Single Chip Lua Prospects

[John Hind]

When I subscribed to this list half a year ago I confidently expected to be able to use a single chip Lua device in my projects by now. I wanted to use many such chips in a seriously distributed architecture with the processing kept near to the IO to simplify wiring. To avoid buying an endless succession of development boards I set a simple threshold specification of 512K on-chip flash, 64K on-chip RAM, Ethernet capability (I was prepared to compromise with a second chip for this). With this type of distributed architecture, I need lots of IO versatility (analog, PWM, UART, I2C, SPI, counters etc.) but multiplexed on relatively few pins since if you run out of IO you just add another chip. It should therefore be possible to have a simple 20-40 pin package which could be available in hand-solderable form.

 

It is very disappointing that after many months still the only chip that comes near is the LPC2888 which seems to be buggy and expensive (the very minimalist LPC-H2888 prototype board is over $150). Looking at the schematics for this board really reinforced my intuition about the importance of on-chip memory – a ridiculous number of pins (and therefore PCB traces) is wasted on the external memory interface. And the economics are terrible - for just $20 more, you can get the new Acer Aspire Revo R3600 nettop (in Linux rather than Vista trim). The latter has both wired and wireless networking, 1GB RAM and an 8GB SSD!

 

IO is the problem of course with an X86 platform. However in my research I found this incredibly poorly marketed chip:

 

http://delcomproduct.com/products_USBIO.asp#DemoBrd

 

A versatile USB to IO pins chip for only $8! This looks like great universal IO problem solver. Unlike most other DIY USB solutions I’ve seen which require a driver and/or serial port emulator, this is HID based and therefore driverless on all major OS’s. I have ordered some of these chips and will report back how it goes.

 

I am beginning to think it might just be better to cooperate with market forces here since I suspect that X86 computing is going to keep widening the functionality gap and reducing the price gap. Rather than trying to squeeze Lua onto a cramped chip with no operating system, would it maybe be better to produce a custom cut-down Linux distro with Lua built in (and perhaps acting as the shell)?

 

What do others think? Will there be cheap, single chip ARM devices meeting my spec available in the next few months?

 

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[BogdanM]

This idea has certainly crossed my mind. A stripped down (uC)Linux distribution running eLua and a sort of I/O expander like the one you mentioned might do the trick. Hack, you might not even need Linux at all. The standalone eLua distribution for PC is a good start for a fully standalone 'make your PC boot into eLua' distro (and it's enough to visit osdir.com to see that expanding this to handle things like disk I/O, USB or Ethernet isn't an impossible task). And this is something that interests me a lot, but sadly I don't have enough resources to advance on this front. Personally I'll keep on extending eLua for micros. But if you (or anybody else) are interested more in the PC (or netbook or whatever other variation) approach, then by all means, go for it :)

Best,
Bogdan

On Mon, Jun 22, 2009 at 1:42 PM, John Hind <[hidden email]> wrote:

When I subscribed to this list half a year ago I confidently expected to be able to use a single chip Lua device in my projects by now. I wanted to use many such chips in a seriously distributed architecture with the processing kept near to the IO to simplify wiring. To avoid buying an endless succession of development boards I set a simple threshold specification of 512K on-chip flash, 64K on-chip RAM, Ethernet capability (I was prepared to compromise with a second chip for this). With this type of distributed architecture, I need lots of IO versatility (analog, PWM, UART, I2C, SPI, counters etc.) but multiplexed on relatively few pins since if you run out of IO you just add another chip. It should therefore be possible to have a simple 20-40 pin package which could be available in hand-solderable form.

 

It is very disappointing that after many months still the only chip that comes near is the LPC2888 which seems to be buggy and expensive (the very minimalist LPC-H2888 prototype board is over $150). Looking at the schematics for this board really reinforced my intuition about the importance of on-chip memory – a ridiculous number of pins (and therefore PCB traces) is wasted on the external memory interface. And the economics are terrible - for just $20 more, you can get the new Acer Aspire Revo R3600 nettop (in Linux rather than Vista trim). The latter has both wired and wireless networking, 1GB RAM and an 8GB SSD!

 

IO is the problem of course with an X86 platform. However in my research I found this incredibly poorly marketed chip:

 

http://delcomproduct.com/products_USBIO.asp#DemoBrd

 

A versatile USB to IO pins chip for only $8! This looks like great universal IO problem solver. Unlike most other DIY USB solutions I’ve seen which require a driver and/or serial port emulator, this is HID based and therefore driverless on all major OS’s. I have ordered some of these chips and will report back how it goes.

 

I am beginning to think it might just be better to cooperate with market forces here since I suspect that X86 computing is going to keep widening the functionality gap and reducing the price gap. Rather than trying to squeeze Lua onto a cramped chip with no operating system, would it maybe be better to produce a custom cut-down Linux distro with Lua built in (and perhaps acting as the shell)?

 

What do others think? Will there be cheap, single chip ARM devices meeting my spec available in the next few months?

 

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Dado Sutter]

Hello,
   The boundaries between a gronwup embedded environment and a linux box are indeed pretty vague these days. This is one of the reasons we don't have a clear choice for our first eLua Stamp yet. I am personally confused with the linux box solutions trend and feel bad for not having too much time to keep an eye on this.
   But I don't think a X86 platform, enhanced with some more I/O capability and running Lua with peripheral handling extensions will be the best answer for the typical dedicated hardware platforms we all dream of, although it can certainly have it's place too.
   Development boards/platforms are indeed expensive, because they sell in relative low quantities. But it is not too hard to build a nice board for eLua in for under US$ 50, in 10~50 quantities. We are actually buiding an eLua board for some specific projects downhere, arround an LPC, with 8Mb external RAM, 512Kb flash and it would come out about this prices, if we have not to build it in a country like Brazil, where import taxes are a joke and PCB building costs are abusive. I am pretty sure that later this year, with the help of you guys, we will be able to come up with a nice and cheap eLua Stamp that fits well in our pockets :)0
   Our motivations for eLua maybe somewhat different from the view of "let's embedd Lua on the best cost x benefit platform on the market". What we see in the field instrumentation area (downhere) is that the hardware is becoming more "inteligent", moving from 8 bit platforms (that still dominate 82% of the market, according to Jack Ganssle) straight to 32 bits or to hibrid architectures, where the 8 bit devices become "satelites" of a central and more powerfull (32 bit) controller.
   This scenario clearly asks for better software tools and specially for better development environments, where the final user, application/process engineers and sensor technitians can program more functionality in their products without the need of years of experience in Desktop-based development, C/Asm languages etc....  To offer something powerfull and simple as Lua, with peripheral handling libs/extensions/modules not only for the peripheral world devices (ie: spi, uart, ad/da, ...) but for final project "parts" (ie: gps, displays, kbds, ...), with no need for a PC-based development for the upper "layers" of the aplications seems to be a nice goal too.
    I think the X86 architecture will always be a strong "market force", although many low end netbooks (and that's what netbooks are, low end laptops :) are also tasting ARM, for still lower prices and sizes. Linux fits well on them all and may very acomodate Lua with embedded extensions to offer what eLua wants to offer otherwise. But imho, the dedicated sensor/instrumentation/appliance dedicated controllers will not be replaced by these generic and customized linux-box based solutions, even considering their price drop.
   Also, both solutions can coexist very well, leaving the pc-based and software-expert development with the linux boxes and final-user and process/product professionals with eLua.

Best
Dado




  

On Mon, Jun 22, 2009 at 07:42, John Hind <[hidden email]> wrote:

When I subscribed to this list half a year ago I confidently expected to be able to use a single chip Lua device in my projects by now. I wanted to use many such chips in a seriously distributed architecture with the processing kept near to the IO to simplify wiring. To avoid buying an endless succession of development boards I set a simple threshold specification of 512K on-chip flash, 64K on-chip RAM, Ethernet capability (I was prepared to compromise with a second chip for this). With this type of distributed architecture, I need lots of IO versatility (analog, PWM, UART, I2C, SPI, counters etc.) but multiplexed on relatively few pins since if you run out of IO you just add another chip. It should therefore be possible to have a simple 20-40 pin package which could be available in hand-solderable form.

 

It is very disappointing that after many months still the only chip that comes near is the LPC2888 which seems to be buggy and expensive (the very minimalist LPC-H2888 prototype board is over $150). Looking at the schematics for this board really reinforced my intuition about the importance of on-chip memory – a ridiculous number of pins (and therefore PCB traces) is wasted on the external memory interface. And the economics are terrible - for just $20 more, you can get the new Acer Aspire Revo R3600 nettop (in Linux rather than Vista trim). The latter has both wired and wireless networking, 1GB RAM and an 8GB SSD!

 

IO is the problem of course with an X86 platform. However in my research I found this incredibly poorly marketed chip:

 

http://delcomproduct.com/products_USBIO.asp#DemoBrd

 

A versatile USB to IO pins chip for only $8! This looks like great universal IO problem solver. Unlike most other DIY USB solutions I’ve seen which require a driver and/or serial port emulator, this is HID based and therefore driverless on all major OS’s. I have ordered some of these chips and will report back how it goes.

 

I am beginning to think it might just be better to cooperate with market forces here since I suspect that X86 computing is going to keep widening the functionality gap and reducing the price gap. Rather than trying to squeeze Lua onto a cramped chip with no operating system, would it maybe be better to produce a custom cut-down Linux distro with Lua built in (and perhaps acting as the shell)?

 

What do others think? Will there be cheap, single chip ARM devices meeting my spec available in the next few months?

