Magnum TTRAM 256MB boards for the TT 030

[tenox]

Does anyone have a design that uses soldered on memory chips rather than off the shelve SIMMs? If so I could produce it.

[Guest]

we are on the case

ill clean up the stbook repairs this week and stram board 2-8s and ill be back to the magnum

we will fix this
ill decompile the code in the next few days to a simple state based model
so we can then use any platform of ic or ic's....

im just very sad to have the proto board back
as it works here with a pair of 64MB's
i never had 128 pairs to work
with any magnum
with the code from the net as supplied to me...

you never know,.
make our own code we will!! from now on

[Guest]

board rules i found

set the board to 1 wait state
add your ram
set to auto
all jumpers on
except wait is off
let the machine boot
does it yes cool try two banks
no
half the ram try again
if it boots

let it boot up and exit and close down
this saves settings to nvram TT or FALCON

now repower and the older saved nvram settings {background} are applied to the new boot

build on speed

[Guest]

hi tenox...

ok having a good look i see one type of DRam you used is 3.3v i/o its supplied via a regulator to give 3.3 to the chips
so is no good
the other type is ps2 5v ram however its 32X36 parity ram
and wont work properly

so as i say i tried my pair of 64MB ps2 non parity 5v fpm works fine ... i used auto mode
and a rev J TT
ill try match up a pair of 128s lets see

ok i have a cunning plan
http://www.alldatasheet.com/datasheet-p ... 6100B.html
on the parity pair the voltage is 5v i checked the ics KM44C16100AK-6 4k refresh cycle A0-A11
the bad type would be the 8k refresh KM44C16000AK-6 this type has the need of A0-A12
and they are good i think ...{try and see}

so ill look for the actual simms datasheet and see what the non parity type is setout
chances are i can just remove the 8 parity ics per simm ill clip them off and tidy the pads
and not try to hot air them off and risk damage...
to give a perfect pair of 128 non parity 32x32 etc...

so is it ok if i alter them
as they arent mine i thought i must ask first..??

[tenox]

Simbo thanks for looking at it. Much appreciated.

Yes... please do whatever it takes to the SIMMs.

[Guest]

ok all done and the chips removed
pads cleaned up

so ill find a vga cable etc and setup the TT for a test... try find a sysinfo floppy...

[Guest]

eventually found the datasheet for the simm so it wasnt a good idea to remove the parity chips
its wired a different way than usual

anyway ive ordered a pair of the correct 128MB simms
72pin 4K refresh FPM 60ns 5v i/o and supply no regulator on the simms!!
no parity 8 chips per side
the guy has over 10 availible and ill test then post details in a week or so..
quite cheep £25 a pair + £12 post they use the same ic's as the two above
just wired up right...

the chip type works aok i know it does because the pair of i use uses the same samsung ic's
they are one sided with slack pads works aok as 128MB availble as a pair
so what ill do is make my 64 pairs into 128's and move ics over from one of these
so they wont go to waste

[Guest]

http://www.datasheetarchive.com/dlmain/ ... og_526.pdf
found this handy guide for dram ..

here is the pair i will order tomorrow

http://www.ebay.co.uk/itm/128MB-MEMORY- ... 1e67fbc630

the guy has plenty of them @ $20 each a fair price but $20 bucks postage

so.. ill be ordering one pair

till i instead use other than the us postal service
ends up costing a packet import tax.. there is a better way

[tenox]

If they are good and we need more I'll order and organize shipment to UK.

[Guest]

del

[Guest]

If they are good and we need more I'll order and organize shipment to UK.

the simms arrived fast

so ill set up the TT and check out what they do

fingers crossed

[Guest]

http://www.datasheetarchive.com/KM44C16 ... sheet.html

as usual with ebay
i get the dram then discover that the chips are EDO
and not as advertised and by the type quoted
so i am going to send them back too the guy by the worst path and that costs the most
terrible
ill keep looking
i hate these people
so avoid this seller http://www.ebay.co.uk/itm/128MB-MEMORY- ... 1e67fbc630

perhaps

im very unhappy but what can i do about it
order another pair i recon ill need to hunt and ask exact chip number details that will be sent

or dont pay
their is always one... or more terrible dealer

you never know they may send me the right pair yet,..
i dont mind paying for what i get but man,,, terrible
never had bad ram from the states before,.

however its 5v and edo but some types of EDO work with the magnum
as its 5v and checked the i.o and supply are 5v
its no harm to try it out

ill report back

but return it is an option and no damage done anyway
oh well as pakman can help now and ill draw up the board to include the flash socket {3 hrs tops from where i am}
ive decided too keep this pair of EDO as they will be handy to get 5v edo to work also with the fixed core
and flash system dev
as it should
infact EDO should work better as its by nature nibble mode and complient with burst mode ...

the simm board does have the pads for a regulator but the ram is 5v and 5v i/o but EDO not FPM
but may work... yet

the PD 1-4 are all set high

so auto mode should work from pd 1 and 2
3 and 4 are ram type ...

