A314 - The RPi co-processor Official thread

[Robban]

Hello!

Just wanted to thank you for coming up with and actually creating this awesome project!
I ordered a board from AndrewNC and just finished soldering it up yesterday. The voltages seems to be alright and i can not find anything fishy with my soldering using a microscope (yet...), so first trial run will be this evening.

I do have some general questions though:

-Earlier in the thread you stated that INT2 is optional. I have an INT2-pin soldered to the expansion connector on my 500+ so i intend to use that connection. Do you see any conflicts when using a Terriblefire TF530 accelerator? This card uses the INT2 wire as well to get IDE working properly. Should i link the connections together (INT2->TF530->A314)?

-I also have a TF534-card. This one emulates the INT2 interrupt directly on the card. Will this pose any problems if i connect the expansion-connectors INT2->A314?

-Will the A314 behave like a normal 1mb chip-ram expansion + clock if i plug it into the Amiga on its own, without the Raspberry pi (? If so, i should be able to boot into Diagrom and check that the extra chipram is detected and tests ok, right?

-On the raspberry, I suspect that i need to set up the a314d binary as an autostarting systemd service myself and just plonk the a314fs, picmd etc into /opt/a314/ ? What about the rasp-com.py script?

Sorry for all the stupid questions.. I grasp some of the details but not all

[Puggsy]

This is crazy insane good! Looking forward to see what is made with this expansion.

[Niklas]

Originally Posted by Robban

Hello!

Just wanted to thank you for coming up with and actually creating this awesome project!
I ordered a board from AndrewNC and just finished soldering it up yesterday. The voltages seems to be alright and i can not find anything fishy with my soldering using a microscope (yet...), so first trial run will be this evening.

Hi Robban! That's fantastic! I didn't see your post until now, so I'm curious how things went last evening.

Originally Posted by Robban

I do have some general questions though:

-Earlier in the thread you stated that INT2 is optional. I have an INT2-pin soldered to the expansion connector on my 500+ so i intend to use that connection. Do you see any conflicts when using a Terriblefire TF530 accelerator? This card uses the INT2 wire as well to get IDE working properly. Should i link the connections together (INT2->TF530->A314)?

-I also have a TF534-card. This one emulates the INT2 interrupt directly on the card. Will this pose any problems if i connect the expansion-connectors INT2->A314?

Hmm. I haven't tried using A314 with a TF accelerator. Eriond has used his A314 with a HC508 and that worked well. Would you consider getting A314 up and running without the TF first, before trying them together?

Regarding INT2 it should work without problem at least in theory, as the INT2 pin is a shared irq-pin.

Originally Posted by Robban

-Will the A314 behave like a normal 1mb chip-ram expansion + clock if i plug it into the Amiga on its own, without the Raspberry pi (? If so, i should be able to boot into Diagrom and check that the extra chipram is detected and tests ok, right?

Yes, that's the idea. I'm not sure if A314 works as intended if the RPi isn't plugged in at all however; it is possible that the SPI input pins on the FPGA are floating and could toggle back and forth between 0 and 1 which could confuse the logic that some communication is going on. I can try this here, but I think I have always had the RPi plugged in.

Originally Posted by Robban

-On the raspberry, I suspect that i need to set up the a314d binary as an autostarting systemd service myself and just plonk the a314fs, picmd etc into /opt/a314/ ? What about the rasp-com.py script?

I should really do an install script or something like that, but I suppose I was being lazy as long as no one asked for it... The binaries (a314d, a314fs.py, picmd.py) should go in /opt/a314 and the conf files (a314d.conf, a314fs.conf, picmd.conf) should go in /etc/opt/a314. All of these should be owned by root. Then there's a file a314d.service, which I seem to have forgotten to upload, that goes in /lib/systemd/service also as root (a314d.service is here https://pastebin.com/zdhczA5R for now). Then do sudo systemctl enable a314d to automatically start a314d when the RPi starts, or just sudo systemctl start a314d to start it once. In the a314fs.conf there's a reference to a directory /home/pi/a314?shared so you should create that also, it should be owned by the user pi. Btw I have only tried this on Raspbian.

rasp-com.py is not used at all at this point; I was thinking about writing skeleton code for a service implemented in Python, but I didn't finish that (as picmd.py works ok as a template anyway).

On the Amiga side you are probably sorted already, but a314.device goes in DEVS:, a314fs goes in L: and pi goes in C:.

I'm signed in (on and off) to the Amiga discord server mentioned here http://eab.abime.net/showthread.php?t=85960 and my user name is Niklas; if questions come up it might be more responsive to chat that way. If you prefer IRC there's a channel #a314 on Efnet.

[Robban]

Thanks for the detailed answer!

I did not really get much done yesterday, I had almost forgotten that the FPGA needs to be programmed as well..
It took the better part of 2 hours to download and install Quartus, and I had planned on using my Segger J-link as JTAG-interface. Unfortunately I could not get it recognized no matter what i tried.
Then I spent the rest of the evening looking for the usb-blaster i know i have laying around somewhere, gave up, and ordered a new one for 8$

So i have to wait a few days until i continue.

