Raspberry Pi + Arduino Shields + BBB Cape + FPGA to hook them all up

[thinlayer]

My board is called PeX which stands for Pi eXpansion.



I'm considering a Kickstarter to launch it. More info here: http://bit.ly/1kwvXez. Last year I ran this http://bit.ly/1pQQ5IT on IGG it was a poor turn out. Only got 10% funded. So I'm trying to collect feedback before risking a Kickstarter face-plant and the expense of prototyping a board that no one wants.

PeX is mostly a respin of my IGG campaign board, but with a socket to attach a RasPi host, a socket to attach an Arduino Uno host, RasPi expansion sockets, and Xbee wireless modules. Each 'wing' of the board can host ONE of: an Arduino Uno Shield OR a RasPi Expansion card OR Xbee modules. Only one wing will be a little different in that it'll also be able to host a Beagle Board cape and/or host depending on what connector is fitted. The main reason for this isn't the array of Capes, but specifically the range of well priced LCD capes. RasPi and Arduino just don't have good embedded LCD options (I'm not holding my breath waiting for RasPi MIPI-DSI drivers and options). I'd leave the BB socket unpopulated and ship loose sockets you could solder on. This allows you to solder wires direct if your application prefers, or IDC cable, or whatever. But by fitting female the sockets to the top side of PeX it'll support a Cape. Fitting male pins to the bottom side would allows a host BB or BBB to attach from beneath, albeit a bit awkward. You could also use a dual male/female part (http://www.gradconn.com/2-54mm/pdf/BB02-KT.pdf) to allow both BB host and cape at the same time.

Everything will be tied together with an FPGA (of course.) The FPGA can be invisible to the user; coming pre-programmed with standard loads depending on your choice of host/shields etc. This board will not force you to learn FPGA unless it's what you want. It'll be either an Altera Cyclone IV EP4CE10 (10KLE) or a Cyclone V 5CEA2 (25KLE). It'll have SRAM, possibly DDR3 and a companion MCU for bootloading, USB Slave, ADC, SD Card, etc... FPGA source code will be supplied in SCHEMATIC form, as well as Verilog. Templates will be supplied for Pi host, Arduino host, etc... and the host will tell the FPGA what expansion card is connected to each wing with an I2C write to the FPGA so it knows where to route the IO. Data would then be steered by driving a mux in the FPGA logic. There can be infinite latitude here if you edit the FPGA schematic of course. For example if you have a serial shield you can steer the UART to that wing, while supplying I2S to another wing, while using GPIO PWM to drive motors on the 3rd etc, all quickly configurable in schematic form on Altera Quartus II's IDE (which also has a built in Logic Analyser called SignalTap that no one knows about.) To RasPi software, a setup like this would look like you had three Pi expansions stacked on top of one another simultaneously, but there will be driver development done for RasPi/Arduino asap.



The thinking behind PeX is that Arduino has far away the BEST set of cheap accessories. 300+ shields and there are some really great ones. RasPi has sold a zillion boards, has far away the best community support and depth of projects/development. BBB is completely open source and has the coveted stronger, cheaper LCD support. Digi Xbee is a quick way to bolt on some wireless action, with a variety of radios. So it would be cool to see people use something from each system based completely on its merits, not it connector and host. PeX will be Open Source under FreeBSD, and Altium files will go up on GitHub just like they have for all my other projects (helix_4 FPGA Module and it's dev boards.)

So this means:
- You'd probably save money because you can shop around between systems to get what you need for less.
- You can use a couple of expansions/shields at a time which is important if you need to see them all at once for control, etc
- You probably won't have a big arse Rube Goldberg ball of wires as your project: the FPGA cleans all that up
- You can use Arduino and Pi together. Sooooo if you have existing Arduino projects/code but need to add a GUI or internet-working etc you can just strap it on by moving your Arduino setup to PeX and then plugging a Pi in on the same board.
- Ditto for Beagle Board
- Xbee wireless hookups x 3
- You'll get a 4.3" LCD on your project for $60 https://www.sparkfun.com/products/12085 vs $145 https://www.sparkfun.com/products/11742
- You get to use the FPGA if you want/need, including free Nios2, SignalTap, schematic editor, blah blah blah

So I need your help! Answer this for me:

1. Should PeX bother with the BBB Cape socket?
Is it a waste of effort just for an LCD? Should I just put a 4.3" LCD down on this board and charge +$80? Cutting the cape might save 2 layers on the PCB, so $10 or so.

