strings. Depending on where in the room you are - these strings are extended and retracted by various amounts. It is possible to work out X/Y/Z coordinates of the point at the user's HMD where the strings come together. The article detailed a lot of the math, as well as possible pitfalls - I would love to republish that article.
If I am unable to get in contact with JG, I might just willy-nilly go ahead and publish - if he has a beef, he can see me then (hey, perhaps this will bring him out of the woodwork!). I may pay a lawyer to get my legal position first.
I wish I could say the contact info in the magazine was valid, but it isn't. You have to remember that this magazine was published in a one-man fashion - a good ole' DIY thing - after getting the magazine wrote up, he went to a printer, printed up some copies (I would guess less then 200 per issue), and mailed them off.
Let's move this conversation off of
Worldcom [worldcom.com] - Generation Duh!
What are you asking, really? (Score:3)
Cheap VR hardware isn't built from the ground up in this fashion. While it is feasible to build cheaper HMDs in this manner, you would still be using mostly OTS (off the shelf) hardware for the display electronics. Building your own display (I am not talking the optics, etc - just the display) is a possibility, but you will spend a bit of money and time doing it (it is simply the scale of manufacturing - it is always more expensive to build just one, than it is to build a whole mess, and this savings really is passed on to the customer, in an OTS item - like an LCD TV - that you would use in a VR system).
I am not trying to discourage you - if you have the money and the time, by all means go for it. But what you are asking would be akin to asking "Hey guys, I have this bolt here - how do I go about building a car with it?".
In closing, I would like to add that if you are really interested in cheap, homebrew VR techniques, check out my site. I have a ton of collected information there, that you would probably find very useful in pursuit of low-cost VR.
Worldcom [worldcom.com] - Generation Duh!
What software to run? (Score:1)
Re:What software to run? (Score:1)
So you can run JOS in 2 versions: on top of Linux as a JRE or as a standalone OS.
Their website has problems. You might take a look at http://sourceforge.net/projects/jos [sourceforge.net] .
I think there's a misunderstanding in here. You can run Java code natively on your chip, but you'd still need the functionality of an OS. I guess you want to manage your resources somehow. What you can leave out is the translation software, but there are still things like thread implementation etc. So I still think that porting JOS is a good idea. Well I guess most work would go into the adaption of classpath. Replacing the native peerings with Java based peerings that interface with your hardware. Still a lot of work. Ciao, Andreas
Re:What software to run? (Score:1)
ftp://jos.sourceforge.net/pub/jos/grub.img
They have much more than the website indicates, but you have to read the JOS mailing list to find out. There are also detailed instructions on how to run this. I have not managed to retreive a recent CVS snapshot, though. So I gess my JOS sources are also a bit dated here. If you are interested in the messages from the JOS list, you can mail me:
a_rueckert@gmx.net
I've tried to mail you a copy of my posting, since I was not sure if you would read my posting, but it seems the mail addy from your website doesn't work...
HTH,
Andreas
PS: If you actually manage to run any JOS or Classpath related stuff on this chip, please post to these list. I and a lot of other folks would be very interested in hearing about this (or even to help out a bit).
This chip vs. MAJC? (Score:1)
Re:What software to run? (Score:1)
I know SUN has their Java FEE-based OS out as well, but I do not think we would need to use it since Java runs natively -- and their OS would convert the Java bytecodes into local machine code.
This chip will bootstrap and start running your code natively. I am sure that there would have to be some wierd stuff done to make GUI work -- but we should not need an "Operating System" to run the Java.
What I want to do is write my own software (using perhaps Java3D API) that would just get executed directly on the chip. Even if I wrote my own Java OS that sat between the chip and the VR software -- at least I would not be writing i386 to run Java. However, in theory, any Java program should run on it.
Re:This chip vs. MAJC? (Score:1)
Although the PSC1000a is a 32-bit chip, it has hardcoded to 8-bit operands (because Java does). The MAJC chip on the other hand seems to be more like the IA-64 (according to articles) and uses the VLIW (Very Long Instruction Word) architecture. Considering the fact that all Java bytecodes use 8-bit operands, I am not sure why SUN went with the VLIW -- but that pops out as the biggest difference. How do we even compare processing speed when the instruction sets are different sizes?
