How was the Propeller chip made?

Microcontrolled

I just saw that thread on how many transistors the Prop has and I got thinking, how did Parallax make the Propeller?

Then being the curious kid I am came up with the following questions.

1. How did Parallax design the Propeller? Is there software or did they do it transistor by transistor?

2. What was the idea that sparked the creation of the Propeller?

3. Why does it take so long for a run of chips to be made?

4. What makes the transistors work if the Prop is just a design etched into silicon?

5. Why did they call it a "Propeller Chip" and why did they come up with the logo? It dosn't really make since.....

Well, there are my questions, even the non technical ones. Hope you can answer them.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Computers are microcontrolled.

Robots are microcontrolled.
I am microcontrolled.

But you·can·call me micro.

Want to·experiment with the SX or just put together a cool project?
SX Spinning light display·
Want cheap wholesale electronic parts?
Transistor parts wholesale

rough_wood

I would guess it's called Propeller due to the "Spinning Hub" that selects cogs.

The logo, I take it as a beani propeller hat. It's sort of a light hearted geeky smart nerdy kid symbol I suppose.

As far as why, from what I've seen, Chip has a real passion for this and it's his form of art. He can see what's wrong with other processors and programming languages, or more likely he wishes to put his own "spin" on things. This guy is posting in forums over Thanksgiving break, it shows he loves his job and probably does it for the joy and art, and not the money (though I'm sure that's nice too) [noparse]:)[/noparse]

As far as the rest of the questions, I'm third year Mechanical Engineering so I have no idea.

rough_wood

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

Bob Lawrence (VE1RLL)

Re: 1. How did Parallax design the Propeller? Is there software or did they do it transistor by transistor?

Chip used a Altera FPGA development system. Check out the Altera website with this link.

www.altera.com/

As a matter of fact, if Chip keeps delaying Prop 2 I'll just download the Altera software and make my own, just as soon as I get that Parallax Asm training on Dec 10th and a little practise.

Post Edited (Bob Lawrence (VE1RLL)) : 11/29/2009 4:46:33 AM GMT

Alsowolfman

1)The Propeller was designed in an HDL, or hardware descriptive language. I believe chip uses AHDL, but the most common ones are verilog and vhdl if you want to find more information. They do not work on a transistor by transistor level, it is done by describing areas where information can be stored called registers, and describing the logic the connect those registers, this is the register transfer level http://en.wikipedia.org/wiki/Register_transfer_level . from there a type of compiler converts your program into a series interconnected gates, then another program will map those gates onto a field programmable gate array for testing, and finally produce a file that can be used in the making of the silicon circuits i believe. you can get the development tools for free from altera or xilinx that can edit for vhdl ahdl, and veirlog. digikey sells a few fpga demo boards that are within hobby range 150-200 for an altera device, you would not be able to do a whole prop, but you could definitely program a single cog.

2)my understanding is that it is a reaction to the general move towards faster chips that had more and more complex interrupts.


4)this question is sort of vague. are you asking how transistors function and how millions of them become a processor, or how a processor is manufactured? or are you asking what the transistors do in the time between them receiving power, and when they start running your code? here a pdf on the manufacturing process from intel http://download.intel.com/pressroom/kits/chipmaking/Making_of_a_Chip.pdf . For how they work i think this requires you to do research to get an answer that is meaningful to you, but i can give you starting points. transistors are used to make up simple logic gates such as and, or, xor... as well as flip flops, flip flops are like a single bit of data storage. (http://en.wikipedia.org/wiki/D_flip_flop#D_flip-flop) the d flip flops can be used in parallel to store groups of information like Bytes and longs, this is what a register is. registers combined with the logic gates then give us our math functions. This is not meant to be comprehensive, but should help you ask a more specific questions if that is what you were asking. and if you were asking what it does at power on i think the answer to that would require investigating how a processor works in general, and reading the manual on the propellers specific boot protocols.

a good way to get to understanding the prop more is learning assembly [noparse]:)[/noparse]

Drone

The process involved with building and actually selling a new micro-controller is complex. In my understanding (may be wrong)... You start with an idea for a new micro-controller, then code the combinatorial/sequential logic in xHDL for the likes of an FPGA. But the xHDL design includes a "core" micro-controller base in also in xHDL which you either license from someone or start with a free open-source core or build it from scratch (ouch); see www.opencores.org.

