Raspberry Pi

[Stoney3K]

http://en.wikipedia.org/wiki/Raspberry_Pi



http://www.raspberrypi.org/

This could be a very interesting base board for new DAC development projects.

The Raspberry Pi is a single-board computer developed in the UK by the Raspberry Pi Foundation. Scheduled for public release in January 2012, the foundation plans to release two versions that range from US$25 to $35 (GBP £16 to £22). It is intended to stimulate the teaching of basic computer science in schools.

The design includes an ARM1176JZF-S 700 MHz processor, 128 or 256 megabytes (MiB) of memory, intended to run Linux or RISC OS.[2][3] The design does not include a built-in hard disk or solid-state drive, instead relying on an SD card for long-term storage.

The Model B board has onboard Ethernet and both boards are powered off standard 5V. Dropping one of these into a projector and doing cool stuff would be a definite snap.

I2C and SPI buses are available for connection to DAC chips (such as the Microchip 4922), but using USB, it could be dead simple to "Ethernetize" existing sound card DACs.

[dilbert]

He he he I've already got my eye on one of these in Jan when they are released for a different network project...

[steve-o]

That looks pretty nice. I used an old "basic-stamp" for a show or 2 a while back .

[DTR]

I am very excited about these. I hear the XBMC devs are going to do a port for it which will be sweet. Here is a video that showed up the other day showing off the 1080P playback.

[chad]

Those are really neat!! I will have to get one of those to add to my collection of things that I can't program!

You could make a daughter board for it with dacs and buffers and all of the stuff you need for it, then write a bunch of software or you could just buy one of these.

http://ether-dream.com/

Not as cheap but he isn't making 25000 of them, the price will come down with volume. These are already built to do laser specifically. They have a 100mhz 32bit ARM, dacs, safety interlock,SD, Ethernet.

More importantally, firmware that is working and an interface through ethernet to a robust and getting better, pc laser software, LSX. The hardware is the easy part. The tool chainn is free, the code is open source.

I am all for reinventing the wheel, but that processor and the linux overhead are way overkill for streaming points to galvos, in my opinion.

anyway, neat little board!

chad

[Stoney3K]

Those are really neat!! I will have to get one of those to add to my collection of things that I can't program!

You could make a daughter board for it with dacs and buffers and all of the stuff you need for it, then write a bunch of software or you could just buy one of these.

http://ether-dream.com/

Not as cheap but he isn't making 25000 of them, the price will come down with volume. These are already built to do laser specifically. They have a 100mhz 32bit ARM, dacs, safety interlock,SD, Ethernet.

More importantally, firmware that is working and an interface through ethernet to a robust and getting better, pc laser software, LSX. The hardware is the easy part. The tool chainn is free, the code is open source.

I am all for reinventing the wheel, but that processor and the linux overhead are way overkill for streaming points to galvos, in my opinion.

anyway, neat little board!

chad

The Etherdream is a good board for people who want to get started with a ready made DAC right away. The Etherdream is powerful enough for most applications, but it may lack the processing power for 'really cool stuff', and with that I mean stuff that you could only do on a board like the Lasergraph DSP.

With a 700MHz CPU at your disposal you can even run your laser display software on the projector, and have only a remote terminal at your front-of-house, eliminating risk of failure because the image data does not have to be streamed real time to multiple targets. Also, with a fully fledged CPU, you can do procedural dynamic frames, real-time 3D rendering to vector graphics and software synthesis on your projector, which would make for some really fluent and beautiful images in the future, not limited to sets of 'frames' being streamed one after another. For example, the OpenLase API, which is already developed for Linux, uses an OpenGL-like graphics library to generate graphics on the fly, and supports USB sound cards out of the box.

Just for streaming points to galvos, it may be a bit overkill (but still cheaper than a RIYA Multibus or an Ether Dream!), but for moving more intelligence into the projector and to make laser shows more dynamic and interactive, I'd be all for developing for it.

[masterpj]

This is very interesting, the 3D realtime rendition is also very very good!

[Stoney3K]

This is very interesting, the 3D realtime rendition is also very very good!

The 3D rendering is done on the Broadcom VideoCore chip which takes care of all of the nitty gritty such as shaders and all of that stuff you need for displaying 3D on a raster display.

It would be quite pointless to render 3D vector graphics to a frame buffer just so you can take advantage of the speed of the graphics chip, and at the same time waste time by having to trace it back into vector data again (with all possible errors to boot) instead of just applying a simple mathematical transform to project 3D vectors into 2D.

I think the community would need to start agreeing on some standards for things like frame/point transmission (generator to DAC) and on formats for API's and rendering graphs. If there's some agreement on that, it's easier for other developers to invent new stuff without having to start from scratch every time.