2. Should I use Cyclone IV? Or Cyclone V?
PeX with Cyclone IV will be about $140 (shipped). With Cyclone V it'll be closer to $160. It'll take an extra month to delivery as well (no design reuse) but the Cyclone V is 2x faster than Cyclone IV and more useful in every way. Do people need more FPGA power?

3. Is this a sh#t idea? Is $140 too much for a Pi accessory?

4. Does anyone want DDR3 DRAM? probably 256MB @ 333MHz, 16bit wide physical (32bit wide virtual.) Again, this kind of ties in with the BB LCD Cape... video is a memory pig and the FPGA would abstract the LCD provide a frame buffer interface to the RasPi and Arduino, SPI or something parallel perhaps...

5. Would a few PMOD sockets around the periphery be a good idea? I might find some extra IO for that.

I really think that you could get some very professional results developing with a system like this, and I am trying really hard to keep it cheap and cheerful so everyone can have a bite. Please post your thoughts no matter how vile. In fact last time I posted to EEVblog I got a good bi^ch slapping. Come on EEVblog. Slap me again - I fu@king love it.

Brent.

PS. No this is not timed to coincide with the Papillio Duo launch. It's just a coincidence.

[Legit-Design]

Ok large pcb shaped in funky way looks nice.
Why not show what it is capable? How does this do something better? Ok for prototyping just plug and play, do you offer some pre-existing projects and libraries and all that for people who want the simplicity of just plug and play things? Make a nifty video showing off the capabilities in a realistic manner?

[miguelvp]

At work so keeping it short.

Why the PSoC3 instead of the 5LP you used before? cost?
Why not the PSoC42xx since you can get them for $1 each and free shipping?
Do you need the CAN bus or the 8051 core instead of the Cortex M3?
Nios II is not free, you can use it but always tethered to your computer.

How many dedicated direct lines to the FPGA from the Arduino and the Pi?

Why only 10K LE, even the Nano has way more. if it's cost the BeMicro CV goes for $50 and it has a Cyclone V E (with LVDS pins)

Is all communication to the FPGA via I2C? What if I need 150MHz or just 1MHz?
How are the digital blocks on the PSoC 3 configured to talk to the FPGA, will any communication be delayed by the 8051 core? Can the arduino or the Pi dynamically reconfigure the PSoC3 digital block to change the communication from the Arduino to either the FPGA or the Pi and all combinations of all three? will the 8051 arbitrate collisions?

What if I take my BeMicro CV $50 and a PSoC 42xx prototype board ($4) and hook it to an arduino or a pi?

Why is the EPCS16 not connected directly to the FPGA and going via the PSoC?

What speed rating on the Cyclone IV or V are you contemplating?
 

[Harvs]

I just don't really see a market for it...

I can see what you're doing, but it appears a bit too niche.

Most of the "maker" crowd I meet these days aren't phased by getting a custom PCB made, whether it's by themselves or getting someone to help them out.

[miguelvp]

I just don't really see a market for it...

I can see what you're doing, but it appears a bit too niche.

Most of the "maker" crowd I meet these days aren't phased by getting a custom PCB made, whether it's by themselves or getting someone to help them out.

There is a market:
Logi-Pi andd Logi-Bone
https://www.kickstarter.com/projects/1575992013/logi-fpga-development-board-for-raspberry-pi-beagl

and the ongoing Papilio Duo that got funded in under one week.
https://www.kickstarter.com/projects/13588168/papilio-duo-drag-and-drop-fpga-circuit-lab-for-mak

This board offers and Altera FPGA compared to the Xilinx of other kickstarters. But the price has to be competitive.

And Cyclone logic elements are not the same as a Xilinx logic cells so it should be similar in capabilities. I would go for >20K LE but keeping the price competitive to the 9K LC xilinx counterparts.

