Terrible Fire Accelerators

Thanks guys, yes I mean the A500 socket.
Round pins sounds good, thanks for the suggestion!

Hi Plasmab, have to admit i've not started watching the videos yet! FAQ's would be awesome if you have the time.
Also thank you for making this project and making it open, really appreciate it!

Do you have a tips paypal or a patreon i can donate to?

[project23]

The videos are an awesome resource, although admittedly there is an awful lot to watch.

I recommend as a bare minimum you watch the videos where he actually constructs the 530.

I've been listening to the videos myself in work in the background - there really is a wealth of information in them, its just that by their nature it isn't indexed information.

EDIT: oh, and grab yourself a diagROM when you can - i've been using it in a 600 to troubleshoot a Furia (don't get me started), and it really is great.

Just looked up diagROM, didn't know about that, so bought one straight away! Thanks for the advice!

[project23]

Originally Posted by Guchion

Just looked up diagROM, didn't know about that, so bought one straight away! Thanks for the advice!

Yeah if you didn't know about that then you really need to watch his videos.

[plasmab]

Originally Posted by project23

Yeah if you didn't know about that then you really need to watch his videos.

Indeed. Almost every video features DiagROM

[plasmab]

Here is a TF520 Backport of the current TF534 firmware bus timings...

https://www.dropbox.com/s/1q8rlu51v5...A_top.zip?dl=0

Usual disclaimers etc.

[8 Bit Dreams]

Originally Posted by plasmab

Here is a TF520 Backport of the current TF534 firmware bus timings...

https://www.dropbox.com/s/1q8rlu51v5...A_top.zip?dl=0

Usual disclaimers etc.

Thank You!

Originally Posted by plasmab

Will do this evening. Can’t guarantee it’s perfect though.might be interesting to see how it works in the A2000.

will test and report soon! (still waiting on rev.3 pcb's and cpu's from USA - took Freescale with the hope to be able to overclock them)
i am more interested to test it on CDTV though!

PS
Have reWatched old TF520 videos from 10.2016 - deamn time passes by too quick!

have some questions: in video U are using ceramic 68020 cpu, however sysinfo shows it as 68EC020 - i was under impression these was two different type of cpu's (different pinout)?
How about frequency on this new alpha FW? Should 16Mhz oscillator be used or i can go higher and grab a 24Mhz one? What do You think, what is the highest speed this card can handle atm?

[plasmab]

Same FW code works at 35mhz so yes.

[plasmab]

Originally Posted by 8 Bit Dreams

Thank You!

have some questions: in video U are using ceramic 68020 cpu, however sysinfo shows it as 68EC020 - i was under impression these was two different type of cpu's (different pinout)?
How about frequency on this new alpha FW? Should 16Mhz oscillator be used or i can go higher and grab a 24Mhz one? What do You think, what is the highest speed this card can handle atm?

Those are gold top ceramic packages. They are all EC versions because its pointless in context to have a 32bit address bus.

[PR77]

@Plasmab, circling back on the RAM issue between K2K RAM expansion and your TF530 card, during one of your live steams I asked the question about the Autoconfig chain requests. How are you (if in fact you are) preventing both expansion cards from decoding and responding to bus cycles at base address 0x00E8xxxx at the same time? Could you point me to the logic?

I would like to add a mechanism to my FW so I can play with expansion ROMs and an External HD, but I can't see in your FW how you could be handling this.

[plasmab]

I answered on stream. I dont allow acceses to E8xxxx to get to the 68000 bus until all cards are configured. its pretty simple.

[PR77]

Originally Posted by plasmab

I answered on stream. I dont allow acceses to E8xxxx to get to the 68000 bus until all cards are configured. its pretty simple.

Still doesn't help me understand unfortunately. Here is my understanding, it would be good to have the gap filled;

intcycle_dout is a product of "zii_access" (a latched state of "zii_decode"). "zii_decode" is active whenever !configured and A[23:16] = 0xE8.

So, whenever TF530 is not configured, you will be responding to 0x00E8xxxx with data on bus. So K2K's board will also not be configured and responding on the bus. I can't the mechanism which is suppressing 68000 bus cycles while the TF530 is not configured.

[plasmab]

Originally Posted by PR77

Still doesn't help me understand unfortunately. Here is my understanding, it would be good to have the gap filled;

intcycle_dout is a product of "zii_access" (a latched state of "zii_decode"). "zii_decode" is active whenever !configured and A[23:16] = 0xE8.

