myStorm, BlackIce and the future

[lawrie]

I have two BlackIce II boards, two TinyFPGA BXs and an HDMI IceBreaker and a Fomu ordered.

I tend to use whichever of BlackIce II or TinyFPGA is easier for that project, and they have strengths and weaknesses.

So when I tried out Pawel Marczewski's Chip-8 project , I did it on the TinyFPGA.

Look no wires:

That would have needed much more messing about making PMODs to make it work on BlackIce or IceBreaker, and the breadboard adapters wouldn't have helped much.

Whereas when I used @uXe's HDMI connector (which he cross-posted on this and the TinyFPGA forum), I chose to use BlackIce as that was easier for a PMOD.

So there are advantages in both approaches.

I see the Hackaday has posted an article on ULX3S board today, and that takes yet another approach:

ULX3S: An Open-Source Lattice ECP5 FPGA PCB

[daveshah]

The big IO connectors on the ULX3S left and right side follow PMOD pinout with IO also in the "gap". This way you can solder PMOD connectors and use with PMODs if you want, otherwise you just have a load of IO pins for whatever purpose without so much wasted space as in a typical PMOD setup.

Almost the best of both worlds, although the need to solder sockets if you want to use PMODs (for a otherwise plug-and-play kit) is perhaps not ideal.

[Folknology]

The big IO connectors on the ULX3S left and right side follow PMOD pinout with IO also in the "gap". This way you can solder PMOD connectors and use with PMODs if you want, otherwise you just have a load of IO pins for whatever purpose without so much wasted space as in a typical PMOD setup.

Strangely enough I was working on something similar to this with the 'Ice' project :

Although I made about 30 of these I never actually released them. The 30 pin dual row R/A female socket for the Pmods isn't populated in this pic but you can see where it goes

[erikolson]

While PMODs are versatile, I'm working on BlackIce embedded in a larger project (working name: IceTea). I use PMODs 7-12 to connect to my board. Yep, its a triple double PMOD. It would be nice to have a simple row of pins to connect to, for lower part count. However, I am encouragng end users to fill PMODs 1-6 and 11 with their own add-ons. That's where the PMOD form factor becomes a feature again.

Price point is important. I figure I can direct people to buy my kit for $44 in a plain case, and separately order a $55 BlackIce to plug in, keeping it at $99 before shipping. That feels like a critical price point for a hobby purchase.

I'm two years into this project and learning Verilog, it has a declining number of bugs and growing number of working features and looks like shipping this year.

Whichever comes after BlackIce, I will consider adapting my hardware to use that. Probably the lower end board would be a better fit.

[davidp]

I have just started learning about FPGAs. I have a DE10-Nano and an IceStick, and acquired a BlackIce in 2017 but haven't managed to make initial progress till recently. The problem has been haphazard and impenetrable documentation and tools which have made getting started very difficult. You have to install gigantic software tools and invest enormous amounts of time to figure out how to do anything that it is a major undertaking to get started.

The enthusiasm of people such as on this site and Lawrie's excellent ebook on Open Source FPGAs have motivated me to do something. But it was only when I happened on Ken's post "Update on getting started with Windows" and four simple paragraphs on uploading a bin file to the BlackIce using TeraTerm that I made some progress.

I needed to know how to load and run an app before moving on to learning how to develop one.

My point is what makes something accessible to a beginner is not the feature set but the infrastructure around it - the documentation and tools.
David

[davidp]

Physical interfaces
My preference is for a small set of generic interfaces with carrier boards to add more capability. For example, I hardly use Arduino Uno but much prefer the Pro Mini/Nano which can easily plug in to breadboards or in to a CNC or Servo carrier board.

I think provision to mount directly to a Pi is a useful feature as the density of connections could be used for other boards but I don't think Arduino shield compatibility would be useful at all. Shield functionality is available as separate breakouts anyway if needed and it's geometry is not helpful.

The PMOD approach is vastly superior to shields in terms of wider application and re-usability. PMOD on the board versus being on a carrier is a topic for discussion.

USB is a good thing for simplicity of power and interfacing but I don't know whether more than one integrated vs PMOD would be better.

