Terrible Fire Accelerators

[plasmab]

Originally Posted by 8 Bit Dreams

is this rev.2 or rev.3 TF card?

Rev 3.

[plasmab]

Originally Posted by FREEMILK

Is that a VARTA on your 2k. that going to explode and take out your TF328 and CF card.

I'll take my chances.

[AwoLStill]

Originally Posted by plasmab

I so love naked computer :-)

[AwoLStill]

Originally Posted by FREEMILK

Is that a VARTA on your 2k. that going to explode and take out your TF328 and CF card.

There's NOTHING wrong with Varta, it's simply a case that no-one designed them to have a 20+ year lifespan.

(Though I do prefer Yuasa myself...)

[nogginthenog]

I can't believe a professional EE like plasmab has the same cheap Chinese hot air station as me

[plasmab]

Originally Posted by nogginthenog

I can't believe a professional EE like plasmab has the same cheap Chinese hot air station as me

I'm not a pro EE. I write software for a living.

EDIT: I used to teach EE at uni.

[8 Bit Dreams]

Originally Posted by plasmab

I'm not a pro EE. I write software for a living.

EDIT: I used to teach EE at uni.

today stream was very interesting, learned a lot and enjoyed it, however, there is one thing that i did not understand:
why this FW works on rev.3, but rejects on rev.2... Or is it really because this board is running @48Mhz..
i have one spare 48Mhz oscillator, will try to solder it @ weekend and give it a go

[PR77]

@Plasmab,

Did you ever come across a #0000004.00001970 Guru during your TF520 / TF530 development? It seems to happen with my MC68000 accelerator on the first cold boot (after caps have completely discharged). Subsequent warm reboots (or even short cold reboots) are OK and stable (even at 18.4MHz with my super dodgy jumpers).

I found this thread;

http://eab.abime.net/showthread.php?t=5653 here and some other German stuff, but the EAB entry seems to be the most detailed.

The failure address is ChipRAM (I suspect the exec.library as 00001970 is still in this region on my KS1.2) so I think I can rule out the FastRAM interface. The Guru comes on the white colour boot screen phase so AutoConfig has run by then and /OVL has been asserted.

I am at a bit stumped on this! Any thoughts? A map or elf file would be awesome!

[plasmab]

Originally Posted by PR77

@Plasmab,



Did you ever come across a #0000004.00001970 Guru during your TF520 / TF530 development? It seems to happen with my MC68000 accelerator on the first cold boot (after caps have completely discharged). Subsequent warm reboots (or even short cold reboots) are OK and stable (even at 18.4MHz with my super dodgy jumpers).



I found this thread;



http://eab.abime.net/showthread.php?t=5653 here and some other German stuff, but the EAB entry seems to be the most detailed.



The failure address is ChipRAM (I suspect the exec.library as 00001970 is still in this region on my KS1.2) so I think I can rule out the FastRAM interface. The Guru comes on the white colour boot screen phase so AutoConfig has run by then and /OVL has been asserted.



I am at a bit stumped on this! Any thoughts? A map or elf file would be awesome!

Nothing like this happened to me.

EDIT: but I don’t use /OVL

[plasmab]

Originally Posted by 8 Bit Dreams

today stream was very interesting, learned a lot and enjoyed it, however, there is one thing that i did not understand:

why this FW works on rev.3, but rejects on rev.2... Or is it really because this board is running @48Mhz..

i have one spare 48Mhz oscillator, will try to solder it @ weekend and give it a go

My Rev 3 was running @ 37Mhz.

There is more timing slack with 37mhz than at 50Mhz

[plasmab]

@PR77 are you are asserting /OVL or /OVR?

If you suspect your ram isn’t appearing at boot then you’ve likely got bugs in you sync bus code. Just use mine.

EDIT: I made my stuff open so people wouldn’t reinvent the wheel.

[PR77]

Originally Posted by plasmab

@PR77 are you are asserting /OVL or /OVR?

If you suspect your ram isn’t appearing at boot then you’ve likely got bugs in you sync bus code. Just use mine.

EDIT: I made my stuff open so people wouldn’t reinvent the wheel.

I'm not asserting /OVL (ROM Overlay). I am also not using /OVR (Override) as I simply don't assert /AS on the Amiga MB when accessing internal FastRAM. What I mean about the /OVL being asserted was the ROM has been shifted to the end of memory and not the top. The error address is 0x00001970 so I think a ChipRAM decode went bad or RAM resident code. I might disassemble the KS to see if there are any direct accesses to 0x00001970 (but this is a looooooong shot for obvious reasons).

