New MiSTer hardware board

[noel]

Hello,

I'm a french hardware engineer and I love MiSTer project. I built my own setup using the DE-10 nano board with some RAM and custom USB boards and I would love designing a custom MiSTer hardware based on the same Cyclone V chip.

I know that we couldn't beat the current price of the DE10 Nano board, as the chip 5CSEBA6U23I7N cost more than 200€ by itself !! On the other side, when new users have to buy the DE10 Nano board with all additional USB, IO, SDRAM, ... boards the total price is quite high too.

I could spend some free time designing a custom board and share this design via CC / openhardware license. I would be interested to manufacture those boards and sell them through my company for user who may be interested.

Maybe is a bad idea, as we could spend those effort into building a new hardware based on a more powerfull chip, but it would require a lot of work too to port all the fpga core to the new platform and to maintain core for more than one hardware...

Before going any further in this kind of huge development, I would like to get your feedback. I'm quite open-minded so please give me your positive and negative feedback about this kind of idea.

[ExCyber]

At some level, I'd love to see an all-in-one MiSTer-compatible board, preferably with a standard form factor (e.g. Mini-ITX), soft power control, and HDMI CEC. However, considering that the price is a huge part of what makes MiSTer viable as a platform, I think a more expensive version is a hard sell in general. If your target market is people who are willing to spend more, you might also think about how your price and features will compare to the FPGA Arcade Replay2, which isn't in production yet (as far as I know) but has a pretty big head start.

[dshadoff]

I would consider this to be worthwhile only if you could bring a more powerful set of features together - like a bigger or faster FPGA... and that may not be possible right now, but may be possible in future.

A second consideration is availability of units; with the DE10 Nano, supply easily meets demand, and there is no visible end of life yet; for a project built by an individual, there is always the possibility of supply being insufficient, or drying up altogether... I believe this happened to the MiST board.

In any case, we haven't quite exceeded the capabilities of the DE10 Nano, so there is no rush... but it is always possible that bigger/better would be useful in future (but also possible that Altera brings out a new board...)

[Slipard]

This is a good idea if you really go open source with an open licence.

A mini-ITX form factor with everything included on a single board is a really great idea and would broaden the audience.

Toss in a nice case and you will be competing with Analogue.

[aberu]

Instead of a regular alternative board, you should compete with the Analogue Pocket with your idea. Analogue needs a real competitor in this space for sure. If you can do the equivalent of the RetroUSB AVS, but a handheld competitor to the Pocket, then please do.

[NegSol]

I for one welcome any idea for making CC hardware. That was a huge part of what I liked about the MiST in the first place.
Also the availability of the hardware played into it. Price not so much. So I would say, as long as it is open and really available, you could get a return on your investment. Make sure to not underestimate the love people have for there present MiSTer setup - the DIY aspect that comes with it is part of it imho.

However my understanding is that MiSTer is not about new hardware at all, but more about preserving the old hardware in HDL and enabling software to be run like it was meant to be run.

[noel]

thank you for taking the time to share some ideas about this project.

First of all, I am very attached to creative common or open hardware licenses. I'm using Eagle cad for almost all my design until now, but i'm thinking of switching to Kicad for all new projects. I admit not being completely sure what this software can give for fast and complex components like fpga and DDR3, but someone does need to clear the way. I used Alitum, Orcad/allegro in the past so the switch is a bit rough but I don't see the point of offering a CC license design based on software that costs more than € 10,000.

The mini-ITX form factor is an excellent suggestion, as it make sense for computer cores like ao486. I still have my old 486 stuff, and I would dream plugging my 3dfx voodoo and s3virge video cards on a mini-itx mister board !! It would allow us to get some nice computer cases to install the mister board.

I don't think that the Analogue Pocket MiSTer alternative is the way to go, as it seems quite hard to deal with thermal dissipation of the MiSTer fpga in such small enclosure. It would only focus on some console core too, as I could not imagine plugging a keyboard on the Analogue Pocket to play some computer based games.

