On 2017-Aug-23, at 6:17 AM, Dominique Carlier via cctalk wrote:
> Here is the situation.
>
> The + 5V 12A that collapses comes out from pins 10-14 / 36-40, the other + 5V
> 12A (pins 15-22 / 36-40) never goes down.
>
> All that I surrounded in green on that image (1950×2361, zoomable) are the
> components that I tested on the power supply G2, A3 motherboard, regulation
> board and heatsink A1 + A2:
> http://www.zeltrax.com/classiccmp_forum/psu_g2_test/g2_tested_components.jpg
>
> To eliminate some doubts and because I do not have some spare parts on hand,
> I switched the modules A1 and A2, same result. Idem with the transistors
> 2N2905, same result.
> The result is always the same : it's always the +5V on pins 10-14 / 36-40
> that collapses, never the other output.
>
> Some resistors are not yet tested is because these must be de-soldered for a
> valid test, but the printed circuit is very fragile and many component have
> legs bent into the weld.
>
> Except for the not tested components (among others the LM376, the rectifier
> diodes) At this stage I start again to suspect a little everything. The
> famous large capacitors of the power supply (C1 to C4). But also a possible
> problems on the boards of the computer itself.
>
> As one of you mentioned, the hypothesis of shorted decoupling capacitors on
> the boards could put the power supply in default.
>
> Note that the machine runs normally with the CPU board, three core memory
> boards (400w each) + two multiplexing boards for terminals + the printer
> board.
> If I add only one of these remaining board:
> - Disk Pack Controller
> - 9-track tape Controller
> - "scanner" board (also for terminals)
>
> -> Power Fail.
>
> Note that : if I only connect the CPU and the disk pack controller card:
> Power Fail too !!
>
> What makes me doubtful about this scenario is that I can not imagine that
> these three boards, each causing the Power Fail, could fail simultaneously.
> Remember that the first time I powered up the beast (one big hour), the
> machine was working with all the boards and Power Fail appeared at once.
> I have not retested since but also note that by adding an external power
> supply just for the deficient + 5V , the machine has restarted and even
> booted the operating system.
>
> If you have another ideas? LM376?
Am I correct in inferring that this machine (the processor) has four +5V
regulators?: it appears there are 2 power supply chassis (rear photo), each
with 2 regulators,
or are those 2 chassis not identical and there are only to two +5 regulators?
Either way, is it documented or has it been mapped out how the regulators are
distributed to the bus/backplane slots?
If not, I would suggest doing so to start with, so you know what slots & boards
each regulator is supplying power to.
Two possibilities come to mind:
- One board, perhaps the CPU board from what you describe, has a fault
increasing it's current draw.
Without other boards plugged in on the same regulator, it's within the
current capabilities of the regulator and 'appears' fine.
But with another board, the current draw is excessive for the regulator
and current limiting kicks in.
- How is the backplane for this machine organised: as a pure bus, or
with dedicated slots for specific boards?
Even if it is a bus structure, there might be requirements/limits on
board combinations to distribute the load amongst the regulators.
Might the machine have been misconfigured by someone moving boards
around, so that 2 heavy load boards end up on the same
regulator and send it into current limiting?
A minor comment regarding the components in the power supply: Q13 is a stage in
the regulator drivers and as consequential as other components such as Q2.
