It's wild to me that people knew it was a serial protocol for 30 years and nobody actually tried to use it (except that one guy who wrote the document) until now.
used to use a techstep back in the days of being a certified mac tech, however, our board-level repair techs could spot problems faster than the tool could expose them so rarely used the thing. You helped bring back a lot of fond memories of those years. it was the dawn of the PPC and the clone wars; we we had a lot of interesting hardware rolling through the shop. i loved apple's toolless industrial design back in those days, everything was so easy to work on.
I'd love to know more about Mac board level repair happening back then. Who was doing this work and where did the spare parts come from? All I ever knew is Apple didn't allow the certified dealers to do this kind of work, or was this a bad assumption?
@@adriansdigitalbasement It's true that Apple discouraged board-level repairs. By then, it was a matter of how quickly you could get each broken Mac back out in the field working. So, if an analog board was the main problem, just swap it out for a new one. Quite a costly fix, at around $260 USD back then, unless covered under AppleCare. Although, there were some repair techs (or shops) who might have quietly stolen the bad parts or "borrowed" them to try and fix the problem. We can thank individuals like Larry Pina and his book "The Dead Mac Scrolls" for narrowing down specific symptoms and what is likely causing it. Sure, you can do the noble thing of getting your Mac repaired through an Apple Authorized Repair Technician. But for those of us looking to better understand or maybe save some money (especially when the AppleCare is long expired), it seems to be the better route to fix it yourself in some cases.
Hello Adrian. It gave me much pleasure watching this SE/30 saga. My conclusion is that you took the right step, making this your full time job. You do it with an abundant amount of pleasure. You face tells it all. Not even mentioning your smart way of working. It freaking works.
Great find. I absolutely love complicated obscure inbuilt test modes like this. Thanks for exploring this and showing us. I hope the archive comes back up soon.
This “test mode” is called the Serial Test Manager (STM), and the source code to a later version of it is included in the SuperMario sources (OS/StartMgr/UST*.a). So it should be possible to figure out exactly how all these tests work. 😉
I"m glad for you and all the Mac mavens out there that you captured this before the Archive went down. They're still down as I type, and this is an excellent example of they kind of stuff we'll be missing without it.
Yeah and I even saved the archived page locally which is something I've learned to do from experience. You just can't always rely on things to be up when you need them.
Enjoyed watching this video even though I do not own a MAC. Your detective work and internet researching skills are top notch. This video teaches your viewers how to problem solve. Two thumbs up!
It took me months to get a loan Techstep recreation to try and diagnose my Colour Classic - then comes Adrian Black with some serial commands I can run! Amazing! This will open so many doors. I still haven't diagnosed the CC, I suppose if the TechStep will come up with some unhelpful errors like the Bus Error, I might be able to dig a bit more into it. Thanks for that Sir! I'll mention your work in my video of course.
What is interesting -- is I never take for granted that documentation I use in a repair will be available in the future, so I always save a local copy and file it away. I really hope the Internet Archive comes back online soon.
The Archive got hit by a DDOS attack. They're currently offline while they upgrade their security. According to their Twitter channel, they expect this to take a few days to get everything right
This series is excellent and confirms many of the frustrations I have encountered over the years when working on these boards. Thank you Adrian, you are an incredible resource to the community!
As a former systems admin/network admin, you made me nostalgic for null modem cables. When I was young using those to transfer files and even play Doom with another person was possible. As an admin, we used Serial all the time to configure devices, and these days you use usb to serial adapters. I still have a frankencable similar to your laplink cable, but it allows me to turn "null modem" off and on, and has both male and female adapters on both sides, so I have a 1-cable solution for any serial ports I come across. I even have an RJ45 adapter for all the networking equipment that uses serial over an RJ45 port.
Absolutely amazing video! I never knew about the built-in diagnostic mode in the Mac SE/30, and your detailed explanation was super engaging. It’s like a treasure trove of knowledge for vintage Mac enthusiasts. Keep up the great work!
Sir, never dissapointed by your videos. Good stuff. Appreciate the content. I have a working VIC-20 now (just passed to my youngest son,) because of the motivation your videos gave me.
Great work Adrian! I had tried similar reverse engineering of the boot diagnostic sequence a couple years ago, but you actually got it done and figured out. Congrats!
I gotta say Thank You for explaining yourself so well in your videos. I admit I probably only understand half of it as I have 0 engineering or electrical experience but I still learn so much with every one posted
One of the reasons Apple implemented board level swaps was because of guys like me! I was harvesting the original Mac and Mac Plus ROMs for use in the Magic Sac and Spectre cartridges for the Atari ST. It was a really cool device which would give you a faster Mac with a larger screen, all for far less money than a Mac (a LOT less!). Floppy support was a problem at first, but by the time the Spectre GCR cart came out, the ST, which used a slight variation of MS-DOS's FAT format, could now read and write to Apple's GCR format. A similar device came out for the Amiga, but I feel like the Atari implementation was better, if not cheaper overall. I had another device that was an internal mod that had an NEC V20 (Intel 8086 and 8080 compatible) so my ST could run Mac, CP/M, MS-DOS, Windows 2 and GeoWorks). I was a pretty happy young man! Maybe someone could loan Adrian a Spectre GCR so he could explore it in a future episode?
You say it would make this video 2 hours long like that's a bad thing, it wouldn't be. Lucky for me I got a bit behind watching your videos so I could straight on to part 3, but now I need to wait for part 4... Excellent videos.
I remember in the last video you mentioned in passing that there was some activity on the serial controller as if you could attach a serial debugger, and I wondered what would happen if you tried-I never imagined it would be this sophisticated though! Amazing work!
What an amazing discovery! How great that you were able to find it and share with the rest of us! Also, that pattern is called Simasimac, and it's frequently associated with screwed up capacitors.