I was aware of the FPGA Arcade Replay2 based on a more powerfull architecture, but I prefer to focus on a 100% compatible DE10 nano board for now. I don't want FPGA developers to have to take a new board into account and waste development effort just to maintain compatibility between all boards available on the market. I know that the current DE10 nano board may never be powerfull enough to play N64, PSX games, but there is still plenty of room for older stuff like the cave arcade games. My PS3 seems a better way to go to play Saturn, PSX1, PSX2 games so far.

I completely agree with the fact that the DE10 Nano is widely available now and a I understand that some users love building theirs MiSTer setup. For my part I would have preferred to buy a ready-made set if I had had the choice, and users will always have the choice.

I think it would only made sense if it will add some functionality the DE10 could not afford due to current board design limitation (not enough gpio available, no onboard SDRAM, ...).

[Sorgelig]

I know that we couldn't beat the current price of the DE10 Nano board, as the chip 5CSEBA6U23I7N cost more than 200€ by itself !! On the other side, when new users have to buy the DE10 Nano board with all additional USB, IO, SDRAM, ... boards the total price is quite high too.

It's quite far from true. For budget solution DE10-nano + SDRAM is enough which is below 200EUR.
Besides the FPGA chip itself there are many other components on board and with small quantities they may cost another 100USD in total. The PCB itself is not 2 and not even 4 layers. I think it's about 8-10layers design. So check the price for such PCB - you will see another large cost addition.
Overall if you want to clone de10-nano design with extras then expect BONE cost not less than 400USD.
I even don't talk about cost of such board development which will require several revisions to build till all bugs will be eliminated. There are many high-speed buses on board requiring a special routing with equal resistance/capacitance/length. So it's quite complex task which eventually will require several K$ just to build the samples.
Do you want to take that adventure - good luck

DE10-nano has pretty much every free GPIO outside. So with current FPGA you won't get more GPIOs even if you make your own board. There are other versions of the same FPGA chip with more GPIOs, but it's basically useless route as any modification in pin definition will require a special core build. So there is no point to stick to the same series of FPGA. Probably if you want a real boost then Xilinx Ultrascale+ should be a better choice. Just the original point of MiSTer will be lost. MiSTer is a project for an affordable board and DIY friendly.

One case I think could be really interesting (although it will require all i've wrote in first paragraph). It's RPi form-factor. Basically the whole DE10-nano + SDRAM on RPI cloned board. In this case MiSTer could be embedded in any case/system designed for RPi. There is exactly the same FPGA chip in smaller (UFBGA) package which could be more suitable for RPi board.
This is pretty much the only interesting useful route for custom board. At least it will be possible to join the RPi ecosystem (although with quite expensive custom board).

[noel]

Your budget solution price tag is true for users that only want DE10 + SDRAM, and I understand that this is the main reason why the DE10 nano was chosen. The RPi or ITX-mini form factor are some interesting ideas, i'll have to think about it.

I agree with you about the overall development and board cost. I already designed 8-10 layers pcb boards and I don't expect the overall cost to be lower 400 USD for small batch quantity. Thats why I found interesting to gather all positive and negative opinions before going any further into this project. Cost development of such board is huge, and I already done some less ambitious project that already cost more than 100K€ in term of R&D and prototype boards manufacturing. You didn't even mention the manufacturing and certification costs

[Sorgelig]

The problem is that with 400 USD (very optimistic price) you will play in area where Ultrascale+ boards are cheaper and provide much more resources.
Special feature such as RPi form-factor may give advantage...
I remember i've saw some FPGA board in RPi form-factor.

[noel]

Replay 2 board is based on the Ultrascale parts and should be ready in next months, and I don't want to competing unnecessarily with the nice job he/they done.

We could try to lower the overall design cost by using a Cyclone V system on module, and try to design a nice motherboard that'll use no more than 4/6 layer PCB stack. For example the microSOM uS02 seems great and it seems to cost lower than 200$ with some core ICs (DDR3, eMMc, PMIC, ...). Unfortunately, that's not a way to design a RPi form-factor board.

