Another BBC B Needs Fixing.

[bagpuss22]

Hi All,

I have an Issue 7 with the standard no boot, single beep! The story so far...……...

PSU has been recapped and voltages are good.

Moved S25 SOUTH and hey presto! I have a 16K Beeb, so CAS1, socketed IC53 - IC60, is OK.
Memory test confirms no stuck bits high or low...

Used the 74LS04 trick so that I could boot to CAS0. I get a corrupted screen, Mode 0 allowed me to
run a memory test that returned 4. soldered IC63 was duly replaced!

Rebooted and the screen was still corrupt, weird, but it now passed the memory test, both at 0 and 255.

Current state of play is that it will still only boot with S25 at south. In S25 North mode I get a continuous
beep and the CAPS light comes on, followed by the SHIFT LOCK.

IC14 and IC15 were commented about recently so I have replaced them..

Any further pointers appreciated.

BP

[Kazzie]

Given that everything works with S25 south, I'd expect it's still a RAM issue. You could have had two faulty RAM chips: one blatantly so, and another with a more subtle fault (e.g. writing to address X also overwrites address Y) that isn't being picked up by a write/read memory test.

Could you post a screenshot of the corruption you get when you power up? The pattern of corrupted characters can indicate which bit is playing up.

[bagpuss22]

Sorry for the delay, this is the display WITH the hex inverter in place.

[bagpuss22]

One thing I just noticed is if I run the basic write 0 to ?&4000 memory test (Mode 7), if I hold my finger on the changed IC53
the screen corrupts further and the test returns a value as though it failed the memory test! BUT it returns what I
am writing to ?&4000, which is 0.

Seems to be an earthing thing as if I earth myself it stops lol

If I now change this to 255 it continually returns 255......

[1024MAK]

If you compare the characters displayed in the MODE 7 ‘boot up’ screen with the characters that should be displayed, you can work out which bit or bits are incorrect. In the back of the manual there is a chart of the MODE 7 / teletext character set.

Which memory test program are you using? How long did you leave it running? And was this with the lid on, or with the lit not fitted? Only some DRAM chip failures are heat sensitive. So if the memory test works, leave it running for at least one hour.

Which regards to holding your finger on the ‘changed IC53’, is this chip in a socket? If yes and you have other DRAM chips in sockets, try swapping two around to see if the fault moves (preferably to a different bit in a different bank).

If the fault does not move, it could be a poor connection due to a problem with the socket, or a dry solder joint, or a cracked PCB track.

Mark

[1024MAK]

Normal Acorn BBC B boot screen:

Code: Select all

BBC Computer 32K

Acorn DFS

BASIC

>

Your Acorn BBC B boot screen:

Code: Select all

BBC Koeputer 3:K

Acgrf DFS

BA←IC

>

So the characters that are different are:

K instead of C.

Code: Select all

K in binary is 01001011 and
C in binary is 01000011. Bit 3 is 1 instead of 0.

e instead of m.

Code: Select all

e in binary is 01100101 and
m in binary is 01101101. Bit 3 is 0 instead of 1.

: instead of 2.

Code: Select all

: in binary is 00111010 and
2 in binary is 00110010. Bit 3 is 1 instead of 0.

g instead of o.

Code: Select all

g in binary is 01100111 and
o in binary is 01101111. Bit 3 is 0 instead of 1.

f instead of n.

Code: Select all

f in binary is 01100110 and
n in binary is 01101110. Bit 3 is 0 instead of 1.

← instead of S.

Code: Select all

← in binary is 01011011 and
S in binary is 01010011. Bit 3 is 1 instead of 0.

Mark

[bagpuss22]

Mark,

I was just running the 3 line bit of code which reads and writes to ?$4000.

10 A=0 (or 255)
20 ?&4000=A: IF ?&4000<>A THEN PRINT ?&4000
30 GOTO 20

Left it running for quite a while and the lid was off.

CAS0 is soldered to the mainboard and when I originally ran the code above I was getting
a constant 4 which I equated to IC63, I removed this, inserted a new socket and put a new IC in that.

It was IC63! not 53, apologies..
IC63 was definitely defunct as if I placed it in CAS1 with S25 South, the machine would not boot.

I am currently having to run a new 64k x 1 as I have no 16k x 1 and this ran OK in CAS1.

BP

[DutchAcorn]

These short memory test programs are often non-conclusive.

A mode 7 screen is a more thorough test. In your case IC64 (bit 3 in CAS0) sometimes fails, but not always.

The best check is a mode 0 screen.

[1024MAK]

Mark,

I was just running the 3 line bit of code which reads and writes to ?$4000.

10 A=0 (or 255)
20 ?&4000=A: IF ?&4000<>A THEN PRINT ?&4000
30 GOTO 20

This type of short test program is designed only to pick up the simple most common faults with DRAM chips. That is where the suspected DRAM chip always fails to provide the correct data output regardless of which address (in the range that that chip covers) is accessed.

So for example, if the output driver in a DRAM chip is defective, this fault would be picked up.

However DRAM chips are complex devices (which is why the cost so much back then) and hence there are many ways they can fail.

So just like any job, if the first ‘tool’ does not help, we move on and use another ‘tool’.

