"TurboProp" timeframe.

KenBash

My first Propeller project is nearing completion. ( a machine controller ).· I should have my "Production" board in hand later today.· ( is a board ever really finished?)

Before I appear ungrateful, I want to mention how much fun it's been working with the Propeller, lots!·

The incredible capability stuffed into this one little chip necessitates my next question.·· With so much capability, I'm running out of pins for the things I could easily do if I had the extra ports.· I had to slave an SX-48 to the propeller just to use it for the additional I/O.

The documentation itself let's us know that a new chip with more I/O is in the pipeline.·· The big question is:· How long is that Pipeline?··Are we talking years, months... are the beta chips already in the hands of others more blessed than I?··· I know that this might be a difficult question to pinpoint, but a ballpark timeframe would help some of us make better decisions about how much time to invest in the work-arounds that the new "Turboprop" would solve immediately.
Can you give any hints?

Ken B.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
" Anything worth doing... is worth overdoing. "

··············································· ( R.A.H. )
····································

Paul Baker

Don't ask, if you know Parallax, you'll know that they shy away from forecasts on products in development. This is because when they don't live upto a forecasted release date, someone inevitably chews thier head off (you may not, but someone else reading it will, it's happened countless times and is even more predictable than the sun rising in the East every morning), so they just don't do it anymore. I know what thier desired timeframe is but in respect to thier preferred method of doing things Im not going to say, because inevitably something will come up that wasn't an expected issue and delay things. For instance they were hoping the original Props would be availible in the spring, but it ended up being summer before they were availible. The original Prop took 8 years to develop, what I will say is it will not take anywhere near that long before you see the TurboProp.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
...

Post Edited (Paul Baker) : 8/10/2006 4:05:59 PM GMT

LoopyByteloose

My own experience with product development is that the developer thinks that it can be done any day now, but reality creeps in as thousands of details are required to go from concept into actual product.

I suspect that the TurboProp is 90% or more done in hardware, but the real restraint s are that the IDE and the documentation of the Propeller are not even complete.

It would be very self-destructive to release a second generation product when you haven't gotten the first generation fully presentible. There are tons of things that users discover in the real world that can never be found at the drawing board. Feedback is critical to seeing the right direction and right manner of heading that way. This kind of information is needed to make the 2nd generation a quantum leap forward, rather than a mere technically complete jumbo size. Would you rather buy a Jumbo-Propeller or the TurboPropeller you are asking for?

Personally I would love to see a Port B that has a high degree of autonomy in some way to operate concurrently with Port A. One example - from what I have learned so far, some of the 32 Port A Pins are a bit tied down to particular functions. While 8 or 16 bits seem easy, it appears that it would be hard to output 32 parallel bits. Maybe that needs to be resolved, maybe it is already done. I don't know.

And of course, there is a lot of learning how to teach the users in an efficent manner. You write something and you discover an easier way to say it the day after you publish it.

Tutorials have to get down to the beginner's level before many people will enjoy buying the product. Thinking that engineers will just have to figure it out is about the worst marketing approach one can have. Parallax's competators have fallen way behind in sales on that attitude.

With all its speed, making an SX48 Slave may be the best way to use the Propeller's resources and have low cost. After all, that is $25 + $10 for the combo. The Turbo might be $50 or more.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"If you want more fiber, eat the package.· Not enough?· Eat the manual."········

···················· Tropical regards,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan

Paul Baker

An even cheaper route to expansion of the pins than an SX is using a transparent latch, something like the 74LVC16373 provides 16 bits of additional output at the minimal cost of a single control pin when tri-state operation isn't required.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
...

Graham Stabler

Ken, your machine controller sounds interesting can we hear anymore?

Graham

tperkins

I'd like to suggest that the PortB remain virtual, accessible solely to cogs internally.

Have a PortC go to I/O pins 32 to 63.

But yes, please, have pins of the additional port be entirely general purpose.