 

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[John Hind]

Hi Dado, you had me googling “gronwup embedded environment” before realising it was a typo!

 

My posting was more born of frustration that the ARM microcontroller industry seems to have advanced so little in the last six months compared to the rate the X86 market moves on. I was really hoping someone would point me at some hope the next six months would bring me what I was looking for. I do not want to give up on distributed processing architectures yet, but it is hard when just one processing node ends up nearly as expensive as a central machine capable of handling the entire requirement!

 

It continues to baffle me that the industry does not get the market for high-functionality, highly integrated, low pin-count devices – this seems so obvious for intelligent sensors and actuators. Even in the 8-bit field, Microchip makes lovely 6- and 8- pin chips ideal for embedding in, for example, a single limit switch, but they have no serial comms capability! You are up to 16 pin devices before that starts to appear. In the ARM market, large on-chip memory seems to go hand in hand with high pin-count and therefore complex and expensive packages and PCBs.

 

Are the boards you are working on network capable?

 

From: [hidden email] [mailto:[hidden email]] On Behalf Of Dado Sutter
Sent: 22 June 2009 14:00
To: eLua Users and Development List
Subject: Re: [eLua-dev] Single Chip Lua Prospects

 

Hello,
   The boundaries between a gronwup embedded environment and a linux box are indeed pretty vague these days. This is one of the reasons we don't have a clear choice for our first eLua Stamp yet. I am personally confused with the linux box solutions trend and feel bad for not having too much time to keep an eye on this.
   But I don't think a X86 platform, enhanced with some more I/O capability and running Lua with peripheral handling extensions will be the best answer for the typical dedicated hardware platforms we all dream of, although it can certainly have it's place too.
   Development boards/platforms are indeed expensive, because they sell in relative low quantities. But it is not too hard to build a nice board for eLua in for under US$ 50, in 10~50 quantities. We are actually buiding an eLua board for some specific projects downhere, arround an LPC, with 8Mb external RAM, 512Kb flash and it would come out about this prices, if we have not to build it in a country like Brazil, where import taxes are a joke and PCB building costs are abusive. I am pretty sure that later this year, with the help of you guys, we will be able to come up with a nice and cheap eLua Stamp that fits well in our pockets :)0
   Our motivations for eLua maybe somewhat different from the view of "let's embedd Lua on the best cost x benefit platform on the market". What we see in the field instrumentation area (downhere) is that the hardware is becoming more "inteligent", moving from 8 bit platforms (that still dominate 82% of the market, according to Jack Ganssle) straight to 32 bits or to hibrid architectures, where the 8 bit devices become "satelites" of a central and more powerfull (32 bit) controller.
   This scenario clearly asks for better software tools and specially for better development environments, where the final user, application/process engineers and sensor technitians can program more functionality in their products without the need of years of experience in Desktop-based development, C/Asm languages etc....  To offer something powerfull and simple as Lua, with peripheral handling libs/extensions/modules not only for the peripheral world devices (ie: spi, uart, ad/da, ...) but for final project "parts" (ie: gps, displays, kbds, ...), with no need for a PC-based development for the upper "layers" of the aplications seems to be a nice goal too.
    I think the X86 architecture will always be a strong "market force", although many low end netbooks (and that's what netbooks are, low end laptops :) are also tasting ARM, for still lower prices and sizes. Linux fits well on them all and may very acomodate Lua with embedded extensions to offer what eLua wants to offer otherwise. But imho, the dedicated sensor/instrumentation/appliance dedicated controllers will not be replaced by these generic and customized linux-box based solutions, even considering their price drop.
   Also, both solutions can coexist very well, leaving the pc-based and software-expert development with the linux boxes and final-user and process/product professionals with eLua.

Best
Dado




  

On Mon, Jun 22, 2009 at 07:42, John Hind <[hidden email]> wrote:

When I subscribed to this list half a year ago I confidently expected to be able to use a single chip Lua device in my projects by now. I wanted to use many such chips in a seriously distributed architecture with the processing kept near to the IO to simplify wiring. To avoid buying an endless succession of development boards I set a simple threshold specification of 512K on-chip flash, 64K on-chip RAM, Ethernet capability (I was prepared to compromise with a second chip for this). With this type of distributed architecture, I need lots of IO versatility (analog, PWM, UART, I2C, SPI, counters etc.) but multiplexed on relatively few pins since if you run out of IO you just add another chip. It should therefore be possible to have a simple 20-40 pin package which could be available in hand-solderable form.

 

It is very disappointing that after many months still the only chip that comes near is the LPC2888 which seems to be buggy and expensive (the very minimalist LPC-H2888 prototype board is over $150). Looking at the schematics for this board really reinforced my intuition about the importance of on-chip memory – a ridiculous number of pins (and therefore PCB traces) is wasted on the external memory interface. And the economics are terrible - for just $20 more, you can get the new Acer Aspire Revo R3600 nettop (in Linux rather than Vista trim). The latter has both wired and wireless networking, 1GB RAM and an 8GB SSD!

 

IO is the problem of course with an X86 platform. However in my research I found this incredibly poorly marketed chip:

 

http://delcomproduct.com/products_USBIO.asp#DemoBrd

 

A versatile USB to IO pins chip for only $8! This looks like great universal IO problem solver. Unlike most other DIY USB solutions I’ve seen which require a driver and/or serial port emulator, this is HID based and therefore driverless on all major OS’s. I have ordered some of these chips and will report back how it goes.

 

I am beginning to think it might just be better to cooperate with market forces here since I suspect that X86 computing is going to keep widening the functionality gap and reducing the price gap. Rather than trying to squeeze Lua onto a cramped chip with no operating system, would it maybe be better to produce a custom cut-down Linux distro with Lua built in (and perhaps acting as the shell)?

 

What do others think? Will there be cheap, single chip ARM devices meeting my spec available in the next few months?

 

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

 

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Dado Sutter]

On Mon, Jun 22, 2009 at 11:34, John Hind <[hidden email]> wrote:

Hi Dado, you had me googling “gronwup embedded environment” before realising it was a typo!

huahehae :) I'm sorry John !!! Pls (you all) orgive my poor english, not only for silly typing errors like this but for the general ideas. I'm not sure if I was clear enough about this topic, that I consider an important and delicate one at this moment and I'll try to make myself more clear next time :)
 

My posting was more born of frustration that the ARM microcontroller industry seems to have advanced so little in the last six months compared to the rate the X86 market moves on. I was really hoping someone would point me at some hope the next six months would bring me what I was looking for. I do not want to give up on distributed processing architectures yet, but it is hard when just one processing node ends up nearly as expensive as a central machine capable of handling the entire requirement!

This is not the impression I have from here.  Although I would also like to have much (and cheaper) more boards available for eLua, specially with some more internal memory, I have the impression they are comming pretty fast now.

 It continues to baffle me that the industry does not get the market for high-functionality, highly integrated, low pin-count devices – this seems so obvious for intelligent sensors and actuators.

Yes, this is strange and also the trend to include now the high-end peripherals in the same families, like video/audio stream processors. It would certainly be nice for us to see it going in the other direction, the one you mention, low pin-count devices, easy to use (DIP, ....) and cheap.

Even in the 8-bit field, Microchip makes lovely 6- and 8- pin chips ideal for embedding in, for example, a single limit switch, but they have no serial comms capability! You are up to 16 pin devices before that starts to appear. In the ARM market, large on-chip memory seems to go hand in hand with high pin-count and therefore complex and expensive packages and PCBs.

Yes, very true :)
But let's keep our eyes open. The market will never be formed arround a single chip but arround many diferent options. That's why we want eLua to be so portable.
Meanwhile, we'll keep looking for a good candidate for our first eLua Stamp (I propose to start this thread immediately :) and hearing the findings of our members. Snyder had found a while ago a neat little cute board from England (I guess) but we've had no news since.

Are the boards you are working on network capable?

Yes, PNY included.
If everything goes well, we may have 5 working withing 2~3 weeks.

Abraçossssssssssssss
Dado
 

 

From: [hidden email] [mailto:[hidden email]] On Behalf Of Dado Sutter
Sent: 22 June 2009 14:00

To: eLua Users and Development List

Subject: Re: [eLua-dev] Single Chip Lua Prospects

 

Hello,
   The boundaries between a gronwup embedded environment and a linux box are indeed pretty vague these days. This is one of the reasons we don't have a clear choice for our first eLua Stamp yet. I am personally confused with the linux box solutions trend and feel bad for not having too much time to keep an eye on this.
   But I don't think a X86 platform, enhanced with some more I/O capability and running Lua with peripheral handling extensions will be the best answer for the typical dedicated hardware platforms we all dream of, although it can certainly have it's place too.
   Development boards/platforms are indeed expensive, because they sell in relative low quantities. But it is not too hard to build a nice board for eLua in for under US$ 50, in 10~50 quantities. We are actually buiding an eLua board for some specific projects downhere, arround an LPC, with 8Mb external RAM, 512Kb flash and it would come out about this prices, if we have not to build it in a country like Brazil, where import taxes are a joke and PCB building costs are abusive. I am pretty sure that later this year, with the help of you guys, we will be able to come up with a nice and cheap eLua Stamp that fits well in our pockets :)0
   Our motivations for eLua maybe somewhat different from the view of "let's embedd Lua on the best cost x benefit platform on the market". What we see in the field instrumentation area (downhere) is that the hardware is becoming more "inteligent", moving from 8 bit platforms (that still dominate 82% of the market, according to Jack Ganssle) straight to 32 bits or to hibrid architectures, where the 8 bit devices become "satelites" of a central and more powerfull (32 bit) controller.
   This scenario clearly asks for better software tools and specially for better development environments, where the final user, application/process engineers and sensor technitians can program more functionality in their products without the need of years of experience in Desktop-based development, C/Asm languages etc....  To offer something powerfull and simple as Lua, with peripheral handling libs/extensions/modules not only for the peripheral world devices (ie: spi, uart, ad/da, ...) but for final project "parts" (ie: gps, displays, kbds, ...), with no need for a PC-based development for the upper "layers" of the aplications seems to be a nice goal too.
    I think the X86 architecture will always be a strong "market force", although many low end netbooks (and that's what netbooks are, low end laptops :) are also tasting ARM, for still lower prices and sizes. Linux fits well on them all and may very acomodate Lua with embedded extensions to offer what eLua wants to offer otherwise. But imho, the dedicated sensor/instrumentation/appliance dedicated controllers will not be replaced by these generic and customized linux-box based solutions, even considering their price drop.
   Also, both solutions can coexist very well, leaving the pc-based and software-expert development with the linux boxes and final-user and process/product professionals with eLua.