The DE0-Nano Cyclone IV has at least that many so does the BeMicro Cyclone V board. both being well under $100.
I know you are going to say they are modules, but they both have memory and GPIOs.

What is the sweet price range, is hard to know, I know for me spending over $100 on a gadget is not an impulse buy.

My most expensive dev FPGA board was the new terasic Cyclone V GX for under $200 but it has a lot to it, including hardware HDMI and I mean with a HDMI TX chip, and sd card and many more things.

[thinlayer]

Q) Why the PSoC3 instead of the 5LP you used before? cost?
A) Cost, no need for 32bit performance. This MCU is not the focus, it's just a sh%t kicker on the board to do the USB, bootloading images, host an SD card and make it transparent, etc. The most important thing it doesn is handle the ADC to the shields. You can reverse this and use the DAC output of the PSOC to drive INTO a shield which is cool. But I stress the PSOC is not the focus of the board.

Q) Why not the PSoC42xx since you can get them for $1 each and free shipping?
A) Last time I checked PSOC4 had no USB. Since the board will be powered by USB might as well use it for debugging and UI too. Plus I'm not sure what you get for $1. Jsut the cheapy part?

Q) Do you need the CAN bus or the 8051 core instead of the Cortex M3?
A) Nope. Will be cheapy PSOC3 with none of that. It's good to have a few opamps etc, comparitors as you can put an analog filter, modulator, etc.

Q) Nios II is not free, you can use it but always tethered to your computer.
A) NIos2 /e is free. The /s and /f versions are $500 perpetual license. The free version is ALL i've ever used and it's very good, with a few warts but they still excellent. The SRAM on the board will be there primarily for NIos2 users.

Q) How many dedicated direct lines to the FPGA from the Arduino and the Pi?
A) All of them. Each wing will have 21 pins that are sharead between Pi and Shield sockets (since only one or the other can be used at one time). Each wing will be point to point with no sharing. The wing hosting the BB socket will have 69 (I think that's the number) IO since those sockets are monsters.

Q) Why only 10K LE, even the Nano has way more. if it's cost the BeMicro CV goes for $50 and it has a Cyclone V E (with LVDS pins)
A) My experience has told me that people want low cost, not more FPGA fabric. I can get the price down this way. Cyclone IV has smaller parts, lower cost still for at least 2 more years. I can use Cyclone V if people tell me that's what they want. Remember that with 10KLE you can still do re-configuration on the fly; so if you want a different function, you can just load it from the Pi during run-time.

Q) Is all communication to the FPGA via I2C? What if I need 150MHz or just 1MHz?
A) I2C only? Sh*t no. Every pin on the expansion connector will be broken out, point to point and run as fast as you like. Protection diodes and resistors will be the only thing (apart from board traces) to limit the speed. I've acheived 40MHz plus on the Multi-Shield boards to every Arduino pin. I2C would only be for slow things like setting a register to steer the IO within a Wing - switching from Shield to Pi Expansion, etc. I would liek to be able to steer individual busses around, like the SPI and UART, with a register setting. A mux for each peripheral.

Q) How are the digital blocks on the PSoC 3 configured to talk to the FPGA, will any communication be delayed by the 8051 core?
A) Separate bus for ADC readings so nothing is bottle necked here. I have this code on the Multi-Shield working fine already. I2C as well and a separate UART for talking down the USB. Again, already developed. PSOC will need new development for SD Card, but cypress give us this for free with EmFile and their SD Card macro.

Q) Can the arduino or the Pi dynamically reconfigure the PSoC3 digital block to change the communication from the Arduino to either the FPGA or the Pi and all combinations of all three? will the 8051 arbitrate collisions?
A) Bootloader on the PSOC would do this. Again, it's a fresh development but a worthy one. RasPi can bootload an image over UART Cypress could help here, they have some boot loaders already we can port.

