Apollo Mk 3 030 Repair (caution lots of images)

Hello there my fellow Amibayer's

Today I am very lucky to share some work I have been upto.


Over on EAB A chap named Rocs9005, was in need of help, for the last year his APollo card had been intermittent and lots of memory faults.

So I decided that I would help, Rocs sent the card, and the rest is the story.

Problems problems and problems -

When I first tested the card it wouldn't boot, but when I changed the ram to a spare 16mb I have here, she was working perfectly (at 40mhz).... for about maybe 10 minutes... and then the guru's kicked in.

The chip was running on the uncomftably warm side (that means bloody hot) - usually there is a heatsink on the SMD components but this one has been removed - who ever removed it should be hunted down =)

I tried the chip running at 33mhz and she booted with 16mb ram, however it wasnt' 2 minutes before alerts kicked in and guru's

I tried again on 25 mhz and nothing, and now she wont boot with ram in at any frequency.

I tested the unit on a known good 1D4 with 512mb CF card.
__________________

After I inspecteded the card, and I discovered that the second port was put in by a non-third party, probably an enthusast, allthough a little messy, it checks out fine under the optics and mr.continuity meter.

___________________

So what do I do ?

Well theres no doubt that I need to remove the SMD based 030 and fit a socket for a PGA 030.

My plan is simple, after removed the chip and putting a PGA socket in place, I have a spare PGA unit rated for 40mhz that I bought this for my GB1000 mobo, but I wont get to the project for the next couple of months so Rocs9005 can have this for a replacement, no rush =D


Well chaps and chapesses - lets get to work eh?

Fluxin' it up =D

(click for stupidly huge picture)

HUzzah! hot air.... accept no substitute =D

(click for stupidly huge picture)

Ooo they wernt there before

(click for stupidly huge picture)


Remember - Cleaning is IMPORTANT =D

(click for stupidly huge picture)


WHOA!! whats going on there?

(click for stupidly huge picture)

______ Important Thred News break ________

unfortunately in removing the original CPU, two small tracks were damaged. this is mainly due to the age of the solder (brittle), pcb substrate (soft with heat) and my mistake of droping the cpu as I was picking it up off the board =(

Anyway, after a crazy moment - using some fetching commodore blue 30AWG wire - I re-routed the damaged two tracks.

We appologies for interupting the thread program.
_________________________


The Close up of closeup's

(click for stupidly huge picture)

.... so does it work?

Some Jig'd up WHDLoad love =D

(click for stupidly huge picture)


It wasn't all sunshine and roses thought

After replacing the CPU and attempting to run the card at 40mhz, it would NOT boot with ram in.

I tried stick after stick, from 50ns to 80ns, with WAIT-states and without, if there was memory installed it just would not boot.

I had a hunch and so reduced the clock to 33mhz and the card booted but ONLY with ram in the second port.

"The very strange thing is, out of 20 or so sticks I have at hand here, the card will only boot on 5 of these ram sticks, and ONLY with the WAIT enabled."

Perhaps even more peculiar is the fact that the 50ns stick (very fast refresh) wont work with the card!?

This leads me to conclude that the memory controller has been damaged in some way, either when the 2nd port was added, or just by age and CPU over-cooking and putting stress on the memory latching mechanisims.

Alas it is unfortunate as I cannot change the memory controller as its based in a CPLD, (Mach chip) - as such theres little choice but to work-around it.


However it is resurected from a dead running with a PGA MC68030RC40@33MHz + 32MB in 1 stick ram

settings -

RAM = ON
SCSI = OFF
WAIT = ON

stable running demo / whdload games / sysinfo cpu bench and memory 32bit fast benching

Click for larger images

Just under 5,000 Dhystones, not bad for a 33mhz 030


Click for larger images

ripping up the native IDE at 1.6MB a second on CF.... although it eats all the CPU time.


Click for larger images

Yet another speed test! give that CPU a work out


Click for larger images

I used a program calld MIPS, and ran through 100,000,000 iterations, final score 10.23 MIPS = NICE


Very pleased about the repair, it worked solid for 2 days here, after bench after whdload after bench, being powered on and off, the card removed and inserted.

she is now wigning her way back to Rocs9005, whom shares a passion for synths and tracking - good luck my friend +D

Awesome work again Keith. Truly magnificent work.

Kin

Truly impressive stuff there chap, and another piece of Amiga hardware saved from the big WEEE tip in the sky. A coffee well deserved, I fancy.

Drinkins coffee now =D

Zetr0licious!!!!

Hey, change the caps and try again :nod:

This is exactly what I'm doing at the moment.


I've already replaced the simm slot and added the second simm slot.
I'm just waiting on my pga sockets to come in to upgrade it to a 50mhz.

What nozzle did you use for the removal? Also, the flux I use is liquid, is the paste style you are using alot better or should either work?

Or should I just clip all the leads from the qfp processor and then tackle the leads with hot air?

What are the 2 jumper wires for? Did you need to repair a trace or is this needed to enable the pga socket?
Edit.. doh pays to read everything before posting.

unfortunately in removing the original CPU, two small tracks were damaged. this is mainly due to the age of the solder (brittle), pcb substrate (soft with heat) and my mistake of droping the cpu as I was picking it up off the board =(

Click to expand...

