Upgrading Apollo 1240 to 1260 a definitive How To Guide

[Zetr0]

@Dimlow

that is Beautiful work sir!

!!!!!Elegant and Beautiful!!!!!

how did you get such good quality of pictures?

when i try and take a pic of something close up it generally ends up blurred!

[Dimlow]

I have an old Nikon coolpix 995, it takes great pictures and it should to it cost me £450 many years ago. Also it has a macro mode for close up, i can go much closer, could zoom into a single pad. I some times use it as a microscope for inspecting board.

[rkauer]

Wow! Single sided, gold-plated pins and tantalum capacitors?

That Apollo will be very happy!

Impressive work, m8!

[Dimlow]

@rkauer

People, I've found a seller who asks 2.40USD for the 27C128 (150ns, the very kind EPROM we need), plus US$5.20 for S&H.

He have 7 pieces. If anyone have interest, I'll contact the EPAY seller to add the chips in a "individual chip buy it now" basis.

Can you explain to me why we need these fast EPROMS ? i have been reading through the Amiga Developer CD and the source codes for the Apollo roms. Before the rom code is executed it is first relocated to ram. I can't see that having a slower rom is going to slow things down at all. This could be proved by writing one of the rom images to a slower rom. I expect it will make no difference at all. or will it ????

As far as i can see, the only reason for the rom, is for the card to be able to identify itself in the boot menu, and to load up the SCSI device if installed, and to patch in the 68060 library if the 060 is installed. There may be other things it does, but 90% of the code in the rom is for the SCSI interface.

As a note, now that we have the source code to the card, maybe we can get some talented programmer to look at it and maybe there are some improvements that can be made.

And another note, whilst looking at the code comments i saw this...

; - V5.55 08.04.96 Apollo-neutralisiert im ROM

I pumped this into bable fish and now believe that in versions later than 5.55 there is an option to disable the card in the rom, i had a look but could not see what is required to disable the card.

I realy think others should look at this rom source code, as my german and 68000 assembler is very rusty

Regards
Gary

[DJBase]

Where can the source code be find? I may can look at it.

[Dimlow]

DJBase, i will pm you

[rkauer]

Originally Posted by Dimlow

@rkauer

Can you explain to me why we need these fast EPROMS ? i have been reading through the Amiga Developer CD and the source codes for the Apollo roms. Before the rom code is executed it is first relocated to ram. I can't see that having a slower rom is going to slow things down at all. This could be proved by writing one of the rom images to a slower rom. I expect it will make no difference at all. or will it ????

As far as i can see, the only reason for the rom, is for the card to be able to identify itself in the boot menu, and to load up the SCSI device if installed, and to patch in the 68060 library if the 060 is installed. There may be other things it does, but 90% of the code in the rom is for the SCSI interface.

Since the Apollo original ROM is 150ns, I still think there is timing issues with slower EPROM. Better safe than sorry, I may say.

BTW: any chance of ripping the SCSI part out? It eats memory doing nothing and installs itself even if no SCSI controller at all! 32Mb is not enough for me, since I use RAD devices, some code (cubic needs 64MB!), and some Total Chaos play . Hmm... 64Mb is half of what I use to have when my 030 was installed.

[DJBase]

there already exist 2 kind of roms. you can make one with (ROM_16bit.32K) and one without scsi-device (ROM_16bit.16K).

Maybe there is a way to support 64mb ram per bank.

[Dimlow]

I have 250 ns eproms here i will give it a test later. See what happens. Oh yea, what stopping us from just reprogramming the ones in the card ?

As for the 64Mb slots ram, yes it looks possible. Very interesting bit of code here "Ram_Einbindung:". looks like it just test some ram, then adds it to system memory. Still not sure the 1240/60 board has all the correct addressing lines connected. Maybe it does, i have not checked, plus at this time , i don't know what I'm looking at! I need to educate myself on this.

[rkauer]

The one in the card is PROM, only that.

[Dimlow]

Oh i see, my one is eprom, oh i lie, one is prom the other erpom

[Dimlow]

Oh well, just programmed an eprom by first reading the original, then programming it to a new one. unfortunately it didn't work. The reason i believe is that i only have 27128 eproms and not 27c128. these are pin compatible i think, but i don't think they are electrically compatible. But then again i could be wrong and it could be the speed of the chips that is the problem. I know i have some other eeproms round here i will have to find some and give it another go.

Also i powered up the Amiga with no rom in the socket and it still works fine. Did a CPU and it listed 68040 as expected. The only problem is that there is no fast ram because the addram routine has not been run in the rom. It would be interesting if somone could write a little bit of software to test memory locations for the existance of ram whilst the rom is removed and a 64Mb simm installed, you may find that that ram is in fact present in the memory map. It could be just that the code in the rom does not add more than 32Mb per simm to the operating systems lists. All that would be needed to add that extra ram is a call to AddMemList with the apropriate parameters. once we had some code that could search for ram and add it to the Memlist you could even run that 1240 card with out the rom installed, so no scsi drivers are loaded either. The down side to this would be that the 060.librarys would also have to be loaded too if you had a 060 installed. But im sure for a very talented programmer this would be no problem.