Q) What if I take my BeMicro CV $50 and a PSoC 42xx prototype board ($4) and hook it to an arduino or a pi?
A) Sure, and that's what I'm asking. What price does PeX need to be to stop you doing that? Consider the other features and the ease of use, lower risk than lash-up of your own. What would you pay not to have to break out the soldering iron and just get on with writing software?

Q) Why is the EPCS16 not connected directly to the FPGA and going via the PSoC?
A) PSOC has a CPLD function here, and can just be wired straight through to the FPGA. It would be transparent to the FPGA. OR you could have the PSOC manage the FPGA load, and the Pi can program it to use either the EPCS flash, or an image from the the SD Card, or something bootloaded from the Pi's own memory. So you can share the FPGA's Active Serial programming interface with many devices and schemes. This would make sense for reconfiguration on the fly.

A) What speed rating on the Cyclone IV or V are you contemplating?
Q) Cheapest. The bottle neck won't be the FPGA, it'll be the comms out of the Pi/Arduino. BB can match it with the FPGA, but I still think we'll see more than enough power from the slow '8' grade FPGA part. It's a lot cheaper.
 

[thinlayer]

What is the sweet price range, is hard to know, I know for me spending over $100 on a gadget is not an impulse buy.

I totally agree. I think that low cost will make all the difference and allow people the option to 'play' with it, and not have to get all hot and bothered analyzing and justifying the expense.

My most expensive dev FPGA board was the new terasic Cyclone V GX for under $200 but it has a lot to it, including hardware HDMI and I mean with a HDMI TX chip, and sd card and many more things.

Consider you buy PeX + RasPi for $130 + $45 + shipping, so call it $200. For this you get:
1. HDMI out (of the Pi) with a lot of support from the Pi community
2. Working linux with said superb support
3. SD card on the PeX, and SD card from Pi
4. USB host from Pi, USB device from PeX (for things like USB CDC, USB Audio, USB mass storage
5. You can put more shields on there... Xbee modules, a BB cape, pi expansion, a few of them, not just one. Who doesn't have a few expansion boards lying around too to get start with.
6. Everything is Open Source (except the Pi actually), but you can use Beagle Board host instead and have 100% open source. PeX will be FreeBSD use as you like.

I think I'm making a product here that is design for people who want to segway their development onto a Raspberrry Pi compute module too. BTW I have the Cyclone V GX board too, but haven't actually used it yet.

[Harvs]

There is a market:
Logi-Pi andd Logi-Bone
and the ongoing Papilio Duo that got funded in under one week.

See now maybe that's the problem.  When I read the post at the top, I did not take away "FPGA dev board coupled with RPi".  What I saw was a RPi, multi-shield expansion system, with an FPGA and PSOC being used to make it.

Where as those other project you've just linked to, when I read their name and intro there's no doubt in my mind that they are an FPGA dev platform.

Is mixed marketing messages what's hurting?

[miguelvp]

Thank you for the reply, makes sense other than the NiosII.

The NiosII/e is free but not stand alone, you have to have the programmer running on your PC and it will create a time limited sof, Or have I done this wrong all along?
So pretty much if that's the case and you want to use the NiosII soft core you must be tethered to the PC, and that makes not very useful for most cases.

Quartus (including Signal Tap and yes people that use Quartus know about it) and PSoC Creator are good tools, but the $140 might be too rich, even if I understand that with overhead and making it worth your while you might not be able to bring it lower. (unless you get a deal from Altera)

What probably helped the Logi-Pi and the Logi-Bone was most likely the robotics aspect of it specially the machine vision, but their form factor is way smaller. And their under $100 price probably didn't hurt either.
So you have to offer something concrete that will be non trivial to implement or buy a shield that will do it for you on an Arduino or Pi, something that people will find very attractive, what is that? don't now. But it has to be worth $140 to the user.

I do have a R-Pi my son got me for Xmas just gathering dust, so I'm a maybe but kind of inclined towards the no, since the Logi-Pi is less than half the price and has a better from factor.
And the papilio duo doesn't need an arduino since it's already there and it has a good form factor.

Sure I have not experienced Xilinx toolchain yet but I've seen it in action and doesn't look as horrid as I once thought it was (other than being huge as an install base, slower compiling etc).