Dear Adrian, this video was like a good crime thriller!!! You really are the Sherlock Holmes of computer hardware errors. Great - and many thanks from Germany! Georg
OMG, Adrian! This is awesome! It's extremely deep troubleshooting. I couldn't stop watching it and learned a lot. I mean, seriously, fantastic! Thank you.
Can confirm. The school I was attending in 92-93 had an SE/30 that broke a lot. Like, in the two years I was there and using it, it broke three times and needed a motherboard swap. They asked the dealer what happened to the old boards and the dealer said they had to send them to Apple for credit.
Apple still works the same way. An analogy would be the core charge that car part suppliers use, which gets refunded when you return the old part (core). The biggest difference is that Apple doesn't charge you the full replacement part price, unless you don't send the bad part back in their return window.
I am pretty amazed those early LaserWriter printers might have this too. The Machine Identifier is there but I'm not sure the diagnostics actually work.
@@adriansdigitalbasement I’m not surprised at all. The early LaserWriters had a 12 MHz 68000 (and the IIf/g were 030s), and if I were a hardware engineer building one, I sure as heck would reuse this external diagnostics protocol given that there was all this hardware support to drive it.
@@adriansdigitalbasement maybe it would at least let you be able to check the RAM. the LW IInt runs off a 68000 and only has 1 SIMM slot, for 2MB total, while the IIntx runs off a 68020 and has *12* SIMM slots for a whopping 12MB total (like mine).
I loved the troubleshooting process that you used here - the way that you narrowed-down the fault. Very useful information is all this, and although I'm not a Mac guy, the 68000 architecture info is useful for my 'fleet' of Amiga computers.
Its particularly fascinating that Apple bothered shipping the techstep and diagnostic mode at all, because they were indeed basically useless outside of Apple’s own engineering departments because nobody was authorized to do board level repair anyway, and if they did some anyway they couldnt get parts and thus were limited to fixing solder joints and adding bodge wires (or replacing the occasional commodity chip). They had to build them for internal engineering purposes, but leaving them In the ROM is pretty fascinating.
I would imagine that someone probably could create an inexpensive good enough TechStep using a Raspberry Pi that could send the appropriate commands to the Mac via a touchscreen GUI. It just needs to be good enough for basic troubleshooting like you are doing here. Plus it wouldn't be stuck to only Macs as you could have Boot select as part of turning the device on. Just a thought for more talented people.
Eric (of the Bluescsi project) already hit me up regarding about this. :-) There is still a ton of work to get done before this could be automated since it appears what you can do varies from Mac to Mac, but once all of this is understood and documented then something open source could easily be developed. For now, I'm just happy to know how to do this myself so I won't be sitting around guessing. Guessing is no way to fix something! :-)
@adriansdigitalbasement documenting the input strings like you are seems to be the best first step, and then someone can create scripts for each task.
Yup, it's super simple using Python to access the serial port of a Pi, and from there to send the proper commands over to the Mac, and then read and translate the various responses into something more human readable. You could also run the same Python code, with a GUI for the desktop, compiled into an .exe for any Linux, Windows or Mac. It could also be adapted for the Amiga since the DiagROM the Amiga has can also work over serial.
I have my own opinions of Apple's whole repair/warranty system, but that's neither here nor there. But I always did have a huge fondness for their early and mid-life designs. They were pricey, but once you got into something and started farting around in there, you could see where a lot of the money went. Some of the case designs really were top notch. The old Power Macs I just adored. I had one I kept going long after it was useful just because I found it such a cool thing. I can't tell you how many times I had to repair that motherboard, which I did purely for the fun of it. At one point it did get to be past the point of being repairable, and I set it aside eventually tossing it when I finally sold my other home. But that's just babbling. It's incredibly cool that you were able to poke around and put this together to get out there. I'm sure it was a fair bit of work trying to figure out all that you did even with the help of that older document. But it's really fantastic that you were able to. I'm sure that's going to make a lot of hobbyist's lives MUCH easier being able to do some of the things that you would otherwise have use that ancient, insanely rare and expensive diag system to do. I always find this stuff incredible interesting, especially in the case of a company like Apple who has always kept their ecosystem played very close to the chest. That's why I really enjoyed the episodes on that dual processor Unix system. That whole thing of just happening to unknowingly acquire one of the company's internal systems that just happened to have one little typo somewhere in the past that accidently preserved so much information about everything on the hard drive was just an awesome thing to watch and learn about, even if the system is largely irrelevant in the grand scheme of things. So digging this diag stuff up is super fascinating to me. Plus, it's a huge win for everyone out there still tinkering with these machines.
37:58 as for the 24-32 bits being used for 8 bit devices comes straight from the processor being big-endian. It will store its MMSB in the 1st byte and its LLSB in the 4th. So an ldr.b instruction thus reads a byte from the 1st byte adressed and has thus to be on the bus as bits 24-31 as this is mapped to the LLSB. One consequence is that if you read a 8 bits from a 16 bits stored value you get the MSB where for little endian machines this would still give you the LSB part.
Thanks Adrian for putting this video together. I have a Mac classic that starts with zigzag line across screen. Hopefully I can find out more with a working pc running terminal software using null modem cable. Thanks again🥂
I had a friend that was an Apple dealer in a fairly small market, and what you are describing about warranty repair work, , is the exact reason he got fed up, closed his business, and went to work for Home Depot.
