32K SRAM board (MB1732) troubleshooting

[Witchy]

Hi folks,

After a successful exhibition the other week which involved lots of questions about my 'bee running Emu Joust I said I'd fix up the other 3 ROM machines I have. It seems like they all have the same problem in that the RAM board in each is completely dead, bus wise. If I swap in the 32K board from my working machine they all spring to life, which is good in that it limits what could be wrong.

Going over the RAM board with my scope I can see that compared to the working one which has active address bus lines on X1, the other 3 are silent apart from 5V where you'd expect it. I pulled IC20 which tested OK in my Minipro but IC22 just wouldn't give me a consistent test so I suspect that may be bad. The way the schematic is laid out it hints that IC22 is the gateway to the board so if it's dead than nothing will happen.

Before I scour the schematic and buy some more 74LS138s is there a well known death for these boards that stops them in their tracks?

Cheers!

Witchy

[Ernest]

Hi Witchy,
RESET, if this is LOW or rather not high it will hold your system in RESET and not fire up.

Once out of RESET the CPU will go through all the addresses from 0000h all the way up to 8000h where IC22 will output on pin 11 a Low ROMSEL to select your ROM BASIC.
Check C21 if leaky will hold the system in reset. 

Ernest

[Witchy]

Hi Ernest,

I did think of that but all my RAM boards have tantalums and they check out at ~230nF. I swapped C21 from good board to bad and it's still failing. You're right though, this is probably RESET based but for a different reason. This is where my lack of formal electronics training lets me down. On the good board at pin 3 of the inverter I get ~4.8V but on the bad board pin 3 is receiving ~0.4V. I've also noticed that the good board has /KBRST at ~4.8V but the bad board it's only at ~2.8V. After lunch I'll pull the resistors to check them.

Cheers

Witchy

[Ernest]

Hi Ernest,

I did think of that but all my RAM boards have tantalums and they check out at ~230nF. I swapped C21 from good board to bad and it's still failing. You're right though, this is probably RESET based but for a different reason. This is where my lack of formal electronics training lets me down. On the good board at pin 3 of the inverter I get ~4.8V but on the bad board pin 3 is receiving ~0.4V. I've also noticed that the good board has /KBRST at ~4.8V but the bad board it's only at ~2.8V. After lunch I'll pull the resistors to check them.

Cheers

Witchy

Yep Witchy,
You have your point of problem with the 2.8V at /KBRST this needs to be higher. I would start by lifting one end of D1 to rule out the Power on RESET circuit. The input pin 3 of the inverter may also be pulling down the voltage. There have been a number of engineering changes around the RESET circuit during the life of the Microbee. Check out the resistor value markings are matching that of the working board.
I assume none of these boards have a backup battery fitted, if so remove that until you get normal function then look at re-establishing battery backup as a secondary action. Just incase you have some dry soldier joint issues or even partial shorts a good visual with a magnifying glass and good light source wouldn't go astray.

Ernest

[Witchy]

Hi Ernest,

Pulling D1 fixes the problem on all 3 'dead' RAM boards and has brought 2 more machines to life. The question is, what is happening to the left of D1? VBB is powering the SRAM while the main power is off, later inverter loops are to introduce a startup delay but my crap maths head can't work out the twin voltages and voltage divider even though I've watched Big Clive and Dave Jones explain it countless times. I should have Ohm's law embedded in my skull by now.

I should also say that none of the boards have batteries because they'd all gone bad. I'm thankful they were all off-board.

Cheers

Witchy

[Ernest]

Hi Witchy,
Your not the first to be scratching their heads over the reset circuit, there were quite a few attempts over the years to deal with the very basic design.
If you have a look at the last revision of the SRAM Coreboards MB8328-6, you will see the same basic design, but some significant changes over time.
the 8328 is flipped from the drawing on your 1732 boards D1 (on the 1732) splits the Reset into two sections.
All the left of D1 is the power on Reset, with that in the middle is the Reset circuit driven from the Reset Key on the Keyboard, finally the three inverters (starting at pin3)  through to D3 are buffer / driver for pulling the reset line low in multiple areas of the computer.