Anyway I'll keep my eye on updates to see if it grabs my interest enough to change my mind.

[thinlayer]

The NiosII/e is free but not stand alone, you have to have the programmer running on your PC and it will create a time limited sof, Or have I done this wrong all along?
So pretty much if that's the case and you want to use the NiosII soft core you must be tethered to the PC, and that makes not very useful for most cases.

No - you dont need to tether Nios2/e. Nios2/e is free and you can use it anywhere untethered for eternity. the /s and /f have more features (that no one uses) and they are free if tethered, or $500 to untether.

It think that this isn't a board that you put into a robot. But its where you might start if you were going to research some algorithms and bring pieces of the robot together (like Accelerometer shield + Motor drive shield + Zigbee radio) for testing and development with relative ease, before shrinking to your own custom board; which is going to be sized up for the robot.

Perhaps I should just drop the Arduino chip on the board as well? I argue against this because you can just buy a Seeeduino or clone for like $20 if you need that. I will bundle them as an option too to save shipping.

[thinlayer]

Is mixed marketing messages what's hurting?

That's a good insight... and something really hard to measure when you are the guy selling.

I kind of understand why it's mixed. When I did my previous board no one gave a sh*t. I put this down to selling it as an FPGA board and not focusing on what people were looking for which was the functionality and capability for Arduino and shields and tools. I went a bit technology crazy and spouted numbers. So I am toning it down and focusing on the RasPi + Shield + other boards function.

Should I spruke it as an FPGA board? Do people really want to learn FPGA? Do they care about the technology? Is it a means or an end? Would they prefer to ignore it if there are alternatives?

I also have the option of ripping the FPGA + MCU off the board, putting a simple Max V CPLD there. This would take >$40 off the price straight away. It would be the data mux, and you'd steer the data around the place the same, but you couldn't do any of the FPGA'ish things like DSP, filtering, embedded processors, etc. I might consider two boards actually, PeX with FPGA, and PeX light (with CPLD), for $140 and $100 or less. Which would you buy?

[Legit-Design]

1) Articulate the idea. Describe it. If you aren’t sure how, focus on answering these questions:

• What is the problem you want to solve
• Who experiences that problem
• How you want to solve that problem
• Why this is a better solution

[sunnyhighway]

I could really see a use for a board like this. But what bothers me is the form factor.
It screams: "don't use me outside the realm of prototyping".
Now it's basically a board with 4 wings. Hard to put it in a nice case if I decided to use it in a one-off real world application.

Why not just replace these wings with connectors where you could stick in the wings you actually want or need?
And by offering both 180 and 90 degree angled connectors there is more flexibility in the type of case a finished project can be put in.

[jancumps]

That's a cool idea  . and the perks could be to ship it with one, two, three or four side boards.

[Rasz]

1) Articulate the idea. Describe it. If you aren’t sure how, focus on answering these questions:

• What is the problem you want to solve
• Who experiences that problem
• How you want to solve that problem
• Why this is a better solution

Very much this.

Currently its a kludged mess of everything in ugly and unfriendly form factor. Not to mention its big and probably expensive (looking at the previous failed one you linked).

Arduino people can hardly walk, they are barely crawling. Something like Papillo will work for them because it already has ecosystem build around it (gameduino, tutorials, videos, forum, new diagram input soft etc).

BBB people are too smart to pay >$100 for <$30 worth of stuff

Raspi, now there is an opportunity, but you might be too late to the party (LOGi-Pi). Altera alternative at same or lower price would probably get funded IF you can get some social media buzz, or if you can make compelling demo project for it. For example source cheap hdmi receiver/lvds transmitter chips and include HDMI-to-_any LVDS LCD_ converter. Raspi people are used to paying sick prices for small LCD screens (up to $200). $100 shield that would let them connect any LCD from a broken laptop + give them FPGA to play with could be a winner. Not to mention HDMI receiver on board means they could to video input stuff (detecting movement, ambilight clones etc)

[thinlayer]

Ok, many thanks to all to have helped me so far. I think I understand what EEVblog people think:

1. Cheaper than $100. Target would be $90.
2. Keep the 25KLE FPGA
3. Fix the form factor, make it fit a box/prototype, i.e. NOT just a dev board. I have a neat solution here that takes inspiration from 'sunnyhighway's suggestion but is a just little different.
4. Fix the marketing message, do market as an FPGA extension to Pi, as well as an IO extension to add more shields and Pi expansion boards.