Best
Dado




  

On Mon, Jun 22, 2009 at 07:42, John Hind <[hidden email]> wrote:

When I subscribed to this list half a year ago I confidently expected to be able to use a single chip Lua device in my projects by now. I wanted to use many such chips in a seriously distributed architecture with the processing kept near to the IO to simplify wiring. To avoid buying an endless succession of development boards I set a simple threshold specification of 512K on-chip flash, 64K on-chip RAM, Ethernet capability (I was prepared to compromise with a second chip for this). With this type of distributed architecture, I need lots of IO versatility (analog, PWM, UART, I2C, SPI, counters etc.) but multiplexed on relatively few pins since if you run out of IO you just add another chip. It should therefore be possible to have a simple 20-40 pin package which could be available in hand-solderable form.

 

It is very disappointing that after many months still the only chip that comes near is the LPC2888 which seems to be buggy and expensive (the very minimalist LPC-H2888 prototype board is over $150). Looking at the schematics for this board really reinforced my intuition about the importance of on-chip memory – a ridiculous number of pins (and therefore PCB traces) is wasted on the external memory interface. And the economics are terrible - for just $20 more, you can get the new Acer Aspire Revo R3600 nettop (in Linux rather than Vista trim). The latter has both wired and wireless networking, 1GB RAM and an 8GB SSD!

 

IO is the problem of course with an X86 platform. However in my research I found this incredibly poorly marketed chip:

 

http://delcomproduct.com/products_USBIO.asp#DemoBrd

 

A versatile USB to IO pins chip for only $8! This looks like great universal IO problem solver. Unlike most other DIY USB solutions I’ve seen which require a driver and/or serial port emulator, this is HID based and therefore driverless on all major OS’s. I have ordered some of these chips and will report back how it goes.

 

I am beginning to think it might just be better to cooperate with market forces here since I suspect that X86 computing is going to keep widening the functionality gap and reducing the price gap. Rather than trying to squeeze Lua onto a cramped chip with no operating system, would it maybe be better to produce a custom cut-down Linux distro with Lua built in (and perhaps acting as the shell)?

 

What do others think? Will there be cheap, single chip ARM devices meeting my spec available in the next few months?

 

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

 

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Louis Mamakos]

On Jun 22, 2009, at 10:34 AM, John Hind wrote:

Hi Dado, you had me googling “gronwup embedded environment” before realising it was a typo!

 

My posting was more born of frustration that the ARM microcontroller industry seems to have advanced so little in the last six months compared to the rate the X86 market moves on. I was really hoping someone would point me at some hope the next six months would bring me what I was looking for. I do not want to give up on distributed processing architectures yet, but it is hard when just one processing node ends up nearly as expensive as a central machine capable of handling the entire requirement!

 

It continues to baffle me that the industry does not get the market for high-functionality, highly integrated, low pin-count devices – this seems so obvious for intelligent sensors and actuators. Even in the 8-bit field, Microchip makes lovely 6- and 8- pin chips ideal for embedding in, for example, a single limit switch, but they have no serial comms capability! You are up to 16 pin devices before that starts to appear. In the ARM market, large on-chip memory seems to go hand in hand with high pin-count and therefore complex and expensive packages and PCBs.

In the embedded space, parts get built to support specific (hopefully volume) applications.  This is why you might not see parts with a large number of on-chip peripherals, or lots of memory, etc.  In a previous professional gig, we built VoIP devices that shipped in moderate volume (1M-2M units/year).  This is a nice "sweet spot" for a part manufacturer who would build a device targeted to that application which also happens to be price sensitive.  You don't want to ship parts out with unused peripherals or lots of extra memory because an extra $2 of cost per unit turns into a pretty noticeable numbers at 1,000,000 units shipped. 

If you're only talking about building a few hundred units, then the extra cost isn't really as much of a consideration, especially when you might be able to reduce the one-time engineering expense by optimizing time spent vs. "extra" hardware.

I think that Marvell is building some interesting  higher-end parts, with perhaps a lack of GPIO I/O, A/D, etc. capability but with networking, etc.  Otherwise, you're looking to repurpose devices built for some targeted applications.

louie

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Asko Kauppi]

Atmel has a new AVR32 chip that _should_ be suitable. We're still  
waiting for promised samples to Dado and Bogdan, though.

See A3 series: http://www.atmel.com/dyn/products/devices.asp?family_id=682

-asko


John Hind kirjoitti 22.6.2009 kello 13:42:

> When I subscribed to this list half a year ago I confidently  
> expected to be able to use a single chip Lua device in my projects  
> by now. I wanted to use many such chips in a seriously distributed  
> architecture with the processing kept near to the IO to simplify  
> wiring. To avoid buying an endless succession of development boards  
> I set a simple threshold specification of 512K on-chip flash, 64K on-
> chip RAM, Ethernet capability (I was prepared to compromise with a  
> second chip for this). With this type of distributed architecture, I  
> need lots of IO versatility (analog, PWM, UART, I2C, SPI, counters  
> etc.) but multiplexed on relatively few pins since if you run out of  
> IO you just add another chip. It should therefore be possible to  
> have a simple 20-40 pin package which could be available in hand-
> solderable form.
>
> It is very disappointing that after many months still the only chip  
> that comes near is the LPC2888 which seems to be buggy and expensive  
> (the very minimalist LPC-H2888 prototype board is over $150).  
> Looking at the schematics for this board really reinforced my  
> intuition about the importance of on-chip memory – a ridiculous  
> number of pins (and therefore PCB traces) is wasted on the external  
> memory interface. And the economics are terrible - for just $20  
> more, you can get the new Acer Aspire Revo R3600 nettop (in Linux  
> rather than Vista trim). The latter has both wired and wireless  
> networking, 1GB RAM and an 8GB SSD!
>
> IO is the problem of course with an X86 platform. However in my  
> research I found this incredibly poorly marketed chip:
>
> http://delcomproduct.com/products_USBIO.asp#DemoBrd
>
> A versatile USB to IO pins chip for only $8! This looks like great  
> universal IO problem solver. Unlike most other DIY USB solutions  
> I’ve seen which require a driver and/or serial port emulator, this  
> is HID based and therefore driverless on all major OS’s. I have  
> ordered some of these chips and will report back how it goes.
>
> I am beginning to think it might just be better to cooperate with  
> market forces here since I suspect that X86 computing is going to  
> keep widening the functionality gap and reducing the price gap.  
> Rather than trying to squeeze Lua onto a cramped chip with no  
> operating system, would it maybe be better to produce a custom cut-
> down Linux distro with Lua built in (and perhaps acting as the shell)?
>
> What do others think? Will there be cheap, single chip ARM devices  
> meeting my spec available in the next few months?
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[jbsnyder]

On Jun 22, 2009, at 5:42 AM, John Hind wrote:

> When I subscribed to this list half a year ago I confidently  
> expected to be able to use a single chip Lua device in my projects  
> by now. I wanted to use many such chips in a seriously distributed  
> architecture with the processing kept near to the IO to simplify  
> wiring. To avoid buying an endless succession of development boards  
> I set a simple threshold specification of 512K on-chip flash, 64K on-
> chip RAM, Ethernet capability (I was prepared to compromise with a  
> second chip for this). With this type of distributed architecture, I  
> need lots of IO versatility (analog, PWM, UART, I2C, SPI, counters  
> etc.) but multiplexed on relatively few pins since if you run out of  
> IO you just add another chip. It should therefore be possible to  
> have a simple 20-40 pin package which could be available in hand-
> solderable form.

What about in stamp form factor?

It would seem to me that this does a lot of what you're asking for  
(aside from ethernet):
http://www.futurlec.com/ET-STM32_Stamp.shtml

It does, however, have built-in CAN.

On Jun 22, 2009, at 9:51 AM, Dado Sutter wrote:
> Yes, very true :)
> But let's keep our eyes open. The market will never be formed  
> arround a single chip but arround many diferent options. That's why  
> we want eLua to be so portable.
> Meanwhile, we'll keep looking for a good candidate for our first  
> eLua Stamp (I propose to start this thread immediately :) and  
> hearing the findings of our members. Snyder had found a while ago a  
> neat little cute board from England (I guess) but we've had no news  
> since.