[plasmab]

Originally Posted by PR77

I'm not asserting /OVL (ROM Overlay). I am also not using /OVR (Override) as I simply don't assert /AS on the Amiga MB when accessing internal FastRAM. What I mean about the /OVL being asserted was the ROM has been shifted to the end of memory and not the top. The error address is 0x00001970 so I think a ChipRAM decode went bad or RAM resident code. I might disassemble the KS to see if there are any direct accesses to 0x00001970 (but this is a looooooong shot for obvious reasons).

Not a hope... 0x00001970 will be the stack during early start up unless I’m mistaken

[plasmab]

Originally Posted by plasmab

Not a hope... 0x00001970 will be the stack during early start up unless I’m mistaken

Most likely thing that is happening is a cycle overlap.

[PR77]

Originally Posted by plasmab

Most likely thing that is happening is a cycle overlap.

I will have a look and check for irregular /AS assertions on the external space (i.e., not FastRAM). It's issues like this I wish I had a 48 channel LA. My 8 channel 24MS/s Saleae just doesn't cut it.

[PR77]

So... I changed my AUTOCONFIG to clock on the negedge of DS = UDS & LDS (in 68000 land DS doesn't exist... but you knew that) and it seems to have fixed the initial Guru after a cold boot. Incidentally, this strategy is straight from your "autoconfig.v". Perhaps that is why you never saw this. I was originally clocking on posedge CPUCLK.

Side ball question, I am looking into the capabilities of DirtyPCB for my next PCB run and I'm wondering if they can handle the TQPF100 (XC95144XL ) package OK. Have you experienced any PCB manufacturing issues with this package?

[plasmab]

Originally Posted by PR77

I will have a look and check for irregular /AS assertions on the external space (i.e., not FastRAM). It's issues like this I wish I had a 48 channel LA. My 8 channel 24MS/s Saleae just doesn't cut it.

I just bought a HP1661CS

[kgc210]

Well these boards are driving me nuts!
Built the REV 3's up bar the RAM and RAM CPLD with the correct 3.3v Oscilators and still no joy.
I have tried the RC2 and SPI firmware.

All I get is the half brightness LED on the A2000 and no serial output on the Diagrom.
Could I have a duff batch of CPLD's?

Anyway I have programmed my Logic Analyzer and can see the clock is doing something. But I am reluctant to probe much just yet till I build the 5v to 3.3v adapter as I don't want to fry the board.

It's a 16 Chanel 96MSPS one (https://blackmesalabs.wordpress.com/...alyzer-for-22/)

[8 Bit Dreams]

Originally Posted by kgc210

Well these boards are driving me nuts!

have no problems at all soldering them, just try to add second CPLD and program it, card should start reporting RAM failure, just click continue then

Originally Posted by kgc210

Built the REV 3's up bar the RAM and RAM CPLD with the correct 3.3v Oscilators and still no joy.

can You describe more?
What exactly is happening?

Originally Posted by kgc210

I have tried the RC2 and SPI firmware.

try Beta FW if You wanna run it on A2000, RC2 has some issues on A2000 atm

Originally Posted by kgc210

All I get is the half brightness LED on the A2000 and no serial output on the Diagrom.

if CPU does not resetting after power on - check your board for shorts and dry joints under microscope

Originally Posted by kgc210

Could I have a duff batch of CPLD's?

if You have purchased them on Ebay (adeleparts) than it can be fakes, last batch i bought from them was sanded and rebranded, i was able to program them, however these was non-working, since this happened to me i'm ordering my parts from Digikey and Mouser only - no issues since then, soldered more than 20 TF cards already

and remove JP5, no need to solder it

[kgc210]

The CPLD's are from UTSource, the fact they came sellotaped on a piece of card doesn't give you much faith.
They program and ID ok, I guess they must be factory rejects?

When you power the system on the LED lights but at half brightness and you get a pure black screen and that's it. With the Diagrom in there is no flicker of the LED as it's loading and no serial output either.

I have had all the boards under my microscope and can't find any faults such as loose pins or bridges.

Out of the 5 boards (3 Rev 2's and 2 Rev 3's) I only have one that sort of works, with a normal ROM it yellow screens and loops like that. With the Diagrom I can get it to load but it displays 3.5MB Fast RAM even with the RAM CPLD blank and when you test the CHIP RAM it shows it's corrupt.

I am going to order a new batch of CPLD's from digikey and go from there.