About the Arduino

Arduino people can hardly walk, they are barely crawling. Something like Papillo will work for them because it already has ecosystem build around it (gameduino, tutorials, videos, forum, new diagram input soft etc).

That's a bit harsh, but perhaps Arduino users needs are better met with Jack's Papillio Duo. I have battled to attract that crowd but they are resistant and it might be that FPGA is a bit too much. And anything you can do on an Arduino you can do in a Pi. I think there's a version of the Wiring language and Arduino for Pi? Does anyone know about this?

Q) So I'm going to drop the Arduino HOST support (But keep the shields of course). Is this a good idea?



About the Spec

BBB people are too smart to pay >$100 for <$30 worth of stuff

Price is a function of volume. I make 100 units, they cost me $100 each to make. I make 1000 units they cost me $30 each to make. So I'll let the volume decide the spec. What I'm going to do is stagger the spec with stretch goals:


1. Funding goal :$2000 (or ~20 boards)
- I will just reuse the helix_4 FPGA modules. I can use my existing stock of about 30 modules (EP4CE22) that are good and tested. This means I can do the assembly in house with very low risk of China board stuffers and purchasing going pear shaped. The spec would be 4 'blocks/wings', with one that takes a Pi, and another that has the FPGA. Each will take an Arduino Shield, and at least one more will take a Pi Expansion, if not all of them. I will see if they can fit an Xbee socket too, goal of at least one of these for wireless bolt-ons. 
- option of EP4CE10 for lower cost system, perhaps $10 cheaper? $80 or so?
- because of IO restrictions there won't be capability for a BB socket, so no LCD.


2. Stretch goal 1 - funding of $6000 ( about ~60 boards)
- A custom board that has the EXISTING FPGA design dropped onto the board. This is a low risk option again, because I won't need to prototype anything, but I'll be using a China board stuffer and that requires some micro management and risk though.
- can only offer EP4CE22 because I'll need to purchase a tray of 84pcs and must consolidate.
- this will have a BB socket because I have access to a lot more IO. This will give a cheap LCD capability to Pi.


3. Stretch goal 2 - funding of about $12000 (about ~120 boards)
- accessory pack that includes a complete set of loopback test plugs. Adding this capacity helps you tell if you've fried your board and this is super important to know if and when you are stuck on a bug that won't seem to un-stick.


4. Stretch goal 3 - $25000 (~250 boards)
- A fresh design using Cyclone V 5CEA2 and/or 5CEA2.
- SD Card for FPGA (through MCU)


5. Stretch goal 4 - $???
- An ultra low cost version developed and offered in parallel with a CPLD (Max V) for people who are only interested in the Pi Expansion, and don't want to touch the FPGA.
- It would look like a big Mux and just switch the data around the place.
- Do able for $65 in volume I hope.


Anyone got an idea for another stretch goal?

I'll make an update to the web page, and post some concept pictures back here this week. I think I have a good solution to the form factor issue.

[miguelvp]

Now I like it a lot!

So keep the shield headers (but think about the ADC pins too) promote it as in a Pi FPGA glue for Arduino shields and/or as a Pi configurable hardware co-processor.
Needs memory that can be modified by the Pi (via the FPGA) and read by the FPGA, frame buffers etc. True dual ported will be best.

Since the FPGA you are talking about has LVDS it opens up driving LCDs directly.

About the Max V CPLD, it will be important to note how the Max V is closer to an FPGA than a CPLD, so it's like an FPGA with non-volatile storage making it an instant-on device.
But with the limits of the LE count not sure about it.

Which Max V do you have in mind?

Edit: $65 might be too much for the Max V.

[sunnyhighway]

Q) So I'm going to drop the Arduino HOST support (But keep the shields of course). Is this a good idea?