So, whenever TF530 is not configured, you will be responding to 0x00E8xxxx with data on bus. So K2K's board will also not be configured and responding on the bus. I can't the mechanism which is suppressing 68000 bus cycles while the TF530 is not configured.

Yes.. and INTCYCLE is set when zii_access is low. and the BUS CPLD doesnt assert AS when INTCYCLE is low.

So once configured is set INTCYCLE no longer gets set low on access to E8xxxx.

EDIT: Its too complex to just point at a single line of code. Its distributed across 2 chips. You need to not let the amiga see the E8xxxx access until all your cards are configured.

[PR77]

Originally Posted by plasmab

Yes.. and INTCYCLE is set when zii_access is low. and the BUS CPLD doesnt assert AS when INTCYCLE is low.

So once configured is set INTCYCLE no longer gets set low on access to E8xxxx.

EDIT: Its too complex to just point at a single line of code. Its distributed across 2 chips. You need to not let the amiga see the E8xxxx access until all your cards are configured.

Thankyou! I missed the distributed INTCYCLE and for some reason was thinking (for example) MOVE $0x00E80000, $someRAMBuffer would occur in a single bus cycle! . I found this nice reference;

http://oldwww.nvg.ntnu.no/amiga/MC68...s/timmove.HTML

[plasmab]

I think i've figured out what is going on with the Kipper2k card. the refresh cycle gets wound up in a main cycle every now and then and corrupts ram. TBH the ram card assumes perfect 4 clock cycles one after the other.. and that simply doesnt always happen.

EDIT: The refresh cycle also assumes a bus cycle end every now and then .. there is no guarantee that will happen with fastram

[8 Bit Dreams]

Originally Posted by plasmab

I think i've figured out what is going on with the Kipper2k card. the refresh cycle gets wound up in a main cycle every now and then and corrupts ram. TBH the ram card assumes perfect 4 clock cycles one after the other.. and that simply doesnt always happen.

EDIT: The refresh cycle also assumes a bus cycle end every now and then .. there is no guarantee that will happen with fastram

can't You just rewrite that FW, to make it work realible?

[plasmab]

Originally Posted by 8 Bit Dreams

can't You just rewrite that FW, to make it work realible?

I thought about that but the design of the card is flawed. It doesnt have a DTACK signal wired to the CPLD so we cant insert wait states while a refresh is happening.

[8 Bit Dreams]

Originally Posted by plasmab

I thought about that but the design of the card is flawed. It doesnt have a DTACK signal wired to the CPLD so we cant insert wait states while a refresh is happening.

so these problems are not visible because this card was used with stock CPU?
However, if You try to use it on higher frequencies - You run in trouble?

PS
May i ask You about this 4MB mod, Cezar done on TF530 (to add another 2MB piggyback):
what do You think about it? I am pretty sure there will be problems because of voltage regulator, so probably You'll need to add one more LDO1117 (also piggyback)
otherwise i don't see any reason why it will not work.. Actually it's very nice feature - to give users a chance to upgrade their existing cards.. i am sure many of us will be happy to use that chance (if there will be any), but it needs another FW correction

[plasmab]

Originally Posted by 8 Bit Dreams

so these problems are not visible because this card was used with stock CPU?

However, if You try to use it on higher frequencies - You run in trouble?



PS

May i ask You about this 4MB mod, Cezar done on TF530 (to add another 2MB piggyback):

what do You think about it? I am pretty sure there will be problems because of voltage regulator, so probably You'll need to add one more LDO1117 (also piggyback)

otherwise i don't see any reason why it will not work.. Actually it's very nice feature - to give users a chance to upgrade their existing cards.. i am sure many of us will be happy to use that chance (if there will be any), but it needs another FW correction

No issue with it. The RAM isnt powered by the regulator.



I wouldnt recommend anyone build a TF530 right now. I'd wait for the TF534. There are many many stability fixes in it.

EDIT: but I absolutely do not support this mod. The TF530 is a 2Mb card

[8 Bit Dreams]

Originally Posted by plasmab

EDIT: but I absolutely do not support this mod. The TF530 is a 2Mb card

oh..such a pity...i was hoping for a chance to upgrade some existing TF530 cards