56:11 Learning that BERR is externally generated is not obvious, but an important thing to know about the 68K hardware. Normally the 68K waits until it is told that memory is ready. This is done by asserting DTACK after whatever wait state time that the memory needs to respond. Since some memory is slow, and I/O ports even slower, the CPU only has to wait as long as is necessary for a particular address. But if the address is in some wasteland that has nothing there, and the chipset doesn't handle it, the 68K will wait forever for either a DTACK or a BERR. In order for BERR to happen, something external to the CPU either has to know all the invalid memory ranges, or count a bunch of idle clocks with no DTACK. For instance, on the Sega Genesis, there is no circuitry at all to generate BERR. If you try to access certain invalid memory ranges, the system locks up waiting for DTACK. In the case of the Sega, it could probably be bodged in with one or two chips, but is only necessary for a development system. Real games don't have a problem with this because they've already been debugged to not access invalid addresses. Conversely, there is a meme of "DTACK grounded", where the DTACK pin is hardwired to always be asserted. This causes the 68K to run as fast as possible. For the original 8MHz 68000, if you use good fast SRAM and fast I/O chips, you can get away with this.
Thanks, great insight there. And the Genesis analogy is perfect -- it's not like it printing a BUS ERROR crash message on a production system would be useful, so I can see why they didn't even bother with it. The 68k really did bring some cool advancements. I hardly ever work on any system based on them, so I'm just not familiar with the system design but indeed, it is kind of essential to learn about these signals. Reading Apple docs on the Mac II line (which the SE/30 is) it introduces wait states for particular regions of the memory map, so I guess the GLUE does this by holding back the DTACK and counting cycles. Awesome!
Came here to basically say this. And on 32 bit systems _DTACK got split into _DSACK0 and _DSACK1, to acknowledge the various transfer sizes that could be initiated. RAM itself does not generate any kind of acknowledge signal, so it was up to the DRAM controller to do this. With static RAM and ROM it was pretty much up to the address decoding hardware to guess based on the part timing used. Quick and dirty is to just OR (active-low) your _CS and _AS together and bump it through an inverter a couple times for the propagation delay. SRAM at least will likely have data ready long before the bus cycle is complete. For a bus error on a machine like this, the first place I'd look is the address decoding.
"and on that cliffhanger..." NOOOOOOOOOOOOOOOoooooooooooooooooooooooooo!!!! ;-) This is an awesome vid and I love the serial diagnostic discovery/work! This is super interesting!!!
Inside Macintosh IS ONLY a programmers reference. It illustrated the applicable Macintosh OS APIs to support software development on the early generations of the Macintosh. Once Apple moved to N.E.X.T object oriented OS as the foundation of the newer Macintosh OS, Inside Macintosh was effectively obsolete. And you are correct, you had to send Macintosh hardware back to Apple, and they destroyed what they could not repair. Worked for Apple for a few years in the mid to late 80s at the high tide of Apple's hardware inventory control methods.
Something that came to mind when you started talking about the 24 bad vs 8 good bits, and didn't go away but instead just got further reinforced with the more details you revealed: Might the memory management chip be sticking in 8-bit mode, AND only reporting 1MB available, because something's stopping it from accessing those other three SIMMs? Maybe a RAS or CAS line that gets as far as the first socket but is broken after that (would prevent the strobe signals needed to select row and column addresses from getting through), or a set of chip/module select lines where three have been damaged but the fourth one is still OK? Because of the way that the 680x0 is signalled to use different data bus widths, it may then also be chucking data out to memory, and reading it back, over what is effectively an 8-bit bus instead of a 32-bit one. It's doing it slowly, but successfully, and seeing each chunk in turn building up to a full 32-bit width, so as far as it's concerned all the bits are good. And as you're doing an extremely simple, yet unfamiliar thing, over a slow serial connection, you're never going to discern any actual difference in speed. So it completes in 16 milliseconds instead of 4... you'll neither notice nor care about that. What will be a problem is any complex memory-bound operation taking 4x as long, and running out of RAM so much faster. (the processor doesn't have any "permanent" idea of what width bus it's on, or that it's "supposed" to be on, only what it's currently seeing an assertion signal to use. 68k's are basically bus width agnostic for 8, 16, and later 32 bits. The 68008 only differs from the 68000 by losing some legs and having the bus width tied internally to "8-bit" permanently, and there's no reason you couldn't do the same with an 020, 030, 040... certainly there are enough examples of systems using an 020 or 030 on a 16-bit bus with only limited or even no 32-bit transfers ever happening. Plus, all of the I/O is memory mapped - it has no idea if a commanded write is going to a peripheral, or to RAM, and same for reads. So long as there's something at the given address that can accept data, the data will be sent there, at the width it's been told to. Thus if all you can see is an 8-bit wide strip of 1MB of memory, that's what you're going to write into, then later read back out of with no trouble at all. It doesn't mean the memory, or sockets, are all working...) Like I guess you can test by removing the other three SIMMs and seeing what happens (if that happened in the video, my apologies, my memory is bad), or certainly two of them. Switch them around musical chairs style just to be sure. Write a routine that writes some known block of data across the 1MB boundary and see if it succeeds, and whether all that can be read back again properly. Maybe it'll get to 1MB plus 1 byte and die. Start the processes at or just before different megabyte breakpoints. Etc etc. Also have you tried putting the memory in the second set of banks and seeing if it then boots? Gives better diagnostic results? You could / can do that with a lot of computers these days, so long as it's present somewhere it's usable and can be logically reassigned in a sensible order.
Also remember that if you're loading and running a tiny 18 byte program, you're likely just running from the 030's cache. It will be writing to RAM, and incorrectly if the SIZ* signals are bad, but read and execute won't actually hit RAM unless the cache is flushed. A RAM test writes/reads too many addresses to hold in cache. There is a CDIS input to the 030 to disable the cache, and the GLUE asserts it for I/O ranges, but you can also ground it to force reads to go to the bus.
That's a great find, some enterprising soul could write a piece of software that would communicate directly with the serial port on a PC or a Mac, and be able to decode the error messages and present them in a graphical format, such as showing what bank and specific SIMM module of memory is bad.
