A540 (or R260) 4MB card

discuss the archimedes & risc pc, peripherals and risc os/risc os on pi
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

I've just noticed my A540 is an Issue 1 board, whereas Andy's R260 is an issue 2.
IMG_E7739.JPG
I first noticed differences around the spade power connectors (not fitted on iss 1, fitted on iss 2).
Then the 1772 is socketed on iss1 but soldered directly on iss 2.
Vice-versa, iC51 is soldered on iss 1, but socketed on iss 2.

There's also these two resistors on the underside of the board near IOC on iss 2 which appear factory fitted.
(Blue wires are not my handiwork btw)
Not sure if these are present on iss 1 and not inclined right now to tear down the A540 to check ;)
IMG_E7740.JPG
Still looking for other differences. There's clearly something significantly "better" about my A540 than anyone else's with respect to working with IanS's cards. Too weird.
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

IanJeffray wrote:
Sun Oct 10, 2021 5:56 pm
I've never seen any combination cause errors for slots 0 and 1
Scratch that one. With IanS cards in slots 1,3 and a Simtec card in slot 2, I'm now seeing very occasional errors for slot 1 (only) on the R260.
AndyMc1280
Posts: 729
Joined: Sat Aug 27, 2011 11:50 am
Contact:

Re: A540 (or R260) 4MB card

Post by AndyMc1280 »

IanJeffray wrote:
Sun Oct 10, 2021 10:58 pm
Scratch that one. With IanS cards in slots 1,3 and a Simtec card in slot 2, I'm now seeing very occasional errors for slot 1 (only) on the R260.
I suspect my R260 is borky, and I also suspect its something to do with the "hotwires" on the back of the board. Just cos I want to lay a massive poop in your ointment :lol:
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

IanJeffray wrote:
Sun Oct 10, 2021 5:56 pm
I've yet to notice any failure ever of the physical memory test with any combination of cards.
And scratch that too. (Maybe). Yes, it may be that this R260 is flaky (but it's running with Acorn cards perfectly, and no other issues when running on a lab supply....)

But here we see an IanS card in slot 1 that's completely fallen over. Sorted by a simple power-cycle tho!
IMG_7745.JPG
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

Another possibly irrelevent but interesting observation - when running steve's memory test, the board is drawing up to 3.1A @ 5V during the logical tests, but only 2.7A during the physical tests.

There's not as much bypass capacitance on IanS's boards as there is on the Acorn boards (by far) and whilst the Simtec board is very similar, it has a pair of SMT electrolytics vs tants on IanS's boards (same capacitance) though if anything you'd expect tants to be better?

Worth testing some additional bypass on an IanS board in slot 2 ? Easy to tap a 2nd 47uF leaded electrolytic across each tant? Can't harm to try?

For shiggles, I also just ran some testing with Acorn, Simtec, Acorn. No issues in 25mins.
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

I had a crappy idea -- take the MEMC from IanS's Simtec card and put it on an IanS card in slot 2 - to see if there was anything 'special' about the MEMC on the Simtec card. With Ian's permission, I've tried this - using AndyMc1280's R260, since my A540 essentially never shows any issues.
(BTW this is only readily possible with the Simtec card, rathe rthan an Acorn card, as MEMCs are soldered on Acorn cards)

Initial results looked good! Down from 95%+ errors per mem11 test line (1:Acorn, 2:IanS, 3:Acorn configuration) to maybe 5% ... but once it warmed up after 10-15mins it had a whole slew of tests that failed maybe 20 lines in a row, then back to passing again for 10 mins before anothe clutsch of errors. Feels "better" but still completely useless really.

So then we try putting the MEMC from the IanS card in the Simtec card.
Configuration is now 1: Acorn, 2: Simtec with your MEMC now in it, 3: Acorn.
No errors in 30mins of running... 'conclusive'-enough for this level of test IMO.