Re:What are you asking, really? (Score:1)
I realize that VR hardware would need a lot more than a CPU. That is one of the reasons I was looking at this chip. This quote from the datasheet is part of the reason I thought it might be a little easier:
By incorporating many on-chip system functions and a "glue- less" bus interface, support chips are eliminated, further lowering system cost. The CPU includes an MPU, a Virtual Peripheral Unit, a DMA controller, an interrupt controller, bit inputs, bit outputs, and a programmable memory interface. It can operate with 32-bit-wide and 8-bit-wide memory and devices, and in- cludes hardware debugging support. A minimum system consists of a PSC1000 CPU, an 8-bit-wide EPROM, an oscillator, and optionally one x8 or two x16 memories.a total of 4 or 5 active components.
As far as the HMD goes, I am looking into what options are cheaper than LCD -- possibly OLED, etc. Most of the homebrew VR hardware I have seen so far uses outdated hardware (LCD, shutter glasses, etc). Since I would not need to be compatable with any current technologies (x86, software, etc) -- I figured I might as well try to optimize it cheaply. My thought was that a lot of VR hardware is expensive because it is kewl ("cutting edge"?) not because it costs a lot to make.
I agree that the scale makes a big difference in costs (even these chips are cheaper in bulk) -- but I would like something better than a 486 running w/Sega & powerglove -- and I think it would be FUN to make and test.
But, possibly to clarify my original question, how would you drive an HMD and Glove with the PSC1000a? How would you drive a network connection with it? I don't want to create a motherboard -- but possibly an HMD that natively handles OpenGL instead of SVGA input for example... or perhaps something that handles the triangulation sensors for tracking? or something that handles finger movements and force feedback -- without my PC...
Wouldn't it be possible to cheaply make hardware running VR code (in Java) that is tailored to the task (VR) instead of being designed to do anything and just happens to be running some VR software on top of Windows on top of DOS on top of the i386?
Re:What software to run? (Score:1)
However, I was thinking about your comment - specifically the part about threading. It looks like WindRiver systems has a port of PersonalJava for the PSC1000a that they licensed to PTSC. I will know when my package gets here whether that comes with the chip.
The peerings is a good point. If the WindRivers stuff DOES in fact replace all of the peerings, I wonder how they replaced the AWT ones? It would seem that all of the peerings would have to be tailored to the specific way in which you decide to hook up your hardware... Even pinouts should change the peering setup, right?
Re:What software to run? (Score:1)
But, in regards to your P.S....
I wonder if there would be enough interest out there for an Open-Architecture Open-Source type group for developing hardware around the PSC1000a.
Anyone? I would, personally.
Re:What are you asking, really? (Score:1)
I am actually looking at a distributed concept. I was thinking of VR decks connecting via RMI over fiber (dedicated bandwidth per frequency -- p2p via RMI). I was considering a chip in the HMD so that it could handle OpenGL or Java3D instead of SVGA input. Java3D already has the 3D Audio as well, so it might work for sound too. I was also thinking that RMI'ing the decks would be like distributing the database as well.
I would like to keep the costs low, so I obviously don't want 10-20 of these chips in a single VR unit... But I was thinking perhaps one in the HMD, one in the deck... Perhaps also one for the gloves/body-suit?
Of course, a lot of this would change if we were looking at doing Waldern-style pods. Then we might want one for the "gun" or whatever and one for the tracking system on the ring around the waist. Not sure.
I realize it would be a lot of work to build our own VR from scratch -- but, would it be cheaper? Better? I am not really sure.
Regarding the $1000+ HMDs... See, that is way out of my league for what I am looking at. I want to make something that at least appears to be high-res for $25-30 (total system cost of about $100). I am considering giving the finished systems away, so I can't really spend $1000 per HMD. That's one of the reason I was looking into OLEDs... They are supposed to be cheaper than LCDs... I have looked around for some small LCDs, but all of the ones I found were kinda expensive... Again getting back to the overall costs....
The problem I had with the powerglove was just the bulkiness, really. I was thinking of something more open to air. Besides, they seem to be a bit harder to come by now.
Where is the REAL costs of designing VR systems from scratch? Obviously it is in the time and the LCDs... But is that it? Is almost everything else fairly inexpensive?