Lots of usb equipment on PCs uses serial Protocols. I often have to snoop the starting sequence and commands with 2 teensys and the arduino serial monitor.( I m talking about controllers for stepper motors or spectrum analyzers and stuff.)
That analogy of someone asking what's wrong with their car without having seen it, that actually happened a lot when I repaired PCs for a living, someone called the shop and asked if I could tell them what was wrong with their computer over the phone without having seen it, and of course, I couldn't, had to explain to them that without having the computer in front of me to diagnose any faults with it that I couldn't tell them what was wrong as it could be anything from a bad OS install to failed hardware which I couldn't guide them through over the phone...
@@mikebarushok5361 for me it was a complaint of a key not working. After fighting the keyboard case and plate, eventually locating the pins, I find it was fine! Turned out to be a locale issue in the OS.
FYI I wanted to try this on my SE/30 but it boots normally. I couldn't get the interrupt switch to work. Putting some DeoxIT on my programmer's switch fixed it and now I can explore this!
I like the concept of keeping the repairs relatively simply with off the shelf equipment. That being said, I would also be interested in advance repairs, using reverse engineered test setups or 'test jigs' that were originally used by Apple. I've found that manufacturers have deployed some interesting methods for troubleshooting, that can be used to our advantage, when servicing legacy equipment. Motorola had Lab Tools, while challenging to get, was very handy when servicing some of their radio equipment.
Funny sidenote when you say star-r that reminds me of the boss in first resident evil that goes STARS. Also gives flashbacks from airplanes in a real airplane they have a delay. So it says Stall-All kind of like a echo effect. If you pull fuse 17 or 19 I think in the fuse box of the 747 it doesn't chime several warnings and stall sounds normal. Not like standard stall-all Loving your videos. Your brilliant.
The laplink cables I remember were bodged parallel cables with 4 bits going in one direction and 4 bits going in the other. I'm not surprised serial ones existed, it makes sense.. just never seen that until you held one up!
MOD3 ram test rang a dim and distant bell in my memory. Those 32 bit values are special. COnvert them to binary and you'll see ther are a repeating pattern of 110. Turns out old ram ocassionally leaked bits in close columns. Running write read cycles with the DB6D etc rotating will flush out flaky ram I wrote a ram test for a proprietary machine back in the ealry 80s and this was one of the recommened tests for 4116 and 4164 ram chips
![Blox Fruits Dragon Rework Update [Full Stream]](http://i.ytimg.com/vi/EqDAp8udhm0/mqdefault.jpg)
This is by far one of your most revolutionary videos for Mac II/SE diagnostics thus far. Kudos to you Adrian for this investigation!
Those "TechStep Replica Guys" will be pissed... but, they're out of stock anyway 😎
@@mwk1 lmao
love these types of moments when you just happen to stumble across an old txt file that unlocks forgotten knowledge
Given Apple's tendencies, this is more like forbidden knowledge.
Sure beats stumbling across an old txt file that unlocks an infohazard
@@HaveYouTriedGuillotineswell it is the forbidden fruit: the poor soul who took that bite did not live to regret it
It's wild to me that people knew it was a serial protocol for 30 years and nobody actually tried to use it (except that one guy who wrote the document) until now.
used to use a techstep back in the days of being a certified mac tech, however, our board-level repair techs could spot problems faster than the tool could expose them so rarely used the thing. You helped bring back a lot of fond memories of those years. it was the dawn of the PPC and the clone wars; we we had a lot of interesting hardware rolling through the shop. i loved apple's toolless industrial design back in those days, everything was so easy to work on.
I'd love to know more about Mac board level repair happening back then. Who was doing this work and where did the spare parts come from? All I ever knew is Apple didn't allow the certified dealers to do this kind of work, or was this a bad assumption?
@@adriansdigitalbasement It's true that Apple discouraged board-level repairs. By then, it was a matter of how quickly you could get each broken Mac back out in the field working. So, if an analog board was the main problem, just swap it out for a new one. Quite a costly fix, at around $260 USD back then, unless covered under AppleCare. Although, there were some repair techs (or shops) who might have quietly stolen the bad parts or "borrowed" them to try and fix the problem.
We can thank individuals like Larry Pina and his book "The Dead Mac Scrolls" for narrowing down specific symptoms and what is likely causing it. Sure, you can do the noble thing of getting your Mac repaired through an Apple Authorized Repair Technician. But for those of us looking to better understand or maybe save some money (especially when the AppleCare is long expired), it seems to be the better route to fix it yourself in some cases.
Hello Adrian. It gave me much pleasure watching this SE/30 saga. My conclusion is that you took the right step, making this your full time job. You do it with an abundant amount of pleasure. You face tells it all. Not even mentioning your smart way of working. It freaking works.
Great find. I absolutely love complicated obscure inbuilt test modes like this. Thanks for exploring this and showing us. I hope the archive comes back up soon.
This “test mode” is called the Serial Test Manager (STM), and the source code to a later version of it is included in the SuperMario sources (OS/StartMgr/UST*.a). So it should be possible to figure out exactly how all these tests work. 😉
SuperMario sources?
I"m glad for you and all the Mac mavens out there that you captured this before the Archive went down. They're still down as I type, and this is an excellent example of they kind of stuff we'll be missing without it.
Yeah and I even saved the archived page locally which is something I've learned to do from experience. You just can't always rely on things to be up when you need them.
@@adriansdigitalbasement Archive of an archive, truly the way of the archivist!
Raise your hand if you wouldn’t mind a 2+ hour video from Adrian…. ✋
Will for sure make me non-active at work, but I love every second of it
Or a live video, no cut scenes, uncensored.
I'll just watch it in multiple chunks, so bring it on!
✋🏼
I made an audible "Noooo!" sound when Adrian cut the video.
