New A3000 4MB memory

Arc/RPCs, peripherals, RISCOS operating system & ARM kit eg GP2x, BeagleBoard

New RAM for you?

Poll ended at Mon Sep 24, 2018 8:05 am

Yes - Quality connectors please!
9
56%
Yes - Much cheapness connectors
1
6%
Maybe - at some point...
2
13%
Narp - I have 4MB already
4
25%
Narp - I have 2MB already and it is plenty
0
No votes
Narp - I have 1MB already and it is adequate
0
No votes
 
Total votes: 16

User avatar
mattyj
Posts: 45
Joined: Sun Nov 08, 2009 1:17 am
Location: South West - Bristol
Contact:

Re: New A3000 4MB memory

Post by mattyj » Wed Sep 12, 2018 11:38 pm

Looking good!

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Thu Sep 13, 2018 5:22 pm

mattyj wrote:
Wed Sep 12, 2018 11:38 pm
Looking good!
Thanks Matt. Have you pulled the trigger on a new microscope yet?

User avatar
mattyj
Posts: 45
Joined: Sun Nov 08, 2009 1:17 am
Location: South West - Bristol
Contact:

Re: New A3000 4MB memory

Post by mattyj » Thu Sep 13, 2018 7:17 pm


cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Thu Sep 13, 2018 7:53 pm

mattyj wrote:
Thu Sep 13, 2018 7:17 pm
Tad expensive at present.
Ring light and barlow lens on top of that too! :S

User avatar
mattyj
Posts: 45
Joined: Sun Nov 08, 2009 1:17 am
Location: South West - Bristol
Contact:

Re: New A3000 4MB memory

Post by mattyj » Thu Sep 13, 2018 11:07 pm

What magnification of Barlow lens do you use?
Also is the ring light an AmScope part?

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Fri Sep 14, 2018 6:10 am

0.7 for greater FOV and working distance. I bought the Amscope ring light so I knew it would fit (wasn't expensive).

User avatar
mattyj
Posts: 45
Joined: Sun Nov 08, 2009 1:17 am
Location: South West - Bristol
Contact:

Re: New A3000 4MB memory

Post by mattyj » Fri Sep 14, 2018 11:40 pm

If i was to buy a setup, these are the three items required?
Attachments
IMG_7013.PNG

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Sat Sep 15, 2018 7:50 am

That's what I got. I didn't want to spend a lot of time researching so I bought the Louis Rossmann (youtube) recommended setup.

They have other microscopes & cheaper stands and I'm sure you can get a ring light from eBay for cheaper but I wanted quick and known good for hours use at a time. 0.7 (or maybe 0.5) is a must or the field of view on minimum zoom is small. With 0.7 I get a 24 pin (narrow) DIP just filling the view fully zoomed out IIRC.

-=--=--=--=--=-

Back on topic:
Connectors ordered and arriving Monday. I'm going to send the first unit to IanS for testing. If that goes well I'll build up the first batch :)
Last edited by cmorley on Sat Sep 15, 2018 7:52 am, edited 2 times in total.

Nutter
Posts: 33
Joined: Sat Sep 08, 2018 11:18 pm
Contact:

Re: New A3000 4MB memory

Post by Nutter » Sun Sep 16, 2018 9:16 pm

Just want to make my interest known :wink:

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Mon Sep 17, 2018 2:43 pm

Production PCBs on order from JLCPCB. I thought I'd try them after their brand ambassador danielj has been doing such a good job :D

I made a slight adjustment to the dimensions (0.025") to make the connector fit better - the Acorn design of 3x SIL connectors is dreadful. :S

I'll post an update when they are here.

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Thu Sep 27, 2018 7:59 pm

JLCPCB screwed the order up & cancelled it. I'd put X*1:Y*5 panel on their system and they complained it was <70mm tall... humm... 167mmx32mm boards... 5 off... 5x32>70 AFAIK. Anyway turns out they'd decided to do it 5x1 so an 850mm x 42mm panel. Geniuses. Ask the customer? Nah just cancel it and ask me to reorder when I contacted support.

