Open source A3020 RAM

Arc/RPCs, peripherals, RISCOS operating system & ARM kit eg GP2x, BeagleBoard
Post Reply
cmorley
Posts: 988
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Open source A3020 RAM

Post by cmorley » Tue Sep 03, 2019 5:23 pm

I've drawn and ordered some PCBs for new A3020 RAM. Zarchos is funding the prototypes & once tested I will release all the design files & Gerbers.

I might make some up for forum members too if there is any interest.

I am using a sister chip to the one I used on the A3000 RAM I did with a small adapter board to the pin headers. Hopefully I got the dimensions right - if not I will have to redo the board.
Photo0791.jpg
Testing.. I don't have an A3020. IanS came up with a good test method of copying large (>90MB) files from the supervisor which uses all the RAM when thesting the A3000 RAM. Any volunteers? It can be scripted with an OBEY file but needs to run for hours :S

Chris

steve3000
Posts: 2085
Joined: Sun Nov 25, 2012 12:43 am
Contact:

Re: Open source A3020 RAM

Post by steve3000 » Tue Sep 03, 2019 6:07 pm

As for the A5000, I'd be happy to volunteer...great excuse to find my A3020 in the loft!

I also have the Acorn RAM test ROMs, which offer a fairly rigorous cyclic RAM test...

User avatar
IanS
Posts: 1010
Joined: Mon Aug 31, 2009 6:02 pm
Contact:

Re: Open source A3020 RAM

Post by IanS » Tue Sep 03, 2019 6:12 pm

steve3000 wrote:
Tue Sep 03, 2019 6:07 pm
I also have the Acorn RAM test ROMs, which offer a fairly rigorous cyclic RAM test...
They never found an issue with the A3000 RAM, which showed up errors via large file copying. The file copying was the first thing I could think of to check the RAM, I'd hope there are better options. Anyone fancy porting MEMTEST86?
Last edited by IanS on Tue Sep 03, 2019 7:41 pm, edited 1 time in total.

steve3000
Posts: 2085
Joined: Sun Nov 25, 2012 12:43 am
Contact:

Re: Open source A3020 RAM

Post by steve3000 » Tue Sep 03, 2019 6:20 pm

IanS wrote:
Tue Sep 03, 2019 6:12 pm
steve3000 wrote:
Tue Sep 03, 2019 6:07 pm
I also have the Acorn RAM test ROMs, which offer a fairly rigorous cyclic RAM test...
They never found an issue with the A3000 RAM, which showed up errors via large file copying. The file copying was the first think I could think of to check the RAM, I'd hope there are better options. Anyone fancy porting MEMTEST86?
That's interesting about the ROMs... thinking about it, they're also 32pin 8-bit ROMs so would need some work to get into the A3020... so file copying sounds best.

User avatar
IanS
Posts: 1010
Joined: Mon Aug 31, 2009 6:02 pm
Contact:

Re: Open source A3020 RAM

Post by IanS » Tue Sep 03, 2019 6:26 pm

steve3000 wrote:
Tue Sep 03, 2019 6:20 pm
That's interesting about the ROMs... thinking about it, they're also 32pin 8-bit ROMs so would need some work to get into the A3020... so file copying sounds best.
Mine were only 28-pin - viewtopic.php?f=16&t=12252&p=156202#p155802
Do you have a different version?

cmorley
Posts: 988
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Tue Sep 03, 2019 6:47 pm

File copying was necessary to get the failure because of the extra noise on the A3000 ground.

Without the IDE activity the poor signal ground on the A3000 MEMC caused all low signals to be raised on falling edges of MEMC output pins... with the IDE going this tipped it over the TTL threshold causing the DRAM to detect a RAS edge. I never did investigate if it was the A3000 PCB layout or insufficient ground pins on the MEMC - would have been easy enough to solder to the MEMC GND pin and scope that but I ran out of enthusiasm!

I am also pretty sure it was only memory reads too - so the machine ran and never crashed but we got bit errors in the file copy. I can't quite remember.

So... IDE activity is a good memory test.

cmorley
Posts: 988
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Fri Sep 06, 2019 7:25 am

Two boards complete. The hole is to allow access to LK32 on the motherboard. Hopefully the dimensions are correct.
Photo0793.jpg

User avatar
trixster
Posts: 924
Joined: Wed May 06, 2015 11:45 am
Location: York
Contact:

Re: Open source A3020 RAM

Post by trixster » Thu Sep 12, 2019 7:25 am

Looks great! Is this using faster ram to allow for an A3020 overclock?
A3020 | A3000 | A420/1 | BBC B | Master Turbo | ZX48K | NeoGeo
Atom | Amiga A4000 | A3000 | A1200 | A500 | PC Engine | Enterprise
Falcon | Saturn | PS1 | SNES | CPC6128 | C64 | 3DO | MD | Jaguar | X68000 | CD32

cmorley
Posts: 988
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Thu Sep 12, 2019 8:00 am

Potentially... if it works in the system. :?

