new FPU completely blocks startup

[Hancock]

Hi everybody,

I have here a Performa 450 in which recently I wanted to install a FPU. (CPU is an overclocked 68030-16, FPU is 68882-33)

With the FPU installed the P450 is completed blocked, not even the initial chime, and the PSU is "ticking".

I finished by checking the pins of the FPU socket: when the PSU is connected, GND and VCC is shorted on the socket (only!),

This explains the behavior.

When the FPU is taken out, the Mac starts normally.

My conclusion is that there must be a blcoking capacitor between VCC and GND which is broken. But which capacitor is that?

Does anyone have an idea?

The board is rather clean, no leakages or other damages of this kind are visible.

Btw: on "normal" startup the Mac is screaming and howling. I suspect the capacitors  C2 and C4 are the reason (if I touch them, the noise goes down)

I thought to fix this later.

 

[AwkwardPotato]

If you're having sound troubles, all of the electrolytic capacitors need to be replaced. Replace those on the logic board and in the power supply, and clean the logic board well. If a bypass cap for the FPU was shorted, the system probably wouldn't turn on even without the FPU. Since you mention there's a short between VCC and GND with the FPU installed, it's also possible that the FPU is dead.

 

[Hancock]

I meant VCC and GND is shorted ON SOCKET as soon as the power supply is connected..

The FPU does not show any shortages and I tested with several FPUs. 

 

[Bolle]

Are you putting the 68882 into the socket the right way?

 

[superjer2000]

Where did the 68882 come from?  I've heard (haven't personally experienced) that some of the CPUs/FPUs on eBay from China-based sellers may not always be the real thing.

 

[Hancock]

@Bolle: I was very careful in inserting the FPU - it was correctly done

@superjer2000: hmm yes, ebay and China and already thought about that, but I bought 5 and they all behave the same - see below.

But: VCC connected with GND on the FPU-socket is much more critical: I compared it with my P460 - there they are not connected (just a little bit of charging current - which proves that there is a capacitor in the game)

 

[AwkwardPotato]

Have you tried swapping the P460 power supply into the P450 to see if it makes any difference?

There are some tantalum caps on the flip side of the board from the FPU, almost certainly used for decoupling. I seriously doubt they've failed, but when they do, they fail shorted.

 

[Bolle]

Well, it’s still possible to put it in backwards even if you’re careful

The FPU will fit the socket in all directions.

Got another machine to verify the chips itself are no fakes? I was lucky everytime I ordered 68882s from China but I tend to not buy the cheapest ones available and like to buy from smaller sellers that don’t have a ton of feedbacks.

I am not sure if the bypass caps connect directly to the power pins on the socket or if they just go to their respective power planes. Got a high res image of the board area in question?

 

[Solvalou]

Also worth making sure pins on the FPU and / or the socket doesn't have bent pins or ones shorting on eachother. Quite easy to have two pins touch on a PLCC chip.

 

[Hancock]

Thanks for the feedback so far.

Sorry for not having mentioned the obvious:

Yes, I have cross-changed all components (PSU, battery) and did the PRAM-reset command

I have checked the FPU-Pins  and I have checked the socket-Pins.

I have also checked the board for dust or anything that could cause an unwanted connection.

As already stated: there is no connection between GND and VCC on the socket as long as the PSU is not connected to the board.

When I connect it, there is no more resistance between GND and VCC

On a comparable box this resistance remains whether or not the PSU is connected.

The only explanation for this behavior that I can see is that the FPU-specific capacitor is broken.

For the next step it would be very helpful to find which one this is.

I would hope that someone can provide some further information.

Thanks in advance.

 

[AwkwardPotato]

Checked the schematic, the references for the tantalums that are bypass caps for the FPU are C100 and C121. If they were actually shorted, though, you would read a short across VCC and GND whether or not the PSU was plugged in (unless I'm very much mistaken).

By what method did you find there was a short? Type of multimeter, mode you're using, etc? Pictures of both sides of the board, especially around the FPU, will be helpful.

 

[Hancock]

Here is the exact testing acitivies:

1) Power supply is not connected to the board: infinite resistance between Pin 9 (GND) and Pin 61 (VCC) of the FPU socket

2) Power supply connected to the board (NOT switched on): 0 resistance for 1-2 seconds, then infinite (same pins)

3) Lift one probe for a moment, and put it on again: same behavior

Can be repeated for GND (pin 9) to 16,17,18 (size),27,43,52,53,61 (all VCC)

GND to GND is always 0 resistance.

On a P460

#2 from above: 0 resistance just for a fraction of a second, then infinite

#3 from above: resistance remains infinite.

Hypothesis:

On the P450 there is a capacitor slowly charging and losing the charge immediately?

On the P460 the capacitor charges quickly and keeps the charge.

Please don't forget:

The Mac is running fine with 36Mb of Ram, SCSI2SD, Floppy, keyboard/mouse, System 7.5, Symantec CPP 7... except C2 and C4 need probably to be replaced.

MacCheck is reporting "all well".

That's why I am still puzzled with this "exotic" issue, it must be "local" to the socket.

Rgds

Btw:

- The Multimeter is TE.Electronic MS8221C

- Pictures will come later

 

[AwkwardPotato]

As far as I can tell, your P450 description sounds normal. Nothing is shorting VCC straight to ground. I'm still suspicious of the power supply; try taking out everything unnecessary but leave the FPU in and see if that changes anything, e.g. remove RAM/VRAM SIMMs, SCSI2SD, keyboard, etc.

 

[Hancock]

With some more testing, i.e. in other boards and on a network card, the conclusion is that the FPU(s) are dead.

I excluded hypothesis this so far because I bought a package of 5 and they all cause the same problem.

Yes, I bought them on ebay, they were coming from China and they have already been in use.

In the same way I recently bought a "Quadra 630"-board which turned out to be a Performa 638 (68LC040) and a 68040 which prevents the board from starting.

And this is not the first time I am getting dead components.

Never again.

Still I need to see that a new FPU will be working properly in the Mac to close the story.

Thanks for the feedback.

 

[johnklos]

If you bought all five FPUs from the same seller, then you likely have five identical relabeled non-FPU chips.

Likewise, the m68040 is probably relabeled. Shame the seller, get a refund, and try different sellers.

 

[Hancock]

Final update:

Received a new FPU today (from a local dealer) - it works as expected.

The Mac was innocent   []

(Sorry for having opened this topic, I just could not believe that a package of 5 FPUs contains only dead ones)