Edit : microSOM uS02-0001 (Cyclone V SoC - 110 KLE) may cost a bit more than expected (found 267$ on the web).

[aberu]

I remember i've saw some FPGA board in RPi form-factor.

I have one kinda like this. It's called the Seeed Spartan Edge Accelerator Board (https://www.seeedstudio.com/Spartan-Edg ... -4261.html). It's FPGA (Xilinx Spartan-7 XC7S15), ESP32, and Arduino shield in one. I guess it's more of an Arduino UNO form factor though technically, cheap alternative to Arduino MKR Vidor 4000.

You might be talking about a different one though.

EDIT: Found the one you are probably talking about. The ZynqBerry

https://shop.trenz-electronic.de/en/TE0 ... orm-Faktor

https://www.digikey.com/eewiki/display/ ... +ZynqBerry

[noel]

As the overall cost of the card is the major sticking point, I opened some discussion with Exor embedded team to get some rough price of their Cyclone V SoM. I will let you know if I can get some solution at acceptable prices.

[Newsdee]

Doesn't the Vampire 4 run the same FPGA as the DE-10? That retails for 500-something Euros, I think.

[ExCyber]

Vampire V4 uses the same family (Cyclone V), but I doubt it's the same chip. I didn't see any references to the exact part, but one preorder page says 77K LEs, which is smaller than MiSTer's fabric and is a size only available in the non-SoC parts (i.e. parts without the integrated ARM system that runs Linux and the main MiSTer program).

[andrezheng]

The cheapest price for this SoC 5CSEBA6U23I7N, I can get from some china local vendor is around 100$. And this chip is not in stock and the lead time is very long. So at this point, it is still not a good point to redesign the DE10-Nano now.

[noel]

Doesn't the Vampire 4 run the same FPGA as the DE-10? That retails for 500-something Euros, I think.

It seems that they use Cyclone V A5 chip, like this one 5CEFA5U19C8N

[noel]

I have received some interesting feedback from Exor. I will seriously consider using their SoM to do a fully featured DE-10 Nano clone.
I'll consider the mini-itx size first as we could easily fit in some really nice case like this one : PHNXTOEM

Here is the SoM I plan to use. This module will allow us to reduce the overall board complexity (number of required layers, buried vias, ...) as it already embed the DDR3 memory and some core ICs.

Untitled-1.png (641.97 KiB) Viewed 17574 times

To further reduce the cost of this SoM, they can customize it.
We could do away with the built-in 4GB flash and reduce the amount of DDR3 RAM.

What do you think ? I could try to reduce the cost with no embedded flash and only 256 or 512MB DDR3 ram.

[NegSol]

Not really sure about reducing the DDR3 ram part. I have the feeling the MiSTer framework is already using the 1GB ram to its limit. At least when Soundfonts for Midi output is used. Also everything framebuffer related needs ram - some more is the way to go not less imho.

[ExCyber]

I'll consider the mini-itx size first as we could easily fit in some really nice case like this one : PHNXTOEM

One thing I'm not sure about for Mini-ITX is the I/O shield. Standard Mini-ITX cases take the same I/O shield size as ATX, but I think the linked case is Thin Mini-ITX which uses a 25 mm tall shield. I know there are some SuperMicro I/O shields that have their own part numbers and might be orderable in bulk, but I think those are all full-height. Beyond that approach, I wonder how hard/expensive it is to get an I/O shield that would be suitable for the usual MiSTer ports.

[noel]

Not really sure about reducing the DDR3 ram part. I have the feeling the MiSTer framework is already using the 1GB ram to its limit. At least when Soundfonts for Midi output is used. Also everything framebuffer related needs ram - some more is the way to go not less imho.

Ok I will then only ask to remove the embedded 4GB flash if we only consider using the external SD card.

One thing I'm not sure about for Mini-ITX is the I/O shield. Standard Mini-ITX cases take the same I/O shield size as ATX, but I think the linked case is Thin Mini-ITX which uses a 25 mm tall shield. I know there are some SuperMicro I/O shields that have their own part numbers and might be orderable in bulk, but I think those are all full-height. Beyond that approach, I wonder how hard/expensive it is to get an I/O shield that would be suitable for the usual MiSTer ports.