Also, if systems like this become more commonplace, I think there's a good chance we're going to move from a traditional setup of 1 'generator' device feeding multiple projectors into a more decentralized setup where each projector generates its own data on demand.

[drlava]

The Etherdream is a good board for people who want to get started with a ready made DAC right away. The Etherdream is powerful enough for most applications, but it may lack the processing power for 'really cool stuff', and with that I mean stuff that you could only do on a board like the Lasergraph DSP.

The 'really cool stuff' that you're talking about is already possible inexpensively with LSX laser show software, you should really check it out some time. With this you can use with the Etherdream or any DAC of your choosing. There are advantages to having the processing engine on the computer, that, while at first it may seem like a good idea to have it on the DAC, eventually the flexibility of the computer interface (and the ability to upgrade computers as they come out instead of redesigning your DAC as new hardware comes out) gives you greater control, and this is what LSX is all about - dynamic control. In fact your desire 'procedural dynamic frames' is right off my LSX announcement page title

I've created it as a way for the community to express themselves through whatever images come to their minds, and also as a tool where amazing things can be created just through playing around. If that sounds like something you'd like to work on, let's chat!

[masterpj]

The 3D rendering is done on the Broadcom VideoCore chip which takes care of all of the nitty gritty such as shaders and all of that stuff you need for displaying 3D on a raster display.

It would be quite pointless to render 3D vector graphics to a frame buffer just so you can take advantage of the speed of the graphics chip, and at the same time waste time by having to trace it back into vector data again (with all possible errors to boot) instead of just applying a simple mathematical transform to project 3D vectors into 2D.

I think the community would need to start agreeing on some standards for things like frame/point transmission (generator to DAC) and on formats for API's and rendering graphs. If there's some agreement on that, it's easier for other developers to invent new stuff without having to start from scratch every time.

Also, if systems like this become more commonplace, I think there's a good chance we're going to move from a traditional setup of 1 'generator' device feeding multiple projectors into a more decentralized setup where each projector generates its own data on demand.

I want one of these, for real!
And yes your right

[Stoney3K]

The 'really cool stuff' that you're talking about is already possible inexpensively with LSX laser show software, you should really check it out some time. With this you can use with the Etherdream or any DAC of your choosing. There are advantages to having the processing engine on the computer, that, while at first it may seem like a good idea to have it on the DAC....

One of the problems with having a 'dumb' DAC (worst case: a sound card) depending on a host system is performance. Especially when your host is running a poorly programmed operating system with a dodgy driver model (e.g. Windows), you may very well run into some timing issues, or even crashes.

When you have the Raspberry Pi on board the projector, the DAC *is* the host system. Think of it as having multiple computers with their own instance of your favorite laser show package running simultaneously, one in each projector. They're not dependent on each other or on a specific host to get their frame data, only on the network for synchronizing them together and for interfacing with the laserist.

I'm not saying you should get rid of the host system at the front of house altogether and have some dumb terminal at that end, on the contrary -- I only suggest moving the frame generation logic to where it is needed most (the projectors), freeing up more processing power on your host for things like new user interfaces and really processor intensive stuff that is not timing critical (as opposed to the frame generators, which need to be fairly real-time), like modeling or rendering in 3D, processing video and other complex operations. When that's done, a new bunch of vector data can be streamed to the projectors which can use them on demand.

The only thing the DACs have to do is crunch a bunch of points (sets of X, Y coordinates), which is an operation that can be done on that ARM 700MHz processor at a rate of millions per second. So I doubt we're going to run out of processing power to do complex frame transforms there in the future.

Just for the sake of comparison, how much processing power is there on high end boards like the QM2000.NET or Lasergraph? Judging by the specs, the Raspberry Pi could run rings around those boards.

[j4cbo]

The argument for super-intelligent, Linux-based network DACs is well-founded - a Raspberry Pi or Beaglebone is more powerful than the desktop PC that a QM2000 might have been put into around when it was released, which means that math and effects that would have been state of the art back then can run handily on a modern ARM.

The counterargument, though, is that PCs have sped up an awful lot too. Effects that a Rasperry Pi would have its hands full with are basically epsilon on a modern CPU/GPU combination, and processing things all the way down to a raw point stream on the PC makes engineering the whole system a lot easier. The network protocol is just "here are some points", not "here's a bunch of backing data; now do your translations on it". (GPUs require large amounts of bandwidth to the host memory. If you replace the ultra-low-latency 64 Gbit/sec PCIe link between a modern GPU and its host system with Gigabit Ethernet, it won't work quite so well...)

There's also an upgradability issue - see the troubles people have had looking for hard-to-find old DIMMs to upgrade the memory on a QM2000. Buying a new, faster PC isn't a big deal - the engineering effort on computers is amortized across basically the entire world, as opposed to our little laser niche. Modern OS kernels are pretty good at real-time handling, and in any case, moving the real-time code to a "big" embedded system doesn't solve all problems magically either. These boards just runs Linux, after all, a grown-up desktop-oriented OS like any other.