Yours, TDP, ml, msl, & pfpp

helloseth

Paul Baker said...
I know what thier desired timeframe is but in respect to thier preferred method of doing things Im not going to say, because inevitably something will come up that wasn't an expected issue and delay things.

Well if you know, and you're outside the company, the next chip must be soonish. I think what Ken is asking for is a P8X64A. Same prop but with more i/o pins, not the future piplined controller that would be a second version, or one with more cogs.

Seth

KenBash

Thanks for the responses everybody!

Honestly. for right now, the "Jumbo Prop" with the extra I/O's would be fine!

The machine that I'm currently controlling is just a single axis machine, pretty simple stuff.

I managed to put everything easily into the demo board. However, phase II of this particular machine is going to have at least two more axis of motion and I'm fresh out of I/O pins on the Prop. ( needing analog input, keyboard, mouse, VGA, RS-232, + machine control )

I have a 24 bit 4 axis linear-interpolative driver I stuffed into the SX, so I grafted my last SX board onto the Prop.

I can use the Prop to talk to the operator and do all the cool "Prop Stuff" while the SX does all the lowly machine control stuff... the SX is acting as a "Motion Coprocessor" (tell the SX where to position 4 axis steppers and it drives in a smooth, straight line from where it is, to where you're going, plus a few things like home to limits, step and repeat and such. My SX board also has a FTDI USB interface on it that gives me a way to talk to the Prop over USB without using the programming pins. )

The reason that I was asking for more information on the Propeller is that I was getting ready to market my SX-48 chip/board for motion control applications. ( send it the positions, i/o's, speed, ramprate etc, from any USB port.) Then the Propeller was announced and I realized how simple it would be to use the Prop for multi-axis control and figured my SX design was dead in the water.

I recently had to bid several thousand boards for an unrelated job and realized very quickly the difference an $11.00 chip verses a $3.00 can make in the bottom line. ( I had to bid another processor with onboard atod…sorry Parallax) So from a price standpoint in some limited applications there still might be a crack open, but I have to decide how seriously to follow this niche.

After running into the limitations on the pin count for this current project, I realized that there still might be a window to introduce my motion controller/ Propeller-slave into, but if we’re talking weeks rather than (several) months, the timing of the NEXTGEN PROPELLER ( whatever we’re going to call it ) could negate several more months of work.

I understand Parallax’s concern completely, (this new machine was due two weeks ago…. ) but even having a “General” idea might save a few of us out here a tremendous amount of wasted time.

If anyone’s interested in seeing my new board I’ll try to get some pics up soon. It was AWESOME seeing it talk to me for the first time last night!

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
" Anything worth doing... is worth overdoing. "

··············································· ( R.A.H. )
····································

RinksCustoms

I agree with Kramer, and unerstand Kens disposition. The Propeller is an amazing tool if used prop-erly, lol. One would imagine the propeller evolving first into 16 COG's @ 160MHz w/64 I/O's & 128MB system RAM and eventually end up with something like 256 I/O's, an 80MHz crystal @ PLL16X (1.28GHz!) looking more like a Pentium4 and probably interfaced with an external 1GB DDR Memory and maybe a small hardrive.

Look @ what the Basic stamp evolved into! The PROpeller can run circles (literally) around the BS1 and in RCSLOW mode doing so! I was in the midst of having a touchscreen music/MPU-controlled crossover interface with a bit of crafty coding when i just hit the perverbial brick wall. The prop (once i get to know the lingo) should be able to do that whole interface in three or four cogs @ 4X the speed, replacing a text-quadrant driven menu system on 2K of program memory and an obsolete SX Video card!
I say the boys @ parallax have deffinately done their job, have a good reason for waiting for feedback so the next version can be spent creating less bugs, if any, possibly improoving the IDE, definitely will be writing new "commands" for the inevitable added features, reducing the overhead of having three Prop's available and only two are being bought. Why buy a BS1 when the BS2sx is a bit more and X times faster and more capable.

