Faulty Atom

discussion of games, software, hardware & emulators relating to the Acorn Atom and Acorn System machines.
PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Faulty Atom

Post by PhilYoung » Wed Jan 04, 2012 5:49 pm

Edit - seems to be fixed (fingers crossed) see post 17


Hi,

Apologies, this ended up quite long.

My Atom has developed an intermittent fault (although saying it's intermittently working would be more accurate), and I was looking for some advice.

Until recently it was working fine with a homemade 8k video memory card, a homemade 32k lower memory card and one of Sir Morris' MMC boards. All was well until I decided to try and build a noise-killer board. I'm intending to get one of Primes RAM/ROM boards if there is another run, but I'd started so I wanted to finish. It was actually all working last night (once I'd realised that the pcb layout was putting two diodes the wrong way round compared to the schematic).

At the moment I've reverted to the original Atom - so all add-ons have been removed, and I'm back to 2k of RAM. The only non-original items at the moment are some turned pin sockets replacing a few slack old ones, and a replacement 3.58MHz crystal which is a lot smaller than the original one and leaves more room for the noise-killer board.

The symptoms are that most of the time, on power up or BREAK, the ACORN ATOM, '>' prompt and cursor come up, but there's no response to the keyboard (other than the BREAK key). Occasionally this will be followed by the display gradually starting to shimmer and eventually breaking up into noise. 'BREAK' normally restores the screen to normal (if it's disrupted) but still remaining unresponsive, and occasionally some random graphics mode appears.

About 25% of the time the Atom powers up normally (so the keyboard works) and runs whatever short program you can now type in. However it return to the unresponsive keyboard after BREAK (or a few BREAKs).

I'm thinking that the video area (6847, video RAM and 3.58MHz clock) must be working since I get a display (leaving aside the screen breaking up occasionally).

I get the start up prompt and cursor so the 6502 and it's clock must be working, as well as the video buffers. If the keyboard works, I can enter a program and list it so presumably the lower memory (lowest 1K only at the moment) must then be working as well.

I suppose it could actually be a fault around the keyboard (the 8255 and the 7445, IC25/26) - I don't have spares of these (yet). I don't have a scope either so won't be able to pin it down that way.

I have been removing and inserting ICs and a small PCB on headers meaning that the board has been handled quite a lot, so I'm thinking it's likely to be a recently provoked dry joint, a newly broken track or possibly a short between tracks. Nothing visible though.

Any thoughts ?

Cheers,

Phil Young
Last edited by PhilYoung on Sat Jan 14, 2012 4:53 pm, edited 1 time in total.

User avatar
danielj
Posts: 7264
Joined: Thu Oct 02, 2008 4:51 pm
Location: Manchester
Contact:

Re: Faulty Atom

Post by danielj » Wed Jan 04, 2012 6:38 pm

The 8255 is responsible for the keyboard and the VDU controller (both on port A) so if it's being flakey for whatever reason that would reasonably cause the symptoms you're seeing. Clean pins, reseat, swap socket etc.... (which I assume you've already done). I think beebmaster has some 8255s in his beebshop?

edit: IC9 is also involved in the clock circuit for the the display, but seeing as it's display and keyboard on the fritz the PIA sounds (to my somewhat naive mind) more likely.

d.

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Wed Jan 04, 2012 7:05 pm

Thanks for the reply.

IC9 has had a new turned pin socket because you need the 3.58MHz out of it for the noise killer, and the original socket was becoming loose due to the header stretching it. In fact, thinking back I replaced that socket first since I noticed that roughly the same symptoms earlier could be fixed by pressing IC9 down in one corner. Maybe it was more than a loose socket, the board looked okay though..

I've not started looking at the 8255 yet, I'll try looking around there next. I've noticed an additional symptom - often when the keyboard is working okay, the sound isn't (so no beep on error 94 or whatever). Sound is generated by the 8255 as well, of course, which points more firmly in that direction. It's all intermittent thought (of course).

Cheers,

Phil Young

User avatar
danielj
Posts: 7264
Joined: Thu Oct 02, 2008 4:51 pm
Location: Manchester
Contact:

Re: Faulty Atom

Post by danielj » Wed Jan 04, 2012 7:30 pm

PhilYoung wrote:Thanks for the reply.

