GURU Boot issue when power cycling

[kolla]

In general, I don't power cycle, I use ctrl-a-a (for up to 30 sec), or use a hardreset program.

[paul1981]

Thanks guys, very helpful and thoughtful posts and I do appreciate it.

I seem to have much better luck if I wait longer to switch the power on/off.  If I wait about 5 seconds, it seems to have the fail rate I mentioned in the first post.   If I wait a full 30 seconds, it seems to boot properly.  

I would think that 5 seconds is enough time to clear memory and "start over", but maybe not?   I haven't switched it on/off enough to make a decisive claim that this fixes the problem, but I have yet to see it fail if I wait 30 seconds to turn it back on and I think I've done it about 15-20 times now.

How long do you guys wait between power-cycling your Amigas?

-Ted

5 seconds ought to be fine.

[VirtualTed]

Well, I guess I'll give it a shot with the new PSU.  I cracked it open once and it "looks" great in there, so I dunno.  Don't think I'll replace caps until I tried the other.

[VirtualTed]

Just thought I would update everyone on this, even though it's been a while maybe it will be helpful for someone else.

This problem continued to remain even after using another power supply.  Otherwise, the machine is perfectly fine.  I just have to wait longer to power-cycle it.  If anyone run across a solution, please keep me (and this thread) in mind, because even if I don't fix it, I would love to know what the real problem was and I'm sure someone out there may have something similar.

Thanks for all the help.

-T

[HammerD]

Just thought I would update everyone on this, even though it's been a while maybe it will be helpful for someone else.

This problem continued to remain even after using another power supply.  Otherwise, the machine is perfectly fine.  I just have to wait longer to power-cycle it.  If anyone run across a solution, please keep me (and this thread) in mind, because even if I don't fix it, I would love to know what the real problem was and I'm sure someone out there may have something similar.

Thanks for all the help.

-T

I've seen GURU's like that on my A4000 system if I don't have the filesystem in the RDB.  For example if I have my boot partition setup to use SmartFileSystem and SmartFileSystem is not in the RDB I will get GURU's.  Once it's in the RDB it is fine, the GURU's go away.

[Matt_H]

Gurus at power cycling are not uncommon - just make sure to give the machine enough time between flipping the switch off and then back on again. This used to happen with my dad's A2000 even back when it was "new" in 1990/91. Waited a few more seconds before power on and all was well.

[VirtualTed]

Gurus at power cycling are not uncommon - just make sure to give the machine enough time between flipping the switch off and then back on again. This used to happen with my dad's A2000 even back when it was "new" in 1990/91. Waited a few more seconds before power on and all was well.

Thanks Matt.   I think I'm coming to that realization.   The machine is perfect otherwise and I'm forgetting how these older devices need a little more time between flipping on/off the switch.

[Shamron]

I would consider 5 secs very short time between power off to power on.

It's probably enough to drain the caps in the PSU, but there are other caps in the machine as well, and some caps in circuit with the ram, could maybe need longer to drain fully.

For example, the X1000 manual specifically says to wait at least 30 seconds to power on again after a power off. This is probably a huge safety margin, but i would still consider 5 seconds to be very "hasty" as Threebeard would have said. :-)

When resetting a router, you often pull the plug for 10-20 seconds.

Amigas are delicate, old ladies. Treat them like an old, vetaran diesel car. You take your time to glow, then start, then let it run a min or two before driving, to let the engine warm up a bit.

[Ral-Clan]

I remember that a lot of computer manufacturers in the 1980s recommended a full 30 seconds between power cycling.  Two seconds or even five seconds would be too short for machines from this era.

[psxphill]

and some caps in circuit with the ram, could maybe need longer to drain fully.

The dram will also take a while to reliably lose their contents. AmigaOS keeps some data in ram that survives resets. It may not be happy if some of it has been corrupted by a couple of seconds without being refreshed.

[A4000Bear]

I have an A4000T which has an extreme manifestation of this problem.