So I did... with Elecrow... and the PCBs are enroute. I should have them next week some time :)
Last edited by cmorley on Thu Sep 27, 2018 7:59 pm, edited 1 time in total.

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Wed Oct 03, 2018 9:43 am

Elecrow delivered. The production batch of PCBs arrived and they look pretty good.
Photo0807.jpg
I sent one to IanS to try out & see if there are any issues. I've got appropriately sized conductive bags on order (should be here tomorrow).

Batch 1 is built so shipping back end of this week :)

One thing to note is that the 1MB on the A3000 is not used anymore since all 4MB are provided by this module when installed. So it should work even if the A3000 has a bad RAM chip on board - theoretically...

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Fri Oct 05, 2018 8:05 am

IanS has a board and is testing it. He's experiencing an issue I can't replicate here.
Edit: I may have replicated it

Another tester would be helpful - someone who is a bit of an expert, has an IanS IDE interface and knows their way around the system, doing an md5 etc.

Voulunteer? Please PM me.

Once the problem is resolved I'll post on the for sale thread requesting payment. Many thanks to those who've paid up front already it really helps with the cash flow.
Last edited by cmorley on Fri Oct 05, 2018 8:13 am, edited 1 time in total.

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Fri Oct 05, 2018 5:33 pm

I investigated the problem and reflowed the RAM on those boards and now they work fully.

I don't think I failed to solder any pins (lead free solder :-& ) but instead think maybe I hadn't baked them properly after cleaning and there was moisture trapped under the chips on some of the boards. Either way reflowing the DRAM made them work properly.

I'll send another to IanS for testing which I definitely think is OK and see if he concurs.
Last edited by cmorley on Fri Oct 05, 2018 5:34 pm, edited 1 time in total.

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Sat Oct 06, 2018 4:37 pm

The test that threw up some errors was a large file IDE->IDE copy using the desktop with NEXT slid to max.

Comparing the results was sloow so I managed to get C working on arculator and wrote a noddy compare tool (attached). Works on RiscOS 3.11 - dunno about other versions. Attachment is an ADF disk image.
Attachments
cmp.zip
(2.55 KiB) Downloaded 4 times

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Sun Oct 07, 2018 5:00 pm

Investigations are still ongoing. I'll post when I have an update. I need to solder wires and take lots of measurements with the scope. :S

Sorry for the delay everyone but I really want to get to the root cause before I take payments and post out boards. My dev board works flawlessly! :S
Last edited by cmorley on Sun Oct 07, 2018 5:01 pm, edited 1 time in total.

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Tue Oct 09, 2018 9:35 am

Progress!

I am very confident I've found the root cause of the occasional errors that IanS discovered. I've also tested a fix which I am confident has cured it. I will modify a board and send that to IanS for further testing as I only have a sample size of one A3000.

It took me a while to find with a few head scratching moments and wrong turns - like thinking it can't be the soldering surely and indeed it wasn't! #-o

Once Ian has had a chance to test the board out properly I'll start collecting money and post the boards out. Another tester would still be helpful. Bueller? Bueller?

Chris

Technical:
The row address select (RAS) on the A3000 is a bit noisy. By the time it reaches my RAM chips it has even more noise coupled from the RAM data lines. The noise was close to the 0.8v threshold which the DRAM manufacturer guarantees as being read as low. Running the IDE increases the noise in the A3000. I think these DRAM chips are fast and sensitive. So hammering the IDE creates enough noise that the chip sees RAS as not being low anymore and terminates the cycle/starts a new RAM cycle - corruption happens on the RAM read. The solution is just to add capacitance to the RAS line.

The wild goose chase with the soldering and decoupling is that both of these affect the current supplied to the chip when it transitions the outputs - better decoupling and lower inductance on the power pins mean the IO can drive more current, so faster edges, so more noise! Hence why removing the decoupling capacitors actually "fixed" the problem. Or resoldering RAM on a new PCB "fixed" or reduced the error rate.

Cutter
Posts: 13
Joined: Sun Feb 11, 2018 2:32 pm
Contact:

Re: New A3000 4MB memory

Post by Cutter » Tue Oct 09, 2018 4:40 pm

Probably a dumb question, but would it make any difference if you were using an internal or external podule?