Enjoyed watching this video even though I do not own a MAC.
Your detective work and internet researching skills are top notch. This video teaches your viewers how to problem solve.
Two thumbs up!
It's "Mac", short for "Macintosh", and not "MAC" - it's not an acronym :)
It took me months to get a loan Techstep recreation to try and diagnose my Colour Classic - then comes Adrian Black with some serial commands I can run! Amazing! This will open so many doors. I still haven't diagnosed the CC, I suppose if the TechStep will come up with some unhelpful errors like the Bus Error, I might be able to dig a bit more into it. Thanks for that Sir! I'll mention your work in my video of course.
16:19 glad you got to IA before they went down. I guess this video might not have happened without it, goes to show how critical of a service it is
What is interesting -- is I never take for granted that documentation I use in a repair will be available in the future, so I always save a local copy and file it away. I really hope the Internet Archive comes back online soon.
@@adriansdigitalbasementI really hope IA is put on Hyphanet, or Pirate Bay torrents.
@@adriansdigitalbasement yeah, it's almost crazy to see the internet archive down
that's one place everyone took for granted will always be there
The Archive got hit by a DDOS attack. They're currently offline while they upgrade their security. According to their Twitter channel, they expect this to take a few days to get everything right
@@adriansdigitalbasement I do that too. I also save docs from Retroweb.
This series is excellent and confirms many of the frustrations I have encountered over the years when working on these boards. Thank you Adrian, you are an incredible resource to the community!
As a former systems admin/network admin, you made me nostalgic for null modem cables. When I was young using those to transfer files and even play Doom with another person was possible. As an admin, we used Serial all the time to configure devices, and these days you use usb to serial adapters. I still have a frankencable similar to your laplink cable, but it allows me to turn "null modem" off and on, and has both male and female adapters on both sides, so I have a 1-cable solution for any serial ports I come across. I even have an RJ45 adapter for all the networking equipment that uses serial over an RJ45 port.
So much
Absolutely amazing video! I never knew about the built-in diagnostic mode in the Mac SE/30, and your detailed explanation was super engaging. It’s like a treasure trove of knowledge for vintage Mac enthusiasts. Keep up the great work!
Sir, never dissapointed by your videos. Good stuff. Appreciate the content. I have a working VIC-20 now (just passed to my youngest son,) because of the motivation your videos gave me.
Well, who knew!! What an incredible discovery. I can see a future full of Raspberry Pi Mac diagnostic tools. Let the revolution begin!
I passed on a complete Apple TechStep at a flea market for $40 about 15 years ago. Still kicking myself to this day for that one.
I sent my Quadra 840AV to Salvation Army about 15 years ago
Hey dude! Very nice meeting you at PRGE. You seem like a super cool dude and I love your videos. 🤘
Thanks! Same! I just watched your PRGE wrap up last night. I forgot to really take any pictures of the whole event. (Typical of me, LOL)
Great work Adrian! I had tried similar reverse engineering of the boot diagnostic sequence a couple years ago, but you actually got it done and figured out. Congrats!
This is Adrian working at a whole new level! Excellent forensic work!
I gotta say Thank You for explaining yourself so well in your videos. I admit I probably only understand half of it as I have 0 engineering or electrical experience but I still learn so much with every one posted
One of the reasons Apple implemented board level swaps was because of guys like me! I was harvesting the original Mac and Mac Plus ROMs for use in the Magic Sac and Spectre cartridges for the Atari ST. It was a really cool device which would give you a faster Mac with a larger screen, all for far less money than a Mac (a LOT less!). Floppy support was a problem at first, but by the time the Spectre GCR cart came out, the ST, which used a slight variation of MS-DOS's FAT format, could now read and write to Apple's GCR format.
A similar device came out for the Amiga, but I feel like the Atari implementation was better, if not cheaper overall.
I had another device that was an internal mod that had an NEC V20 (Intel 8086 and 8080 compatible) so my ST could run Mac, CP/M, MS-DOS, Windows 2 and GeoWorks). I was a pretty happy young man!
Maybe someone could loan Adrian a Spectre GCR so he could explore it in a future episode?
You say it would make this video 2 hours long like that's a bad thing, it wouldn't be. Lucky for me I got a bit behind watching your videos so I could straight on to part 3, but now I need to wait for part 4... Excellent videos.
I remember in the last video you mentioned in passing that there was some activity on the serial controller as if you could attach a serial debugger, and I wondered what would happen if you tried-I never imagined it would be this sophisticated though! Amazing work!
What an amazing discovery! How great that you were able to find it and share with the rest of us! Also, that pattern is called Simasimac, and it's frequently associated with screwed up capacitors.
Thanks!
Nooooo… “To be continued”. Nice diagnosis, skills, and storytelling Adrian - can’t wait for episode III!
This. THIS is why we watch ADB. Fabulous stuff, Adrian!
You might want to turn local echo on in your terminal program. It makes life easier when the remote machine doesn't echo.
Watched from beginning to send, great work sir! I feel you are on the road to a booting system. Looking forward to the next installment
Dear Adrian,
this video was like a good crime thriller!!!
You really are the Sherlock Holmes of computer hardware errors.
Great - and many thanks from Germany!
Georg
OMG, Adrian! This is awesome! It's extremely deep troubleshooting. I couldn't stop watching it and learned a lot. I mean, seriously, fantastic! Thank you.
Loving this series. Was hoping for a quick bodge wire and a test at the end but I can't wait for the conclusion.
This is some brilliant work Adrian. What a result for the community. Cliffhanger is brutal though, a 2 hour video would definitely be ok by me!
Can confirm. The school I was attending in 92-93 had an SE/30 that broke a lot. Like, in the two years I was there and using it, it broke three times and needed a motherboard swap. They asked the dealer what happened to the old boards and the dealer said they had to send them to Apple for credit.
