Blizzard PPC Repair Process

[BYTEMAN]

Hi everyone,

About half and a year ago I owned faulty Blizzard PPC accelerator board. It had 040@40 and PPC@240 onboard and 128Mb ram. PCB was bent, some traces were broken because previous owner tried to desolder 040 with heatgun hehe I understood that repair of this shit will be a big pain, but I'm risky and decided to buy it for approx. 120usd. I believe that was the only PPC accelerator in Belarus (and now maybe too)

So, first of all I've desoldered 040, oscillator sockets and SIMM slots. Then I've put board to coocking oven (to fry some chips haha) as I knew that bent PCBs can be restored that way (I've restored some PCB and Xbox360 mobos that way). 80 degrees C, for 4 hours were enough. Board became flat.

Then I've inspected board with microscope, fount two broken traces and restored them.

After that, I've soldered acquired 060 socket to it, put 040 in it and tried to start my 1200. No luck, no start. Worth to mention that this was my first 040+ accelerator in my life (I've had 1230Mk4). So I didn't know that sockets for 060 and for 040 are "a bit" different So I've removed some pins from second 060 socket (bought to build A3660 along with handmade A4000). Soldered missing pins and...... CARD ALIVED!!! Yes, I was able to boot system, WhichAmiga showed me 266MHz PPC and 40MHz 68040. But it was way too hot of course

After watching Starstruck on real Amiga for the first time, I've found that board cannot boot anymore. I see Kickstart "flying diskette" when removing HDD, but cannot boot workbench, system completely hangs.

Further investigation showed me completely cracked balls under PPC (it is seen even on photos). Here I want to stop and tell some investigations I've found on this accel. board. When PPC is desoldered (and when it had cracked balls), PPC board was not functioning well. I was able to boot sytem only when SIMMs are not installed. With SIMMs, first access to FastROM (I believed so) screwed system and everything was hung. I dont know if it is a common operation of PPC accelerator (I know there are people who have PPC boards with PPC part not working, in my case without PPC card was not functioning properly).
So, PPC was reballed and successfully soldered back. Card alive again, hoorah, that could be the end, BUT......

As you know, full 060 are quite expensive now, so for me it was just a dream to get 060. I've tried to use PPC/040 in a classic Amiga case, but very hot operation....

One day a good man gave to me full 060 Rev5 chip!!! Dreams come true

Okay, I've installed 060 to BPPC, thanks Chucky and other people describing how to do that. Okay, installed, trying to boot for reflash, and ..... dark power led. Putting DiagROM gave solid start, 060 was detected well.

But there were issues with FastRAM. "Complete memorydetection" hangs on "Detecting z3 memoryspace". "Test detected FastRAM" also hangs without printing current address (but printing that "detected z3 ram 1000000kb"). Ok, my first assumption was that some traces related to 060 are broken on this board. But I got information, that many-many years ago this board was upgraded to 060 and then back to 040 because 060 was not working from initial... So non-working 060 was not my fault hehe)))

Behavior is the same with or without SIMMs: Kickstart cannot start (dark power led, dark grey screen), DiagROM starts but hangs on any operations with testing FastRAM (with or without SIMM I get the same behavior). So I've tried to use "Manual memory edit" function of DiagROM. Starting from $40000000 address it seems that BPPC FastRAM starts there. And I'm getting very interesting results. When trying to write anything to ram, I get following: for each byte of longword, bits 0-3 are always written as 'b0, but bits 4-7 are shifted one bit left. Read seems to work well because random RAM data after poweron contains here different data. For example, I'm typing $11, but getting $20, typing $FF gives me $E0, typing $77 gives me $E0.

My assumption is that latches between FastRAM and 060 (74FCT2543, four chips near 060) are not latching in a proper way. It seems VERY STRANGE that data gets shifted, I cannot imagine what kind of fault may lead to such thing.

I would like to ask, maybe someone working with BPPC repairs has struggled with the same or similar issue and has some information, how to fix it? Currently my plan is to check if without SIMM instlled, data can pass well from SIMM pins to 060. Then I think it might be useful to desolder those buffers and check how it behaves.

PPC oscillator is 66MHz with x3 multiplier, 060 oscilltor is 50MHz.

Here are some photos: https://photos.app.goo.gl/zA9B9eJn5rfjcPpy5

Thanks,
Alexander

[Leon Besson]

Bromigo! 1st good work on the BPPC repair.