[pakman]

however its 5v and edo but some types of EDO work with the magnum
as its 5v and checked the i.o and supply are 5v
its no harm to try it out

ill report back

Have checked the datasheet, should work fine.
So what is your test result?

[Guest]

well jumper set simm PD i get 131MB with the pair of 128 EDO 32X32 5v {all jumpers on} and the mem test shows '=1027'
with simm PD set bypassed i get 64MB from the pair {all jumpers off}
a pair of 64MB FPM give me 128MB and shows the same.... memory test details
all test done in wait state 0 rev j ,,,

i used the board proto i made ages ago and sent over to tenox....
made from the older rework i did last
ill check out what the logic prob says about the pd lines and post tomorrow
rule out bad PD resistors on the simm type seen this before

i get exactly the same result using an original magnum with original controller
infact i tried two different ones
starting too think an address line isnt right ... as it runs perfectly stable @ 131MB
its gota be the code.... defo memory addressing or more over auto detect and address decoding
gona study the design a bit and see if i can spot anything up like a missing tag etc ....
gota be something easy now...
looking at the M address select i see A24 is selected with A25 to MAD 11 {bit 12}
row low = A25 row high= A24 selected for MAD11
11 I/os for Address input i assume for interpliration
I suspect a high up address line by the way it behaves
I also see A24 goes to pin 40 I/o 28 pin A25 goes to pin i/o 0 on a different bank
i assume its used as some sort of banking enable...
gona look at the code and see what this input does

draw a few diagrams

gona get an early night im bushed...

so will i go ahead and route up the rest of rev1 with flash socket etc...
?? any more jumpers needed ?? perhaps put a pad on the spare i/os and mount a connection pin ?s etc..

i can do it and get a batch of four made for around the same cost as one... {slack comes cheep}

perhaps some others wish to help fix the edo issue.. and size issue
as this type is far easier to get than hens teath FPM big non parity

[pakman]

well jumper set simm PD i get 131MB with the pair of 128 EDO 32X32 5v {all jumpers on} and the mem test shows '=1027'

1. Under the above condition, please check these 2 (undocumented) system variables:
- LONG $5A4 (ramtop)
- LONG $5A8 (ramvalid)

2. What happens with only one 128MB SIMM installed?
- Try different jumper settings
- Every time check LONG $5A4 (ramtop) and LONG $5A8 (ramvalid)

[Guest]

well jumper set simm PD i get 131MB with the pair of 128 EDO 32X32 5v {all jumpers on} and the mem test shows '=1027'

1. Under the above condition, please check these 2 (undocumented) system variables:
- LONG $5A4 (ramtop)
- LONG $5A8 (ramvalid)

2. What happens with only one 128MB SIMM installed?
- Try different jumper settings
- Every time check LONG $5A4 (ramtop) and LONG $5A8 (ramvalid)

found this detail
ramtop LONG 0x5a4 Pointer to the end of the Fast-RAM in Atari TT (not officially documented).
ramvalid LONG 0x5a8 Magic value that shows whether ramtop holds a sensible value. Must be 0x1357bd13. Not officially documented.

[Guest]

well jumper set simm PD i get 131MB with the pair of 128 EDO 32X32 5v {all jumpers on} and the mem test shows '=1027'

1. Under the above condition, please check these 2 (undocumented) system variables:
- LONG $5A4 (ramtop)
- LONG $5A8 (ramvalid)

2. What happens with only one 128MB SIMM installed?
- Try different jumper settings
- Every time check LONG $5A4 (ramtop) and LONG $5A8 (ramvalid)

found this detail
ramtop LONG 0x5a4 Pointer to the end of the Fast-RAM in Atari TT (not officially documented).
ramvalid LONG 0x5a8 Magic value that shows whether ramtop holds a sensible value. Must be 0x1357bd13. Not officially documented.

ill set up a boot floppy tomorrow as the hdd is a bit old scsi on this TT

[Guest]

ok ill spend tonight and dump off the values and experiment
came across frankB on my way home so now know what too do.. !! thanks
it amazes me when i find the exact person to tell me out the blue on a bus jorney
he is a mega walker so i had to get off the bus a stop before and hijak him.....
... needle in a haystack....

lets see what i find as ive made a boot disk with some tools on it sysinfo etc
as he recons sysinfo dumps all this info i didnt know this never needed this detail before
i always ask for help when i need it

i think 1 simm on its own perhaps wont work
it has a note about it at the bottom of the texts for the magnum
ill get too it in the morning

sore head and due too the cold ive developed the flu
so bad a summer end of june and its <15deg at night
worst summer ever and the trees and plants are more than a month late to bloom

so off to bed after a very hot bath...

eek... middle of summer its a muddle of low temp good for bugs
bad for health coughing like a drain

[Guest]

ive broken the code to blocks for each core

so ive also written the jed code block fuses out in bytes rather than bits


too tired too do more now but easy to map more easily now

couple of days..