Apple still works the same way. An analogy would be the core charge that car part suppliers use, which gets refunded when you return the old part (core). The biggest difference is that Apple doesn't charge you the full replacement part price, unless you don't send the bad part back in their return window.
This video not just for the SE/30, but for all those early 90s Mac machines...
Even the freakin printers.
I am pretty amazed those early LaserWriter printers might have this too. The Machine Identifier is there but I'm not sure the diagnostics actually work.
@@adriansdigitalbasement I’m not surprised at all. The early LaserWriters had a 12 MHz 68000 (and the IIf/g were 030s), and if I were a hardware engineer building one, I sure as heck would reuse this external diagnostics protocol given that there was all this hardware support to drive it.
@@adriansdigitalbasement I would love to see a vid experimenting on all the various 90s apple stuff! =P
@@adriansdigitalbasement maybe it would at least let you be able to check the RAM. the LW IInt runs off a 68000 and only has 1 SIMM slot, for 2MB total, while the IIntx runs off a 68020 and has *12* SIMM slots for a whopping 12MB total (like mine).
I’m fine with 2 hour videos! Looking forward to the next part. Great work!
I'm fine with two hours. Can't wait for part 3! Awesome video!
This is really a fascinating video and some of your best work yet! Well done and thanks ever so much for sharing with the rest of us.
I loved the troubleshooting process that you used here - the way that you narrowed-down the fault. Very useful information is all this, and although I'm not a Mac guy, the 68000 architecture info is useful for my 'fleet' of Amiga computers.
Really fascinating process. Love it. Looking forward to part 2
Hi, Adrian. Thankyou for a nice video. I loved it. Stay safe to you and your loved ones.👍🏻
I have no particular interest or love for old macs, but I do appreciate your walk through of the troubleshooting process. Thanks.
This will definitely save people hundreds of dollars! Great diagnostics and investigation as usual from you.
That was some awesome investigations. Kudos! Really looking forward to then next episode.
Outstanding video and analysis. Kudos, Adrian! What a cliffhanger though! 😆
Its particularly fascinating that Apple bothered shipping the techstep and diagnostic mode at all, because they were indeed basically useless outside of Apple’s own engineering departments because nobody was authorized to do board level repair anyway, and if they did some anyway they couldnt get parts and thus were limited to fixing solder joints and adding bodge wires (or replacing the occasional commodity chip).
They had to build them for internal engineering purposes, but leaving them
In the ROM is pretty fascinating.
I would imagine that someone probably could create an inexpensive good enough TechStep using a Raspberry Pi that could send the appropriate commands to the Mac via a touchscreen GUI. It just needs to be good enough for basic troubleshooting like you are doing here. Plus it wouldn't be stuck to only Macs as you could have Boot select as part of turning the device on. Just a thought for more talented people.
Eric (of the Bluescsi project) already hit me up regarding about this. :-) There is still a ton of work to get done before this could be automated since it appears what you can do varies from Mac to Mac, but once all of this is understood and documented then something open source could easily be developed. For now, I'm just happy to know how to do this myself so I won't be sitting around guessing. Guessing is no way to fix something! :-)
@adriansdigitalbasement documenting the input strings like you are seems to be the best first step, and then someone can create scripts for each task.
Yup, it's super simple using Python to access the serial port of a Pi, and from there to send the proper commands over to the Mac, and then read and translate the various responses into something more human readable. You could also run the same Python code, with a GUI for the desktop, compiled into an .exe for any Linux, Windows or Mac. It could also be adapted for the Amiga since the DiagROM the Amiga has can also work over serial.
Actually, even a esp8266/32 can handle this and you can do this via WiFi
@@VitorFMyou beat me to it A Pi is massive overkill for this project….
Very fun stuff. I love your patience and persistence. Would love to hear how you plan your repairs. I’m a hope for the best expect the worst type.
I have my own opinions of Apple's whole repair/warranty system, but that's neither here nor there. But I always did have a huge fondness for their early and mid-life designs. They were pricey, but once you got into something and started farting around in there, you could see where a lot of the money went. Some of the case designs really were top notch. The old Power Macs I just adored. I had one I kept going long after it was useful just because I found it such a cool thing. I can't tell you how many times I had to repair that motherboard, which I did purely for the fun of it. At one point it did get to be past the point of being repairable, and I set it aside eventually tossing it when I finally sold my other home.
But that's just babbling. It's incredibly cool that you were able to poke around and put this together to get out there. I'm sure it was a fair bit of work trying to figure out all that you did even with the help of that older document. But it's really fantastic that you were able to. I'm sure that's going to make a lot of hobbyist's lives MUCH easier being able to do some of the things that you would otherwise have use that ancient, insanely rare and expensive diag system to do. I always find this stuff incredible interesting, especially in the case of a company like Apple who has always kept their ecosystem played very close to the chest. That's why I really enjoyed the episodes on that dual processor Unix system. That whole thing of just happening to unknowingly acquire one of the company's internal systems that just happened to have one little typo somewhere in the past that accidently preserved so much information about everything on the hard drive was just an awesome thing to watch and learn about, even if the system is largely irrelevant in the grand scheme of things. So digging this diag stuff up is super fascinating to me. Plus, it's a huge win for everyone out there still tinkering with these machines.
Very impressive diagnostic work! I wish I had your level of knowledge and understanding for how to diagnose stuff, but for C64 and 1541 diagnostics.
One of your better Videos I was spel bound to the end.
9:58 Commandant Eric Lassard joined the chat(many many many)
Best ADB in a while, bravo
This is Magic - I'm using it right now to diagnose!
