Amiga 3000 cold power on issues

[Castellen]

Doesn't sound like any of the symptoms I'd attribute to capacitor problems, which usually don't happen unless they start leaking and the electrolyte causes corrosion problems; which is rarely an issue in the A3000.

Could be many different causes behind it.  I'd suggest a few measurements when it's powered on and in the 'dead' state to understand where to look next:
1. What's the state of any of the reset lines while the system is 'dead', are they at run state (5V) or is the system being held in reset state (0V)?

2. If the reset lines are in run state, do you see the system trying to run or not?  Easiest way is measure ROM_EN (pin 12 of either ROM) with an oscilloscope and see if there's activity there or not.  That will narrow down if  the CPU is running and there's valid address bus data.

3. Are you getting a valid 50MHz clock (around 4Vp-p) from the oscillator module on the CSmk2?

[Castellen]

Solder joint issues on the KEL connector are rare, but worth checking.  Cleaning the contacts may help.  Visually inspecting contacts to see if any are not aligned and re-forming any misalignments is usually better then attempting to replace the connector.  Replacement connectors are expensive, difficult to find, and it's a major soldering job.

The clue that DiagROM doesn't show the same issue at power on suggests that it's not related to the KEL connector.  If it was the connector at fault, DiagROM would likely present the same issues at power on as the standard ROM does.  I wouldn't be concerned with the fast memory not appearing with DiagROM, I see that fairly often with correctly working machines, suspect there's a few imperfections in DiagROM.

Given the clues that the A3000 is fundamentally running in the 'dead' state, and DiagROM works reliably, then most likely the system is getting hung up somewhere while booting from the standard ROM.  I've seen scsi.device do exactly this when things aren't happy with the SCSI controller or some of the connected SCSI devices, possibly even bus termination.

If you have the capability to program EPROMs, and a null modem cable to connect to another computer's serial port, I can send you a specially modified ROM to diagnose the power on issue, which will definitively show what's actually running and where it's getting hung up.  Contact me by Email (the PM notification Emails from on this site don't appear to work these days).
http://amiga.serveftp.net

[Castellen]

But what are your thoughts on it just booting fine with the 68030 on board versus having issues with the CSMkII installed?

There's not enough information to say exactly what might be causing the problem.  You mentioned that the BFG9060 works reliably, but then you also said that the CSMkII "is now booting fine all the time".  There's obviously some kind of intermittent issue somewhere.

You'd need to collect more clues with the system in the fault state to draw more solid conclusions, which you can't easily do without some diagnostics tools such as a special ROM to see what's happening under normal boot conditions.

Intermittent problems are often caused by issues with connectors, and the A3000 is full of these in the form of horrible IC sockets with tin plated contacts.  In these situations it's often worth removing each of the socketed ICs, including the PLCCs, carefully brushing all contact surfaces with isopropyl alcohol and re-inserting them.  Be sure to use a PLCC extractor tool since anything else will damage the sockets.  Otherwise leave the PLCCs alone and just focus on the DIP devices.

The KEL connectors are at least gold plated, so don't suffer the same tarnishing and oxidisation problems that tin plated contacts do.  They're usually reliable except when individual pins somehow get over-compressed or bent out of alignment.  You can also brush the contacts with isopropyl to remove any surface contamination.

[Castellen]

An update, heating RP109 makes the system boot. So it may be that component.

Good find to narrow it down that far.  The resistor networks are usually fairly reliable, I've only seen them give problems as a result of corrosion damage.

You can test it easily enough.  With the system in the fault state, and powered off, measure the DC resistance between 5V (or pin 1) and pins 2 - 9 of the resistor network, they should should all be around 1k Ohms.  Some variants use 2.7k or 3.3k for address bus pullups.

To further test the heating theory, you can confine heat to a single component by putting your soldering iron against the component body for 5 - 10 seconds.  That means you'll be heating only the suspect component and not anything else in the area around it as the hot air does.

[Castellen]

how many layers is the 9.2 rev A3000 board? Could it be a bad via or a trace in the board that's wonky?

It's a 4-layer board.  A majority of the signals are routed on the external layers, with the internal layers mainly used for 5V and ground.  Internal layers do contain signal routing in congested areas, such as around the 68030.  It's a decent quality PCB and faults are rare except when there's physical damage or corrosion.

Since you were able to make the fault go away with localised heating, that's still a good lead to follow.  Though I'd suggest trying to use heat in much smaller areas.  For example, heat a component body using a soldering iron.

Once the system is in a state where it's not showing the problem, you can try using an aerosol can of freeze spray to provide localised (spot) cooling to see if you can make the fault re-appear on demand.  That might get you a lot closer to whatever is causing the problem.