(Asking as I'm using an IanS ide podule on my A3000, but the 16bit version plugged into the external expansion port on the back, instead of an internal 8bit ide minipodule (as I want to retain the internal minipodule slot for a user port add-in card))
Last edited by Cutter on Tue Oct 09, 2018 4:40 pm, edited 1 time in total.

User avatar
IanS
Posts: 664
Joined: Mon Aug 31, 2009 6:02 pm
Contact:

Re: New A3000 4MB memory

Post by IanS » Tue Oct 09, 2018 4:46 pm

Nope, the podule used doesn't make any difference. It's not even just my IDE cards, the same issues were seen using a Simtec podule.

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Fri Oct 12, 2018 4:12 pm

Ian and I have been running a lot of tests and the cap mod seems to have fixed the issue. The cap mod just acts as a low pass filter & filters out some of the runt pulses coming from the A3000.

We have repeatability between my and Ian's A3000s so the test results are genuine. My test copies 190MB and checks for bit errors (takes just over 30 mins with my 8-bit IDE board from Ian).

I modded the entire batch and tested them for a greater sample size. I have 11 boards here which all pass on my machine now.

I have some more experiments planned and parts on order (due Monday) to do those experiments (didn't have suitable components already in my lab). So a little more investigation to go because I want to be sure these things are reliable.

Once again hats off to IanS for his time and testing. :)
Last edited by cmorley on Fri Oct 12, 2018 4:26 pm, edited 1 time in total.

aotta
Posts: 79
Joined: Fri May 26, 2017 8:57 am
Location: Italy
Contact:

Re: New A3000 4MB memory

Post by aotta » Fri Oct 12, 2018 4:29 pm

Good news! My complimenta to you and ians!

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Wed Oct 17, 2018 3:32 pm

There is good news and bad news.

The good news is my experiments have yielded success and I can tame the out of spec TTL DRAM signals in the A3000. The test card has the most sensitive RAM on I've found and works fine it seems.

The bad news is this means a redesign of the board to add a load of schmitt trigger buffers. This is going to cause another delay.

User avatar
myelin
Posts: 443
Joined: Tue Apr 26, 2016 9:17 pm
Location: San Francisco, CA, USA
Contact:

Re: New A3000 4MB memory

Post by myelin » Wed Oct 17, 2018 8:42 pm

cmorley wrote:
Wed Oct 17, 2018 3:32 pm
The bad news is this means a redesign of the board to add a load of schmitt trigger buffers. This is going to cause another delay.
That sounds painful! What signals need all the buffering?
SW/EE from New Zealand, now in Mountain View, CA, making BBC/Electron hardware projects for fun.
Most popular: fast serial port, FX2+PiTubeDirect Tube/Cartridge adapter, USB cart interface.

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Thu Oct 18, 2018 7:11 am

RAS, A0-9 are all out of TTL spec on my machine. The four CAS look better. So basically almost everything that comes from the MEMC is dodgy! RAS, A3 & A8 were the worst IIRC - out of spec with no memory card fitted i.e. stock machine. It is all high frequncy synchronous rubbish - if you turn on BW limit on the scope (or look on a 20MHz scope) it will all look fine.

Many signals take a positive offset (+100mV or so) too with IDE activity which means that there is poor 0v in the A3000 as well. This offset is the straw which breaks the proverbial camel's back with modern brand new DRAM. Hence why it worked flawlessly before the IDE but you get random fails when hammering the IDE.

cmorley
Posts: 662
Joined: Sat Jul 30, 2016 7:11 pm
Location: Oxford
Contact:

Re: New A3000 4MB memory

Post by cmorley » Fri Oct 19, 2018 9:16 am

Issue 2 sent off for manufacture. Wow this has been a time vacuum for something which should have been straight forward!

This one has buffers on all the MEMC lines (optional ones on the 4xCAS which I don't need). So if this doesn't work check the For Sale forum for an A3000 with boot prints on. :lol:

Top:
issue2_top.png
Bottom:
issue2_bottom.png

Post Reply