IC9 has had a new turned pin socket because you need the 3.58MHz out of it for the noise killer, and the original socket was becoming loose due to the header stretching it. In fact, thinking back I replaced that socket first since I noticed that roughly the same symptoms earlier could be fixed by pressing IC9 down in one corner. Maybe it was more than a loose socket, the board looked okay though..

I've not started looking at the 8255 yet, I'll try looking around there next. I've noticed an additional symptom - often when the keyboard is working okay, the sound isn't (so no beep on error 94 or whatever). Sound is generated by the 8255 as well, of course, which points more firmly in that direction. It's all intermittent thought (of course).

Cheers,

Phil Young
Do you have a multimeter? If so it's worth pulling the ICs out and checking continuity between all the socket pins in that area if you've not done so already. Should at least show up any tracks/vias that may have been damaged swapping the sockets. Again, you may have done this already - I'm just running through what I'd do...

d.

d.

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Wed Jan 04, 2012 8:13 pm

Yes, I have a multimeter, I'll have to go through it all systematically. I don't think damaged anything de-soldering the old sockets, they looked clean when they came out. Plenty of scope for screwing it up removing the 40-pin PIA socket of course, so I think I'll try a replacement first.

Of course, since it's intermittent, anything you do looks like it's worked for a few seconds. Life is cruel.

I'll order a 8255 (or two) as you mentioned to be on the safe side as well.

Cheers,

Phil Young

User avatar
1024MAK
Posts: 8802
Joined: Mon Apr 18, 2011 4:46 pm
Location: Looking forward to summer in Somerset, UK...
Contact:

Re: Faulty Atom

Post by 1024MAK » Thu Jan 05, 2012 1:53 am

PhilYoung wrote:I have been removing and inserting ICs and a small PCB on headers meaning that the board has been handled quite a lot, so I'm thinking it's likely to be a recently provoked dry joint, a newly broken track or possibly a short between tracks. Nothing visible though.
I don't think this is a short. First, have you checked the 5V supply to check that it is okay?

This sounds like it could be an intermittent connection to a "common" connection, like a supply or ground pin / or / feed to a chip.
Could be a dry joint or a broken track <just> touching where it is next to a soldered joint. Check between the +5V and 0V/ground pins of each and every chip. Then as suggested, test the circuits with a meter on resistance.

But before you do, one question, well two actually :mrgreen:

If when the symptoms are present, does it make any difference how hard you press break and for how long?
Also does gently tapping or flexing the board make any difference?

Mark K.

P.S. this type of fault can sometimes be a bit of a pig to find and fix - but don't give up :wink:

User avatar
oss003
Posts: 2977
Joined: Tue Jul 14, 2009 11:57 am
Location: Netherlands
Contact:

Re: Faulty Atom

Post by oss003 » Thu Jan 05, 2012 7:37 am

PhilYoung wrote:I've noticed an additional symptom - often when the keyboard is working okay, the sound isn't (so no beep on error 94 or whatever). Sound is generated by the 8255 as well, of course, which points more firmly in that direction. It's all intermittent thought (of course).
Hi Phil,

while renovating my old Atom I encountered the same problem: No sound and a keyboard that sometimes worked and sometimes not after a break. Replacing the 8255 did the trick for me but to be sure, check if all signals are comming through at the 8255 socket.

Greetings
Kees

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Thu Jan 05, 2012 11:02 am

1024MAK wrote:
PhilYoung wrote:I have been removing and inserting ICs and a small PCB on headers meaning that the board has been handled quite a lot, so I'm thinking it's likely to be a recently provoked dry joint, a newly broken track or possibly a short between tracks. Nothing visible though.
I don't think this is a short. First, have you checked the 5V supply to check that it is okay?
It looks okay. Using two multimeters I get 4.93/4.89V at the Atom input, and 4.90/4.86V across the 8255 (pins 26 and 7).

1024MAK wrote: This sounds like it could be an intermittent connection to a "common" connection, like a supply or ground pin / or / feed to a chip.
Could be a dry joint or a broken track <just> touching where it is next to a soldered joint. Check between the +5V and 0V/ground pins of each and every chip. Then as suggested, test the circuits with a meter on resistance.
Yes, I'll check all those, as well as further visually checking the board through a magnifying glass.