I still have that board, and still need to finish the port for it :-)

It's an interesting board, and provide some nice interfaces for  
working with the various onboard peripherals:
http://mbed.co.uk/beta

They provide on-line access to a RealView ARM compiler, which I don't  
think has any code size limits on it.  It also has some rather nice  
usability features like behaving like mass storage on a computer (just  
drop the bin files onto the drive and hit the reset button to upload  
firmware) and it presents a USB CDC for serial interaction over USB.

-jsnyder
_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Dado Sutter]

Thank you Louie,
   The truth, and I think you made it very clear in your message, is that that chips are made to profitable targeted applications and not to fullfill our dreams on a general-purpose complete and cheap platform.

Best
Dado

  
  

On Mon, Jun 22, 2009 at 12:21, Louis Mamakos <[hidden email]> wrote:

On Jun 22, 2009, at 10:34 AM, John Hind wrote:

Hi Dado, you had me googling “gronwup embedded environment” before realising it was a typo!

 

My posting was more born of frustration that the ARM microcontroller industry seems to have advanced so little in the last six months compared to the rate the X86 market moves on. I was really hoping someone would point me at some hope the next six months would bring me what I was looking for. I do not want to give up on distributed processing architectures yet, but it is hard when just one processing node ends up nearly as expensive as a central machine capable of handling the entire requirement!

 

It continues to baffle me that the industry does not get the market for high-functionality, highly integrated, low pin-count devices – this seems so obvious for intelligent sensors and actuators. Even in the 8-bit field, Microchip makes lovely 6- and 8- pin chips ideal for embedding in, for example, a single limit switch, but they have no serial comms capability! You are up to 16 pin devices before that starts to appear. In the ARM market, large on-chip memory seems to go hand in hand with high pin-count and therefore complex and expensive packages and PCBs.

In the embedded space, parts get built to support specific (hopefully volume) applications.  This is why you might not see parts with a large number of on-chip peripherals, or lots of memory, etc.  In a previous professional gig, we built VoIP devices that shipped in moderate volume (1M-2M units/year).  This is a nice "sweet spot" for a part manufacturer who would build a device targeted to that application which also happens to be price sensitive.  You don't want to ship parts out with unused peripherals or lots of extra memory because an extra $2 of cost per unit turns into a pretty noticeable numbers at 1,000,000 units shipped. 

If you're only talking about building a few hundred units, then the extra cost isn't really as much of a consideration, especially when you might be able to reduce the one-time engineering expense by optimizing time spent vs. "extra" hardware.

I think that Marvell is building some interesting  higher-end parts, with perhaps a lack of GPIO I/O, A/D, etc. capability but with networking, etc.  Otherwise, you're looking to repurpose devices built for some targeted applications.

louie

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[jbsnyder]

This is true, however I think the non-custom parts actually go a  
pretty good distance towards what we're after.  Many of the chips on  
these devel boards are kind-of "kitchen sink" chips that have way more  
peripherals than are needed for any one single application.

It seems like the problem is that we're asking for something that's  
even a bit more general, but not that much more.  Since these general  
parts already have most of the stuff we need and they're still not  
much more than $10 by themselves, I think we should be able to find  
something that meets a reasonable minimum set of peripherals, and,  
maybe, requires one or two added external chips at the max.

It might be a good idea to start by speccing out some sort of common  
set of peripheral needs that would be satisfactory for a Lua stamp,  
and then go from there?

Here's what I, personally, would like:
- somewhere near 40 pins out, most/all of which can be used for GPIO,  
or configured for various peripherals
- output pins are dip, compatible with standard-sized breadboards
- 8+ A/D channels (10-bit, or 12-bit if possible)
- 2+ D/A (10 or 12 bit)
- Ethernet w/ PHY  or CAN
- 2+ USART
- 2+ I2C
- can be powered over USB
- 64+K of SRAM
- 512+ K of Flash
- optionally powered and programmed over a USB link

I'd be happy with more, but I think this should work for a wide range  
of applications.  My first inclination if any one particular thing  
could be added would be: as much RAM as possible.

-jsnyder

On Jun 22, 2009, at 1:05 PM, Dado Sutter wrote:

> Thank you Louie,
>    The truth, and I think you made it very clear in your message, is  
> that that chips are made to profitable targeted applications and not  
> to fullfill our dreams on a general-purpose complete and cheap  
> platform.
>
> Best
> Dado
>
>
>
>
> On Mon, Jun 22, 2009 at 12:21, Louis Mamakos <[hidden email]>  
> wrote:
>
> On Jun 22, 2009, at 10:34 AM, John Hind wrote:
>
>> Hi Dado, you had me googling “gronwup embedded environment” before  
>> realising it was a typo!
>>
>> My posting was more born of frustration that the ARM  
>> microcontroller industry seems to have advanced so little in the  
>> last six months compared to the rate the X86 market moves on. I was  
>> really hoping someone would point me at some hope the next six  
>> months would bring me what I was looking for. I do not want to give  
>> up on distributed processing architectures yet, but it is hard when  
>> just one processing node ends up nearly as expensive as a central  
>> machine capable of handling the entire requirement!
>>
>> It continues to baffle me that the industry does not get the market  
>> for high-functionality, highly integrated, low pin-count devices –  
>> this seems so obvious for intelligent sensors and actuators. Even  
>> in the 8-bit field, Microchip makes lovely 6- and 8- pin chips  
>> ideal for embedding in, for example, a single limit switch, but  
>> they have no serial comms capability! You are up to 16 pin devices  
>> before that starts to appear. In the ARM market, large on-chip  
>> memory seems to go hand in hand with high pin-count and therefore  
>> complex and expensive packages and PCBs.
>
> In the embedded space, parts get built to support specific  
> (hopefully volume) applications.  This is why you might not see  
> parts with a large number of on-chip peripherals, or lots of memory,  
> etc.  In a previous professional gig, we built VoIP devices that  
> shipped in moderate volume (1M-2M units/year).  This is a nice  
> "sweet spot" for a part manufacturer who would build a device  
> targeted to that application which also happens to be price  
> sensitive.  You don't want to ship parts out with unused peripherals  
> or lots of extra memory because an extra $2 of cost per unit turns  
> into a pretty noticeable numbers at 1,000,000 units shipped.
>
> If you're only talking about building a few hundred units, then the  
> extra cost isn't really as much of a consideration, especially when  
> you might be able to reduce the one-time engineering expense by  
> optimizing time spent vs. "extra" hardware.
>
> I think that Marvell is building some interesting  higher-end parts,  
> with perhaps a lack of GPIO I/O, A/D, etc. capability but with  
> networking, etc.  Otherwise, you're looking to repurpose devices  
> built for some targeted applications.
>
> louie
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev
>
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Dado Sutter]

I think this is a good start for a wish list :) and I would change just a few details:

128 KB RAM minimum (or as much as possible)
SPI instead of I2C
2 Free UARTs (typically the USB link will use one more)
I think Ethernet w/ PHY is a must
As much PWM outputs as possible

Best
Dado

On Mon, Jun 22, 2009 at 15:42, James Snyder <[hidden email]> wrote:

This is true, however I think the non-custom parts actually go a
pretty good distance towards what we're after.  Many of the chips on
these devel boards are kind-of "kitchen sink" chips that have way more
peripherals than are needed for any one single application.

It seems like the problem is that we're asking for something that's
even a bit more general, but not that much more.  Since these general
parts already have most of the stuff we need and they're still not
much more than $10 by themselves, I think we should be able to find
something that meets a reasonable minimum set of peripherals, and,
maybe, requires one or two added external chips at the max.

It might be a good idea to start by speccing out some sort of common
set of peripheral needs that would be satisfactory for a Lua stamp,
and then go from there?

Here's what I, personally, would like:
- somewhere near 40 pins out, most/all of which can be used for GPIO,
or configured for various peripherals
- output pins are dip, compatible with standard-sized breadboards
- 8+ A/D channels (10-bit, or 12-bit if possible)
- 2+ D/A (10 or 12 bit)
- Ethernet w/ PHY  or CAN
- 2+ USART
- 2+ I2C
- can be powered over USB
- 64+K of SRAM
- 512+ K of Flash
- optionally powered and programmed over a USB link

I'd be happy with more, but I think this should work for a wide range
of applications.  My first inclination if any one particular thing
could be added would be: as much RAM as possible.