The Power on Reset looks at the 12V supply via the resistor divider R12 / R11, as this has changed over the years to reduce power dissipation in the onboard regulators so has the need to change the value of the resistors in the divider. The final design incorporated a Zener diode to reduce the effect of a range of power sources.
R10 is to limit the input current & D2 is to ensure that voltage doesn't exceed 5v quite important when you consider the design has the inverters running on the Vbb power source (The backup battery).
Now that is why I asked about the battery, if a battery is fitted or even just removed the best the power to the inverter will be your normal 5V - 0.6 V so at best 4.4V this will give you a threshold as an absolute minimum of around 3V required on the input pin 13 to take the output pin 10 out of the Power on reset state to allow the system to run. changing the Power voltage into the Microbee to lower values (some operate down to 8.5V) you may not be able to get to the threshold on the input of Pin13 and the system remains in perpetual reset. So if your power in is significantly below 12 V  you may need to address the resistor divider values. However as you have one board operational on you supply and if the resistor values are the same across the three Coreboards then you should be able to stay with them as they are.
First up confirm your D1 is actually behaving as a diode and not leaky in the reverse direction before reconnecting it back into circuit.
As the voltage at pin 3 is getting to around 2.8 volts when D1 is connected I would say that the output from pin 10 is not all the way to a Low output state, otherwise your voltage on pin 3 should be pulled to about 0.6v. Based on that, my suspicion would be the 4584 CMOS Schmitt inverter, if you have trouble getting one of these you can use a 74HC14 Pin for Pin compatible but only suited to 5 Volt logic, but as we are running on 5 Volt logic, will be suitable in this location.

The final caveat is "No Board Damage" from historic battery leakage, which could be causing partial shorts etc.

Ernest

[Witchy]

Hi Ernest,

Well now I feel a bit embarrassed. 

I'm running my bench PSU at 9V 1.5A which gives me an input on the 12V rail on the RAM board of ~8.5V, exactly on the threshold you mention. Because one board is quite happy at this voltage I naturally assumed the other 3 were bad, but just by increasing the input voltage to 12V the previously dead boards have come to life. I wonder if this also relates to the program loading issue I had a few weeks ago. I'll test that after work.

The reason I was running that low is posts elsewhere that said people were using lower voltage PSUs to reduce the core temp of the regulators, and of course I had a working machine at that voltage too.

Now I've got 3 workers and 1 very dead one that appears to be missing some components on the main board.

Thanks for pointing me in the right direction!

Cheers

Witchy

[Ernest]

That's great Witchy,
You could try increasing to a 10V input, many have found that this is a good compromise.
If you want to stick with the 8.5 V, I suggest changing R11 to a 2k2 Ohm resistor instead and that should give you the extra needed for the Reset circuit. However make sure D2 is working as the importance of D2 as a protection to the input at pin 13 becomes more important if you happen to use a 12V supply at some point in the future.

What components are missing from the other main board you mentioned?

Ernest

[Ernest]

That's great Witchy,
You could try increasing to a 10V input, many have found that this is a good compromise.
If you want to stick with the 8.5 V, I suggest changing R11 to a 2k2 Ohm resistor instead and that should give you the extra needed for the Reset circuit. However make sure D2 is working as the importance of D2 as a protection to the input at pin 13 becomes more important if you happen to use a 12V supply at some point in the future.

What components are missing from the other main board you mentioned?

Ernest

Further info, just remembered there are two different versions of the MB1732 Coreboard, the resistor numbering in the reset circuit is different between Mark 1 and Mark 2.
So if you can't find R11 and your R10 is 47K, then the one you should change from 1K5 to 2K2 will be labeled as R9 not R11 Mark 2 board.

Sorry if this created any confusion.

Ernest

[Witchy]

Hi Ernest,

I'll keep them as standard I think. The only times they will be used is when I'll be in reach of a bench supply (even at exhibitions) so I just need to put a label on them telling me what voltage to use so I don't forget. I'm intrigued about the machine code loading problems I had at the reduced voltage so in a couple of hours I can check if it makes a difference.

The 4th machine has some component legs lifted like they were testing things before giving up, first one I spotted was under the keyboard on the right. Hah, it's R26 and all 4 of them have it lifted so it clearly isn't a problem 

How did Microbee test the main boards before adding the RAM?

Cheers

Witchy