1024MAK wrote: But before you do, one question, well two actually :mrgreen:

If when the symptoms are present, does it make any difference how hard you press break and for how long?
Not that I've noticed. I've tried both 'quick stab', and 'deliberate press' without either seeming to be better than the other. Another just remembered data point, my AtoMMC board only re-initialises correctly on BREAK about half the time, but always comes up OK on power up. Or it does once I added a 10uF capacitor across the 0.1uF one originally fitted at C11 to slow up the Reset. The reset goes via two inverters on IC9, which is the first loose socket that I replaced, so I'll double check that area (again) in case of a broken track or whatever.
1024MAK wrote: Also does gently tapping or flexing the board make any difference?
Not that I've noticed. Once it's in the 'keyboard not working' state it only (sometimes) recovers on BREAK or power off and back on again.
1024MAK wrote: Mark K.


P.S. this type of fault can sometimes be a bit of a pig to find and fix - but don't give up :wink:
Thanks for the suggestions. I'd like to get it back to working reliably, I've had it from new.
I get the feeling it might take a while as well.....

Cheers,

Phil Young

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Thu Jan 05, 2012 11:10 am

oss003 wrote:
PhilYoung wrote:I've noticed an additional symptom - often when the keyboard is working okay, the sound isn't (so no beep on error 94 or whatever). Sound is generated by the 8255 as well, of course, which points more firmly in that direction. It's all intermittent thought (of course).
Hi Phil,

while renovating my old Atom I encountered the same problem: No sound and a keyboard that sometimes worked and sometimes not after a break. Replacing the 8255 did the trick for me but to be sure, check if all signals are comming through at the 8255 socket.

Greetings
Kees
Thanks for the suggestion, I've ordered a couple of 8255s (and 7445s and turned pin sockets) so when they arrive that will be a quick job to check.

I don't have a scope so I can only really look at DC voltages and continuity.

I think what I'll do is:

1. Check 5v/0v on all the implicated ICs, plus continuity and visual inspection in case of a broken track.

2. Re-solder in case of a dry joint

3. Swap in an new 8255, see if that cures it. (and 7445 as well)

4. Replace the 8255 socket with a new one (but the original seems firm enough).

By the way, the trouble started when I tried to fit a no-snow board that you posted the design for somewhere (BBMMC google groups I think), so really it's all your fault.... :D

Cheers,

Phil Young

User avatar
oss003
Posts: 2977
Joined: Tue Jul 14, 2009 11:57 am
Location: Netherlands
Contact:

Re: Faulty Atom

Post by oss003 » Thu Jan 05, 2012 11:58 am

PhilYoung wrote:By the way, the trouble started when I tried to fit a no-snow board that you posted the design for somewhere (BBMMC google groups I think), so really it's all your fault.... :D
Ok, maybe I forgot to mention that your keyboard stops working and your screen start flickering #-o
but the the main thing is .... without snow!! \:D/

You don't need a scope to check if a track is broken, just check if the socket pin of the 8255 is connected with the right signal on e.g. the processor with your universal meter.

Greetings
Kees

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Thu Jan 05, 2012 12:39 pm

oss003 wrote:
PhilYoung wrote:By the way, the trouble started when I tried to fit a no-snow board that you posted the design for somewhere (BBMMC google groups I think), so really it's all your fault.... :D
Ok, maybe I forgot to mention that your keyboard stops working and your screen start flickering #-o
but the the main thing is .... without snow!! \:D/
Well, once I realised that the two diodes were the wrong way round (due some inconistency between the diode symbol footprint in the schematic editor, gschem, and the pcb layout editor, pcb) it worked fine. Until the next day.......
oss003 wrote: You don't need a scope to check if a track is broken, just check if the socket pin of the 8255 is connected with the right signal on e.g. the processor with your universal meter.

Greetings
Kees
Fair enough, I thought you mean looking at the actual signals on various pins. Continuity and resistance I can do.

I might just wait until the replacement 8255 arrives, I'm getting the board layout burnt into my retinas.

Cheers,

Phil Young

User avatar
oss003
Posts: 2977
Joined: Tue Jul 14, 2009 11:57 am
Location: Netherlands
Contact:

Re: Faulty Atom

Post by oss003 » Thu Jan 05, 2012 12:42 pm

PhilYoung wrote:Another just remembered data point, my AtoMMC board only re-initialises correctly on BREAK about half the time, but always comes up OK on power up. Or it does once I added a 10uF capacitor across the 0.1uF one originally fitted at C11 to slow up the Reset. The reset goes via two inverters on IC9, which is the first loose socket that I replaced, so I'll double check that area (again) in case of a broken track or whatever.
Not booting the AtoMMC interface at break is because of key-bouncing pulses on the reset line. At power up, you don't have key-bouncing pulses so that's why booting the AtoMMC at power up always works, even with REPT, SHIFT of CTRL pressed during power up. I did some testing with replacing the second inverter of IC9 with a Smitt-Trigger port to get rid of the key-bouncing. Charlie programmed a delay in the AtoMMC kernal for a clean boot and this worked ok with the AtoMMC at #A000.