If you are considering carrier boards then would you integrate video or phone sockets on the fpga board or leave that to the carrier/PMODs?

Beyond this core you are moving from something that can be a component to being a development board that may waste features when included in an application.
David

[davidp]

Some core features
These are thoughts on high level functions, I don't have enough knowledge to comment on timing etc and the realities of signalling etc may mean you really have to have an integrated LVDS connector, I don't know.

For me even a learning platform needs to enable me to get to a substantial application. My end goal would be to be able to implement a softcore, such as RISC-V but more likely MIPS initially so I can exploit much of the learning material on digital/computer design and maybe implement some retro machines like the Beeb or a Digico M16.

In terms of scale then I might view an entry system (My*) as supporting a RISC-V MPU comparable to an Arduino or mbed ARM, while a more extensive board (BlackEdge) might be a platform for a Linux capable softcore

So a key feature is memory directly connected to the FPGA with the performance necessary for a viable MPU or 6502 or 32-bit CPU. I suspect this means SRAM for the MPU scale and SDRAM for CPU scale but I'm just guessing.

I think it has already been pointed out that the DE10-Nano RAM is connected to the ARM HPS rather than the FPGA which I feel limits how far you can take that platform. The BlackIce however, like the Papilio Duo, has RAM directly connected to the FPGA which for me is a key USP for that platform.

PMODs are essential for prototyping and learning but may be less optimal for application.

Diagnostics and debugging - For me the most helpful role of the integrated (ARM) MPU is to simplify bitstream handling, as BlackIce does really well, and support diagnostics and debugging. I don't want to use it for anything else, I want it to be a stable and robust means to access the FPGA easily. If I need an application processor then I would look to attach an Arduino, mbed ARM or Raspberry Pi or even a second My*/BlackEdge running a softcore system ...

I'd like to emphasis how much I like the approach to bitstream uploading on BlackIce - I only need a simple terminal program instead of proprietary hardware/software that I might only be able to get for Windows. As I gain experience I might use a CL tool via Make or whatever but I can always fall back on the simple serial terminal.
David

[Folknology]

Hi @erikolson, your project sounds cool, I am not sure of it's exact features or if you are willing to share them here, if not please direct message me as I am sure I can help you out on a number of fronts moving forward.

While PMODs are versatile, I'm working on BlackIce embedded in a larger project (working name: IceTea). I use PMODs 7-12 to connect to my board. Yep, its a triple double PMOD. It would be nice to have a simple row of pins to connect to, for lower part count. However, I am encouragng end users to fill PMODs 1-6 and 11 with their own add-ons. That's where the PMOD form factor becomes a feature again.

Love the name 'IceTea' BTW, I am trying to understand how many Pmods you need for 'Your board stuff' vs for users, would 2 double Pmods for each be enough?

Price point is important. I figure I can direct people to buy my kit for $44 in a plain case, and separately order a $55 BlackIce to plug in, keeping it at $99 before shipping. That feels like a critical price point for a hobby purchase.

Prices on the new products have not been set yet, the my* products will likely be lower cost (depending on options) and the BlackEdge products higher cost. I might be able to do something for you pricing wise with a co-operative or OEM agreement if that helps let me know.

I'm two years into this project and learning Verilog, it has a declining number of bugs and growing number of working features and looks like shipping this year.

Hey that's great I know how building hardware products can be daunting

Whichever comes after BlackIce, I will consider adapting my hardware to use that. Probably the lower end board would be a better fit.

BlackEdge will make integration in your own product much simpler, that is one of the projects goals, we will provide documentation and libraries to make this straight forward.

[Folknology]

I have just started learning about FPGAs. I have a DE10-Nano and an IceStick, and acquired a BlackIce in 2017 but haven't managed to make initial progress till recently. The problem has been haphazard and impenetrable documentation and tools which have made getting started very difficult. You have to install gigantic software tools and invest enormous amounts of time to figure out how to do anything that it is a major undertaking to get started.