-jsnyder

On Jun 22, 2009, at 1:05 PM, Dado Sutter wrote:

> Thank you Louie,
>    The truth, and I think you made it very clear in your message, is
> that that chips are made to profitable targeted applications and not
> to fullfill our dreams on a general-purpose complete and cheap
> platform.
>
> Best
> Dado
>
>
>
>
> On Mon, Jun 22, 2009 at 12:21, Louis Mamakos <[hidden email]>
> wrote:
>
> On Jun 22, 2009, at 10:34 AM, John Hind wrote:
>
>> Hi Dado, you had me googling “gronwup embedded environment” before
>> realising it was a typo!
>>
>> My posting was more born of frustration that the ARM
>> microcontroller industry seems to have advanced so little in the
>> last six months compared to the rate the X86 market moves on. I was
>> really hoping someone would point me at some hope the next six
>> months would bring me what I was looking for. I do not want to give
>> up on distributed processing architectures yet, but it is hard when
>> just one processing node ends up nearly as expensive as a central
>> machine capable of handling the entire requirement!
>>
>> It continues to baffle me that the industry does not get the market
>> for high-functionality, highly integrated, low pin-count devices –
>> this seems so obvious for intelligent sensors and actuators. Even
>> in the 8-bit field, Microchip makes lovely 6- and 8- pin chips
>> ideal for embedding in, for example, a single limit switch, but
>> they have no serial comms capability! You are up to 16 pin devices
>> before that starts to appear. In the ARM market, large on-chip
>> memory seems to go hand in hand with high pin-count and therefore
>> complex and expensive packages and PCBs.
>
> In the embedded space, parts get built to support specific
> (hopefully volume) applications.  This is why you might not see
> parts with a large number of on-chip peripherals, or lots of memory,
> etc.  In a previous professional gig, we built VoIP devices that
> shipped in moderate volume (1M-2M units/year).  This is a nice
> "sweet spot" for a part manufacturer who would build a device
> targeted to that application which also happens to be price
> sensitive.  You don't want to ship parts out with unused peripherals
> or lots of extra memory because an extra $2 of cost per unit turns
> into a pretty noticeable numbers at 1,000,000 units shipped.
>
> If you're only talking about building a few hundred units, then the
> extra cost isn't really as much of a consideration, especially when
> you might be able to reduce the one-time engineering expense by
> optimizing time spent vs. "extra" hardware.
>
> I think that Marvell is building some interesting  higher-end parts,
> with perhaps a lack of GPIO I/O, A/D, etc. capability but with
> networking, etc.  Otherwise, you're looking to repurpose devices
> built for some targeted applications.
>
> louie
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev
>
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[John Hind]

Thanks James, that is indeed quite close and it does all seem to be on chip
- I think the second chip is just an RS232 level shifter (RS232 - how last
century!). The LPC2368 is better though - Ethernet and 58kB RAM. The mbed
also looks kinda promising (except the 32k RAM) if they ever get it out of
beta, though I cannot see where they've hidden the Ethernet!

Commenting on the other points, I think the key is building market by
addressing a lot of relatively low volume applications through making really
flexible chips which means putting as much functionality in as possible. It
is also possible to vary the packaging to address different markets with the
same silicon - just put the same chip in a lower pin-count package and only
bring out a fraction of the IO. This would only really work with a full
crossbar switch between the peripherals and the IO though.

> -----Original Message-----
> From: [hidden email] [mailto:elua-dev-
> [hidden email]] On Behalf Of James Snyder
> Sent: 22 June 2009 18:52
> To: eLua Users and Development List
> Subject: Re: [eLua-dev] Single Chip Lua Prospects
>
>
> On Jun 22, 2009, at 5:42 AM, John Hind wrote:
>
> > When I subscribed to this list half a year ago I confidently
> > expected to be able to use a single chip Lua device in my projects
> > by now. I wanted to use many such chips in a seriously distributed
> > architecture with the processing kept near to the IO to simplify
> > wiring. To avoid buying an endless succession of development boards
> > I set a simple threshold specification of 512K on-chip flash, 64K on-
> > chip RAM, Ethernet capability (I was prepared to compromise with a
> > second chip for this). With this type of distributed architecture, I
> > need lots of IO versatility (analog, PWM, UART, I2C, SPI, counters
> > etc.) but multiplexed on relatively few pins since if you run out of
> > IO you just add another chip. It should therefore be possible to
> > have a simple 20-40 pin package which could be available in hand-
> > solderable form.
>
> What about in stamp form factor?
>
> It would seem to me that this does a lot of what you're asking for
> (aside from ethernet):
> http://www.futurlec.com/ET-STM32_Stamp.shtml
>
> It does, however, have built-in CAN.
>
> On Jun 22, 2009, at 9:51 AM, Dado Sutter wrote:
> > Yes, very true :)
> > But let's keep our eyes open. The market will never be formed
> > arround a single chip but arround many diferent options. That's why
> > we want eLua to be so portable.
> > Meanwhile, we'll keep looking for a good candidate for our first
> > eLua Stamp (I propose to start this thread immediately :) and
> > hearing the findings of our members. Snyder had found a while ago a
> > neat little cute board from England (I guess) but we've had no news
> > since.
>
> I still have that board, and still need to finish the port for it :-)
>
> It's an interesting board, and provide some nice interfaces for
> working with the various onboard peripherals:
> http://mbed.co.uk/beta
>
> They provide on-line access to a RealView ARM compiler, which I don't
> think has any code size limits on it.  It also has some rather nice
> usability features like behaving like mass storage on a computer (just
> drop the bin files onto the drive and hit the reset button to upload
> firmware) and it presents a USB CDC for serial interaction over USB.
>
> -jsnyder
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Asko Kauppi]

I'll share some part numbers recommended by my Atmel contact:

> ==> We have three chips with enough internal RAM.
> AT91SAM7X512    512 kB Flash & 128 kB SRAM
> AT32UC3A3256           256 kB Flash & 128 kB SRAM
> AT91FR40162SB    2 MB Flash and 256 kb SRAM.
>       Note that this is a multichip package in a 10 x 10 mm BGA
>       and the periperals set is sort of outdated.
>
> And by not needing to use those pins for external RAM interface, they
> would be spared for other usage.

BGA is perhaps a no-go.  AT32UC3A3256 is what we're trying to get.  
AT91SAM7X512 some of you may have experience on?  (or is it a  
newcomer, since it's marked as "sampling")

AT32UC3A3256
- SRAM 128K
- Flash 256K
- IO pins 110
- UARTs 4
- USB
- package LQFP 144 (or TFBGA 144)
- 66 MHz max.

http://www.atmel.com/dyn/resources/prod_documents/32072S.pdf

AT91SAM7X512
- package LQFP 100
- Integrated Ethernet, USB _and_ CAN
- Flash 512K
- SRAM 128K
- Status: sampling
- 55 MHz max.

http://www.atmel.com/dyn/resources/prod_documents/6120s.pdf

-asko


James Snyder kirjoitti 22.6.2009 kello 21:42:

> This is true, however I think the non-custom parts actually go a
> pretty good distance towards what we're after.  Many of the chips on
> these devel boards are kind-of "kitchen sink" chips that have way more
> peripherals than are needed for any one single application.
>
> It seems like the problem is that we're asking for something that's
> even a bit more general, but not that much more.  Since these general
> parts already have most of the stuff we need and they're still not
> much more than $10 by themselves, I think we should be able to find
> something that meets a reasonable minimum set of peripherals, and,
> maybe, requires one or two added external chips at the max.
>
> It might be a good idea to start by speccing out some sort of common
> set of peripheral needs that would be satisfactory for a Lua stamp,
> and then go from there?
>
> Here's what I, personally, would like:
> - somewhere near 40 pins out, most/all of which can be used for GPIO,
> or configured for various peripherals
> - output pins are dip, compatible with standard-sized breadboards
> - 8+ A/D channels (10-bit, or 12-bit if possible)
> - 2+ D/A (10 or 12 bit)
> - Ethernet w/ PHY  or CAN
> - 2+ USART
> - 2+ I2C
> - can be powered over USB
> - 64+K of SRAM
> - 512+ K of Flash
> - optionally powered and programmed over a USB link
>
> I'd be happy with more, but I think this should work for a wide range
> of applications.  My first inclination if any one particular thing
> could be added would be: as much RAM as possible.
>
> -jsnyder
>
> On Jun 22, 2009, at 1:05 PM, Dado Sutter wrote:
>
>> Thank you Louie,
>>   The truth, and I think you made it very clear in your message, is
>> that that chips are made to profitable targeted applications and not
>> to fullfill our dreams on a general-purpose complete and cheap
>> platform.
>>
>> Best
>> Dado
>>
>>
>>
>>
>> On Mon, Jun 22, 2009 at 12:21, Louis Mamakos <[hidden email]>
>> wrote:
>>
>> On Jun 22, 2009, at 10:34 AM, John Hind wrote:
>>
>>> Hi Dado, you had me googling “gronwup embedded environment” before
>>> realising it was a typo!
>>>
>>> My posting was more born of frustration that the ARM
>>> microcontroller industry seems to have advanced so little in the
>>> last six months compared to the rate the X86 market moves on. I was
>>> really hoping someone would point me at some hope the next six
>>> months would bring me what I was looking for. I do not want to give
>>> up on distributed processing architectures yet, but it is hard when
>>> just one processing node ends up nearly as expensive as a central
>>> machine capable of handling the entire requirement!
>>>
>>> It continues to baffle me that the industry does not get the market
>>> for high-functionality, highly integrated, low pin-count devices –
>>> this seems so obvious for intelligent sensors and actuators. Even
>>> in the 8-bit field, Microchip makes lovely 6- and 8- pin chips
>>> ideal for embedding in, for example, a single limit switch, but
>>> they have no serial comms capability! You are up to 16 pin devices
>>> before that starts to appear. In the ARM market, large on-chip
>>> memory seems to go hand in hand with high pin-count and therefore
>>> complex and expensive packages and PCBs.
>>
>> In the embedded space, parts get built to support specific
>> (hopefully volume) applications.  This is why you might not see
>> parts with a large number of on-chip peripherals, or lots of memory,
>> etc.  In a previous professional gig, we built VoIP devices that
>> shipped in moderate volume (1M-2M units/year).  This is a nice
>> "sweet spot" for a part manufacturer who would build a device
>> targeted to that application which also happens to be price
>> sensitive.  You don't want to ship parts out with unused peripherals
>> or lots of extra memory because an extra $2 of cost per unit turns
>> into a pretty noticeable numbers at 1,000,000 units shipped.
>>
>> If you're only talking about building a few hundred units, then the
>> extra cost isn't really as much of a consideration, especially when
>> you might be able to reduce the one-time engineering expense by
>> optimizing time spent vs. "extra" hardware.
>>
>> I think that Marvell is building some interesting  higher-end parts,
>> with perhaps a lack of GPIO I/O, A/D, etc. capability but with
>> networking, etc.  Otherwise, you're looking to repurpose devices
>> built for some targeted applications.
>>
>> louie
>>
>> _______________________________________________
>> Elua-dev mailing list
>> [hidden email]
>> https://lists.berlios.de/mailman/listinfo/elua-dev
>>
>>
>> _______________________________________________
>> Elua-dev mailing list
>> [hidden email]
>> https://lists.berlios.de/mailman/listinfo/elua-dev
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[John Hind]

Thanks Asko!