When you have a RAM/ROM board it is better to have the AtoMMC ROM at #E000 instead of #A000. Booting from #E000 is software controlled and always stable while booting from #A000 is done by interrupt and not stable due to key-bouncing pulses.

Greetings
Kees

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Thu Jan 05, 2012 2:14 pm

oss003 wrote:
PhilYoung wrote:Another just remembered data point, my AtoMMC board only re-initialises correctly on BREAK about half the time, but always comes up OK on power up. Or it does once I added a 10uF capacitor across the 0.1uF one originally fitted at C11 to slow up the Reset. The reset goes via two inverters on IC9, which is the first loose socket that I replaced, so I'll double check that area (again) in case of a broken track or whatever.
Not booting the AtoMMC interface at break is because of key-bouncing pulses on the reset line. At power up, you don't have key-bouncing pulses so that's why booting the AtoMMC at power up always works, even with REPT, SHIFT of CTRL pressed during power up. I did some testing with replacing the second inverter of IC9 with a Smitt-Trigger port to get rid of the key-bouncing. Charlie programmed a delay in the AtoMMC kernal for a clean boot and this worked ok with the AtoMMC at #A000.

When you have a RAM/ROM board it is better to have the AtoMMC ROM at #E000 instead of #A000. Booting from #E000 is software controlled and always stable while booting from #A000 is done by interrupt and not stable due to key-bouncing pulses.

Greetings
Kees
That's interesting about the AtoMMC behaviour. I'm not up to the very latest firmware yet.

I was intending to try the #E000 version (and patched #F000 to match) once I had built a ROM board, which was going to be the next project but I'm being distracted by the Keyboard problem. It seems more logical to put it in the DOS space, and it leaves A000 free for utility ROMs, assuming you have the hardware so I quite understand why the A000 version is the standard. I've bought a couple of 27SF256 EEPROMs, and I'm using an Arduino as a programmer, though I've not tried programming yet. I can erase them successfully, and read the contents successfully (not that there are any now after the erasing worked) so I don't think programming will be a problem.

I was intending to have one EEPROM as the #C000, #D000, #E000 and #F000 ROMS twice over, one with normal #F000 and empty #E000, the other with the MMC firmware at #E000 and the patched #F000. Link selectable.

The other would be as an #A000 ROMbox, selectable with one of those little binary coded rotary switches, one of those ROMs being the MMC #A000 version.

I think there used to be a patched Atom OS that would search through an #A000 Rom box by paging in the various ROMs with a latch at #BFFF searching for commands, might you have any information about that ?

Thanks for the information,

Cheers,

Phil Young

User avatar
oss003
Posts: 2977
Joined: Tue Jul 14, 2009 11:57 am
Location: Netherlands
Contact:

Re: Faulty Atom

Post by oss003 » Thu Jan 05, 2012 3:21 pm

PhilYoung wrote:I was intending to have one EEPROM as the #C000, #D000, #E000 and #F000 ROMS twice over, one with normal #F000 and empty #E000, the other with the MMC firmware at #E000 and the patched #F000. Link selectable.
Sounds good to me.
PhilYoung wrote:The other would be as an #A000 ROMbox, selectable with one of those little binary coded rotary switches, one of those ROMs being the MMC #A000 version.
You have to do a few things to automatically switch ROM's:

1. Decode #BFFF as rotary switch (read/write)
2. Patch the Floating Point ROM (#Dxxx) to jump to the ROM switch software if an ERROR 94 occurs
3. Install ROM switch software.

In my Atom I've installed the ROM switch software in the AtoMMC ROM free space at #Exxx and patched the Floating Point ROM to jump to this code. On Youtube you can see the software in action in a previous version where I loaded the ROM switch software at #1000 with SHIFT-BREAK and an AtoMMC interface. All commands are from different ROM's.

http://www.youtube.com/watch?v=566r5NvxOUw

So all you need is Phill's Ram/Rom board. It has a decoded rotary switch at #BFFF, space for 16 #Axxx ROM's and you can flash the software in #Dxxx, #Exxx and the #Fxxx ROM and as extra .... the no snow option.