Re:What are you asking, really? (Score:1)
One of the first examples given when learning RMI is a MUD. It works EXTREMELY well for such a situation, since you can transparently be redirected to a different server when you walk through a door. I was thinking of using this model, so that each deck only contained their own part of space -- linking them thus distributing the database.
Yes, the pod I was talking about was Waldern Industries. Unforunately, I only got to play Dactyl nightmare -- but it was far more immersive than the BattleTech/RedPlanet centers.
So, in regards to the tracking for those... So, what it did when I turned around or knelt down was just get stronger/weaker signals off the gun or headset? From what I understand, most of those types of tracking are only really good in 2D (up/down OR left/right)... Although I guess an extra receiver on the pod would fix that.
I agree that it would be kewl wireless. I was only thinking of the pod with the waist-thing to keep people from walking around bumping into things.
I actually have an Edmunds Scientifics, and I agree -- their optics are really expensive (with the exception of things like Frensel). However, I would not be using any of those.
The OLED stuff I was researching showed that a lot of companies are building displays with them right now. The only problem with the EInk version is that it is monochrome -- but the JCPennys chains are all using them. All the pages I went to talked about how OLED displays could be printed with InkJet printers, etc -- and how they are cheaper than LEDs... From what I understand, you get a sheet of polymer/plastic with transistors on it, then spray it with the ink (which contain polarized light-emiting chips)... So, in theory, it would be LESS expensive if I choose the right manufacturer to work with.
The camera/digitizing software has some appeal for the full-body concept, but how expensive would it be to do well? I figure making gloves could be fairly cheap (thin leather motorcycle gloves with the holes in them?)... I remember seeing the schematic for the IR-version.... What ever came of using "Muscle Wires"/piezo? I like the idea of being able to provide force feedback.
If I can design the hardware well (and cheap), and use this PSC1000a chip... I should be able to use Java for the interface to it. I am currently planning on writing my own software (possibly on top of Java3D), but there are already some games out there using Java... There are also utilities to convert different formats (DXF, VRML, etc) into J3D... I realize the software would be a lot of work -- but until I get the hardware done, I don't even know what I am interfacing too....
I have to say though, this discussion feels productive towards those ends.
Re:What are you asking, really? (Score:1)
Regarding the manufacturing.. No, there are a lot of places currently selling OLED displays -- some even as headmounts.. I have just been emailing them all about pricing so I could choose the best deal (cheap yet nice). But, they would only be for the LCD-replacement-part... I would still have to do the drivers and connecting hardware. If I can get the system price low enough, I will probably either give them away or teach classes on how to make them.
What level of power do you think we need for the full-body thing? Perhaps use a camera quick enough to keep track of key-points?
I was looking at some MuscleWires on that link from the Quickies yesterday, and I notice that the name-brand one is quite expensive. Anyone know if there is some generic brand of muscle wires -- or perhaps something like piezo-wiring we can use?
How much did you pay for your HMD? How's the quality?
Ideally, I would have some basic hardware, write some software -- then be able to upgrade my hardware without needing to start over from scratch. I have lots of ideas of what I would like to do (hardware and software) -- but figured I have to start at the hardware because little things like # of pins could drastically change the software.
Re:What are you asking, really? (Score:1)
Re:What are you asking, really? (Score:1)
I think we could write the software for pattern recognition to work with multiple cameras... But I wonder if a suit would be cheaper? It doesn't have to be a full-body lycra thing, just something that could sense changes at the joints or something... Dunno...
Halted looks like they have some kewl stuff. I will have to take a look at their 2Meg catalog :)
I thought I sent it to the correct address... It didn't bounce back... Well, in that case, I will go ahead and post what I found on here.... At http://www.colotechu.edu/main/faculty.html [colotechu.edu] it says:
Joseph Gradecki, Adjunct Associate Professor
Engineer, Raytheon Systems
M.S., Computer Science, University of Wyoming
B.S., Computer Science, Metropolitan State College of Denver
So, I think that is saying that he is a professor at the Denver Campus of Colorado Tech.
Good luck! If you get it up and running, I wanna subscribe!
Re:What are you asking, really? (Score:1)
Hey, I wouldn't mind the CDRom... I bet the information about republication is in the magazine... If the contact information in the magazine for who to contact is not valid...well, there I don't know...
Re:What are you asking, really? (Score:1)