[Zetr0]

@Dimlow

this is very interesting!

the ability to have more than 32MB per SIMM is awesome!!!!!

[DJBase]

Hm, theres something I don't understand. In rom_teil2.s there is 16M as max.?

[Dimlow]

yes, i saw that, but i thought it was just me. maybe there is something we are missing. what i think it is, is that it allocates in 16 meg chunks ?
;* aktivate_ED liest an einigen Stellen des 16M-Blocks, der bei a4 *
;* beginnt, um das ED-RAM beim Einschalten zu aktivieren !!!

[Dimlow]

OK, its all go here except its not.

Removing the 040 was quite easy with the new KADA 852D+ it took quite a lot of heat to get out. but i managed to do it with no damage to the board.

I have now upgraded my 1240 a 1260. I converted the card i had that was faulty when i received it, but repaired by me and working. But remember i thought i had over heated the 68040 chip that was on it. well i thought it was the 040 that was dead, but i now think that there was some other problem as the now i have the 060 in it and it still does not work.

First suspicions was the mach chips, and bad connections to in the sockets, so i removed the sockets and soldered the chips direct. But still no joy. Its a strange problem, i did have activity in the address and data lines and it looked like the CPU was running, but crashing. Now i have no signals at all, i cant understand it.

Id like to get this board running before i attempt to upgrade the other working 1240 i have, but I'm getting low on patients so may just go ahead with it anyway. But first a few questions.

Does anyone know what exactly are the functions of the mach chips ? I'm assuming its address decoding but not sure.

How is the main processor disabled when the accelerator is running?

I like to determine if these signals I'm seeing on the board are coming from the 020 or the accelerator, how do i do that?

last question should i just go ahead and upgrade the working card ? I soldered the 68060 directly to the other one as i had no sockets. Will it survive the removal, it gets quite hot? or should i get another one for the working board ?

Mmm me need to ponder

Oh, and Zetro, that kada soldering station works very nicely, I was a little worried about the quality of the soldering iron as it looks quits cheep, but it does work well, the tips are good too, but no the best i have had. The hot air is great, i like being able to adjust the air flow and the heat independently. My old old i could only adjust the heat its was only a modeling heat gun, not a real solder station. Id say this Kada is worth every penny, lets hope it continues to work as well as this for the future.

[DoogUK]

Not trying to hijack your thread Dimlow but it seems appropriate to put this here.

I've just had a spell without internet..boredom getting the better of me i decided to upgrade my apollo 040@28 mhz with an 040@40 mhz that i got from ic-china and the new xtal i got from Zetr0.

I got this idea off youtube to desolder the original 040 and thought why not.

First i took all the socketed chips off the apollo then took a plumbers heatproof shielding mat and cut a hole in it so it fitted perfectly around the 040 cpu and protected the rest of the card from heat.

I then took a normal household iron <grin> and mounted it upside down and switched it on full heat setting.

Place the card on the iron so only the 040 cpu's ceramic face sat on the iron..5 mins later the cpu lifted out of the card cleanly and easily.

After cooling the card, around 60% of the cards cpu holes were still blocked with solder so half an hour with a solder sucker and iron and i had a clean desolder job.

Soldered in the new cpu, fired it up and i now have a blisteringly fast apollo 1240.

Interesting thing is...i installed the apollo 040 library and benchmarked then replaced it with the 040 library from os3.9 and rebenchmarked and it was a lot faster with the os3.9 library.

i can now play mp3's @ 128kp and cpu usage hovers around the 60% mark allowing me to use other programs.

I've just been reading through this thread with gret interest and i am excited about the possibility of 64mb of ram in the apollo.

keep up the good work!

Edit: Not only clean but blisteringly fast Bas!

[alexh]

Originally Posted by Dimlow

Does anyone know what exactly are the functions of the mach chips?

Very few people know how they do what they do (in terms of logic) but a high level overview of what they do is they convert 68000 bus cycles into DRAM cycles. They do the Autoconfig etc. They interface to the SCSI chip, clock etc.

Originally Posted by Dimlow

How is the main processor disabled when the accelerator is running?

It's complicated. But surrounds the use of the BOSS signal

[Dimlow]

Cheers Alex, is there any documentation around? I have the schematics, but some reading material would be nice. Also how likely am i to kill the CPU if i remove it and stick it in the other card ? I'm a bit temped to do this now, but i don't want two dead cards.

Dooguk, I'm glad you had success, i just thought since you had the processor out why did you not fit an 060 ?


Edit: Seems this post has been quoted that i have the schematics for the apollo, sorry but i was refering the the schemtics of the 1200 here not the schematics of the apollo 1240, i should have been clear about this. Please no more requests for the apollo schematics, i dont have them and i doubt they will ever come into the public domain.

[rkauer]

Originally Posted by alexh

--snip--

It's complicated. But surrounds the use of the BOSS signal

Me! Me!

First, the original processor boots. Then the new one sends a bus_request and waits a bus_granted. When it receives, the new one sends a bus_granted_acknowledge. After that, the glue logic on the accelerator puts the old CPU into a reset loop.

If I'm not right, my accelerator will delay a lot...