Open source A3020 RAM

discuss the archimedes & risc pc, peripherals and risc os/risc os on pi
alexgibson
Posts: 16
Joined: Sun Dec 04, 2016 5:46 pm
Contact:

Re: Open source A3020 RAM

Post by alexgibson » Fri Jan 10, 2020 10:18 pm

cmorley wrote:
Tue Jan 07, 2020 7:01 am
alexgibson wrote:
Tue Jan 07, 2020 12:21 am
I like the cut-out for easy prising!
:shock:
Oh that was straight up... :)

I could 3D print a super simple tool that slots into the square hole, expands and lets you pull the board vertically up - really clean, low wear way to remove the part.

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Sat Jan 11, 2020 3:54 pm

Sure take a look. They aren't fun to remove at the moment because of the friction of all the pins in the ZIP sockets.

You can get the exact dimension from the gerbers.... viewmate will export DXF from Gerbers for example. The hole is there to miss the links on the motherboard - you'll need to work around them. I don't know where they are exactly in relation to the hole. I guesses and made the hole "really big" for plenty of clearance!

For a permanent installation swapping the ZIP sockets for turned pin sockets would make it easier.

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Sun Jan 12, 2020 9:42 pm

I tried an overclock today on an A4000. I don't have a range of crystal oscillators so I hooked up an FPGA to use the PLL. I soldered a buffer onto a SOIC breakout board to get a 5V CMOS output and plugged that in X100. It isn't ideal but it will do for testing and I can adjust the clock speed easily from the laptop.

The computer boots and seems stable at an oscillator speed of 75MHz (25MHz CPU speed). At 76 I get some message on boot and a * prompt but the error message is just off the screen on my TV! 77 won't boot.

I've run a few rounds of the IanS RAM test with 0 errors @75MHz osc. I'm not sure it is entirely stable but that could be the signal integrity of my Heath Robinson oscillator module... I've got dangly wires from the FPGA to the buffer - not ideal.

So it seems this RAM is pretty good for overclocking even with the signal buffers. The FPM version of the chip has a faster page mode (20ns vs 30ns) than the EDO chip but I don't have any of those left to test. (I'd have to buy 117 of them!)

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Mon Jan 13, 2020 12:39 pm

I've done some more testing and looked at some of the RAM signals on the scope. It looks good.

The system isn't stable at 75MHz oscillator but I'm not sure the RAM is the limit... the RAM test at 75 works perfectly every time. It might be the limit of this ARM250 or it might be the quality of my generated clock - perhaps a crystal oscillator would be more stable.

Anyway it works fine at 72MHz and the machine seems stable at that. I think that is a 100% overclock on an A4000. The GUI is certainely snappier. I can't try any games because this monitor doesn't support the refresh rate on the VGA input. Perhaps at ABUG it can run some demos and suchlike.

So I think this is pretty much signed off now. I'll sort out a sales thread this week - I will make 1 batch to order and that is it probably.

User avatar
danielj
Posts: 8241
Joined: Thu Oct 02, 2008 5:51 pm
Location: Manchester
Contact:

Re: Open source A3020 RAM

Post by danielj » Mon Jan 13, 2020 1:31 pm

That's pretty much ARM3 performance minus the the cache. Speedy!

d.

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Mon Jan 13, 2020 3:20 pm

72MHz - I wonder if you could just link the video oscillator over to the CPU clock input?! I looked on mouser and digikey and the DIP type oscillators are not easy to buy anymore.

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Mon Jan 13, 2020 5:47 pm

This thread is to wordy...

RAM in an A4000:
Photo0885.jpg
No oscillators so clock from an FPGA:
Photo0886.jpg
Bodge a buffer onto a SOIC breakout adapter:
Photo0888.jpg
Result!
Photo0883.jpg

User avatar
IanS
Posts: 1318
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: Open source A3020 RAM

Post by IanS » Mon Jan 13, 2020 6:44 pm

cmorley wrote:
Mon Jan 13, 2020 3:20 pm
72MHz - I wonder if you could just link the video oscillator over to the CPU clock input?! I looked on mouser and digikey and the DIP type oscillators are not easy to buy anymore.
Would a solution based on the Si5351 be a possible?
https://www.silabs.com/documents/public ... 5351-B.pdf
Would need initial setup, but could setup a fall-back output, selectable with a link. Make it in-circuit programmable?