Oh well, something else to rule out / add to the mix - no apparent "magic" with Simtec's MEMC and no issue with the 12QC MEMC on IanS's card.
MEMCs swapped back to their original positions.
steve3000
Posts: 2571
Joined: Sun Nov 25, 2012 12:43 am
Contact:

Re: A540 (or R260) 4MB card

Post by steve3000 »

If we're going for random tests, how about taking the RISC OS ROMs from your computer into Andy's? I saw significant differences with different ROM sets...

From the evidence so far it seems like any changes on the address bus are likely to impact the results to a degree.

But how do we make IanS work permanently? Is this a case of modern electronics on the IanS card having less 'slack' than older ICs used on other cards? (differing capacitance? buffering?... I'm guessing now...)
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

steve3000 wrote:
Tue Oct 12, 2021 8:49 pm
If we're going for random tests, how about taking the RISC OS ROMs from your computer into Andy's? I saw significant differences with different ROM sets...
Good call. The 'original' 3.10 ROMs from my A540 are still sat here as I swapped them for 3.11's that 'caused' that single error ever observed on my A540, so I'll bung those in right away. (Don't worry Andy, you'll have your precious 3.11's returned...).
steve3000 wrote:
Tue Oct 12, 2021 8:49 pm
But how do we make IanS work permanently? Is this a case of modern electronics on the IanS card having less 'slack' than older ICs used on other cards? (differing capacitance? buffering?... I'm guessing now...)
I'm assuming RAMs are comparable in terms of setup/hold/drive etc - I've not tried to compare to the parts on the Simtec card myself. I've a fast FET probe, so could try comparing signals, but ... urgh really (Would also ideally need some code to just spin accessing a given RAM slot continually).

Could try adding more local capacitance to IanS's card - it's quite light (Hella light in terms of how we design lower voltage stuff these days!)

You said multiword accesses are always (?) fine so does that mean it'd be possible to isolate the issues to reads vs writes? (By doing STR/LDM vs STM/LDR when testing). Would that even help with understanding?

Whittering now. Will get these ROMs swapped...
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

3.10 ROMs from the A540 made no difference to testing on the R260 .. still absolute mess with errors on almost every iteration of the mem11 test. Useless.
User avatar
daveejhitchins
Posts: 6791
Joined: Wed Jun 13, 2012 6:23 pm
Location: Newton Aycliffe, County Durham
Contact:

Re: A540 (or R260) 4MB card

Post by daveejhitchins »

IanJeffray wrote:
Tue Oct 12, 2021 9:49 pm
3.10 ROMs from the A540 made no difference to testing on the R260 .. still absolute mess with errors on almost every iteration of the mem11 test. Useless.
Just out of interest: What's the decoupling, on the various boards, and where is it placed?

Dave H.
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

daveejhitchins wrote:
Wed Oct 13, 2021 7:54 am
What's the decoupling, on the various boards, and where is it placed?
Simtec vs IanS:
IMG_E7762.JPG
Very similar arrangements. Simtec uses some weird kind of ultra-flat series resistors, Ian uses packs.
There's (unstitched) plane fill on the Simtec board too.
Bulk caps are tants vs SMD Electrolytic - same value.
Local decoupling appears simlar, if not mildly increased on Ian's board.
Ian's board has a PTH socket, Simtec's is SMT, but if we're down in the weeds of signal integrity of through=hole pins at 12MHz, that would be bizarre (normally only worry about this above 100MHz) and trace-length matching would then start being an earlier issue (?) which clearly isn't done on either board.
Trace widths look greater on Ian's board, so even though it's only a JLCPCB (1oz copper it aint) should be no concern?
User avatar
daveejhitchins
Posts: 6791
Joined: Wed Jun 13, 2012 6:23 pm
Location: Newton Aycliffe, County Durham
Contact:

Re: A540 (or R260) 4MB card

Post by daveejhitchins »

I may be "teaching to suck eggs" here (so apologies!), however . . .

1) Not all Tantalum capacitors are the same: The ESR can vary wildly! I've seen values, for seemingly 'similar' capacitors, from 75mR to 10R.

2) The routing of local decoupling is often ignored - including by myself for the much lower frequency boards. The correct way is to track the supply rail to the capacitor first - then, from the capacitor to the device e.g. the device supply pin shouldn't be connected directly to the power plane, but to the power plane 'via' a track from the capacitor.