37:58 as for the 24-32 bits being used for 8 bit devices comes straight from the processor being big-endian. It will store its MMSB in the 1st byte and its LLSB in the 4th. So an ldr.b instruction thus reads a byte from the 1st byte adressed and has thus to be on the bus as bits 24-31 as this is mapped to the LLSB.
One consequence is that if you read a 8 bits from a 16 bits stored value you get the MSB where for little endian machines this would still give you the LSB part.
.....better than an episode of Columbo!......to be continued....cant wait!
Thanks Adrian for putting this video together. I have a Mac classic that starts with zigzag line across screen. Hopefully I can find out more with a working pc running terminal software using null modem cable. Thanks again🥂
I had a friend that was an Apple dealer in a fairly small market, and what you are describing about warranty repair work, , is the exact reason he got fed up, closed his business, and went to work for Home Depot.
I always take a little bit of you on all my repairs , thanks John (uk)
56:11 Learning that BERR is externally generated is not obvious, but an important thing to know about the 68K hardware.
Normally the 68K waits until it is told that memory is ready. This is done by asserting DTACK after whatever wait state time that the memory needs to respond. Since some memory is slow, and I/O ports even slower, the CPU only has to wait as long as is necessary for a particular address. But if the address is in some wasteland that has nothing there, and the chipset doesn't handle it, the 68K will wait forever for either a DTACK or a BERR. In order for BERR to happen, something external to the CPU either has to know all the invalid memory ranges, or count a bunch of idle clocks with no DTACK.
For instance, on the Sega Genesis, there is no circuitry at all to generate BERR. If you try to access certain invalid memory ranges, the system locks up waiting for DTACK. In the case of the Sega, it could probably be bodged in with one or two chips, but is only necessary for a development system. Real games don't have a problem with this because they've already been debugged to not access invalid addresses.
Conversely, there is a meme of "DTACK grounded", where the DTACK pin is hardwired to always be asserted. This causes the 68K to run as fast as possible. For the original 8MHz 68000, if you use good fast SRAM and fast I/O chips, you can get away with this.
Thanks, great insight there. And the Genesis analogy is perfect -- it's not like it printing a BUS ERROR crash message on a production system would be useful, so I can see why they didn't even bother with it.
The 68k really did bring some cool advancements. I hardly ever work on any system based on them, so I'm just not familiar with the system design but indeed, it is kind of essential to learn about these signals.
Reading Apple docs on the Mac II line (which the SE/30 is) it introduces wait states for particular regions of the memory map, so I guess the GLUE does this by holding back the DTACK and counting cycles. Awesome!
Came here to basically say this. And on 32 bit systems _DTACK got split into _DSACK0 and _DSACK1, to acknowledge the various transfer sizes that could be initiated. RAM itself does not generate any kind of acknowledge signal, so it was up to the DRAM controller to do this. With static RAM and ROM it was pretty much up to the address decoding hardware to guess based on the part timing used. Quick and dirty is to just OR (active-low) your _CS and _AS together and bump it through an inverter a couple times for the propagation delay. SRAM at least will likely have data ready long before the bus cycle is complete. For a bus error on a machine like this, the first place I'd look is the address decoding.
Absolutely bonkers!! Just amazing
64KB of dual ported VRAM, providing a monochrome 512x342 display, in 1989, for $4369.. lol. What a racket.
"and on that cliffhanger..."
NOOOOOOOOOOOOOOOoooooooooooooooooooooooooo!!!!
;-)
This is an awesome vid and I love the serial diagnostic discovery/work! This is super interesting!!!
Detecting an error but not giving any indication of what it could be, thus requiring an expensive trip to Apple, is such an Apple thing to do.
LOL, you've not seen a lot of vintage Apple machines. Or pc clones of the same era.
Brilliant!
I knew it existed but did not have an idea how it worked & did not have access to a Mac for long so it was board swapping only!
Am I the only one that would be fine with a 2 hour video?
Inside Macintosh IS ONLY a programmers reference. It illustrated the applicable Macintosh OS APIs to support software development on the early generations of the Macintosh. Once Apple moved to N.E.X.T object oriented OS as the foundation of the newer Macintosh OS, Inside Macintosh was effectively obsolete. And you are correct, you had to send Macintosh hardware back to Apple, and they destroyed what they could not repair. Worked for Apple for a few years in the mid to late 80s at the high tide of Apple's hardware inventory control methods.
Something that came to mind when you started talking about the 24 bad vs 8 good bits, and didn't go away but instead just got further reinforced with the more details you revealed:
Might the memory management chip be sticking in 8-bit mode, AND only reporting 1MB available, because something's stopping it from accessing those other three SIMMs? Maybe a RAS or CAS line that gets as far as the first socket but is broken after that (would prevent the strobe signals needed to select row and column addresses from getting through), or a set of chip/module select lines where three have been damaged but the fourth one is still OK?
Because of the way that the 680x0 is signalled to use different data bus widths, it may then also be chucking data out to memory, and reading it back, over what is effectively an 8-bit bus instead of a 32-bit one. It's doing it slowly, but successfully, and seeing each chunk in turn building up to a full 32-bit width, so as far as it's concerned all the bits are good. And as you're doing an extremely simple, yet unfamiliar thing, over a slow serial connection, you're never going to discern any actual difference in speed. So it completes in 16 milliseconds instead of 4... you'll neither notice nor care about that. What will be a problem is any complex memory-bound operation taking 4x as long, and running out of RAM so much faster.
