My Falcon went to black screen

[derkom]

I'm not 100% sure, maybe someone can check.. But I think the Falcon boots firstly into 8MHz mode and switches into 16MHz later on.

That is correct.

[mikro]

I'm not 100% sure, maybe someone can check.. But I think the Falcon boots firstly into 8MHz mode and switches into 16MHz later on.

That is correct.

Actually this is a good hint what is going on - no 8 -> 16 MHz switch, no RS232 but still Combel seems to be producing the correct clocks. So this perhaps rules out a problem on the big chips (Combel, Videl) and points to something related to TOS/RAM/CPU processing. Maybe it's really just some stupid GAL or something similar.

[sety]

I'm not 100% sure, maybe someone can check.. But I think the Falcon boots firstly into 8MHz mode and switches into 16MHz later on.

That is correct.

Oh, I just had a thought too! You don't still have the diag cartridge in do you? I have a feeling (although I could be wrong) that the diag cart uses the 8Mhz clock to keep things simple (more chance it'll load in a faulty machine) so you won't see a clock change.

You can switch the clock once the diagnostics start. It's in the menu under "Examine/Modify System"

[mikro]

Sety: no, he does have the cart but without any output on serial.

However now I'm reading the diagnostic flowchart, I think Christophe has made a mistake - he didn't follow the "B" chart but "C" instead. And "B" indeed recommends to check whether 030 is halted.

[christophe.griard]

Hello Mikro, hello Sety, hello Exxos,

Thanks for the information about the initial 8MHz frequency on the CPU at boot.

Yes I do have the diagnosis cartridge with no display on my Atari STE connected to the Falcon using a null modem cable and VT52 running on the STE (using Language Disk for Atari STE).

In fact, I followed the paragraph 3.3 "Troubleshooting a dead unit" on page 67 of the service manual and not the diagnostic flowcharts.

And when I follow chart B, it says to check that 16Mhz is present and if not replace the oscillator.

What I do not understand is that I have 32 Mhz at the inlet of the COMBO IC, and both 25 Mhz and 32 MHz at the inlet of VIDEL. But still only 8 Mhz oin the CPU (pin 6) and expansion port (pin 23). I don't know what is the oscillator that leads to the 16 Mhz (there only 3 oscillators on the motherboard: 3,67 Mhz, 25,175 Mhz and 32 Mhz).
Question : is the 32 Mhz oscillator also in charge of delivering the 16 Mhz? Is it the COMBO IC that delivers this 16 Mhz based on the 32 MHZ at its inlet?

I checked again this morning before writing this message that pin 18 of expansion port shows a stable +5V (I also checked that I have also a stable +5V on the pin 129 of the CPU ) : so the CPU is not halted.

Exxos adviced to check the signal on pin 3, 132 and 125 : I will try to do it just after writing this message. Not an easy task as there is the need to trigger on reset to properly see the signals...

I will keep you uptodate. Thanks for your information especially how the 16 Mhz is delivered.

Have a great day.

[derkom]

Question : is the 32 Mhz oscillator also in charge of delivering the 16 Mhz? Is it the COMBO IC that delivers this 16 Mhz based on the 32 MHZ at its inlet?

Yes, the COMBEL accepts the 32 MHz clock input on pin 20, and divides it as necessary to produce the 8 and 16 MHz clocks on pin 191.

[sety]

Question : is the 32 Mhz oscillator also in charge of delivering the 16 Mhz? Is it the COMBO IC that delivers this 16 Mhz based on the 32 MHZ at its inlet?

Yes, the COMBEL accepts the 32 MHz clock input on pin 20, and divides it as necessary to produce the 8 and 16 MHz clocks on pin 191.

Yep! Combel contains the clock dividers, which is fed by the 32.something oscillator and is software selectable. Falcon will run quite happily at 8Mhz if you want, but the instruction to change combel clock divider comes from TOS. So if you're not seeing the clock change it means that TOS isn't loading. If you're using a test cart then TOS isn't loading anyway because the cart takes over.

I'm 90% sure, actually 80% sure, make that 75% sure that the test cart doesn't switch the clock to 16Mhz as it likes to keep things simple... It's only diagnostics anyway, they're not trying to sell a Ferrari at that point. My cart is down as I stole the eeprom for another project and replacement hasn't arrived yet, otherwise I would confirm that for you.

This from the Falcon developer doc...

Screen Shot 2020-04-29 at 7.48.06 pm.png (13.64 KiB) Viewed 3497 times

Cheers and stay safe mate. Keep at it, you'll fix it!

[christophe.griard]

Hello Sety, hello Exxos, Hello Derkom,

Thanks a lot Derkom and Sety for the confirmation from that the COMBEL divides the 32 Mhz signal.

Today, I have checked that:
- 32 Mhz signal arrives on pin 20 of COMBEL (already checked yesterday)
- COMBEL delivers on pin 193 a 4 Mhz signal
- COMBEL delivers on pin 192 a 8 Mhz signal

On pin 191 of COMBEL (clock signal to the CPU), I measured a 8 Mhz signal (coherent with what I have seen at the inlet of the CPU on pin 6).

Thanks a lot Sety for the information about the switch between 8 Mhz and 16 Mhz coming from TOS... As you say, it means that TOS isn't loading...
And thanks for your support to try to fix my baby... .

With these new information, it seems COMBEL is not responsible for the failure as he does his job at this stage of the diagnostic.

After these measurements, I tried to make the measurements proposed by Exxos (maybe these pins/signals are involved in TOS loading with orders sent by CPU).
I set my oscilloscope to trigger on channel 2 (I check that channel 2 in DC coupling and that the signal reacts as intented (that is to say +5V stable and goes to 0 when reset button pushed). On the picture 1, the set of channel 2 as the channel on which channel 1 should trigger). The trigger mode auto was used but I tried to trigger to fixed values of reset signal with the same results.

Then I put channel 1 on the 3 pins 132 (picture 2), 125 (picture 3) and 3 (picture 4) of the 68030 and below are the signals...
I am not sure that I made the measurements correctly... Signals seem strange but I made them several times. On the pictures are some values relative to signals.
When I reset the Falcon, the signal are frozen for some very short period of time and I was not able to something during these first cycle of clock. Only pin 132 seems to be impacted by reset, but I am not sure at all.

I also measured the signal on pin 119 of COMBEL that seems to be the same signal than the one of pin 132 of CPU (picture 5).

Some days ago, as the failure could come from TOS, I replaced it with 2 other TOS in 4.04 version without any success...

[exxos]

Thanks for the images.. I think one of your problems is your scope does not appear to have sufficient bandwidth to capture the signals correctly..

It appears though that /AS & /DTACK0 are pulsing.. So this suggests the CPU is running to some degree...

Do those signals continue for any significant amount of time or do they stop switching a little later ?

[christophe.griard]

Hello Exxos,

Thanks for the feedback.

Yes my oscilloscope is a very small one and its 25 Mhz Bandwidth is probably not enough to properly capture signals at these frequencies...
I've been wanting to buy a good oscilloscope for a long time and may have to invest ... And to do such measurement with a 32 Mhz signal, I think there is the need at least to have a 100 Mhz scope.

Conerning your question, the signals remains like this for at least one minute without changing in shape/level.