2nd. But I don’t wish to bring politics into this. < Political content removed >

[SpeedGeek]

@Thread

Political discussions are not allowed on EAB. Please see the forum rules:

https://eab.abime.net/faq.php?faq=ea...faq_eab_banned

[Leon Besson]

Originally Posted by SpeedGeek

@Thread

Political discussions are not allowed on EAB. Please see the forum rules:

https://eab.abime.net/faq.php?faq=ea...faq_eab_banned

Fair enough Bromigo! But shouldn’t you guys be putting some sort of ban hammer down on those who are clearly showing as from these parts of the world? Just saying Mang!

[Kin Hell]

@ OP

There isn't much this guy doesn't know about these boards....

Sent it to stachu100

[BarryB]

He couldn't repair my PPC, think the MACH chips were faulty and all the info about them was lost by DCE I believe?

[Kin Hell]

Originally Posted by BYTEMAN

Hi everyone,

About half and a year ago I owned faulty Blizzard PPC accelerator board. It had 040@40 and PPC@240 onboard and 128Mb ram. PCB was bent, some traces were broken because previous owner tried to desolder 040 with heatgun hehe I understood that repair of this shit will be a big pain, but I'm risky and decided to buy it for approx. 120usd. I believe that was the only PPC accelerator in Belarus (and now maybe too)

So, first of all I've desoldered 040, oscillator sockets and SIMM slots. Then I've put board to coocking oven (to fry some chips haha) as I knew that bent PCBs can be restored that way (I've restored some PCB and Xbox360 mobos that way). 80 degrees C, for 4 hours were enough. Board became flat.

Then I've inspected board with microscope, fount two broken traces and restored them.

After that, I've soldered acquired 060 socket to it, put 040 in it and tried to start my 1200. No luck, no start. Worth to mention that this was my first 040+ accelerator in my life (I've had 1230Mk4). So I didn't know that sockets for 060 and for 040 are "a bit" different So I've removed some pins from second 060 socket (bought to build A3660 along with handmade A4000). Soldered missing pins and...... CARD ALIVED!!! Yes, I was able to boot system, WhichAmiga showed me 266MHz PPC and 40MHz 68040. But it was way too hot of course

After watching Starstruck on real Amiga for the first time, I've found that board cannot boot anymore. I see Kickstart "flying diskette" when removing HDD, but cannot boot workbench, system completely hangs.

Further investigation showed me completely cracked balls under PPC (it is seen even on photos). Here I want to stop and tell some investigations I've found on this accel. board. When PPC is desoldered (and when it had cracked balls), PPC board was not functioning well. I was able to boot sytem only when SIMMs are not installed. With SIMMs, first access to FastROM (I believed so) screwed system and everything was hung. I dont know if it is a common operation of PPC accelerator (I know there are people who have PPC boards with PPC part not working, in my case without PPC card was not functioning properly).
So, PPC was reballed and successfully soldered back. Card alive again, hoorah, that could be the end, BUT......

As you know, full 060 are quite expensive now, so for me it was just a dream to get 060. I've tried to use PPC/040 in a classic Amiga case, but very hot operation....

One day a good man gave to me full 060 Rev5 chip!!! Dreams come true

Okay, I've installed 060 to BPPC, thanks Chucky and other people describing how to do that. Okay, installed, trying to boot for reflash, and ..... dark power led. Putting DiagROM gave solid start, 060 was detected well.

But there were issues with FastRAM. "Complete memorydetection" hangs on "Detecting z3 memoryspace". "Test detected FastRAM" also hangs without printing current address (but printing that "detected z3 ram 1000000kb"). Ok, my first assumption was that some traces related to 060 are broken on this board. But I got information, that many-many years ago this board was upgraded to 060 and then back to 040 because 060 was not working from initial... So non-working 060 was not my fault hehe)))

Behavior is the same with or without SIMMs: Kickstart cannot start (dark power led, dark grey screen), DiagROM starts but hangs on any operations with testing FastRAM (with or without SIMM I get the same behavior). So I've tried to use "Manual memory edit" function of DiagROM. Starting from $40000000 address it seems that BPPC FastRAM starts there. And I'm getting very interesting results. When trying to write anything to ram, I get following: for each byte of longword, bits 0-3 are always written as 'b0, but bits 4-7 are shifted one bit left. Read seems to work well because random RAM data after poweron contains here different data. For example, I'm typing $11, but getting $20, typing $FF gives me $E0, typing $77 gives me $E0.