Hmmm ... sounds like some advertising campaign !! :lol:

Greetings
Kees

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Thu Jan 05, 2012 5:18 pm

oss003 wrote:
So all you need is Phill's Ram/Rom board. It has a decoded rotary switch at #BFFF, space for 16 #Axxx ROM's and you can flash the software in #Dxxx, #Exxx and the #Fxxx ROM and as extra .... the no snow option.

Hmmm ... sounds like some advertising campaign !! :lol:

Greetings
Kees
Yes, I'm intending to get one when the next batch are made up, plus the Colour board. Of course I'll need a working Atom to install it into. I'd like to get my own ROM board working as well though, out of some perverse sense of pride

Thanks for the info on the OS ROM switching. I didn't realise that was how it was done, and can just about see how it works in principle. I'll stick to a manual rotary switch for mine thought (maybe directly poke a #BFFF latch if I'm feeling energetic).

I'll wait for the 8255 to arrive before doing any more work, that will hopefully be an instant solution to the keyboard problem,

Cheers,

Phil Young

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Sat Jan 07, 2012 1:53 pm

PhilYoung wrote:
I'll wait for the 8255 to arrive before doing any more work, that will hopefully be an instant solution to the keyboard problem,

Cheers,

Phil Young
The 8255s and 7445s arrived this morning courtesy of Beebmaster. Sadly they did not improve matters, so it's back to the magnifying glass, strong light, multimeter and kicking the cat. I'll get it in the end (the fault, not the cat).

Cheers,

Phil Young

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom - Seems to be Fixed

Post by PhilYoung » Sat Jan 14, 2012 4:52 pm

After a lot of messing around, I seem to have fixed my Atom. But the cure seems a bit random.

It turns out that if I use one (and just one) of the two original 74LS04s as IC9 (which is used for the 3.58MHz oscillator, inverts the Break key and inverts it back again, and buffers phi2) everything works. If I use a new 74LS04, or a 74HCT04, or a 74HC04 or even the other original 74LS04 it stops working as before - normally keyboard response, no 'beep' when the keyboard does occasionally work and sometimes unstable display.

This is completely reproducible, works (or not) every time.

The 74LS04 that works is a National Semiconductor DM74LS04N. The other original one that doesn't is a TI SN74LS04N. I have a new 3.58MHz crystal fitted in place of the original if that makes any difference. IC9 socket has been replaced with a new turned pin one.

I don't have a scope so can't check if it's the 3.58MHz clock not working or the phi2 buffering that is altered, but why on earth is it so sensitive to which particular 7404 is used ?

Cheers,

Phil Young

User avatar
danielj
Posts: 7264
Joined: Thu Oct 02, 2008 4:51 pm
Location: Manchester
Contact:

Re: Faulty Atom - Seems to be Fixed

Post by danielj » Sat Jan 14, 2012 5:59 pm

PhilYoung wrote:After a lot of messing around, I seem to have fixed my Atom. But the cure seems a bit random.

It turns out that if I use one (and just one) of the two original 74LS04s as IC9 (which is used for the 3.58MHz oscillator, inverts the Break key and inverts it back again, and buffers phi2) everything works. If I use a new 74LS04, or a 74HCT04, or a 74HC04 or even the other original 74LS04 it stops working as before - normally keyboard response, no 'beep' when the keyboard does occasionally work and sometimes unstable display.

This is completely reproducible, works (or not) every time.

The 74LS04 that works is a National Semiconductor DM74LS04N. The other original one that doesn't is a TI SN74LS04N. I have a new 3.58MHz crystal fitted in place of the original if that makes any difference. IC9 socket has been replaced with a new turned pin one.

I don't have a scope so can't check if it's the 3.58MHz clock not working or the phi2 buffering that is altered, but why on earth is it so sensitive to which particular 7404 is used ?

Cheers,

Phil Young
That is a touch odd. I don't have time right now, but I'll check what I have in mine (it actually needs unscrewing atm), then have a play with a few different makes and scope-it and see if we can see any differences.

d.

User avatar
george.h
Posts: 1030
Joined: Wed Apr 13, 2011 5:32 pm
Location: Chelmsford Essex
Contact:

Re: Faulty Atom

Post by george.h » Sat Jan 14, 2012 7:00 pm

Used to see that "funny" on ye olde Microtan/Tangerine 6502 SBCs. The clock circuit was VERY fussy about who's 74LS04 it would work with reliably (back in the days when "canned clocks" cost a holiday home and boat).