3) The value of the series resistors can also have an influence on reflections and thus operation. It's not necessarily the frequency - but the edges!

So, It may be worth playing with the above!

Dave H.
User avatar
IanS
Posts: 1963
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: A540 (or R260) 4MB card

Post by IanS »

IanJeffray wrote:
Wed Oct 13, 2021 9:27 am
Simtec uses some weird kind of ultra-flat series resistors, Ian uses packs.
The simtec resistors are printed (silk-screen) I believe. The are also on some of their IDE podules, but often hard to see as they are hidden under the endge of sockets.
User avatar
daveejhitchins
Posts: 6791
Joined: Wed Jun 13, 2012 6:23 pm
Location: Newton Aycliffe, County Durham
Contact:

Re: A540 (or R260) 4MB card

Post by daveejhitchins »

IanS wrote:
Wed Oct 13, 2021 10:22 pm
IanJeffray wrote:
Wed Oct 13, 2021 9:27 am
Simtec uses some weird kind of ultra-flat series resistors, Ian uses packs.
The simtec resistors are printed (silk-screen) I believe. The are also on some of their IDE podules, but often hard to see as they are hidden under the endge of sockets.
They could be Ultra Low Impedance types? Normal SM resistors are LASER trimmed, rendering them slightly inductive. Not sure if this applies to the multi-packaged type, though!

Dave H.
User avatar
IanJeffray
Posts: 1806
Joined: Sat Jun 06, 2020 3:50 pm
Location: Scotland
Contact:

Re: A540 (or R260) 4MB card

Post by IanJeffray »

daveejhitchins wrote:
Thu Oct 14, 2021 7:16 am
They could be Ultra Low Impedance types? Normal SM resistors are LASER trimmed, rendering them slightly inductive. Not sure if this applies to the multi-packaged type, though!
Well, given that Acorn's cards just use standard SMT resistors, I don't think suchlike would be the critical factor. All the resistors on the Simtec card appear physically similar, so I'd guess they probably just had some fruity way of applying resistors, rather than doing so here for any special purpose.

IanS - can I try changing those tants for something more exotic on one card?
User avatar
IanS
Posts: 1963
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: A540 (or R260) 4MB card

Post by IanS »

IanJeffray wrote:
Thu Oct 14, 2021 10:26 am
IanS - can I try changing those tants for something more exotic on one card?
Sure, go for it.
User avatar
IanS
Posts: 1963
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: A540 (or R260) 4MB card

Post by IanS »

daveejhitchins wrote:
Thu Oct 14, 2021 7:16 am
IanS wrote:
Wed Oct 13, 2021 10:22 pm
IanJeffray wrote:
Wed Oct 13, 2021 9:27 am
Simtec uses some weird kind of ultra-flat series resistors, Ian uses packs.
The simtec resistors are printed (silk-screen) I believe. The are also on some of their IDE podules, but often hard to see as they are hidden under the endge of sockets.
They could be Ultra Low Impedance types? Normal SM resistors are LASER trimmed, rendering them slightly inductive. Not sure if this applies to the multi-packaged type, though!
I think they were just a convienient way of fitted multiple resistors in one go. See the picture of the IDE podule. Note the two black printed resistors near the LED connector. I'm fairly sure those don't need to be anything special in the way of low-impedance. (There are also multiple printed resistors around the main chip, but they are hidden by the socket.)
Pic https://web.archive.org/web/20071025020 ... k-full.jpg
simtec-ide.PNG
simtec-ide.PNG (323.27 KiB) Viewed 72 times
Same thing on the A540 board. https://web.archive.org/web/20071023005 ... e-full.jpg
User avatar
daveejhitchins
Posts: 6791
Joined: Wed Jun 13, 2012 6:23 pm
Location: Newton Aycliffe, County Durham
Contact:

Re: A540 (or R260) 4MB card

Post by daveejhitchins »

Ah! That type are printed during the manufacture of the PCB - usually where tolerance isn’t an issue.

Dave H.
Post Reply

Return to “32-bit acorn hardware”