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Mon Jan 13, 2020 7:01 pm

IanS wrote:
Mon Jan 13, 2020 6:44 pm
Would a solution based on the Si5351 be a possible?
I imagine so. Put it and a 25MHz crystal on a small 8 or 14 pin DIP footprint PCB & program with an AT-Tiny USB adapter or something? You'd need a 3v3 reg and a 5V buffer on the output for an archie...

That might be a handy widget Ian. Were you thinking of making one? I might tack some clock generators and crystals onto my next Mouser order and make one if not...

I just used the PLL in the FPGA & buffer because I have them to hand!

User avatar
IanS
Posts: 1318
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: Open source A3020 RAM

Post by IanS » Mon Jan 13, 2020 7:05 pm

cmorley wrote:
Mon Jan 13, 2020 7:01 pm
That might be a handy widget Ian. Were you thinking of making one? I might tack some clock generators and crystals onto my next Mouser order and make one if not...
It's not high on my priority list, so go for it. I always thought they would make a good solution for VIDC enhancers.

User avatar
IanS
Posts: 1318
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: Open source A3020 RAM

Post by IanS » Mon Jan 13, 2020 7:59 pm

IanS wrote:
Mon Jan 13, 2020 6:44 pm
Make it in-circuit programmable?
Not sure this is possible. To re-program the outputs you have to disable all the outputs. And the power-up state is only One-time programmable. Still a potentially useful device, but not as good as it could be for this application.

User avatar
1024MAK
Posts: 10090
Joined: Mon Apr 18, 2011 5:46 pm
Location: Looking forward to summer in Somerset, UK...
Contact:

Re: Open source A3020 RAM

Post by 1024MAK » Sun Jan 19, 2020 11:46 am

The following photos show a test fit of a test board in a A3020 and are provided so that Chris can see how the test board fits in practice. This test board may look different to the production boards.

Once fitted and pushed down, the test board is below the level of the metal screening, so there is no issue with the keyboard touching it.

The orientation should be correct (if it is not, let me know, taken using an Apple iPad mini) and ‘up’ is towards the back of the case.
C08CC7B9-EDA5-4B8D-99EA-8C58509B3779.jpeg
Empty sockets
.
5BF26E9E-CA94-4D2E-ADC9-B2F69F476614.jpeg
Test board in left hand sockets
.
49C571AB-F84A-40B9-AF7B-ED38039D4513.jpeg
Test board in left hand sockets
.
059BAEE8-0270-4235-8043-4159D435032B.jpeg
Test board in left hand sockets
.
3B08325F-6C0E-4878-9486-BDF606698216.jpeg
Test board in right hand sockets
.
D478DEE6-9910-4E84-AB50-D30EDEE08EC0.jpeg
Test board in right hand sockets

Fitting is a little tricky in the left hand sockets, as it is more difficult to see if the pins are correctly aligned. Removal is rather fun as well. But both are possible for this test board without having to remove the metal screen. Of course if you want the easy life, remove the metal screen :mrgreen:

Mark

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Sun Jan 19, 2020 12:40 pm

Thanks Mark!

There isn't a lot of room but there is enough it seems... the boards are much harder to install when all the pins are populated - and I left of as many as I could. So I'd say it is a shield off job for all but the most brave soles.

Next weekend at ABUG we'll try a fit & electrical test of a complete set in A3020(s). Since the A3020 & A4000 are so close electrically that should be a box ticking exercise. [-o<

Mark: The solder looks a bit dull around the electrolytic cap in you photo. Not sure if it is the camera angle but it looks like corrosion - have you replaced the caps, they could leaking?

User avatar
1024MAK
Posts: 10090
Joined: Mon Apr 18, 2011 5:46 pm
Location: Looking forward to summer in Somerset, UK...
Contact:

Re: Open source A3020 RAM

Post by 1024MAK » Sun Jan 19, 2020 3:53 pm

I’ve not done much with this machine (apart from play games :mrgreen: ). I have not replaced any capacitors. The board is a bit dirty in places. It could do definitely do with a bit of TLC.