My thought is that it'll probably be about a year (or more) before we see another version of the propeller. But whne we do i feel that it'll leap like the BS2 did over the BS1, nearly doubled in every respect. But you do present us with a very good point! Why spent alot of time developing a touchscreen menu system or 4 axis motion controller squeezing every bit (no pun) out of the RAM to fit your program and interfacing external periphials to do the job? For now, it's a good solution, but i feel that parallax will manage to evolve such a masterpiece of MPUism and would be well worth the wait!

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Definetly a E³ (Electronics Engineer Extrodinare!)
"I laugh in the face of imposible,... not because i know it all, ... but because I don't know well enough!"

Post Edited (RinksCustoms) : 8/12/2006 3:24:33 AM GMT

KenBash

...not to mention my first version of the board was an SX-52....

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
" Anything worth doing... is worth overdoing. "

··············································· ( R.A.H. )
····································

Paul Baker

Those who were invited for the original Prop's unveiling (back in Feb) were told about the next generation already in progress, and were told when they hoped to have it done. That info is now 6 months stale, and no further "inside" information has been provided on it's status since then, so there is no telling if the current projected date has shifted since then, or how much more it will shift in the time between now and when they finally release it.

Seth Vanderdrift said...

Paul Baker said...
I know what thier desired timeframe is but in respect to thier preferred method of doing things Im not going to say, because inevitably something will come up that wasn't an expected issue and delay things.

Well if you know, and you're outside the company, the next chip must be soonish. I think what Ken is asking for is a P8X64A. Same prop but with more i/o pins, not the future piplined controller that would be a second version, or one with more cogs.

Seth

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
...

RinksCustoms

P16X64AB

1.26GHz 80MHz @ PLL16x, 64bit proccesing, 1GB ram, probably DVI support.

OK, i'm just dreaming, but taking the prop & raising the proccessing power 4 fold is more difficult than one might imagine. There's the heat issue, an almost required ATX power supply just to turn it on, but i tell ya if parallax ever created such a beast i think the boys @ apple & Microsoft would be aiming the laser pointers, lol

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Definetly a E³ (Electronics Engineer Extrodinare!)
"I laugh in the face of imposible,... not because i know it all, ... but because I don't know well enough!"

Lawson

huh sounds like all the "extra IO" tricks that the BS2 folks use would be usefull for you. A descrete 8 input ADC hanging off of the I2C pins with the EEPROM, one of the serial to lots of inputs/outputs circuts i've seen floating arround that uses 74HC logic, I think i've seen IO expanders that live on I2C, Paul Baker's suggestion of a transperent latch, etc... Just keep all the pins that need speed on the Propeller and farm out all the masses of "slow" response pins to some of the above circuits, shouldn't take more than one cog to scan 'em all then dump it into a nice cosy warm spot in Hub ram. With the speed of the Propeller it should be possable to check and update the "slow" pins at 1000Hz or more. (certainly more than enough to take care of limit switches)

Tperkins: Having an extra Virtual IO pin port is certainly an interesting idea for Cog-Cog communications. It'd provide a simple and familiar way for all of the Cogs to talk at the same time without frying anything, but it'd require the creation of a software communication protacol so it may not be faster than Hub-ram for a lot of things.

Cap'in Obvous,
Marty

KenBash

Actually, I think that Chip/Parallax is already on the correct course to make the "Big Boys" stand up and pay attention.

The Von Neumann bottleneck has been creeping slowly tighter so that increases in over-all computing ability on a single processor gets more and more difficult to achieve, but let's face it, a single cog, with perhaps an order of magnitude more memory or so ( AND SOME MORE I/O !!! Lol ) should be able to do 95%+ of all the control functions I can think of, from running a precision machining center to landing a space shuttle.

This ability to do parallel processing without having to glue a bunch of computers together is wonderful! The ability to "cut and paste" different functions like video, keyboard, communications, etc, is even more significant.

These are all wonderful things, but the fact that a critical mass of individuals is willing to contribute to the over-all capabilities of this technology ( and each other) is probably the most significant factor in the potential growth of the Parallax company/community.