Dropping support for the Arduino host would be a bad idea in my opinion.

Let's just describe some properties of both the Raspberry Pi and Arduino people.

• The chances of Raspberry Pi people having a number of Arduino shields lying around will at least be more than one order of magnitude less than would be the case with Arduino people.
• Raspberry Pi people who already do have these Arduino shields, most certainly have an Arduino host as well.
• The Raspberry Pi people who don't have Arduino shields, need to invest in those. Invest as in spending money on these shields but also in learning how to uses them.

I'll leave it to the reader to find out what the implications are.

[rob77]

who is the target audience for the project ?
i see it useful to some extent for people playing with everything at the same time - but the majority of the community will make no use of it.
to have a successful kickstarter project, you should rather come up with something unique and useful for a bigger audience, something other than connecting everything
no offense - none at all. i'm just sharing my point of view.

you know there is a huge Rpi community, arduino community, fpga community (i mean a hobbyist fpga community) but there are only few people in 2 of them at the same time and even fewer in all 3 of them.

[sunnyhighway]

you know there is a huge Rpi community, arduino community, fpga community (i mean a hobbyist fpga community) but there are only few people in 2 of them at the same time and even fewer in all 3 of them.

You are absolutely right on this one. There must be a reason for this.
Could that reason be that there is currently no product available capable of bridging this gap as the PeX does? (with Arduino host support that is...)

[Rasz]

Still no vision. Still no description of what you are trying to solve/accomplish. Still no simple example of usage.
Also less is more - this rule also works in marketing. 100-in-1 thingmajigy only sells on late night TV shopping network, or in one of the Walmart isles full of backward rednecks. Things that promise to do lots of things do all of them poorly.

Basically WHY would I buy this instead of $30 FPGA board from ebay. No, im not handicapped, I can connect 8 pins together. Is there any form of instant gratification? Any must have?
 HW spec is just a small part of decision making process. Others are price, usefulness, ecosystem, marketing, killer features.

It looks like you are building this thing because its cool, not because there is a market for it.

Could that reason be that there is currently no product available capable of bridging this gap as the PeX does? (with Arduino host support that is...)

No. People that play with FPGAs have ZERO need for slow 8bit mcu with crippled IDE. Usually you already have one of those on FPGA board in a form of helper mcu (USB I/O).

[rob77]

you know there is a huge Rpi community, arduino community, fpga community (i mean a hobbyist fpga community) but there are only few people in 2 of them at the same time and even fewer in all 3 of them.

You are absolutely right on this one. There must be a reason for this.
Could that reason be that there is currently no product available capable of bridging this gap as the PeX does? (with Arduino host support that is...)

actually that's not the reason... the reason is the learning curve - someone doing stuff on 8 bit MCUs will not learn VHDL or Verilog unless he really needs to. and if someone once masters the fpga, he would rather do everything in fpga - even the 8bit MCU (if he needs it) as a soft-core (and many fpgas have a cpu core by default).
same for AVR vs. Rpi - no one will start learning linux programming unless he really needs to, and once he masters it - he will rather do everything on the Rpi in a comfortable high level language.
simply there is no need to connect AVR to anything else (other than having a AVR to collect sensor data and send it over serial line)... AVR and arduino are for small projects or simple projects.

the only exception is a small group of people playing with everything at the same time - but even those will probably rather connect it together themselves, than buying a board - so the target audience for such a project is very very limited.

[miguelvp]

You can use modules in schematic diagram, sure you can assign input and output pins and glue the hardware you want to drive to a Pi. Supply your clock (up to 4 clocks per pll?) line addresses data in/out or in/out, define output levels etc.

You don't need HDL for that, for simulation you would but Quartus will convert schematic to VHDL, Verilog or SystemVerilog for you so you can do your simulation and timing analysis.

Need LVDS signals, the fpga will provide those for you (at least the cyclone V he is targeting).

Buy yeah, the marketing part is what he lacks. How to convey to a user that they need this and that it would be hard otherwise.