I know how typical vendor tools are gigantic but IceStorm (And open source tooling) is pretty small in comparison, that's one of it's benefits

The enthusiasm of people such as on this site and Lawrie's excellent ebook on Open Source FPGAs have motivated me to do something. But it was only when I happened on Ken's post "Update on getting started with Windows" and four simple paragraphs on uploading a bin file to the BlackIce using TeraTerm that I made some progress.

The enthusiasm of people such as on this site and Lawrie's excellent ebook on Open Source FPGAs have motivated me to do something. But it was only when I happened on Ken's post "Update on getting started with Windows" and four simple paragraphs on uploading a bin file to the BlackIce using TeraTerm that I made some progress.

I needed to know how to load and run an app before moving on to learning how to develop one.

My point is what makes something accessible to a beginner is not the feature set but the infrastructure around it - the documentation and tools.

I am glad you got started with Kens blog piece on Windows. When we ship BlackIce now we also email a 'Getting Started link' which has a simple to follow guide, perhaps this was before you purchased your BlackIce which meant you had to rely on finding Ken's blog, apologies if this is the case, we have made a lot of effort to improve documentation over the last few years and have lots of input from our owners. We intend to keeping this up to continually improve these things..

[Folknology]

I think provision to mount directly to a Pi is a useful feature as the density of connections could be used for other boards but I don't think Arduino shield compatibility would be useful at all. Shield functionality is available as separate breakouts anyway if needed and it's geometry is not helpful.

The PMOD approach is vastly superior to shields in terms of wider application and re-usability. PMOD on the board versus being on a carrier is a topic for discussion.

In the current new product designs and layouts I have moved away from Arduino shields and RaspberryPi Hats, I believe, like you and others, That Pmods represent a better and more flexible connectivity/expansion, particularly as the Pmod options grow in availability and choices. BTW the BlackEdge design offers both carrier and Pmods as it comes with a fitted Pmod carrier as standard..

USB is a good thing for simplicity of power and interfacing but I don't know whether more than one integrated vs PMOD would be better.

Not sure exactly what you are asking here as USBs and Pmods are not mutually exclusive in my mind. For my* it's likely there will be a single USB connector, on BlackEdge there will likely be more than 1 USB option.

If you are considering carrier boards then would you integrate video or phone sockets on the fpga board or leave that to the carrier/PMODs?

For the BlackEdge boards, integrated video is the plan although it may be limited by the chip set capability. Not sure what you mean about 'Phone' sockets, perhaps you can expand this..

Beyond this core you are moving from something that can be a component to being a development board that may waste features when included in an application.

It depends on the product, each product will offer different features,I hope that the range of products offered enable the best choices tro be made, hence this conversation among others

[Folknology]

These are thoughts on high level functions, I don't have enough knowledge to comment on timing etc and the realities of signalling etc may mean you really have to have an integrated LVDS connector, I don't know.

Yes for BlackEdge this is necessary and it will include higher speed connectivity especially for differential signalling.

For me even a learning platform needs to enable me to get to a substantial application. My end goal would be to be able to implement a softcore, such as RISC-V but more likely MIPS initially so I can exploit much of the learning material on digital/computer design and maybe implement some retro machines like the Beeb or a Digico M16.

This is something that is key, I and others are focusing on softcores to make them more approachable for both for my* and BlackEdge.

In terms of scale then I might view an entry system (My*) as supporting a RISC-V MPU comparable to an Arduino or mbed ARM, while a more extensive board (BlackEdge) might be a platform for a Linux capable softcore

This isn't far off my thoughts to be honest, their may be a bit inbetween for the lower end BlackEdge..

So a key feature is memory directly connected to the FPGA with the performance necessary for a viable MPU or 6502 or 32-bit CPU. I suspect this means SRAM for the MPU scale and SDRAM for CPU scale but I'm just guessing.

For BlackEdge this will likely be SDRAM or DDRRAM for random access (in addition to internal memory) along with external Flash. For my* it will be SRAM and internal embedded blocks along with external Flash.

I think it has already been pointed out that the DE10-Nano RAM is connected to the ARM HPS rather than the FPGA which I feel limits how far you can take that platform. The BlackIce however, like the Papilio Duo, has RAM directly connected to the FPGA which for me is a key USP for that platform.