Best guess at the time was that something about the clock design made it very sensitive to the difference in operating parameters from one chip manufacturer to another. Possibly choice of loading capacitor values or bias resistor or even PCB layout.

I think (could be very wrong about this) that most TTL XTAL clock circuits using inverters ('04 '00 with inputes tied etc) will usually only work with the family they were design for. So a one designed for LS-TTL won't usuaully work if you substitute ALS/HC/HCT etc due to differences in input and output characteristics when used as analogue amps.

So worth checking the passives around that chip and the XTAL itself.
Pic Caption: "One day son, this will all be yours..."

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom - Seems to be Fixed

Post by PhilYoung » Sun Jan 15, 2012 8:44 am

danielj wrote:
That is a touch odd. I don't have time right now, but I'll check what I have in mine (it actually needs unscrewing atm), then have a play with a few different makes and scope-it and see if we can see any differences.

d.
Well, I said it was working, I didn't say it made any sense......

If you have time, let me know what you conclude with your Atom.

I'll be able to borrow a scope fairly soon so I should be able to see what the differences are between the various 7404s. By the way, I should have said that a new 74LS04 works as IC45 (the 4 MHz oscillator etc.) So something is different there, even though it's the same circuit. Maybe I'll try putting the old 3.58MHz Xtal back. Also, one lead of C1 (the 10nF that's part of the 3.58MHz oscillator) runs under the end of the body of PL7 (which is a 64-way DIN socket on my Atom, so SK7 to be pedantic), maybe it's starting to short some tracks after all this time ? Seems unlikely.

Cheers,

Phil Young

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Sun Jan 15, 2012 9:00 am

george.h wrote:Used to see that "funny" on ye olde Microtan/Tangerine 6502 SBCs. The clock circuit was VERY fussy about who's 74LS04 it would work with reliably (back in the days when "canned clocks" cost a holiday home and boat).
I never saw a Microtan. I think they did a series on them in ETI, maybe around the time the Atom and ZX80 started to appear. I see that you can get 4MHz oscillator modules easily, but I've not found 3.58MHz ones yet.
george.h wrote:Best guess at the time was that something about the clock design made it very sensitive to the difference in operating parameters from one chip manufacturer to another. Possibly choice of loading capacitor values or bias resistor or even PCB layout.

I think (could be very wrong about this) that most TTL XTAL clock circuits using inverters ('04 '00 with inputes tied etc) will usually only work with the family they were design for. So a one designed for LS-TTL won't usuaully work if you substitute ALS/HC/HCT etc due to differences in input and output characteristics when used as analogue amps.

So worth checking the passives around that chip and the XTAL itself.
I didn't expect the HC or HCT versions to work, but it's a bit surprising that a new 74LS04 works as IC45 (4 MHz ocs) but not as IC9, and that only one of the originals works as IC9. But once I've borrowed a scope I'll be able to have a more detailed look, maybe put it onto a breadboard and start experimenting with the passives as you say.

It must have worked with a new chip at one point though, my no-snow board (with new 74LS04s as IC9 and IC45) worked for a while reliably and then suddenly stopped working overnight.

Maybe my Atom is haunted......

It's still working fine at the moment.

Cheers,

Phil Young

User avatar
oss003
Posts: 2977
Joined: Tue Jul 14, 2009 11:57 am
Location: Netherlands
Contact:

Re: Faulty Atom

Post by oss003 » Mon Jan 16, 2012 12:11 pm

Hi Phil,

the only thing I can think of is that the BREAK-key inverters of IC9 are causing problems because the reset line of the 8255 is after the first BREAK-key inverter. The 8255 is responsable for the beeper, the keyboard and video-modes so I think it's worth checking/measuring the signals around the BREAK-key/8255 reset line.

Greetings
Kees

Prime
Posts: 2719
Joined: Sun May 31, 2009 11:52 pm
Contact:

Re: Faulty Atom

Post by Prime » Mon Jan 16, 2012 12:44 pm

It's worth noting that unlike most of the rest of the system, the 8255 is active
*HIGH* reset, so should normally be low and go high when break is pressed.

Cheers.

Phill.