In an earlier post, I showed how working out which characters are actually shown compared to what should be shown on a MODE 7 / teletext MODE screen can give clues about what is happening.

There are also more complex and comprehensive RAM test programs. This thread has better DRAM test program.

More information on DRAM faults in this post.

Mark

[bagpuss22]

Thanks Mark,

Will have a look a these links.

I have replaced U14 with a new one so hopefully now its a case of maybe of hunt the
RAM. Looking at you previous calculations on the display that would point to IC64
playing up.

I might consider socketing all of CAS0 as it would make life simpler.

Thanks

Russ

[bagpuss22]

OK still none the wiser.

I ran Martins test program and that sometimes passes and sometimes fails but does
not return an IC! It returns the address buy no IC number.....?

[1024MAK]

OK still none the wiser.

I ran Martins test program and that sometimes passes and sometimes fails but does
not return an IC! It returns the address buy no IC number.....?

Part of the test is the displayed image during testing. Is the pattern drawn consistent? If it shows anything odd, take a photo.

Mark

[MartinB]

I ran Martins test program and that sometimes passes and sometimes fails but does
not return an IC! It returns the address buy no IC number.....?

That’s confusing What exactly do you get as an output? The earlier versions just report the failing location but v3 converts the bit-level failure to a specific chip with its schematic IC number and I just can’t see how you can have the address error without the chip detail?

Where from/how did you download the program and get it onto your Beeb?

Questions questions....

[bagpuss22]

OK the last 2 days have been really weird, sometimes the test will pass sometimes it will fail!

Yesterday it ran for 2 hours without a failure but the system will not boot normally

Martin I am using the V3 WAV file that was on your post, pointed too by Mark, hopefully using the latest!


My RAM Area as is.


Boot screen with S25 North inverted.


Loading the MemTst program in MODE0.


Listing Memtst!! The screen behind was after a failed run yesterday.


Failure 1 today


Failure 2 today.


There have also been runs without failure....

BP.

[marcusjambler]

If the fault isnt constant its worth taking a look at the power connector solder lugs.
Sometimes they are loose in the PCB causing all sorts of weird stuff

[DutchAcorn]

If the idea is to fix the BBC, I'd recommend replacing IC64; the mode 7 screens very consistently show an issue with bit 3.

That said, it would be good to know why the memory test does not confirm this.

[1024MAK]

Are you keeping note of all the addresses that the memory test program shows as failures?
If yes, please post them here. If not, please start taking notes.

It certainly is a good idea to swap the DRAM chip for IC64.

It’s possible that you have multiple issues with this machine. However, the normal way to proceed is to fix the problems as they are found. Or try to. A lot of fault finding is a process of elimination. So even if you think a chip is good, it is often worthwhile if it is still a suspect, changing it or swapping it with another used elsewhere in the machine just to see if doing this affects the symptoms.

Is the MODE 7 ‘boot-up’ screen always the same at start-up? Or is it different every time?

Mark

[MartinB]

Last night, I thought Russ was suggesting that he didn't even get the 'Bit Lo/Hi..' dialogue (which was somewhat impossible) but I see now that he's quite rightly querying the fact that there's no actual IC number reported after the failure detect dialogue. The reason for this is that the memory fault must be very intermittent, even down to a write-by-write level. The main program iteration performs an initial byte read-back test using values of $55 and $AA and if an inequality is detected, it then performs a further write and read-back using $FF and $00 in order to establish the presence of individual stuck hi or stuck lo bits. What's clearly happening here then is that following the initial failure detection, the given 'location' is then passing the subsequent stuck bit tests resulting in no IC number being reported. This is why the program as a whole includes a ram visual pattern 'window' so that additional clues are available in the event of intermittent faults. On that basis, I'd agree with Paul and suggest you consider replacing IC64.

[bagpuss22]

To answer Marks question,

I have only started collating results today. I will now keep a log of the results.

The initial MODE 7 boot screen is the same for every power up.

When the test passes the screen does have a lovely pattern, on several runs
I did get some artefacts on the screen too put the test found no failures.

I will socket up IC64. I only have 64k x 1 at the moment until a new batch of 16k x 1s
arrive. They are the same pinout except for pin9 which is A7.

BP

[bagpuss22]

OK this evenings update.

Socketed IC62, by mistake, and replaced with new 64k x 1! Ooops, seemed to make things worse.

So did some testing with the new IC62.....
MODE 7 boot screen was better but had characters changing, and in MODE0 artefacts moved up the screen.
As for testing I had the following failures:
&43FF No IC, &4005 IC56, &4005 No IC, &4078 No IC and &401A No IC.

Moved to MODE7 and had these failures
&FE0 ICs 63 64, &EE7 ICs 67 68, &E01 No IC, &E04 ICs 54 56 57 60.

Next I reinstalled the original 16k x 1 back into IC62.

I then socketed IC64 and used the original16k x 1 IC.

Machine booted OK.....!

Been running memory tested in MODE 7 and MODE0 without issue..... Result.

BP

[KenLowe]

Sounds like you had a bad connection when the RAM was soldered directly into the board, and explains the very intermittent behaviour.

Glad you got it fixed. Well done.