My assumption is that latches between FastRAM and 060 (74FCT2543, four chips near 060) are not latching in a proper way. It seems VERY STRANGE that data gets shifted, I cannot imagine what kind of fault may lead to such thing.

I would like to ask, maybe someone working with BPPC repairs has struggled with the same or similar issue and has some information, how to fix it? Currently my plan is to check if without SIMM instlled, data can pass well from SIMM pins to 060. Then I think it might be useful to desolder those buffers and check how it behaves.

PPC oscillator is 66MHz with x3 multiplier, 060 oscilltor is 50MHz.

Here are some photos: https://photos.app.goo.gl/zA9B9eJn5rfjcPpy5

Thanks,
Alexander

Originally Posted by BarryB

He couldn't repair my PPC, think the MACH chips were faulty and all the info about them was lost by DCE I believe?

I know I didn't miss you saying stchu100 had already looked at it....

Sounds like.... Game over then eh!

[BarryB]

Yeah, it doesn't recognise any installed RAM

[jbenam]

Originally Posted by BarryB

Yeah, it doesn't recognise any installed RAM

I had the same issue and it turned out to be microcracks in one of the internal layers. I just didn't have much hope for the card left so I thought of doing the unthinkable, reflow the entire card. And well, it worked It then proceeded to properly detect RAM.

The BPPC is a very complex card with plenty of layers, there is only so much you can do externally.

[cpiac64]

check 2 chips near xtal

[BYTEMAN]

Hi everyone,

Finally I've made it work with 060!!!

Let me describe what was checked and what was the issue. First, my assumption was that data buffers between SIMM and 68k bus are doing some mess. I've checked it with scope and saw that when I perform writes to SRAM (using "Manual memoryedit" feature of DiagROM), then data is passing well to SIMM lots (I've checked using FF pattern and see that every byte out of 4 performs write 'b1 to all bits). It seemed to be OK.
So, but after write there were different data... BINGO! Let's check WE! Ok, WE is coming from MUX under 68k XTAL, input 1 is 68k R/W pin, input 2 is +5V. Checking with dual-channel scope, seems to be OK, MUX is working. Switching signal is coming from one of CPLDs. But there are two consecutive WE triggering on SIMM..... Let's check what data is coming during second WE. Ok, CPU overwrites normal data with wrong ones during second WE.

Hmmmmm..... So something is wrong with Write cycles of CPU... Looking through 68060 datasheet, I've fond information about "Extra write hold" mode set using IPLx pins. Whow, that is the difference from 68040!!!!! For 68040 and 68060 IPL pins during Reset phase are doing different things! Found information for Cyberstorm MK1, where it is needed to modify IPL behavior when changing CPU to 68060....

Well, let's check what's on IPL during reset. All 'b1!!! IPL is driven from cPLD, no MUX or some other glue logic.... So, it seemed that accelerator was not working on 060 mode....

BINGO!!! 060 JUMPER!!! Checked it, there was short between trace coming to pad, and another pad using for PPC multiplier. It was under resistor, and I dodn't tuch that place before (so it came this way from previous owner). Short eliminated, power on - NO DATA CORRUPTION!!!!!!!!!!

So, here the story ends, I've successfully reflashed ROM to 060 version, launched Starstruck, checked PPC operation using some demos from WarpOS distribution.

I hope that my story might be useful for someone

Thanks,
Alexander

[Jope]

Originally Posted by BYTEMAN

I hope that my story might be useful for someone

Thanks for your report, this will most likely be useful to someone in the future. :-)

[BarryB]

Originally Posted by jbenam

I had the same issue and it turned out to be microcracks in one of the internal layers. I just didn't have much hope for the card left so I thought of doing the unthinkable, reflow the entire card. And well, it worked It then proceeded to properly detect RAM.

The BPPC is a very complex card with plenty of layers, there is only so much you can do externally.

I'm not that advanced to do reflow, nor how to do it!!

Originally Posted by cpiac64

check 2 chips near xtal

What exactly should I be looking for?

[cpiac64]

tomorrow send more info


edit

verifidy chips 257 under the xtal

[jbenam]

Originally Posted by BarryB

I'm not that advanced to do reflow, nor how to do it!!