[mixedgas]

Run rings around a Quad Transputer with Mathcos, maybe not?. (Lasergraph DSP)

QM2000 is a Coldfire with a novel fast DRAM accessing scheme that was so unique 'Rola let Bill Benner post it as a Motorola applications note world wide. BB has authorship credit on a AP note, and that is a rare thing for a outsider.

I think Mr Pi only has pure speed over a Coldfire.

On the other hand, look what some very clever code lets a Coldfire do, pretty much independent of its host.

QM32 was a 40 Mhz 68040, and my QM32 has continued playing shows when the Host 98 or ME system has crashed BSOD.
QM32 was famous for doing that, the PC was just a user interface, power supply, and storage.

Only with " Beyond" has the QM2000 become pretty much a "DAC", before that the 2000 did perhaps 90% of the work if you ask Mr Benner.

Whats amazing is how little firmware space is used on a Pangolin product. My QM32 has two 64K eproms if I remember correctly, I'm away from home so I cannot go check.

So what does the 700 Mhz Tiny ARM really have, other then pure speed, gained from Moore's law?

You could probably almost "emulate" the Coldfire on the Pi. However, the real magic is in the years of time spent writing the code that does a laser show with such precise timing and repeatability, and functionality. Without that code even a state of the art board like Pi is a mere paperweight.

I have a board that plays laser images on a 4 Mhz Z80, its not the processor, its the man hours of code that is unique. A playback frame dumper is easy, I can do that with simple counters, sram or eprom, and glue logic, the real magic (tm) is in the system code.
Since the only freeware code is LFI, even if you get a Pi, what are you going to run on it?

Steve

[White-Light]

There's also an upgradability issue - see the troubles people have had looking for hard-to-find old DIMMs to upgrade the memory on a QM2000.

I believe the hardware has recently been upgraded on the newer QM's to try to alleviate that problem going forward. For existing QM owners, I noticed on their web site that you can buy DIMM's from Pangolin although whether they have any stock I'm not sure.

[Stoney3K]

You could probably almost "emulate" the Coldfire on the Pi. However, the real magic is in the years of time spent writing the code that does a laser show with such precise timing and repeatability, and functionality. Without that code even a state of the art board like Pi is a mere paperweight.

Oh, I will defnitely have to agree with that! I'm all for developing a completely new platform based on the Pi (or any other ARM board running Linux), together with a desktop host running Linux, since it opens up a lot more possibilities. Sure, you can use the Pi as a 'dumb' DAC if you want and if you don't need anything more, so the integration with existing systems is fairly seamless. But if you want to do additional features on the DAC, that's perfectly possible because you have plenty of processing power left.

I can also see it moving from the realm of 'players and displayers' into the realm of modular software synthesizers where you can define an arbitrary rendering graph for every projector or even build/change it on demand at show time. So if you have a show that's being played back from a linear time line, it'll just play straightforward, but if you want to intervene by adding live frames, manipulating parameters or adding abstracts, you could add filters for that on the fly.

The FOSS community needs to get off their bum and start writing software for us laserists.

[mixedgas]

The FOSS community needs to get off their bum and start writing software for us laserists. [/QUOTE]

The guys who have done it have been so abused by needless complaints that they all have quit.
Laser show software is a demanding real time application. Real time control on a Windows platform requires programmers that know hardware, are demented, dedicated, and special. Thats rare.

Steve

[Stoney3K]

The guys who have done it have been so abused by needless complaints that they all have quit.
Laser show software is a demanding real time application. Real time control on a Windows platform requires programmers that know hardware, are demented, dedicated, and special. Thats rare.
Steve

The FOSS developers I know don't even worry about complaints from mere users. (We're too busy developing, not tech support), and most of the FOSS stuff that's in the real time realm is in permanent alpha or beta stage until there's enough backup from other developers to provide support to people who need it.

Most of the building blocks that are needed have already been built. For example, if you use a package like Pureata together with JACK, you can run a live laser show without any additional coding, but there's the drawback of not having an intuitive interface.

At the very least I'd like to take a top-down look at the laser display infrastructure: What do we as laserists need or want in terms of user interface, creative opportunities and hardware features, instead of developing bottom up through improving on the existing "DAC/tape-player with a projector attached" concept.

[masterpj]

The Raspberry Pi sounds so good
Especially for development, also in doing some 3D calculations for say safety which might help if your using one of those not good intergrated video card laptops (which are of a low vertex and pixel shader version).

The only thing i'm wondering is where to attach a DMX in and out and the ilda itself to