It is probably best to start with a normal 68000, yes. I'll make an adf in WinUAE containing workbench and the a314.device etc and then boot it using a gotek. With the Pi in place, just in case.

Do you know if there are any components that differ between the beta-2 pcb and V1.0? I looked through the Eagle files and saw that there was a lot of changes in the inner layers, but not too much on the surface ones. Right now i have used the BOM from the V1.0-folder on a beta-2 board and it seemed to work out.

I'll look into the IRC-channel

[Eriond]

There are no component differences between b-2 and 1.0. Just some mechanical placement adjustments to make the JTAG connector fit better. (see issue #5 and #9 on GitHub)

[Eriond]

Originally Posted by project23

Please don't think I was making a negative criticism; just trying to manage expectations of people requesting completed boards, that's all.

We welcome all comments and criticism. I believe those trying to wrap their mind around what kind of beast this actually is, deserve to hear all kind of input, despite it being "good" or "bad".
Of course, keeping the criticism on a constructive level helps a lot. I still haven't heard the "Your project sucks, and the whole idea is doomed" that other developers seem to encounter every now and then .


There are a number of things that could be considered as negative, but that must be seen in the light of what this project is all about:

• It's not you typical accelerator. Meaning, it won't make your 68k code run any faster. It can however make your A500 more responsive in certain aspects, when processor-intensive tasks are handed over to the RPi.
• It's not a Harddrive, because you can't boot from it. But it has a disk device (that shows up on the workbench as any other disk) with theoretically unlimited storage. Even connected to the cloud if you would like that!
• It's not a graphic/RTG board. But it can do some fancy graphic tricks by having access to Amiga chip mem. Playing video, for instance.
• It's not a network card/adapter per se, but it can already now serve your Amiga data and files from the Internet. It could become a SANA adapter too, if someone is willing to spend some time coding for it.
• It's not a sound card, but it could play MP3 using Paula audio, by offloading the tedious decoding to the RPi.
• It's not a cheap and readily available device, but it has the potential if someone is ready to invest in it. I made a quick estimate for what it would cost to make 50 complete populated and assembled boards in china: apx 85-90USD. Not cheap, but not unreasonably expensive either, considering what you get.

The nice part is that it's completely open, both hardware and software wise. That means it will only become more valuable over time when more features are added both on the Amiga and the RPi side. See it as a bridge between two different computer epochs, where the newer is given the possibility to serve the older, in order to leverage the Amiga experience into our modern age. [Fanfares and drum-rolls]

[Robban]

You are correct that the inner layers are the same. I manged to export .PNGs where one only showed the gnd vias and the other one only the vcc vias. Doh!

With some help I have managed to get the card working as a 512kb expansion + RTC Both Diagrom and Amiga Test kit detects the extra memory and tests it ok.

The communication Amiga<->Rpi does not seem to work yet though. We did some basic troubleshooting and SPI seems to be working at least. I have probably screwed something up on the raspberry side so more troubleshooting will follow.

We tested out a way to use a separate Rasberry pi to do the programming of the FPGA. This has the Pi emulating an Ethernet blaster, while flashing the generated .pof like normal from Quartus. It was hacky, had poor documentation and was somewhat glitchy. I love it!


The instructions are here https://weekly-geekly.github.io/arti...524/index.html

Some newbie things i encountered:
-It is the .pof file that should be flashed to the fpga, not the .sof. Quartus selects the .sof-file as default, so make sure to change it.
-If using the rpi to program, you need to disable SPI in raspi-config. This because the program for some reason uses GPIO0 (pin 27) and GPIO1 (pin 28) for two of the connections. These are unusable if SPI is enabled. If you use the same pi in the a314 afterwards the SPI needs to be enabled again.
-If you google "jtag pinout" and select the first image, you will see a 10-pin Arm Jtag connector. This is not the correct pinout. It should be the one you get when googling "altera jtag header" or similar (pin1=tck, pin3=tdo etc).

Edit: 512kb slow only for now, as i changed to a rev6 500 for testing. This has a 8372A agnus and is limited to 1MB of ram total. I have not done the 1MB chip-ram mod.

[Niklas]

Originally Posted by Robban

The communication Amiga<->Rpi does not seem to work yet though. We did some basic troubleshooting and SPI seems to be working at least. I have probably screwed something up on the raspberry side so more troubleshooting will follow.

Hey Robban! I just tried taking the binaries from GitHub and running, and I got the same result that you did -- it seems I had forgotten to commit the latest compiled a314d to the repo! Very sorry about that. I have now committed the latest compiled version of a314d (only that file is different) and pushed to GitHub, and with that version it works here, so download and try that.

Having both source code and compiled binaries in the same repository seems like a bad idea; I was slightly worrying something like this could happen, and the reason it was like that in the first place is probably laziness. I added an issue for this: https://github.com/niklasekstrom/a314/issues/15.