Mouser & DigiKey have the AT91SAM7X512 100 pin LQFP at $15. It looks
absolutely ideal for our purposes, especially with the 128kB RAM.

I'll look into this in more detail. Anyone got anything against this chip?

- John

> -----Original Message-----
> From: [hidden email] [mailto:elua-dev-
> [hidden email]] On Behalf Of Asko Kauppi
> Sent: 22 June 2009 21:44
> To: eLua Users and Development List
> Subject: Re: [eLua-dev] Single Chip Lua Prospects
>
>
> I'll share some part numbers recommended by my Atmel contact:
>
> > ==> We have three chips with enough internal RAM.
> > AT91SAM7X512    512 kB Flash & 128 kB SRAM
> > AT32UC3A3256           256 kB Flash & 128 kB SRAM
> > AT91FR40162SB    2 MB Flash and 256 kb SRAM.
> >       Note that this is a multichip package in a 10 x 10 mm BGA
> >       and the periperals set is sort of outdated.
> >
> > And by not needing to use those pins for external RAM interface, they
> > would be spared for other usage.
>
> BGA is perhaps a no-go.  AT32UC3A3256 is what we're trying to get.
> AT91SAM7X512 some of you may have experience on?  (or is it a
> newcomer, since it's marked as "sampling")
>
> AT32UC3A3256
> - SRAM 128K
> - Flash 256K
> - IO pins 110
> - UARTs 4
> - USB
> - package LQFP 144 (or TFBGA 144)
> - 66 MHz max.
>
> http://www.atmel.com/dyn/resources/prod_documents/32072S.pdf
>
> AT91SAM7X512
> - package LQFP 100
> - Integrated Ethernet, USB _and_ CAN
> - Flash 512K
> - SRAM 128K
> - Status: sampling
> - 55 MHz max.
>
> http://www.atmel.com/dyn/resources/prod_documents/6120s.pdf
>
> -asko
>
>
> James Snyder kirjoitti 22.6.2009 kello 21:42:
>
> > This is true, however I think the non-custom parts actually go a
> > pretty good distance towards what we're after.  Many of the chips on
> > these devel boards are kind-of "kitchen sink" chips that have way
> more
> > peripherals than are needed for any one single application.
> >
> > It seems like the problem is that we're asking for something that's
> > even a bit more general, but not that much more.  Since these general
> > parts already have most of the stuff we need and they're still not
> > much more than $10 by themselves, I think we should be able to find
> > something that meets a reasonable minimum set of peripherals, and,
> > maybe, requires one or two added external chips at the max.
> >
> > It might be a good idea to start by speccing out some sort of common
> > set of peripheral needs that would be satisfactory for a Lua stamp,
> > and then go from there?
> >
> > Here's what I, personally, would like:
> > - somewhere near 40 pins out, most/all of which can be used for GPIO,
> > or configured for various peripherals
> > - output pins are dip, compatible with standard-sized breadboards
> > - 8+ A/D channels (10-bit, or 12-bit if possible)
> > - 2+ D/A (10 or 12 bit)
> > - Ethernet w/ PHY  or CAN
> > - 2+ USART
> > - 2+ I2C
> > - can be powered over USB
> > - 64+K of SRAM
> > - 512+ K of Flash
> > - optionally powered and programmed over a USB link
> >
> > I'd be happy with more, but I think this should work for a wide range
> > of applications.  My first inclination if any one particular thing
> > could be added would be: as much RAM as possible.
> >
> > -jsnyder
> >
> > On Jun 22, 2009, at 1:05 PM, Dado Sutter wrote:
> >
> >> Thank you Louie,
> >>   The truth, and I think you made it very clear in your message, is
> >> that that chips are made to profitable targeted applications and not
> >> to fullfill our dreams on a general-purpose complete and cheap
> >> platform.
> >>
> >> Best
> >> Dado
> >>
> >>
> >>
> >>
> >> On Mon, Jun 22, 2009 at 12:21, Louis Mamakos <[hidden email]>
> >> wrote:
> >>
> >> On Jun 22, 2009, at 10:34 AM, John Hind wrote:
> >>
> >>> Hi Dado, you had me googling "gronwup embedded environment" before
> >>> realising it was a typo!
> >>>
> >>> My posting was more born of frustration that the ARM
> >>> microcontroller industry seems to have advanced so little in the
> >>> last six months compared to the rate the X86 market moves on. I was
> >>> really hoping someone would point me at some hope the next six
> >>> months would bring me what I was looking for. I do not want to give
> >>> up on distributed processing architectures yet, but it is hard when
> >>> just one processing node ends up nearly as expensive as a central
> >>> machine capable of handling the entire requirement!
> >>>
> >>> It continues to baffle me that the industry does not get the market
> >>> for high-functionality, highly integrated, low pin-count devices -
> >>> this seems so obvious for intelligent sensors and actuators. Even
> >>> in the 8-bit field, Microchip makes lovely 6- and 8- pin chips
> >>> ideal for embedding in, for example, a single limit switch, but
> >>> they have no serial comms capability! You are up to 16 pin devices
> >>> before that starts to appear. In the ARM market, large on-chip
> >>> memory seems to go hand in hand with high pin-count and therefore
> >>> complex and expensive packages and PCBs.
> >>
> >> In the embedded space, parts get built to support specific
> >> (hopefully volume) applications.  This is why you might not see
> >> parts with a large number of on-chip peripherals, or lots of memory,
> >> etc.  In a previous professional gig, we built VoIP devices that
> >> shipped in moderate volume (1M-2M units/year).  This is a nice
> >> "sweet spot" for a part manufacturer who would build a device
> >> targeted to that application which also happens to be price
> >> sensitive.  You don't want to ship parts out with unused peripherals
> >> or lots of extra memory because an extra $2 of cost per unit turns
> >> into a pretty noticeable numbers at 1,000,000 units shipped.
> >>
> >> If you're only talking about building a few hundred units, then the
> >> extra cost isn't really as much of a consideration, especially when
> >> you might be able to reduce the one-time engineering expense by
> >> optimizing time spent vs. "extra" hardware.
> >>
> >> I think that Marvell is building some interesting  higher-end parts,
> >> with perhaps a lack of GPIO I/O, A/D, etc. capability but with
> >> networking, etc.  Otherwise, you're looking to repurpose devices
> >> built for some targeted applications.
> >>
> >> louie
> >>
> >> _______________________________________________
> >> Elua-dev mailing list
> >> [hidden email]
> >> https://lists.berlios.de/mailman/listinfo/elua-dev
> >>
> >>
> >> _______________________________________________
> >> Elua-dev mailing list
> >> [hidden email]
> >> https://lists.berlios.de/mailman/listinfo/elua-dev
> >
> > _______________________________________________
> > Elua-dev mailing list
> > [hidden email]
> > https://lists.berlios.de/mailman/listinfo/elua-dev
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Dean Hall]

I found the SAM7 series peripherals much more complicated to program  
than an AVR, but this becomes meaningless when a high-level API is  
provided for the peripheral.

!!Dean


On Jun 22, 2009, at 16:45 , John Hind wrote:

> Thanks Asko!
>
> Mouser & DigiKey have the AT91SAM7X512 100 pin LQFP at $15. It looks
> absolutely ideal for our purposes, especially with the 128kB RAM.
>
> I'll look into this in more detail. Anyone got anything against this  
> chip?
>
> - John

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[Tony-12]

If you're looking for a PC-level ARM development board, there are at least four interesting choices:
1. BeagleBoard: $150 600MHz OMAP3530 Cortex A8, VLIW DSP, graphics, 256M flash, 128M DRAM, HS USB 2.0, no Ethernet.
2. Gumstick Overo (OMAP35xx board) and Verdex (PXA270), starts at $129
3. Leopard board: $99, 210MHz 320DM355, camera, Ethernet, USB, 256M flash, 128M DRAM
4. Marvell Sheeva Plug, 1.2GHz Marvel Sheeva, 512M flash, 512M DRAM, Gigabit Ethernet, USB

All of these are smaller than a nettop, and don't require any cooling fans.  With the x86 PC, you still have to add any interesting peripherals (ADC, DAC, PWM, encoder, etc) via USB or similar.  BTW, USB interfaces using HID will be very limited in speed.  And I think the simplicity of the eLua approach is a major attraction; Linux is still Linux, even on a reduced-fat system.

There have been plenty of exciting ARM announcements recently, ranging from low end (Cortex-M3 such as the TI/Luminary 9B and STM32 Connectivity) up to Gigahertz Cortex-A8 systems (TI, Freescale, etc).