You just need a temperature-controlled hot air station - cover the plastic parts with foil, and give everything a good heating at around 350°C, while circling around the board. Probably best to keep it as a last chance when everything else fails though, these old cards don't take kindly to thermal stress

[BarryB]

Originally Posted by cpiac64

tomorrow send more info

OK

Originally Posted by jbenam

You just need a temperature-controlled hot air station - cover the plastic parts with foil, and give everything a good heating at around 350°C, while circling around the board. Probably best to keep it as a last chance when everything else fails though, these old cards don't take kindly to thermal stress

Don't put it in the oven then like i've read some people do!!

[jbenam]

Originally Posted by BarryB

Don't put it in the oven then like i've read some people do!!

Indeed, that should only be done when plastics and caps are not present on the board, otherwise you will most certainly risk melting plastics and/or drying caps

[Kin Hell]

Originally Posted by BYTEMAN

Hi everyone,

Finally I've made it work with 060!!!

Let me describe what was checked and what was the issue. First, my assumption was that data buffers between SIMM and 68k bus are doing some mess. I've checked it with scope and saw that when I perform writes to SRAM (using "Manual memoryedit" feature of DiagROM), then data is passing well to SIMM lots (I've checked using FF pattern and see that every byte out of 4 performs write 'b1 to all bits). It seemed to be OK.
So, but after write there were different data... BINGO! Let's check WE! Ok, WE is coming from MUX under 68k XTAL, input 1 is 68k R/W pin, input 2 is +5V. Checking with dual-channel scope, seems to be OK, MUX is working. Switching signal is coming from one of CPLDs. But there are two consecutive WE triggering on SIMM..... Let's check what data is coming during second WE. Ok, CPU overwrites normal data with wrong ones during second WE.

Hmmmmm..... So something is wrong with Write cycles of CPU... Looking through 68060 datasheet, I've fond information about "Extra write hold" mode set using IPLx pins. Whow, that is the difference from 68040!!!!! For 68040 and 68060 IPL pins during Reset phase are doing different things! Found information for Cyberstorm MK1, where it is needed to modify IPL behavior when changing CPU to 68060....

Well, let's check what's on IPL during reset. All 'b1!!! IPL is driven from cPLD, no MUX or some other glue logic.... So, it seemed that accelerator was not working on 060 mode....

BINGO!!! 060 JUMPER!!! Checked it, there was short between trace coming to pad, and another pad using for PPC multiplier. It was under resistor, and I dodn't tuch that place before (so it came this way from previous owner). Short eliminated, power on - NO DATA CORRUPTION!!!!!!!!!!

So, here the story ends, I've successfully reflashed ROM to 060 version, launched Starstruck, checked PPC operation using some demos from WarpOS distribution.

I hope that my story might be useful for someone

Thanks,
Alexander

Originally Posted by Jope

Thanks for your report, this will most likely be useful to someone in the future. :-)

Oven Reflowing.... or not....

Put it on a baking tray covered in good quality kitchen foil, Pre-heat oven to 240 DegC or Gas Mark 9 & bake for 31 minutes & 16.25 seconds...

Anyhow, the OP has fixed it so no need to go on eh!

@ BYTEMAN

Well done fella. - Great work!

[Jope]

Yeah, I wasn't specifically thanking for any cooking oven reflowing tips, but more like the 040/060 jumper details which might go overlooked by many.

Whatever.

[Jimbeam]

Hello!
I'm new to this but quite old to Amiga .
A year ago start to restore all my Amigas back to life. All seems to works fine.
One A1200 I have with BlizzardPPC603 and 68040 with 32mb ram in 2 sims, behave some weird. I remember this behave back in 2000 when its new.

When I start the A1200 it boots OK. All works fine. No crush, no wormup too hi. Scsi controller works too.
When i reset the Amiga continious to reboot and work fine.
When turn off and on again the power.... Black screen..
After a few minutes or even half hour wen I turn on works fine again!
Has enyone this issue with this card?
And risently I found that when hold reset keys during cycle the power it boots most of the times.
I have knowledge of electronics and seems to me that issue whose something with reset cyrcuite.
So im going to scope reset signal and report back.
But if enyone has some idias please tell me to check.
Thanks a lot and sorry for my terrible English.