Indeed, I had not noticed this detail. This is a point to keep aside, but it is still very tempting to use the smallest possible ITX case available.
I found some specifications (see chapter 5.5). We should try to found some Thin Mini-ITX I/O Shield that would suit the MiSTer ports.

[mattsoft]

Rather than making an entire replacement board for the DE-10, why not design a Mini-ITX board that can host the DE-10? The Mini-ITX board could have a powered USB hub with audio, video, and USB ports on the back, and a way to use an SFX power supply (or DC adapter) to power everything. Then the DE-10/MiSTer could be mounted on the board and placed inside any Mini-ITX case.

[grizzly]

The problem with doing any kind of board where the de-10 would plug in is mainly the plugin in part.
For example the SD card is on one side and is (at least as far as i know) the hardest input/output to move so would probably need to be exposed to the io shield/port for ease of plugin in/out.
Then we have power/hdmi on another side, and then network/usb on a third side, and then all IO pins (that have for example 3,5audio/toslink/VGA and so on) on a fourth side (the top).

ALL (or at least most) will have to be connected and then routed/cabled to the IO shield/port on a ATX board.
How is that happening? By cables/flatflex/interconnecting boards, HOW are they going to fit as in needs to be thought out how to connect placement/orientation of pins/connectors.
Do we need to special order any of them or are for example normal length usb/hdmi panel mounts ok (probably will have to much cable that need to be coiled around).

[noel]

Rather than making an entire replacement board for the DE-10, why not design a Mini-ITX board that can host the DE-10? The Mini-ITX board could have a powered USB hub with audio, video, and USB ports on the back, and a way to use an SFX power supply (or DC adapter) to power everything. Then the DE-10/MiSTer could be mounted on the board and placed inside any Mini-ITX case.

I'm not ruling out any idea but I don't think it's interesting to use the DE-10 directly.

For the moment I am focusing on the mini-ITX format and I am trying to think about the different needs you have in relation to Mister (Computer, console, arcade, ...) and the impacts that this can have on the io shield and the need of specific hardware (Jamma interface, video amplifier, DAC, bluetooth, ...).

[MikeJ]

MIkeJ (Replay2) here - I wish I could find a cheap module supplier for the ultrascale part, it would make my life so much easier!
Good choice with the Exor module - although have you considered the TSoM module that Terasic do as well?

[noel]

Hi MikeJ ! It makes me very happy to read you here, I discovered fpgarcade.com in 2007 when I was working on some spartan3E for video processing product development, and have been a huge fan of your work ever since

I get some interesting feedback from Exor. I have seen the modules from Terasic but have not yet contacted them.
With a little luck we will be able to lower the displayed prices a little.

[MikeJ]

Ah thanks! Your biggest problem will be any changes to the pinout etc will mean forking all the cores. I might be able to target your board with a new backend for my framework depending what memory choices you make. Send me an email.

[Sorgelig]

reducing DDR3 amount will affect some cores. MiSTer starts to use more and more DDR3 as project evolutions. Less than 1GB of DDR3 will seriously limit current MiSTer potential.
Bandwidth of DDR3 is also very important. It must be at least 32bit. 16bit won't be enough.
There was similar to DE10-nano board from NovTech where you can see how FPGA board should NOT be designed. They simply ruined the whole idea (and sold recently remaining for $30 per board) . Basically same FPGA chip where one board takes everything possible with highest possible performance while another is utter trash.

[noel]

The Exor module has 1GB of DDR3 in 32bit, so I will stay on it for the moment. I will try to not design this board like NovTech did !!
As I would like to be compatible with current cores, I would take any advice on the elements that are the most critical and that should not be changed

[Sorgelig]

SOM you want to use definitely has its own pin configuration and you won't be able to convince the manufacturer to change it to de10-nano pin config. It means cores won't be able to work without special compilation anyway.
Hard to tell what is critical unless you try to port the cores to a new board.