I prefer directly connected FPGA memory..

PMODs are essential for prototyping and learning but may be less optimal for application.

see above comments re Pmods/carrier

Diagnostics and debugging - For me the most helpful role of the integrated (ARM) MPU is to simplify bitstream handling, as BlackIce does really well, and support diagnostics and debugging. I don't want to use it for anything else, I want it to be a stable and robust means to access the FPGA easily. If I need an application processor then I would look to attach an Arduino, mbed ARM or Raspberry Pi or even a second My*/BlackEdge running a softcore system ...

I'd like to emphasis how much I like the approach to bitstream uploading on BlackIce - I only need a simple terminal program instead of proprietary hardware/software that I might only be able to get for Windows. As I gain experience I might use a CL tool via Make or whatever but I can always fall back on the simple serial terminal.

I like yourself (and other BlackIce owners) love the co-processor bitstream handling, it makes work with the FPGA much more agile and faster, particularly in testing and tweaking stages, it is also good for rapid experimentation. So this will persist with BlackEdge and my*. The programming model will be simpler moving forward such that the functions provided by the co-processor become more black box like and can be called upon from the FPGA side, rather than having seperate dev tools to develop directly concurrently on the co-pro. This will be achieved by having special modes for the co-pro that can be called upon from the FPGA, it will be simpler and much more effective and transparent as possible.

I intend keeping the CL and make compatibility, along with simple terminal features etc.. these are great for integrating into other software projects and tools BTW

[davidp]

The USB aspect is having the absolute minimum integrated sockets if a PMOD-USB is just as feasible.

Sorry, "Phone" was a typo, I meant "phono".

My thoughts here are that phono would take up real estate, incur costs and effect the overall height of all the boards where only a few applications might actually use audio. Personally I think an alternative such as PMOD or a development carrier board are better options. The same argument could be made of video as there is already a VGA PMOD unless there are practical/performance reasons that demand a video socket on the FGPA's board.
David

[daveshah]

HDMI is not ideal as a PMOD for signal integrity reasons, and the lack of a standard diff pair pinout for PMODs. There are PMODs with a HDMI transmitter chip on them too, which work but are only useful in one direction.

A mini HDMI connector takes up very little space but is very handy

[Folknology]

Pmods are great for the lower bandwidth applications, audio for example is easily channelled over a Pmod connection. So for BlackEdge products higher bandwidth applications will use built in connectors where practical e.g HDMI,Video Cameras and higher speed USB..

[erikolson]

I started my own thread for IceTea.

IceTea: interface BlackIce to TI-99.4A

I explain there the difficulty of squeezing everything into 24 pins (three double PMODs.)

One thing that would really help me is the ability to multiplex the SRAM pins. I need an address bus and a data bus, BlackIce has one in use, if only it could come to the edge of the board.

-Erik

[davidp]

Thanks to you and Alan below for clarifying that aspect.
David

[lawrie]

I just listened to the Hackaday podcast with @daveshah etc. I see Tim 'mithro' Ansell wants to get to a price point of $5 or $10, so my suggestion of $35 seems quite modest:

Hackaday Podcast Ep2 – Curious Gadgets And The FPGA...

[Folknology]

Yeah it was interesting listening to that podcast, (@ken.boak pointed me at it also), where Tim is talking about a low cost Raspberry Pi zero FPGA project just like my 'Ice' board I posted above, not quite sure about his $5 price point though!

[ken.boak]

With the Pi Zero costing around £5 but capable of supporting the whole FPGA development system, Tim Ansell's idea of putting a very low cost Lattice FPGA as an I/O board for a Pi Zero makes some sense.

The Pi Zero can program the FPGA via SPI, without any requirement for an FTDI chip that costs almost as much as the Pi Zero.

3 years ago, when BlackIce was first conceived, Alan and I tried to provide the best platform at minimum cost, but that market has moved on to support high, middle and entry point FPGA boards. 2019 will be an exciting ride for open source FPGA.

[daveshah]

Well the Upduino launched at $7, but that was with the support of Lattice and so many corners cut that the PLL wouldn't lock! I think the $5 target for a decent product is a way off though.