User avatar
MartinB
Posts: 5206
Joined: Mon Mar 31, 2008 9:04 pm
Location: Obscurity
Contact:

Re: Faulty Atom

Post by MartinB » Mon Jan 16, 2012 12:57 pm

Phil wrote:Maybe my Atom is haunted......
When presented with symptoms such as variable functionality (or even a complete lack-of) between seemingly identical chips, I would expect there to be an operating parameter that is very marginal in terms of the chip’s specified envelope. Power supply is a common one for LS and ALS chips so I would look more closely at the DC supply, not in terms of a DVM showing roughly 5v but more closely still, ideally with a scope, in terms of are we seeing 4.7v, 4.8V or 5v at the chip and is there any DC ripple or noise? Noise immunity of the different types varies and the transfer functions tend to be quite steep over a small operating voltage range. Also. It’s often worth adding a pull-up (sometimes pull-down) to affected outputs and see if that makes any difference, some circuit applications can be very sensitive to load and capacitive effects can often bring a chip to the edge of its specified operating capability.

All very vague I know but like all paranormal activity in circuitry, there’s usually a scientific explanation :wink:

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Mon Jan 16, 2012 5:47 pm

Thanks for the replies, lots of useful information.

oss003 wrote:Hi Phil,

the only thing I can think of is that the BREAK-key inverters of IC9 are causing problems because the reset line of the 8255 is after the first BREAK-key inverter. The 8255 is responsable for the beeper, the keyboard and video-modes so I think it's worth checking/measuring the signals around the BREAK-key/8255 reset line.

Greetings
Kees
I think any more investigation will have to wait until I can borrow a scope and have a detailed look around that area.

However, I've realised that the AtoMMC which used to initialise correctly about half the time is now always starting correctly. You mentioned earlier that it may not start correctly due to bounce on the BREAK key, which is inverted by IC9 (for the 8255) and then inverted again for the 6502. So IC9 is implicated here as well, with the 'good' IC9 giving me correct video, correct keyboard, correct beeper and now correct AtoMMC.
Prime wrote: It's worth noting that unlike most of the rest of the system, the 8255 is active
*HIGH* reset, so should normally be low and go high when break is pressed.

Cheers.

Phill.
yes, I see that, it get's inverted for the 8255 and then inverted again for the not-RST to the 6502.

Can I ask if your latest version RAM/ROM/no-snow board retains the old IC9/IC45 or are these incorporated into a CPLD or anything like that?
MartinB wrote: Phil wrote:
Maybe my Atom is haunted......
When presented with symptoms such as variable functionality (or even a complete lack-of) between seemingly identical chips, I would expect there to be an operating parameter that is very marginal in terms of the chip’s specified envelope. Power supply is a common one for LS and ALS chips so I would look more closely at the DC supply, not in terms of a DVM showing roughly 5v but more closely still, ideally with a scope, in terms of are we seeing 4.7v, 4.8V or 5v at the chip and is there any DC ripple or noise? Noise immunity of the different types varies and the transfer functions tend to be quite steep over a small operating voltage range. Also. It’s often worth adding a pull-up (sometimes pull-down) to affected outputs and see if that makes any difference, some circuit applications can be very sensitive to load and capacitive effects can often bring a chip to the edge of its specified operating capability.

All very vague I know but like all paranormal activity in circuitry, there’s usually a scientific explanation :wink:
The 5v is actually more like 4.9V, and 4.85V by the time it reaches the 8255. I'll try a variable PSU and set it at exactly 5V on the 8255 Vcc (maybe a tiny bit more) and see if that makes a difference. I can't detect any ripple on the 5V lines with the multimeter (about 2mVac), another thing to check with the scope. At the moment I'm using a linear PSU I built around the time so it should be okay, it's just 50Hz rectified, smoothed and fed through a 7805 (1.5A type), and then a slight drop through the cable to the DC socket. So there no SMPS switching breaking through. It draws a lot less than 1A with the CMOS SRAM back.