Thanks for the post!

later,
dabone

SUPER WORK!
Here I am de-soldering all the bloody time when I could be using flux and hot air (plenty of that from my mouth as many may already know :roll.

Still, 33MHz is better then a 020 @ 14MHz...

You do so love your flux, Mr Z!
Nice work

The postman just brought me 2 68030 pga sockets.

I'll have fun this weekend.

Hopefully I can show pics of my success.


Later,
dabone

MY good friend, is the QFP 030 still functional?

Very nice work! I bow down to your fluxiness.



Great job Keith!

Heather

@Zeets

I bow down to you again big fella, amazing quality of work

TC

That was a slick and amazing restoration pr0n!
I love these threads, despite making me feel very unskillful

Nice job my friend!

Thanks for you kinds words guys (and girl =D),


even though there was a good result, I could kick myself for droping the chip back onto the board - infact I think I did lol!

what had happend after lifting it with tweasers, I angled it up to look at - in doing so it slipped inthe tweesers, catching my little pinky finger, this giving it a good burn from the legs of the chip.

as my had reacted, it tossed the 030 into the air and it can back down onto the softboard, on its angled edge...

It severed one track completely and the other I wasnt sure so I cleaned up the damage, removed dodgy track(s) and re-routed.

I should of used a smaller amount of 30Awg, but I wanted to see it running as soon as it was cleaned and cooled..... yes I am weak!

rkauer said:

MY good friend, is the QFP 030 still functional?

Click to expand...

It wasn't really functional before the operation my friend, I seriously doubt theres much use left in it, either way, I sent it back to Andrew (Rocs9005).

If you need a QFP030 I have a spare here somewhere I shall pop it in you little box I have here, well, when I say little..... =D

I will PM you later what has gone in it +D

Ok, I got home and decided to go ahead with my cpu replacement also.

You scared me with your chip drop so I went a safer route with the hot air. Heat up a few pins and lift, repeat until 3 sides are done then just drop it off the 4th side.

The CPU looked like this when I was done.



Here is the board after removal and clean up with solder wick / flux remover.



Socket installed.



And CPU and ram.



I was having a very bad problem with the RAM also. Running at 50mhz the card will not boot at all. 40mhz sometimes works with ram. As I was testing I noticed the mach chip next to the cpu was getting quite warm.
I've installed a 33mhz and now I've got a 32meg EDO simm in running aibb over and over.

I bet the mach chip is the limiting factor with these cards, this card came originally at 40mhz but won't run with the new cpu at 40. Anyone want to take a stab as to why? Maybe a cap installed to a pin of the qfp thats not routed to the pga?

Later,
dabone

@dabone

some great work there!!!!


the CPLD (MACH chip) controls the RAM and generates a couple of more signals needed for the 030.

I think the limmiting factor will be the RAM, have you tried booting the card with no ram at 40mhz or 50mhz? what refresh (in ns) are you running and are the 2k or 4k refresh SIMMS... length matters lol =D

y'know, thats a really nice and neat job! - i do hope you are proud of your work.


I have been hitting some books / techspecs lately, and its important to mention under-voltage*, what output from your PSU are you getting on the +5v rail with the Apollo in?


*Although for the most part the Amiga can handle a little under voltage (say 4.75 instead of 5v) its quite likely that accelerator cards will not. see if you can get a solid +5v (above the 4.98 threashhold with load) - thing is the higher in the frequency we go the more of a solid *higher* core voltage we require - its possible that to achieve solid operation at 40 or 50mhz, you may need to juice up the power-supply.

I get no boot at 50mhz with or without ram. (Maybe a bad oscillator, I stole it from a 386 board). At 40mhz the machine is not stable.
(Gurus, corrupted my cf card once)

I was seeing no ram until I reseated the mach chip.

At 33mhz the machine is stable even with the wait jumper removed.
I'm running the ram in the primary slot, the secondardy slot lifts the card up and puts pressure on the simm slot when in use.


oh, and good work yourself, you are more of a masochist than I am
building your own socket, I cheated and bought premade sockets.

This was my first qfp removal ever, I just got my hot air rework station
this week. I practiced by doing a c64 repair.

Later,
dabone

Have you changed the old electrolytic SMT capacitor by a new one ?

Zetr0 said:

@dabone

some great work there!!!!

Click to expand...

Seconded! There is a high level of Win in this thread, I'm loving it!

Zetr0 said:

....I have been hitting some books / techspecs lately, and its important to mention under-voltage*, what output from your PSU are you getting on the +5v rail with the Apollo in?


*Although for the most part the Amiga can handle a little under voltage (say 4.75 instead of 5v) its quite likely that accelerator cards will not. see if you can get a solid +5v (above the 4.98 threashhold with load) - thing is the higher in the frequency we go the more of a solid *higher* core voltage we require - its possible that to achieve solid operation at 40 or 50mhz, you may need to juice up the power-supply.

Click to expand...

Isn't this what we suspect was wrong with my ATX supporting my PPC and Bvision?