Logi-Pi is out there but it's a Xilinx FPGA, his will be the first Altera one and if he provides tools and shows how to interact from the Pi to the FPGA to do something cool that would be impossible to do on the Pi, like sample 30MHz with 16 bits DAC shield (if there is one out there), Or how you can get any module that has data sheets and make them available to your Pi.

Pi is intended for programmers not for hardware people, so FPGAs might be well suited for them to be able to access hardware not targeted for the Pi.

[thinlayer]

So to answer the communicative question:


What is the problem you want to solve

We all want to make our embedded projects go smoother, faster and cheaper. And we shouldn't get 'pidgeon-holed' into one field of technology; it's good to cross borders and learn.


Who experiences that problem

There are four people: 1) RasPi guy, 2) Arduino guy, 3) FPGA guy, and 4) guy with a project to get done (but no preference how).

1) I'm familiar with RasPi and use it as a nice little PC already; but it isn't really an embedded computer. I want to use RasPi for real embedded computing that is more specific to my interests and I have to prototype real hardware. A RasPi project for me isn't just picking a nice case and plugging in the HDMI and RJ45. RasPi has a few little expansion boards, but not nearly enough. They pale in comparison to the varied and excellent (for embedded) Arduino Shield catalog. There are shields for every application and they are cheap, open source and they are simple. I want the RasPi to access Arduino Shields.

2) I'm an Arduino user and I've outgrown the platform. I want to move on and I think that RasPi looks like my next step. I'm familiar with Arduino hardware and if I could use my Shield collection along side a RasPi that would give me a big boost. With PeX I can grow my skills all the way to FPGA; I can ride the whole learning curve without spending more money on hardware.

3) I'm a guy who is familiar and comfortable with FPGA. I have a project that suits FPGA, and there are a few cheap dev boards around but the problem is the stupid money you pay to get peripherals on FPGA. The hardware can be expensive, but the IP/software do drive them is off the chart. USB on FPGA costs a couple of $k, ethernet MAC costs $500, IP stack can cost $4k and the list goes on. I want to slap a RasPi on my board, and bolt on a few shields for things like bluetooth, etc, but I will do the real work in the FPGA. I may eventually use the RasPi compute module along side an FPGA on my custom board but I want to try some software out first.

4) I want to rapidly develop an embedded widget with a stack of peripherals and capability in hardware. I can see 90% of my peripherals on Arduino Shields but I don't want to use Arduino. It doesn't have performance, and I want to write in C/C++ or another language that I have mastered. I want to use a few shields on a processor that isn't 8bit crappy.


How you want to solve that problem

PeX is a way to remove boundaries between our favorite development platforms. It's like one of those power adapters with the plugs for every country. This means you can take the OS from RasPi and mix it with the 300+ Arduino shields that are great for embedded applications. Some will even use the FPGA to offload work, or accelerate performance. PeX is about getting sh1t done quickly, saving time (and money) by not spinning as many custom boards and not risking new circuits where you don't have to. It's about reuse and up-cycling the best parts of each system.


Why this is a better solution

Because in one day of work you can add massive embedded capability to your RasPi. Get that Motor Shield out and put wheels on you Pi. Add a battery shield and a Bluetooth Shield and drive it with your phone. It's not hard to make a custom board to do these things but why not reuse, mix and match, solve and get on with it.

Wouldn't you rather jump straight to writing software?

[thinlayer]

Stretch goal 4: An LVDS output connector for driving LCD Panels.

@ $30,000 I would fit a hirose DF13 connector with LVDS channels to drive a 1080 display, and a generic backlight drive circuit (probably CAT4240). FPGA would drive 8 LVDS channels directly, this is quite easy to do.

This connector would accept any of the Digital View LCD cables (that terminate in a DF13, i.e. most of them) http://www.digikey.com/product-search/en?FV=fff40008%2Cfff80054&k=digital+view&mnonly=0&newproducts=0&ColumnSort=0&page=1&stock=1&quantity=0&ptm=0&fid=0&pageSize=500,

You would be able to drive any laptop replacement screen off eBay this way, and a 10.1" LCD module (1280x600) are about $45 + $11 for the cable. The PeX would be  good fit behind a display of this size.