[pakman]

Hello simbo,

If you don't find the time for further testing, I can do.
Just send the Magnum + the 2 SIMMs to me.

[Guest]

its ok
ill do this this evening
been busy with stbooks now done till i get some ic's

so later tonight ill post the results i also found another pair of 128's behave the same
but are stated as 3.3v type of ic with 5v supply
looking at the ics they are dual 3.3v and 5v i/o and 5v supply fpm
so should work

ill look at it again in an hour or two after food {and a fly whiskey}

[Guest]

i have version 8.34 of sysinfo and it doesnt display these variables
who has a small prg to display them or a later version of sysinfo...??

ive checked the PD jumpers

with all jumpers on its 0000 as expected
however with the jumpers off
its 0101 this is wrong
it should be 1111 ??? sure im right???


with one simm i get 64MB all jumpers on and 32MB all jumpers off
the logic level is 0010 so not 1110 as i would expect...

so i suspect i need 1111 to get the full 128mb each...

to this effect ill do this and bypass the PD lines
too supply 1111 to the pins on the mach

see what happens

sounds like just bad PD setup on this pair

the machine is perfectly stable with 64 single or a pair of 128s seen as 128

so i suspect this so.. ill check it out now... and report back

not sure seems strange...

[Guest]

HA good news

jumper 1234
on my own magnum {old one } i simply cut the PD lines from the jumpers too the simms a long time ago

so they just tell the chip what too do and the simms dont matter what they are set too
i already added this mod to the rev 2 board

i now have 1 1 1 1 at the pd select {5v 5v 5v 5v }
and i get 256MB no problems

so the ram at the last link i posted is AOK EDO
http://www.ebay.co.uk/itm/128MB-MEMORY- ... 1e67fbc630

so go buy some now... as they are quite cheep and arrived here in 4 working days from the states and no tax

attached is the proof



so this leaves us with the problem of what to do
will i just wire up the board as i posted above
or will i just make a run of the finished rev 2 board that has the added jumpers to disable what the simms PD lines say
???
the question packman is ...

if i wire up the latest revision with the flash socket and the added disable pd detect jumpers etc
can i still plug in the older mach 210 and it will work as is with the jumpers and its 1.5 firmware or will i need to modify the older code
to accept the revised i/os dont fancy waisting the ics or getting them flashed again

not sure?? need to clarify
if this is the case ill wire the board tomorrow and get it produced next week

[Guest]

If they are good and we need more I'll order and organize shipment to UK.

texon ill wrap up and send back your unit tested and working tomorrow

one of the pairs you sent the 5v type that is 3.3v i/os works fine

it uses the same pcb as the pair i got and had the same two zero ohm links to remove

all done and tested

what revision of TT do you have again
if its rev d-k leave the wait state jumper on
if not remove it

im just sad this is the last few day ill have this board
so no need to worry more its on its way back working!!

seems since i stopped working ive got more done in the last 6 days than in the last 6 months

[pakman]

so this leaves us with the problem of what to do

if i wire up the latest revision with the flash socket and the added disable pd detect jumpers etc
can i still plug in the older mach 210 and it will work as is with the jumpers and its 1.5 firmware or will i need to modify the older code

MACH210 with 1.5 firmware will work in rev.2 boards.
Even if we combine WE1 and WE2 into WEX.
No modifications to the MACH code required.
Just set the Jumper for MACH, that's all.

will i just wire up the board as i posted above
or will i just make a run of the finished rev 2 board that has the added jumpers to disable what the simms PD lines say
???

Go for the rev.2 board!

The pdf attachment is a proposal for the jumper block, will save space.
(left side: current circuit; right side: my proposal)

MagnumTT Jumper.pdf

Something more to consider:
At each SIMM there are following connections:
Pin 44 connected to pin 34
Pin 45 connected to pin 33
Can you manage to route these connections at the bottom side of the PCB?
I want to have the option to isolate pins 33 and 34 by cutting the lines, even on a fully assembled PCB.
This opens the chance to use SIMMS with 8k refresh or with more than 256MB in total (with some additional wiring..)