VOGONS


Reply 60 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

I have chips for three modules coming, and then 5 PCBs. So if someone has a system, and needs one, there is a chance, but first we gotta see if it works at all 😁

On my Globalyst 510, the 'vlb' socket is in no way under the floppy drive, so clearance was not an issue

Reply 61 of 121, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie
rasz_pl wrote on 2024-02-12, 15:43:

2 RetroSpector78 also got Globalyst 575 system, its another similar but not exact same board with VLB cache slot https://stason.org/TULARC/pc/motherboards/A/A … BALYST-550.html

Globalyst 575 has the same mainboard as Globalyst 550 (not made by FIC). It has the 2nd version of the VLB-cache-stick that I showed some posts earlier.

The attachment ATT-Globalyst_550_575.jpg is no longer available

Reply 62 of 121, by rasz_pl

User metadata
Rank l33t
Rank
l33t
majestyk wrote on 2024-02-12, 16:51:

Globalyst 575 has the same mainboard as Globalyst 550 (not made by FIC). It has the 2nd version of the VLB-cache-stick that I showed some posts earlier.
ATT-Globalyst_550_575.jpg

You mentioned it, but I dont remember seeing this picture. Could you please link/post back side too? Do you happen to know the model of mobo in 550? nvm found it https://theretroweb.com/motherboards/s/atandt … globalyst-550-c interestingly documentation calls cache slot SL2 there, while GAC-2 calls it SL1 so they did mark it differently so no one makes the mistake of plugging wrong Cache module :---)

Last edited by rasz_pl on 2024-02-19, 17:01. Edited 1 time in total.

AT&T Globalyst/FIC 486-GAC-2 Cache Module reproduction
Zenith Data Systems (ZDS) ZBIOS 'MFM-300 Monitor' reverse engineering

Reply 63 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

Chips get:

The attachment IMG20240215133352.jpg is no longer available

(That's some long srams 😆 )

Still waiting on the PCBs, but that's due to the Chinese lunar new year.

Reply 64 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

Pcb's arrived

The attachment IMG20240227123940.jpg is no longer available

Looks like the fingers are covered in mask 🫤
Looking back at the renders, the fingers are under the mask there too... Oops, looks like we all missed that little detail 🫣

Width seems ok, but it's too thick as it is to be able to push it down in the slot

The attachment IMG20240227124314.jpg is no longer available

Will have to see what can be done to get the fingers exposed and make it go in the slot.

Reply 65 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

Tried a few different things to remove the mask, and sanding worked the best, so I'll use another of this batch to do it nicely.

Then I'll need to do a bit of beveling, to see if it will go in the slot any easier. I should be able to do this tomorrow.

If it goes in, then I'll get it soldered, to prove if it works or not. This is targeted for Friday, provided it fits.

Reply 66 of 121, by rasz_pl

User metadata
Rank l33t
Rank
l33t
RockstarRunner wrote on 2024-02-27, 10:50:

Looks like the fingers are covered in mask
Looking back at the renders, the fingers are under the mask there too... Oops, looks like we all missed that little detail 🫣

front%20render.png
we all looked at it and went yes that looks perfectly normal 🤣 at least thats what I tell myself to deflect some of the blame 😜
Mask cutout missing on all Edge Connector pads 😳
Now that Im having a second look CT cap footprints might be too large, those were provisioned for big tantalums, are the caps you bought fitting fine in there?
Are SOJ SRAM footprints ok? Pads look awfully small to me 😀
Im not loving edge connector middle gap being this wide, leaves possibility of bad alignment 🙁 I think I used this https://www.eevblog.com/forum/vintage-computi … 450/#msg3592450

The attachment vlb-edge-conn.png is no longer available

but not sure if I followed it to the letter, I remember fiddling with it, measuring ... mine is even smaller than this drawing 😒 1.805 while drawing demands 1.86-1.96 I think I measured original PCI slot gap and it was way smaller than this so I made mine somewhere in between. Ill change final project to tighter PCI gap, VESA VLB technical docs be damned.

RockstarRunner wrote on 2024-02-27, 10:50:

Width seems ok, but it's too thick as it is to be able to push it down in the slot
Will have to see what can be done to get the fingers exposed and make it go in the slot.

VLB slots are very tight, extra mask will definitely exacerbate it. Beveling will also help. I read n some PCIE card github suggestion to make yourself a 45 degree jig before sanding for that perfect even angle 😀

AT&T Globalyst/FIC 486-GAC-2 Cache Module reproduction
Zenith Data Systems (ZDS) ZBIOS 'MFM-300 Monitor' reverse engineering

Reply 67 of 121, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

The thickness of the original module is 1.6mm , the Tantalums are 5.5 x 3.5mm.

Reply 68 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

Regarding the divider, I would not suggest making it smaller, as it appears to get wider as you go deeper in to the slot, and I think it's about right (non scientific assessment)

The attachment IMG20240228101030.jpg is no longer available
The attachment IMG20240228101111.jpg is no longer available

After much sanding, the board just about goes in, but I would describe it as nearly a death grip. You have to see saw it in, as if you try and push it straight down, it refuses to go, and the mother board just flexes.

The attachment IMG20240228101327.jpg is no longer available

I did manage to get it out, being careful.

Reply 69 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member
The attachment IMG20240228102243~2.jpg is no longer available

The srams and buffers look correct, though soldering may require finesse, this remains to be seen.