I'm sitting here in my engineering "Nest" with 8 different Windows systems staring back at me... I use Windows in it's various guises because I have to, but I've been keeping close watch on Linux over the past couple of years and the momentum that it is developing and speed that its gaining is incredible. It makes me think that the "Switch over" timeframe isn't too far out in the future.

If you want to crank the "Crystal Ball" into high gear about where the Propeller might be going, keep in mind that there is already a "Free" operating system out there designed to take advantage of multiple processors. Some of the fastest computers in the world are nothing more than a bunch of desktop PCs linked together and running applications designed to divy-up certain applications to these individual processors then bring the results back when everybody’s done with their part.

If Moore’s Law keeps going for just a few more years, We’ll be seeing a “Scram Jet (Propeller) with 64 - 256 processors., each with more horsepower than the current Propeller. If you stare just a bit deeper into the crystal, you see that each of these processors has it’s own 50 gig. “thumb drive” ,and wireless networking. You won’t even need to wire chips together, just put them close to each other and give ‘em power. They’ll figure out who’s boss.

The hardware isn’t going to be the tricky part, it’s going to be keeping track of who does what does what where, when and how. There is an “Operating System” in our future if for nothing else than to keep track of what resources are available (both on-board and world-wide) and what needs to be done at any given moment in time.

… but I digress…. Lol . My over-all point is that Parallax has fired the first shot in the parallelism wars ( dual core Pentiums don’t really count ) and this Forum , the object exchange library, open software, and open communications between all of us are the cruise missiles about to take on the breech-loaders of yesteryear.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
" Anything worth doing... is worth overdoing. "

··············································· ( R.A.H. )
····································

Mike Green

Do keep in mind in all this exuberance that Parallax and the Propeller are intended for a particular market, that of low level control functions. We're already seeing multiprocessors on one chip with the latest Intel processors. What's different about the Propeller is that Chip has very nicely optimized the processor instruction set for control functions. In particular, there are all kinds of bit operations (and some more projected for the next generation processor), the counters are very powerful and, with some possible changes to the PLL, even more versatile for waveform generation. Some areas for development include better intra-COG communication. Extra memory and additional COGs is mostly a problem of chip density and layout, not fundamental architecture. Speed will get better with improved processes and materials and some structural redesign (ie pipelining), but there is likely not going to be a lot of scaling up, partly because the advantages of its design is in its simplicity, straightforwardness, and low power for the speed and capability. The Propeller is likely to be able to be its own development platform. There is enough power if you add some external memory. That'll be nice, but there'll always be advantages in using whatever desktop power is available. I've worked with several ancient computers with much less memory and speed than the Propeller that supported their own development environment. At the time, there wasn't much else available and it did work! There was no GUI, no on-line help, no windowing, etc. Anyway, as we dream along, it'll be important to keep remembering what's different about the Propeller compared to most other microcontrollers and computers in general and where those differences have their power. Remember that some of the early cathode ray tube memory computers had similar instruction sets (like conditional execution, optional writing of results, and small program memory spaces).

Paul Baker

Mike is right, the Propeller is squarly aimed at the embedded market. It will never be aimed at the general computing market. Considering how saturated that market already is, its a wise course for Parallax. Bill Gates has recognized the next "frontier" to be the embedded market for more than a decade now. It just that his history has encumbered him to fully capitalize on that realization (primarily because his history is in software, not hardware). Meanwhile those who are already in the market suffer from thier own inertia of incremental improvements on existing architectures. The Propeller is poised to upset that apple cart and force other companies to broaden thier currently narrow outlook. While the Propeller has definite room to advance, there is an eventual practical limit to it's path. One of the principle design concepts behind the Propeller is that unused silicon is wasted silicon. This is a carry over from the SX and the virtual peripheral concept. In the same vain, if a 256 cog Propeller on average uses a 1/4 of those cogs, you've wasted a vast amount of silicon (and money).

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
...