You pick-apart modify and and take the parts you like from the core to meet at the end functional requirement for the new controller. Then you simulate the new micro physically in the target FPGA to make sure it does what you want. The next step is to move from the xHDL code to an actual physical part that is manufactured as an application specific integrated circuit (ASIC) by a "Fab" (typically a Taiwan-based chip Fabricator).

Optimizing the xHDL code is first done to some degree by the FPGA design software provided by the manufacturer automatically and is verified in simulation. But in the end, there is likely a lot of "tweaking" that must be performed by a Human that understands the mathematics and techniques of combinatorial/sequential reduction.

Moving to the Fab is very a complex step as the physical characteristics of the resultant part depends very much on how you specify the type and layout of transistors, diodes etc. on the eventual ASIC chip die to be produced in volume (not to mention choosing substrate technology, physical design rules and nm technology, physical package types and die interconnect, thermal simulation/specification, on-and-on). IC-Layout is complex and is as much technical as it is an art; imagine laying out a multi-layer home-made PC-board times a thousand.

The tools to do the ASIC/IC layout are expensive and have a steep learning curve. For smaller companies, end-run IC layout is often outsourced to to the Fab or experts/consultants/contractors that do nothing but IC layout for a living. Likely some pre-production fabrication runs are done to physically and functionally check the design before going to volume production (likely time-consuming and expensive). Once you have a final design the Fab can go to volume production. Once you start taking chips from the Fab in volume, the design is locked-in. Any hardware/design/logic/interface issues are there to stay unless you iterate for a new part all over again. (I venture to guess Parallax wished they protected the Prop V1 PLL circuitry a bit better in hind-sight).

@microcontrolled, your question is excellent. As the Propeller is used in many educational environments, IMHO it would be nice if Parallax described the process from idea to physical part for all of us to see. I"m sure there would be nothing proprietary in what's revealed, just an overview of what it takes to make a new micro-controller these days.

Caveat; I'm not a micro-controller designer. I'm just stating what I understand it takes to make a new part (corrections most welcome). I do use micros as well as FPGA/CPLD parts in my work (commercial or otherwise). I'm more of a "Systems Engineer", putting the parts others (like Parallax) sell to make "bigger" things.

Regards, David

heater

Have a read of "Why the Propeller Works" by Chip Gracey.
From here www.parallax.com/tabid/442/Default.aspx
A most fascinating story.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
For me, the past is not over yet.

Alsowolfman

Here is a thread linking to an interview with Chip Gracey. around 3:00-5:00 in part 2 of 7 he talks about some of these questions.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
My head is pasming!

Nick McClick

I've never ordered custom silicon before, but I think the reason why it takes so long for a batch of IC's to be made is lead time;

The equipment to make Integrated Circuits is incredibly expensive, so the foundry has to make sure it's always being used (if it's sitting idle, then it can't make money and pay for itself). So the production equipment is booked well in advance. You can't just 'slip' something into the production schedule. There's a lead time like that at multiple steps along the way (packaging, inspection, etc) and it all adds up to a good chunk of time.

Even the lead time for something very simple (say an sensor board) is several weeks. Think about all the steps;
1 - Do a prototype run of boards (5 Days)
2 - Test and validate the design (2-3 Days)
3 - Order a production run of boards (10 Days)
4 - Order parts (2-3 Days)
5 - Order stencil (5 Days)
6 - Get everything assembled (4 Days)
7 - Get everything sent back (3 Days)
8 - Test parts (2 Days)
9 - Package (1 Day)

Some steps can be done simultaneously, but it's still a lengthy process - and that's something very basic! If there's a hiccup anywhere (and there always is), it extends everything out.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Gadget Gangster - Share your Electronic Projects