(the processor doesn't have any "permanent" idea of what width bus it's on, or that it's "supposed" to be on, only what it's currently seeing an assertion signal to use. 68k's are basically bus width agnostic for 8, 16, and later 32 bits. The 68008 only differs from the 68000 by losing some legs and having the bus width tied internally to "8-bit" permanently, and there's no reason you couldn't do the same with an 020, 030, 040... certainly there are enough examples of systems using an 020 or 030 on a 16-bit bus with only limited or even no 32-bit transfers ever happening. Plus, all of the I/O is memory mapped - it has no idea if a commanded write is going to a peripheral, or to RAM, and same for reads. So long as there's something at the given address that can accept data, the data will be sent there, at the width it's been told to. Thus if all you can see is an 8-bit wide strip of 1MB of memory, that's what you're going to write into, then later read back out of with no trouble at all. It doesn't mean the memory, or sockets, are all working...)
Like I guess you can test by removing the other three SIMMs and seeing what happens (if that happened in the video, my apologies, my memory is bad), or certainly two of them. Switch them around musical chairs style just to be sure. Write a routine that writes some known block of data across the 1MB boundary and see if it succeeds, and whether all that can be read back again properly. Maybe it'll get to 1MB plus 1 byte and die. Start the processes at or just before different megabyte breakpoints. Etc etc.
Also have you tried putting the memory in the second set of banks and seeing if it then boots? Gives better diagnostic results? You could / can do that with a lot of computers these days, so long as it's present somewhere it's usable and can be logically reassigned in a sensible order.
I'm supe rexcited about this and I don't even have a compatible machine lol
The whole community is going to have a field day with this!
Very good and comprehensive video
Just imagine the number of boards thats gone to the great e-waste in the sky before this info became more public 😮
I'm hoping that people can dust off their broken boards they have hopefully held onto to try poking around at again.
I love his videos everyday and I wish my school is like that back in my day-Laure paine Tesla
As it happens my SE/30 just got sick (literally while running) and this will be very useful… once I get the whole DIN8 USB chain working.
Also remember that if you're loading and running a tiny 18 byte program, you're likely just running from the 030's cache. It will be writing to RAM, and incorrectly if the SIZ* signals are bad, but read and execute won't actually hit RAM unless the cache is flushed. A RAM test writes/reads too many addresses to hold in cache.
There is a CDIS input to the 030 to disable the cache, and the GLUE asserts it for I/O ranges, but you can also ground it to force reads to go to the bus.
1:00:40 I like the transition to Adrian the Grey, full diagnostic wizard at the end
That's a great find, some enterprising soul could write a piece of software that would communicate directly with the serial port on a PC or a Mac, and be able to decode the error messages and present them in a graphical format, such as showing what bank and specific SIMM module of memory is bad.
biting my nails in anticipation..will be worn down to nubs,
lol
always like your videos
i shall call you Sherlock! that's some good investigation work.
I see no problem with a two hour Adrian repair video 😊
Ooh!! I think my brother might have a Techstep! I’ve seen one in his boxes. He’s sold Macs since the 80s. Imma gonna see if I can find it :)
Lots of usb equipment on PCs uses serial Protocols. I often have to snoop the starting sequence and commands with 2 teensys and the arduino serial monitor.( I m talking about controllers for stepper motors or spectrum analyzers and stuff.)
That analogy of someone asking what's wrong with their car without having seen it, that actually happened a lot when I repaired PCs for a living, someone called the shop and asked if I could tell them what was wrong with their computer over the phone without having seen it, and of course, I couldn't, had to explain to them that without having the computer in front of me to diagnose any faults with it that I couldn't tell them what was wrong as it could be anything from a bad OS install to failed hardware which I couldn't guide them through over the phone...
The one I remember most was a phone call that began with "I can't print!"
It was caused by a full hard drive.
@@mikebarushok5361 for me it was a complaint of a key not working. After fighting the keyboard case and plate, eventually locating the pins, I find it was fine! Turned out to be a locale issue in the OS.
Awesome news. Although I never had Macs, I can imagine some being saved from the scrap heap.
This is really useful information, thank you!
FYI I wanted to try this on my SE/30 but it boots normally. I couldn't get the interrupt switch to work. Putting some DeoxIT on my programmer's switch fixed it and now I can explore this!
Fascinating video. Once you get your teeth into a problem like this, you don't let go.
I like the concept of keeping the repairs relatively simply with off the shelf equipment. That being said, I would also be interested in advance repairs, using reverse engineered test setups or 'test jigs' that were originally used by Apple.
I've found that manufacturers have deployed some interesting methods for troubleshooting, that can be used to our advantage, when servicing legacy equipment.
Motorola had Lab Tools, while challenging to get, was very handy when servicing some of their radio equipment.
Gripping, looking forward to the next video.
Adrian for the Win!!!
You need what?!! An Unoptanium ridiculously expensive piece of diagnostics equipment.
Adrian: "One sec. Hold my Beer"
That tool probably cost the dealer as much as a Macintosh II back then.
Funny sidenote when you say star-r that reminds me of the boss in first resident evil that goes STARS.
Also gives flashbacks from airplanes in a real airplane they have a delay. So it says Stall-All kind of like a echo effect. If you pull fuse 17 or 19 I think in the fuse box of the 747 it doesn't chime several warnings and stall sounds normal. Not like standard stall-all
Loving your videos. Your brilliant.
You could script this very easily and do some really cool tests.
You ARE making a living repairing computers.
The laplink cables I remember were bodged parallel cables with 4 bits going in one direction and 4 bits going in the other. I'm not surprised serial ones existed, it makes sense.. just never seen that until you held one up!
MOD3 ram test rang a dim and distant bell in my memory. Those 32 bit values are special. COnvert them to binary and you'll see ther are a repeating pattern of 110. Turns out old ram ocassionally leaked bits in close columns. Running write read cycles with the DB6D etc rotating will flush out flaky ram
I wrote a ram test for a proprietary machine back in the ealry 80s and this was one of the recommened tests for 4116 and 4164 ram chips
This is awesome thanks Adrian!
This video is really awesome.
It's surprising sometimes how resilient chips can be.. 😅