[Robban]

Haha!

Laziness from both of us then, it is not like it would have taken me too much effort to compile it myself to be sure. I might have time to test tonight, otherwise it will have to be in a few days time.

[Robban]

I tried the new a314d daemon and got better results than ever!
It did not work right away, the daemon stopped with "SyntaxError: Non-ASCII character '\xc3' in file /opt/a314/picmd.py on line 1, but no encoding declared".
Looking at the first och picmd.py line it said "# Copyright (c) 2018 Niklas Ekström". This is obviously some form of copy protection, so i cracked my copy by removing the line

Now i have this thing of beauty:

Yay!!
(Yes, this Amiga is in dire need of a recap, it has terrible image quality even via rgb)

[Niklas]

Originally Posted by Robban

It did not work right away, the daemon stopped with "SyntaxError: Non-ASCII character '\xc3' in file /opt/a314/picmd.py on line 1, but no encoding declared".
Looking at the first och picmd.py line it said "# Copyright (c) 2018 Niklas Ekström". This is obviously some form of copy protection, so i cracked my copy by removing the line

Darn, I thought that would be foolproof I will update the files with the required encoding string.

Originally Posted by Robban

Now i have this thing of beauty:
Attachment 63452
Yay!!

Well done! That's very nice to see.

[aligebes]

I've been lurking for a while reading this and find myself wanting to support this, as I like the idea of easily being able to transfer files from the internet.

I see the boards are on ebay, and would like to know if I purchase one, that I am indeed supporting those that have developed the A314, and have someone who can build this at some point way in the future

If it is possible though at some point to buy a fully built board from someone here ( as I have no experience in soldering whatsoever ) I would very much like to express my interest in obtaining an A314 . Would this maybe be possible please, and what sort of approximate price would I be looking at ? So that I can plan ahead for it.

Thank you, and look forward to any replies .

[AndrewNC]

Originally Posted by aligebes

I've been lurking for a while reading this and find myself wanting to support this, as I like the idea of easily being able to transfer files from the internet.

I see the boards are on ebay, and would like to know if I purchase one, that I am indeed supporting those that have developed the A314, and have someone who can build this at some point way in the future

If it is possible though at some point to buy a fully built board from someone here ( as I have no experience in soldering whatsoever ) I would very much like to express my interest in obtaining an A314 . Would this maybe be possible please, and what sort of approximate price would I be looking at ? So that I can plan ahead for it.

Thank you, and look forward to any replies .

I am the seller on eBay, if you want a PCB it's cheaper to buy one directly from me on here (eBay price is higher due to fees etc).

At some point i'll have a couple of spare completed boards available to buy, but i haven't had time to build them yet. Price will be £90 GBP plus shipping (Pi not included).

Andy.

[aligebes]

I'll happily wait and pick up a completed board in due course. It'll give me time to put some money aside

[hyami]

I was wondering; when this works on an A500, would it be feasible to make it work on a Zorro card? It would be a perfect add-on for my A2000.

[Mark sealey]

Awesome work...i would love one of these.....but could never make one.

[McTrinsic]

+1 for the Zorro version

[hooverphonique]

Originally Posted by hyami

I was wondering; when this works on an A500, would it be feasible to make it work on a Zorro card? It would be a perfect add-on for my A2000.

Some features will not work in that config, especially where the RPi places data directly in 'chip ram'.

[Niklas]

Originally Posted by hyami

I was wondering; when this works on an A500, would it be feasible to make it work on a Zorro card? It would be a perfect add-on for my A2000.

Originally Posted by McTrinsic

+1 for the Zorro version

One of the features of A314 is that the memory on A314, to which the Raspberry Pi can read and write, is chip memory. This means that the Amiga custom chips can access this memory directly without involving the 68000 processor, which is how video playback and RemoteWB can work so efficiently. If there were a version of A314 that sits on the Zorro bus then the memory on that A314 would not be chip memory, and the applications mentioned above would not work as efficiently.

There are however several applications that do not require access to chip memory, such as the a314fs file system and the pi command. For this reason I think a Zorro version could still be interesting.

I don't know of any efforts to design such a Zorro version of A314 today. If someone would be interested in designing the hardware then the software should be fairly easy to port over from the current hardware. The hardware for the Zorro board would consist of level shifters (5V->3.3V), an FPGA (FPGA and level shifters could possibly be replaced with an XC95 CPLD), 1MB SRAM, and the connector for the RPi (the interface between FPGA and RPi is 4 pins SPI and a single GPIO pin for interrupt request). I know for certain that Eriond could design the hardware easily, but I don't think he has the time to do it(?) If someone else feels up to designing and building the hardware then I would gladly help out with doing Verilog and porting the software.

[Methanoid]

Much as I would love to score a board, I'm not a dev of any kind, but I would suggest you consider whether you can also implement using the USB stack on the Pi to give Amiga USB mice.

Examples here for ST [ Show youtube player ]
There is some extra code which makes it Amiga compatible - https://github.com/BigWhale/STuffEmu