It costs a significant amount of money to put a new MCU into production, including design, verification, mask costs, and verification.  If you look at all the functions the high-end MCUs support (e.g. Stellaris 9Bx), you'll see that even 100 pins isn't very many.  Also, with the new chip-scale packages (such as QFN and BGA), larger packages are smaller than lower pin count old packages such as DIP and SOIC.  IIRC, Microchip and ST have some models with very flexible I/O assignments, but there may be reasons why this isn't widespread.  AFAIK, only Microchip is actively making new products in hobbyist-friendly DIP packages; e.g. one of the smallest ARM MCUs, a 36-pin STM32, is only available in QFN.  But Microchip doesn't have large memory MCUs.  And, in summary, I doubt the demand is there for large memory, low pin count MCUs.

Besides, I think it's better to use a Stamp format anyway (like the ET-STM32 Stamp) -- then you put all the required voltage regulators and clock stuff up on the Stamp, and just have a reduced pin count, DIP-style interface to your prototype board.

BTW, the AT32UC3A3256 has a couple other nice features Asko didn't mention: high speed USB 2.0 (still not common in QFP), external SDRAM, and external NAND support.  But there's no Ethernet.  Mouser pricing is $12.86 (1), $8.56 (100).

If you are looking for >=64K SRAM, >=256K flash, there are many choices.  Even at >=96K SRAM, >=512K flash, there are quite a few choices.  I'd say the most interesting candidates for an eLua stamp right now are the UC3 (since it has 128K SRAM, external memory interfaces,  and HS USB) or the Stellaris 9B (if it's available soon -- the $99 dev kits are in stock -- since it has 96K SRAM, Ethernet, USB, external memory interfaces, and cool peripherals like PWM and quadrature encoders).

--Tony


_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[BogdanM]

Yes, this is precisely The Truth. Once we make eLua succesfull enough, they'll be happy to give us chips with 256k+ of internal SRAM. Until then ... :)

Best,
Bogdan

On Mon, Jun 22, 2009 at 9:05 PM, Dado Sutter <[hidden email]> wrote:

Thank you Louie,
   The truth, and I think you made it very clear in your message, is that that chips are made to profitable targeted applications and not to fullfill our dreams on a general-purpose complete and cheap platform.

Best
Dado

  
  

On Mon, Jun 22, 2009 at 12:21, Louis Mamakos <[hidden email]> wrote:

On Jun 22, 2009, at 10:34 AM, John Hind wrote:

Hi Dado, you had me googling “gronwup embedded environment” before realising it was a typo!

 

My posting was more born of frustration that the ARM microcontroller industry seems to have advanced so little in the last six months compared to the rate the X86 market moves on. I was really hoping someone would point me at some hope the next six months would bring me what I was looking for. I do not want to give up on distributed processing architectures yet, but it is hard when just one processing node ends up nearly as expensive as a central machine capable of handling the entire requirement!

 

It continues to baffle me that the industry does not get the market for high-functionality, highly integrated, low pin-count devices – this seems so obvious for intelligent sensors and actuators. Even in the 8-bit field, Microchip makes lovely 6- and 8- pin chips ideal for embedding in, for example, a single limit switch, but they have no serial comms capability! You are up to 16 pin devices before that starts to appear. In the ARM market, large on-chip memory seems to go hand in hand with high pin-count and therefore complex and expensive packages and PCBs.

In the embedded space, parts get built to support specific (hopefully volume) applications.  This is why you might not see parts with a large number of on-chip peripherals, or lots of memory, etc.  In a previous professional gig, we built VoIP devices that shipped in moderate volume (1M-2M units/year).  This is a nice "sweet spot" for a part manufacturer who would build a device targeted to that application which also happens to be price sensitive.  You don't want to ship parts out with unused peripherals or lots of extra memory because an extra $2 of cost per unit turns into a pretty noticeable numbers at 1,000,000 units shipped. 

If you're only talking about building a few hundred units, then the extra cost isn't really as much of a consideration, especially when you might be able to reduce the one-time engineering expense by optimizing time spent vs. "extra" hardware.

I think that Marvell is building some interesting  higher-end parts, with perhaps a lack of GPIO I/O, A/D, etc. capability but with networking, etc.  Otherwise, you're looking to repurpose devices built for some targeted applications.

louie

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[BogdanM]

On Mon, Jun 22, 2009 at 11:44 PM, Asko Kauppi <[hidden email]> wrote:

I'll share some part numbers recommended by my Atmel contact:

> ==> We have three chips with enough internal RAM.
> AT91SAM7X512    512 kB Flash & 128 kB SRAM

Very nice chip, yet virtually impossible to get. I checked many times, and none of the big distributors have it stocked. Which means that it's basically impossible for the "regular" hobbyist to get his hands on it. Still my best candidate for "eLua on a chip" though, mainly because of the RAM.

> AT32UC3A3256           256 kB Flash & 128 kB SRAM

Sounds very good, still could do with a little more Flash, but we don't live in a perfect world.

> AT91FR40162SB    2 MB Flash and 256 kb SRAM.
>       Note that this is a multichip package in a 10 x 10 mm BGA
>       and the periperals set is sort of outdated.

Yes, the peripheral set is almost inexistent, unfortunately. Also the chip is once again very rare, and very hard to work with because of the BGA package.
 

AT91SAM7X512 some of you may have experience on?  (or is it a
newcomer, since it's marked as "sampling")

It's not at all a newcomer, it's listed on Atmel's site since forever (sorry, I don't have an actual date, I just know it's been a long while since I first saw it). It's just very hard to find.

Best,
Bogdan

James Snyder kirjoitti 22.6.2009 kello 21:42:

> This is true, however I think the non-custom parts actually go a
> pretty good distance towards what we're after.  Many of the chips on
> these devel boards are kind-of "kitchen sink" chips that have way more
> peripherals than are needed for any one single application.
>
> It seems like the problem is that we're asking for something that's
> even a bit more general, but not that much more.  Since these general
> parts already have most of the stuff we need and they're still not
> much more than $10 by themselves, I think we should be able to find
> something that meets a reasonable minimum set of peripherals, and,
> maybe, requires one or two added external chips at the max.
>
> It might be a good idea to start by speccing out some sort of common
> set of peripheral needs that would be satisfactory for a Lua stamp,
> and then go from there?
>
> Here's what I, personally, would like:
> - somewhere near 40 pins out, most/all of which can be used for GPIO,
> or configured for various peripherals
> - output pins are dip, compatible with standard-sized breadboards
> - 8+ A/D channels (10-bit, or 12-bit if possible)
> - 2+ D/A (10 or 12 bit)
> - Ethernet w/ PHY  or CAN
> - 2+ USART
> - 2+ I2C
> - can be powered over USB
> - 64+K of SRAM
> - 512+ K of Flash
> - optionally powered and programmed over a USB link
>
> I'd be happy with more, but I think this should work for a wide range
> of applications.  My first inclination if any one particular thing
> could be added would be: as much RAM as possible.
>
> -jsnyder
>
> On Jun 22, 2009, at 1:05 PM, Dado Sutter wrote:
>

>> Thank you Louie,
>>   The truth, and I think you made it very clear in your message, is
>> that that chips are made to profitable targeted applications and not
>> to fullfill our dreams on a general-purpose complete and cheap
>> platform.
>>
>> Best
>> Dado
>>
>>
>>
>>
>> On Mon, Jun 22, 2009 at 12:21, Louis Mamakos <[hidden email]>
>> wrote:
>>
>> On Jun 22, 2009, at 10:34 AM, John Hind wrote:
>>
>>> Hi Dado, you had me googling “gronwup embedded environment” before
>>> realising it was a typo!
>>>
>>> My posting was more born of frustration that the ARM
>>> microcontroller industry seems to have advanced so little in the
>>> last six months compared to the rate the X86 market moves on. I was
>>> really hoping someone would point me at some hope the next six
>>> months would bring me what I was looking for. I do not want to give
>>> up on distributed processing architectures yet, but it is hard when
>>> just one processing node ends up nearly as expensive as a central
>>> machine capable of handling the entire requirement!
>>>
>>> It continues to baffle me that the industry does not get the market
>>> for high-functionality, highly integrated, low pin-count devices –
>>> this seems so obvious for intelligent sensors and actuators. Even
>>> in the 8-bit field, Microchip makes lovely 6- and 8- pin chips
>>> ideal for embedding in, for example, a single limit switch, but
>>> they have no serial comms capability! You are up to 16 pin devices
>>> before that starts to appear. In the ARM market, large on-chip
>>> memory seems to go hand in hand with high pin-count and therefore
>>> complex and expensive packages and PCBs.
>>
>> In the embedded space, parts get built to support specific
>> (hopefully volume) applications.  This is why you might not see
>> parts with a large number of on-chip peripherals, or lots of memory,
>> etc.  In a previous professional gig, we built VoIP devices that
>> shipped in moderate volume (1M-2M units/year).  This is a nice
>> "sweet spot" for a part manufacturer who would build a device
>> targeted to that application which also happens to be price
>> sensitive.  You don't want to ship parts out with unused peripherals
>> or lots of extra memory because an extra $2 of cost per unit turns
>> into a pretty noticeable numbers at 1,000,000 units shipped.
>>
>> If you're only talking about building a few hundred units, then the
>> extra cost isn't really as much of a consideration, especially when
>> you might be able to reduce the one-time engineering expense by
>> optimizing time spent vs. "extra" hardware.
>>
>> I think that Marvell is building some interesting  higher-end parts,
>> with perhaps a lack of GPIO I/O, A/D, etc. capability but with
>> networking, etc.  Otherwise, you're looking to repurpose devices
>> built for some targeted applications.
>>
>> louie
>>
>> _______________________________________________
>> Elua-dev mailing list
>> [hidden email]
>> https://lists.berlios.de/mailman/listinfo/elua-dev
>>
>>
>> _______________________________________________
>> Elua-dev mailing list
>> [hidden email]
>> https://lists.berlios.de/mailman/listinfo/elua-dev
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[BogdanM]

On Tue, Jun 23, 2009 at 12:45 AM, John Hind <[hidden email]> wrote:

Thanks Asko!