Again, thanks for all the suggestions,

Cheers,

Phil Young

User avatar
1024MAK
Posts: 8802
Joined: Mon Apr 18, 2011 4:46 pm
Location: Looking forward to summer in Somerset, UK...
Contact:

Re: Faulty Atom

Post by 1024MAK » Tue Jan 17, 2012 5:08 pm

Most logic chips need a stable +5V dc supply with the voltage +/- 5%.
So at least 4.75V up to 5.25V.
However, it is much better to have the main supply voltage much closer to 5V or a little over (say 5.1V) as this allows for any voltage drops (e.g. like along any fine PCB tracks). As digital chips switch their outputs, for a short time some actually momentarily short out the supply. This can cause no end of problems. This is why decoupling capacitors are used (typically one 100nF ceramic for every one or two IC's).
In your case, what is the output voltage on the pins of the 7805 regulator?
Oh, and most normal 7805 regulators are rated at 1A. Only some are rated at 1.5A
If you are losing voltage due to volt drop, use a higher rated (thicker) cable for both the +5V and 0V/ground wires.

Mark K.

Prime
Posts: 2719
Joined: Sun May 31, 2009 11:52 pm
Contact:

Re: Faulty Atom

Post by Prime » Tue Jan 17, 2012 5:39 pm

1024MAK wrote: As digital chips switch their outputs, for a short time some actually momentarily short out the supply. This can cause no end of problems. This is why decoupling capacitors are used (typically one 100nF ceramic for every one or two IC's).[/i][/b]
I was always taught one for every pair of power pins, e.g. the CPLD on the RAM-ROM board has 3 sets of power pins and so has 3 100nf caps. Though it doesn't have a cap on the CPU because that has caps on the mainboard wher the ram-rom plus in.

Cheers.

Phill.

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Tue Jan 17, 2012 9:03 pm

1024MAK wrote:Most logic chips need a stable +5V dc supply with the voltage +/- 5%.
So at least 4.75V up to 5.25V.
However, it is much better to have the main supply voltage much closer to 5V or a little over (say 5.1V) as this allows for any voltage drops (e.g. like along any fine PCB tracks). As digital chips switch their outputs, for a short time some actually momentarily short out the supply. This can cause no end of problems. This is why decoupling capacitors are used (typically one 100nF ceramic for every one or two IC's).
I've just tried a bench PSU, adjusted to 5.1V across IC9 (the problematic 74LS04). Same result, the only one that works is the NI original one. Completely reproducible.

The original Atom de-coupling capacitors are still in place, lots of 47nF, but good point I'll dab a few 100nF across the no-snow board for luck, especially IC9.
1024MAK wrote: In your case, what is the output voltage on the pins of the 7805 regulator?
Oh, and most normal 7805 regulators are rated at 1A. Only some are rated at 1.5A
If you are losing voltage due to volt drop, use a higher rated (thicker) cable for both the +5V and 0V/ground wires.

Mark K.
About 4.9V across IC9 (now mounted on the no-snow board). About 4.85V across the 8255 (both these when using the 7805 type PSU). Draws about 0.85A, so well under 1A. The regulator is an ST electronics L7805CV "Output Current up to 1.5A" according to the data sheet.

I've been able to borrow a scope for the weekend, so I'll have a look at the the 3.58MHz clock, and around the 8255 and 7445 to see if the keyboard matrix is being scanned since the keyboard not working is the most obvious symptom. Also I'll be able to check if the 5V lines are clean.

Cheers,

Phil Young

PhilYoung
Posts: 201
Joined: Sun Jun 12, 2011 4:55 pm
Contact:

Re: Faulty Atom

Post by PhilYoung » Tue Jan 17, 2012 9:22 pm

Prime wrote:
1024MAK wrote: As digital chips switch their outputs, for a short time some actually momentarily short out the supply. This can cause no end of problems. This is why decoupling capacitors are used (typically one 100nF ceramic for every one or two IC's).[/i][/b]
I was always taught one for every pair of power pins, e.g. the CPLD on the RAM-ROM board has 3 sets of power pins and so has 3 100nf caps. Though it doesn't have a cap on the CPU because that has caps on the mainboard wher the ram-rom plus in.

Cheers.

Phill.
All the original Atom 47nF de-coupling caps are still in place (some on the other side of the board to make room for the 8K and 32K CMOS RAM boards). So when the new boards are removed we're back to a standard 2K Atom and I still need to find out why only one particular 74LS04 works as IC9. I've got a scope on loan this weekend so I can compare the working and no-working versions,and hopefully work out what is going on.

Cheers,

Phil Young

User avatar
oss003
Posts: 2977
Joined: Tue Jul 14, 2009 11:57 am
Location: Netherlands
Contact:

Re: Faulty Atom

Post by oss003 » Thu Feb 09, 2012 2:08 pm

Hi Phil,

were you able to solve the 74LS04 mystery?

Greetings
Kees

Post Reply