I have to keep the power off for up to two minutes in order to prevent the Guru from returning after a bad crash. Needless to say it's quite an inconvenience.

I have tried several different CPU cards. No difference.

Power supply has been checked for both voltage and noise. No problem there. Reset has also been checked. It pulls low cleanly for at least 500ms and returns to a full 5V.

Even more weirdly, I can sometimes cause this problem by adding or removing a Zorro card, and if I do this and turn the machine back on within two minutes I get no boot and a blinking power LED. Wait for more than two minutes and the machine then powers up perfectly!

It even remembers the kind of crash too. If I had a red software failure message, that's what I get again if I power up too soon. Likewise if it's a black screen and blinking LED, or simply no boot at all, even though the power LED changes brightness.

[paul1981]

Is that with the original Commodore/Amiga Technologies PSU?

[guest11527]

I have an A4000T which has an extreme manifestation of this problem.

There is really nothing you can do. AmigaOs tests AbsExecbase and its checksum for consistency to distinguish between a cold start and a reboot. If the RAMs keep their contents well enough - and later and more "modern" RAMs do - then the system will consider your power cycle as a reboot, even though a lot of other resources are damanged - which will then later on just cause a crash. It's a very principle problem that cannot be avoided.

Aminet should contain a couple of programs that simulate a full power cycle. It's actually not that hard, you "only" need to damage execbase "hard enough" to make the system believe that you have power-cycled the machine and go through a complete re-initialization cycle.

The only question I would have is why Ctrl+Amiga+Amiga isn't good enough in your case.

[RiP]

I have the same problem with an A500 rev6a too

[Castellen]

I have the same problem with an A500 rev6a too

I had a look at this with the A500 and A600 a few months back.  The short explanation is that it's a minor problem with the power on reset circuit in both of these models.  If there's a fairly quick power off-on cycle, then the circuit that is supposed to hold the system in reset state for about 470mSec while the 5V supply stablises doesn't get triggered because a capacitor used in the triggering of the reset circuit hasn't discharged, meaning the machine begins program execution before the 5V supply has stabilised because reset isn't asserted at power up, which essentially results in the software failure message because the CPU is trying to talk to other system components before they're ready.

It's only the A500/A500+ and A600 that has the problem.  i.e. if it receives a power cycle that's a bit too fast, then sometimes you'll get a software failure alert on power up.  The reset circuit on all the other models works a bit differently and doesn't have the problem.


I've not tested this solution, but most likely it can easily be corrected by putting a small signal diode (e.g. 1N4148) across R711 (A500/A500+) or R611 (A600) with the cathode (stripe side) towards the 5V end of the said resistor, and the anode connected to the side of the resistor that connects to pin 2 of the 555 timer IC.

What that will do is cause the 100nF capacitor (C611/C711) to discharge immediately at power down.  So during the next power up cycle, the 100nF capacitor will be discharged, meaning that the 555 based timer circuit will get triggered and correctly wait 470mSec before bringing the machine out of reset state.

Hope that helps.

[Castellen]

I have an A4000T which has an extreme manifestation of this problem.

I have to keep the power off for up to two minutes in order to prevent the Guru from returning after a bad crash.

The problem in the A4000T is different to the one I mentioned above; mine does this too from a cold power on.  From what I can tell, it's something happening with the IDE driver in ROM.  If you power up the A4000T, with the standard v40.70 ROM, and something connected to the IDE interface, then you get the guru message with Error=80000025  Task=07C02228.  But it eventually gets over that and boots normally.  And a warm boot is OK.  The problem doesn't happen if you're using SCSI only and nothing connected to the IDE interface.

I was doing some testing recently with the ROM software developer and this A4000T power up error problem is resolved in the Cloanto v45.64 (3.X) ROM.  You'll need the specific A4000T version of this, which I don't see on the AmigaKit.com website.  Not sure if it's officially available or not yet, send them a query.