Mouser & DigiKey have the AT91SAM7X512 100 pin LQFP at $15. It looks
absolutely ideal for our purposes, especially with the 128kB RAM.

I'll look into this in more detail. Anyone got anything against this chip?

A small correction to one of my previous messages: the _LQFP_ variant of AT91SAM7X512 is non stocked (just checked again on findchips.com). You can still find the BGA version, but I can't do too much with that. And I know I'm not the only one.

Best,
Bogdan
 

- John

> -----Original Message-----
> From: [hidden email] [mailto:[hidden email]

> [hidden email]] On Behalf Of Asko Kauppi
> Sent: 22 June 2009 21:44
> To: eLua Users and Development List
> Subject: Re: [eLua-dev] Single Chip Lua Prospects
>
>

> I'll share some part numbers recommended by my Atmel contact:
>
> > ==> We have three chips with enough internal RAM.
> > AT91SAM7X512    512 kB Flash & 128 kB SRAM
> > AT32UC3A3256           256 kB Flash & 128 kB SRAM
> > AT91FR40162SB    2 MB Flash and 256 kb SRAM.
> >       Note that this is a multichip package in a 10 x 10 mm BGA
> >       and the periperals set is sort of outdated.
> >
> > And by not needing to use those pins for external RAM interface, they
> > would be spared for other usage.
>
> BGA is perhaps a no-go.  AT32UC3A3256 is what we're trying to get.
> AT91SAM7X512 some of you may have experience on?  (or is it a
> newcomer, since it's marked as "sampling")
>
> AT32UC3A3256
> - SRAM 128K
> - Flash 256K
> - IO pins 110
> - UARTs 4
> - USB
> - package LQFP 144 (or TFBGA 144)
> - 66 MHz max.
>
> http://www.atmel.com/dyn/resources/prod_documents/32072S.pdf
>
> AT91SAM7X512
> - package LQFP 100
> - Integrated Ethernet, USB _and_ CAN
> - Flash 512K
> - SRAM 128K
> - Status: sampling
> - 55 MHz max.
>
> http://www.atmel.com/dyn/resources/prod_documents/6120s.pdf
>
> -asko
>
>
> James Snyder kirjoitti 22.6.2009 kello 21:42:
>
> > This is true, however I think the non-custom parts actually go a
> > pretty good distance towards what we're after.  Many of the chips on
> > these devel boards are kind-of "kitchen sink" chips that have way
> more
> > peripherals than are needed for any one single application.
> >
> > It seems like the problem is that we're asking for something that's
> > even a bit more general, but not that much more.  Since these general
> > parts already have most of the stuff we need and they're still not
> > much more than $10 by themselves, I think we should be able to find
> > something that meets a reasonable minimum set of peripherals, and,
> > maybe, requires one or two added external chips at the max.
> >
> > It might be a good idea to start by speccing out some sort of common
> > set of peripheral needs that would be satisfactory for a Lua stamp,
> > and then go from there?
> >
> > Here's what I, personally, would like:
> > - somewhere near 40 pins out, most/all of which can be used for GPIO,
> > or configured for various peripherals
> > - output pins are dip, compatible with standard-sized breadboards
> > - 8+ A/D channels (10-bit, or 12-bit if possible)
> > - 2+ D/A (10 or 12 bit)
> > - Ethernet w/ PHY  or CAN
> > - 2+ USART
> > - 2+ I2C
> > - can be powered over USB
> > - 64+K of SRAM
> > - 512+ K of Flash
> > - optionally powered and programmed over a USB link
> >
> > I'd be happy with more, but I think this should work for a wide range
> > of applications.  My first inclination if any one particular thing
> > could be added would be: as much RAM as possible.
> >
> > -jsnyder
> >
> > On Jun 22, 2009, at 1:05 PM, Dado Sutter wrote:
> >
> >> Thank you Louie,
> >>   The truth, and I think you made it very clear in your message, is
> >> that that chips are made to profitable targeted applications and not
> >> to fullfill our dreams on a general-purpose complete and cheap
> >> platform.
> >>
> >> Best
> >> Dado
> >>
> >>
> >>
> >>
> >> On Mon, Jun 22, 2009 at 12:21, Louis Mamakos <[hidden email]>
> >> wrote:
> >>
> >> On Jun 22, 2009, at 10:34 AM, John Hind wrote:
> >>
> >>> Hi Dado, you had me googling "gronwup embedded environment" before
> >>> realising it was a typo!
> >>>
> >>> My posting was more born of frustration that the ARM
> >>> microcontroller industry seems to have advanced so little in the
> >>> last six months compared to the rate the X86 market moves on. I was
> >>> really hoping someone would point me at some hope the next six
> >>> months would bring me what I was looking for. I do not want to give
> >>> up on distributed processing architectures yet, but it is hard when
> >>> just one processing node ends up nearly as expensive as a central
> >>> machine capable of handling the entire requirement!
> >>>
> >>> It continues to baffle me that the industry does not get the market
> >>> for high-functionality, highly integrated, low pin-count devices -
> >>> this seems so obvious for intelligent sensors and actuators. Even
> >>> in the 8-bit field, Microchip makes lovely 6- and 8- pin chips
> >>> ideal for embedding in, for example, a single limit switch, but
> >>> they have no serial comms capability! You are up to 16 pin devices
> >>> before that starts to appear. In the ARM market, large on-chip
> >>> memory seems to go hand in hand with high pin-count and therefore
> >>> complex and expensive packages and PCBs.
> >>
> >> In the embedded space, parts get built to support specific
> >> (hopefully volume) applications.  This is why you might not see
> >> parts with a large number of on-chip peripherals, or lots of memory,
> >> etc.  In a previous professional gig, we built VoIP devices that
> >> shipped in moderate volume (1M-2M units/year).  This is a nice
> >> "sweet spot" for a part manufacturer who would build a device
> >> targeted to that application which also happens to be price
> >> sensitive.  You don't want to ship parts out with unused peripherals
> >> or lots of extra memory because an extra $2 of cost per unit turns
> >> into a pretty noticeable numbers at 1,000,000 units shipped.
> >>
> >> If you're only talking about building a few hundred units, then the
> >> extra cost isn't really as much of a consideration, especially when
> >> you might be able to reduce the one-time engineering expense by
> >> optimizing time spent vs. "extra" hardware.
> >>
> >> I think that Marvell is building some interesting  higher-end parts,
> >> with perhaps a lack of GPIO I/O, A/D, etc. capability but with
> >> networking, etc.  Otherwise, you're looking to repurpose devices
> >> built for some targeted applications.
> >>
> >> louie
> >>
> >> _______________________________________________
> >> Elua-dev mailing list
> >> [hidden email]
> >> https://lists.berlios.de/mailman/listinfo/elua-dev
> >>
> >>
> >> _______________________________________________
> >> Elua-dev mailing list
> >> [hidden email]
> >> https://lists.berlios.de/mailman/listinfo/elua-dev
> >
> > _______________________________________________
> > Elua-dev mailing list
> > [hidden email]
> > https://lists.berlios.de/mailman/listinfo/elua-dev
>
> _______________________________________________
> Elua-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev

[BogdanM]

If you are looking for >=64K SRAM, >=256K flash, there are many choices.  Even at >=96K SRAM, >=512K flash, there are quite a few choices.  I'd say the most interesting candidates for an eLua stamp right now are the UC3 (since it has 128K SRAM, external memory interfaces,  and HS USB) or the Stellaris 9B (if it's available soon -- the $99 dev kits are in stock -- since it has 96K SRAM, Ethernet, USB, external memory interfaces, and cool peripherals like PWM and quadrature encoders).

I'd like to add the STR912FAW44 from ST. Very nice chip, can be found in LQFP package (albeit with a smaller pitch than the "usual" LQFPs, still much easier to use than a BGA variant), 512k of Flash, 96K of SRAM, ARM9 at 96MHz (ARM9 is very rare on the microcontroller market, and 96MHz is very nice), external memory interface, and we already have a port for it :)  $15 at Mouser in small quantities:

http://mouser.com/search/ProductDetail.aspx?qs=A09LVk8FdlJqRV16XEzRIg%3d%3d

Kinda hard to beat this. 96k of SRAM is not 128K, but still is more than 64K :), and one can do lots of stuff in 96k. Plus, adding a small external SRAM shouldn't be difficult, anyway (it can also do PSRAM, but it seems that PSRAM only comes in BGA packages, so I'd stick with SRAM). It has other interesting features, like a very neat way to assign peripheral signals to actual pins. And, unlike the AT91SAM7X512, I could actually _get_ this chip right now :) It could be the base for something like an eLua powered Arduino. I'm not a PCB designer (well, I designed a few boards, but they were all very simple) but I bet a PCB designer could come up a very nice board around this chip, and one that could be easily accesible to a hobbyist.
Of course, the Stellaris 9B is also a very nice alternative to consider. Even the STM32F103 from ST:

http://www.st.com/mcu/devicedocs-STM32F103ZE-110.html

Cortex-M3 with only 64k of internal SRAM, but also with an external (P)SRAM capable memory bus.

Best,
Bogdan

_______________________________________________
Elua-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/elua-dev