The modern replacements for those tantalums are clearly much smaller, and will need some change.
Also, does it matter that they aren't Tantalum caps now?

Reply 70 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

The prototype is under construction... Watch this space!

The attachment IMG_20240228_191008.jpg is no longer available

Reply 71 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

Bad news, the PC refuses to start with the module inserted. No even a hint of activity.

We checked all the soldering and didn't find any issues.

Then we tried switching jumpers from 256k to 128k, and it didn't make a difference.

We then put a probe on the CT1 and saw no activity, like 5v line never comes up.

Run out of time to investigate further, but I think we need to recheck things.

Smells like a short between 5v and ground, and PSU is shutting down. Suspicion is currently if we have the slot pin out correct, as the module did not have any unexpected shorts.

My friend who is helping with soldering work, will probably clarify.

Last edited by RockstarRunner on 2024-02-28, 20:14. Edited 1 time in total.

Reply 72 of 121, by miisalo

User metadata
Rank Newbie
Rank
Newbie

Just small clarification, I added wires to 5V and GND just using CT1 as soldering point and used oscilloscope to check what happens with powerline as computer didn't react at all and it looked like 5V started to come up and shut down immediately. My brain is currently lacking any blood circulation after hard training, so I might have missed something but module itself looks good now.

Reply 73 of 121, by miisalo

User metadata
Rank Newbie
Rank
Newbie

Measured motherboard against schematics:

B56 - 30 Ohms to the ground this is 5V not GND like in schema
B52 - 0 Ohms to the ground, this is GND, not 5V like in schema
B48 - 30 Ohms to the ground this is 5V not GND like in schema
B43 - 0 Ohms to the ground, this is GND, not 5V like in schema

The attachment fix.png is no longer available

So did this dirty fix

The attachment IMG_8559.jpeg is no longer available

And the result is booting machine with cache

The attachment IMG_8557.jpeg is no longer available

This is probably fine as it is but going to add some wires there.

Reply 74 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

We did it!!!
.... Now I sleep

Reply 75 of 121, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

Puuh - happy I got it right in post "2024-02-05, 17:15".

This should be easy to fix in layout "Rev. 1.1".

Congrats to everyone!

Reply 76 of 121, by miisalo

User metadata
Rank Newbie
Rank
Newbie
majestyk wrote on 2024-02-28, 21:25:

Puuh - happy I got it right in post "2024-02-05, 17:15".

This should be easy to fix in layout "Rev. 1.1".

Congrats to everyone!

Would have helped if I had looked back the thread when I realized that there was short between 5V and GND. And of course rockstarrunner sent me the link to that post after I said I found at least one mixed 5V/GND, but I was just measuring every effing pin just then and did not notice. Why do things easily if you can do them hard way, that is where all the fun is 😁

Reply 77 of 121, by rasz_pl

User metadata
Rank l33t
Rank
l33t

👍 Happy new farther feelings 😁

miisalo wrote on 2024-02-28, 21:41:

short between 5V and GND

I was expecting this being the most difficult of the whole endeavor Re: FIC 486-GAC-2 and mystery proprietary cache module - Help needed :

rasz_pl wrote on 2024-02-05, 01:56:

Ironically I think power and ground on the cache connector itself are the only things I wont be able to deduce from images, every 4th pin is a ground stich, but might be power.

Luckily it was just few pins and not the whole thing flipped around 😀
edit: yep Re: FIC 486-GAC-2 and mystery proprietary cache module - Help needed majestyk was on point and I was confused 😀

miisalo wrote on 2024-02-28, 21:41:

but I was just measuring every effing pin just then and did not notice.

I was just thinking right now that tinted vias is not a good idea on a prototype 😀

miisalo wrote on 2024-02-28, 21:15:

Measured motherboard against schematics:

thank you, will incorporate in final version.
You want to hear something funny? While routing b48 (blue) it looked very wrong to me 😀 Wrong to the point I started scratching my head and questioning original designer. b48 trace sharply turns left before stitching to the ground plane, and the via it uses is right next to a50 track (orange) which is also supposed to be ground. Why would anyone design it like that and not make a50 track touch the via, why not just turn right instead (where the cursor is)?

The attachment b48.png is no longer available

but and the end I just deferred to original design and moved on 😐

miisalo wrote on 2024-02-28, 21:15:

And the result is booting machine with cache

Time for some benchmarks, cachechk and doom (fastdoom is easier to benchmark). Im really curious about 128/64KB difference, and if 64KB setting works at all.

miisalo wrote on 2024-02-28, 21:15:

This is probably fine as it is but going to add some wires there.

yeah, I wouldnt bother with wires, this excessive power/ground stitching is for lowering impedance, wires will probably make it worse.

Could I please trouble you for few photos of assembled module, both sides lose and in its final resting place installed in motherboard inside the case?

AT&T Globalyst/FIC 486-GAC-2 Cache Module reproduction
Zenith Data Systems (ZDS) ZBIOS 'MFM-300 Monitor' reverse engineering

Reply 78 of 121, by supercordo

User metadata
Rank Member
Rank
Member

Nice Work!!! How much for one?

Reply 79 of 121, by RockstarRunner

User metadata
Rank Member
Rank
Member

I will benchmark it this evening in 256k mode, and check the 128/64 modes when I get the chance.

I'll also post the photos you asked later today.