The timings for key parameters are faster at typically 60% of the 80ns Toshiba TC514800. So as long as the RAM is compatible then there should be decent headroom for overclocking...

cmorley
Posts: 988
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Tue Sep 17, 2019 5:46 pm

Steve has done a bunch of testing and it doesn't work in an A4000 reliably. I imagine it is exactly the same problem as the A3000 but I was hoping the ARM250 SoC would have had better signal integrity :(

So buffer version ready to test at ABUG. If it works I will redo the PCBs and work out how to fit 11 buffers on the tiny boards :S
Photo0800.jpg
I adapted one of the A5000 boards which has space for all the buffers. Hopefully good enough for a proof of concept.

Buffers will take some of the overclock performance away but perhaps not too terrible because the RAS and A0-9 have good setup time margin IIRC and I don't buffer CAS... time and testing will tell.

steve3000
Posts: 2085
Joined: Sun Nov 25, 2012 12:43 am
Contact:

Re: Open source A3020 RAM

Post by steve3000 » Tue Sep 17, 2019 7:46 pm

Wow, I love that spider RAM... Now that's what proper upgrades look like =D>

cmorley
Posts: 988
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Sat Sep 21, 2019 5:55 pm

spider RAM doesn't work :( I think the long wires are just too poor for crosstalk/signal integrity. I will do a new PCB with the buffers on board.

cmorley
Posts: 988
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Mon Sep 23, 2019 3:33 pm

New PCB design with buffers on. I'll double check it tomorrow and then place an order.
Top:
A3020ramtop.png
Bottom:
A3020rambottom.png

steve3000
Posts: 2085
Joined: Sun Nov 25, 2012 12:43 am
Contact:

Re: Open source A3020 RAM

Post by steve3000 » Mon Sep 23, 2019 5:42 pm

Looks great, is that the same size footprint?

cmorley
Posts: 988
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Mon Sep 23, 2019 5:48 pm

steve3000 wrote:
Mon Sep 23, 2019 5:42 pm
Looks great, is that the same size footprint?
Yes. It all looked very tight at the weekend in XLT's A4000 so I tried hard to keep the PCB outline the same. The "Acorn A3020 RAM" text on the top was a casualty though...

steve3000
Posts: 2085
Joined: Sun Nov 25, 2012 12:43 am
Contact:

Re: Open source A3020 RAM

Post by steve3000 » Mon Sep 23, 2019 5:52 pm

cmorley wrote:
Mon Sep 23, 2019 5:48 pm
steve3000 wrote:
Mon Sep 23, 2019 5:42 pm
Looks great, is that the same size footprint?
Yes. It all looked very tight at the weekend in XLT's A4000 so I tried hard to keep the PCB outline the same. The "Acorn A3020 RAM" text on the top was a casualty though...
Ah ok. Might want to say "A4000/A3020" RAM?

Kazzie
Posts: 918
Joined: Sun Oct 15, 2017 7:10 pm
Location: North Wales
Contact:

Re: Open source A3020 RAM

Post by Kazzie » Mon Sep 23, 2019 8:22 pm

steve3000 wrote:
Mon Sep 23, 2019 5:52 pm
cmorley wrote:
Mon Sep 23, 2019 5:48 pm
steve3000 wrote:
Mon Sep 23, 2019 5:42 pm
Looks great, is that the same size footprint?
Yes. It all looked very tight at the weekend in XLT's A4000 so I tried hard to keep the PCB outline the same. The "Acorn A3020 RAM" text on the top was a casualty though...
Ah ok. Might want to say "A4000/A3020" RAM?
But that'll take up even more space! :)
BBC Model B 32k issue 7, Sidewise ROM board with 16K RAM
Archimedes 420/1 upgraded to 4MB RAM, ZIDEFS with 512MB CF card
Acorn System 1 home-made replica

steve3000
Posts: 2085
Joined: Sun Nov 25, 2012 12:43 am
Contact:

Re: Open source A3020 RAM

Post by steve3000 » Mon Sep 23, 2019 10:15 pm

#-o

User avatar
vanpeebles
Posts: 589
Joined: Wed Nov 28, 2012 10:01 am
Location: UK
Contact:

Re: Open source A3020 RAM

Post by vanpeebles » Tue Sep 24, 2019 7:32 am

Are we any closer to a turbo ram style upgrade? :) I would love one for my old A3010 [-o<

Post Reply