Mark

User avatar
SimonSideburns
Posts: 545
Joined: Mon Aug 26, 2013 9:09 pm
Location: Purbrook, Hampshire
Contact:

Re: Open source A3020 RAM

Post by SimonSideburns » Sun Jan 19, 2020 10:48 pm

I have my two A3020s out of the loft, and I've even cut the batteries out and neutralised with vinegar.

I'm hoping they still work and I'll be bringing them either way.
Just remember kids, Beeb spelled backwards is Beeb!

User avatar
1024MAK
Posts: 10090
Joined: Mon Apr 18, 2011 5:46 pm
Location: Looking forward to summer in Somerset, UK...
Contact:

Re: Open source A3020 RAM

Post by 1024MAK » Sat Jan 25, 2020 8:41 pm

At the ABug meet-up today, Chris provided me with some RAM modules, which I then fitted to my A3020.
You definitely have to remove the metal screen, as very firm pressure is needed to push the modules into the sockets.

Here is a photo showing the two RAM modules in place:
60DE7B92-20C2-4822-B262-DDA0F3D589B7.jpeg
The two RAM modules in place
It may not look any different from the outside...
C03712A9-D03C-4B1C-A3DE-F1A2C521868C.jpeg
And here is the result confirming that the OS can see the extra RAM:
4254F75B-84B6-44E3-BCEB-54318E73939E.jpeg
Ahh, 4096K of RAM :-)
Thanks Chris :D

Mark

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Mon Jan 27, 2020 10:33 am

Thanks Mark. :D

Notes from testing with Mark:
  • Remove the shield so you can get good access & visibility on the pin alignment.
  • Installing in the ZIP sockets requires a scary amount of force - especially if the ZIP have never been populated.
  • Use foam and push down with thumbs on the pins to get vertical force into the pins and to minimise any twisting.
  • Don't try to push holding the edges of the module PCB.

Kazzie
Posts: 1757
Joined: Sun Oct 15, 2017 8:10 pm
Location: North Wales
Contact:

Re: Open source A3020 RAM

Post by Kazzie » Mon Jan 27, 2020 11:34 am

cmorley wrote:
Mon Jan 27, 2020 10:33 am
[*]Installing in the ZIP sockets requires a scary amount of force - especially if the ZIP have never been populated.
Oh, that brings back memories of installing sixteen (actual) ZIP RAM chips into my A420/1: I was glad I bought a bulk batch of chips with a few spares, as I broke two while trying to fit them! :oops:
BBC Model B 32K issue 7, Sidewise ROM board with 16K RAM
Archimedes 420/1 upgraded to 4MB RAM, ZIDEFS with 512MB CF card
RiscPC 600 under repair
Acorn System 1 home-made replica

User avatar
trixster
Posts: 1042
Joined: Wed May 06, 2015 12:45 pm
Location: York
Contact:

Re: Open source A3020 RAM

Post by trixster » Mon Jan 27, 2020 7:24 pm

When it comes to overclocking then, is it a matter of changing an oscillator on the motherboard?

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Mon Jan 27, 2020 7:54 pm

trixster wrote:
Mon Jan 27, 2020 7:24 pm
When it comes to overclocking then, is it a matter of changing an oscillator on the motherboard?
Not change but fit the missing one & the resistor to the motherboard. It is a 22 Ohm resistor not a 220R... 220 written on an 0805 means 22 * 10^0 not 220R. Vectorlight has an overclock page with some good photos on. You should leave the RAM on the motherboard alone - this upgrade is all 4MB.

The A4000 I've been testing with runs stably with a 72MHz oscillator so I am going to try linking the 72MHz VIDC clock across. I'll post if that works. Was going to try it at ABUG but home time came too soon :(

steve3000
Posts: 2251
Joined: Sun Nov 25, 2012 12:43 am
Contact:

Re: Open source A3020 RAM

Post by steve3000 » Wed Jan 29, 2020 11:17 pm

Great to read the progress with overclocking this awesome RAM upgrade. Stable at 24MHz is amazing, especially considering the original ARM3's were spec'd at 20MHz...!

I've not yet had a chance to test any overclocking using this in my A4000, as real life has got in the way of computers over the past couple of months :( but I will do soon, I hope.
IanS wrote:
Mon Jan 13, 2020 7:59 pm
IanS wrote:
Mon Jan 13, 2020 6:44 pm
Make it in-circuit programmable?
Not sure this is possible. To re-program the outputs you have to disable all the outputs. And the power-up state is only One-time programmable. Still a potentially useful device, but not as good as it could be for this application.
You need the ICD2053B, I have this driving the ARM3 on the Monster A3000, and you can re-programme on-the-fly, as it provides a glitch-free frequency modification procedure.

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Thu Feb 27, 2020 3:42 pm

They are officially horrible to remove from the ZIP sockets! The sockets really grip the 39 pins - you have to pry between the socket and the pin headers to get them out without damaging the pins. Lots of patience required.

I am going to swap the ZIP sockets on the project sponsor's A4000 for turned pin sockets.

Talking with RobC last weekend he told me the IOC is the overclock limit on these things. That ties up with my experience that RiscOS will boot fine at the high clocks with no graphical corruption but accessing the floppy will randomly fail & random SWI errors.

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Thu Feb 27, 2020 6:01 pm

There:
Photo0900.jpg
Much better! It is nice and easy to remove the RAM now.

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Sat Feb 29, 2020 5:36 pm

Memory fitting

As per Mark's notes earlier in this thread then getting good access to your motherboard is essential to see what you are doing.

1. LINK31 pin 1-2
2. Remove LINK32
3. Insert modules (be careful)
4. Connect wire from LINK32 pin on module to LINK32 pin2 on the A3020/A4000

When inserting get your eye (or camera) down to the PCB level to check the pins are aligned in all the sockets before pressing home. Then with some foam under your thumbs press down directly over the pins - so the force goes into the pins then into the socket. The board should then go in vertically. It does require a lot of force to insert - which is why you should double or tripple check the pin alignment before applying pressure.

See Mark's excellent photos and comments in this thread.

Removal

This is hard as the pins grip in the sockets really well. Use plastic tools to pry the boards up a fraction at a time & evenly. Or leave them in forever more! :)

Repair

If you do break a pin then They can be replaced with basic soldering skills. It can be easier to replace 2 or 3 pins instead of one because a group of pins is less fiddly to handle. Use a turned pin socket to align the pins when you solder.

User avatar
trixster
Posts: 1042
Joined: Wed May 06, 2015 12:45 pm
Location: York
Contact:

Re: Open source A3020 RAM

Post by trixster » Sat Feb 29, 2020 6:05 pm

Looking forward to getting the modules now :)

stephen_usher
Posts: 129
Joined: Sun Apr 10, 2016 2:47 pm
Location: Oxford, UK.
Contact:

Re: Open source A3020 RAM

Post by stephen_usher » Sat Mar 07, 2020 12:12 pm

OK, the modules have been fitted and working.

To make it (FAR) easier to fit them I opened the ZIP memory connectors by poking my oscilloscope probe into the holes, widening the contacts just enough. The modules then slid into place with only a "nice" level of force.

As I said, no problem. It JustWorks(tm).

stephen_usher
Posts: 129
Joined: Sun Apr 10, 2016 2:47 pm
Location: Oxford, UK.
Contact:

Overclocking.

Post by stephen_usher » Sat Mar 07, 2020 1:11 pm

After the conversation last night I've fitted a 22 ohm resistor between the 72MHz oscillator output to the CPU pad side of R10 and...

It works!

Zarch runs a bit quick!!!

But Pacmaina runs really slowly.

So, some things don't like the timings.

cmorley
Posts: 1277
Joined: Sat Jul 30, 2016 8:11 pm
Location: Oxford
Contact:

Re: Open source A3020 RAM

Post by cmorley » Sat Mar 07, 2020 1:48 pm

stephen_usher wrote:
Sat Mar 07, 2020 12:12 pm
To make it (FAR) easier to fit them I opened the ZIP memory connectors by poking my oscilloscope probe into the holes, widening the contacts just enough.
Huh... good idea.

Post Reply

Return to “32-bit acorn hardware”