HP 9895A Part 2: Epic Controller Board Repair (and a transparent chip surprise)
We restore more HP 9895A dual 8" diskette HP-IB drives, but these will be way more challenging. We'll need a full on logic analyzer assisted repair to debug a really, really nasty fault. It's a long video, but don't miss the intermission where Master Ken inspects a transparent silicon-on-sapphire chip!
Many thanks to HP enthusiast Craig Ruff for the ROM dump and disassembly.
8" Floppy Playlist: • 8" Floppy Diskette Drives
HP 9825 Playlist: • HP 9825 Repair
Ken's Silicon-on-Sapphire chip blog post: www.righto.com/2023/12/HP-sili...
Stuff that supports the channel:
- Amazon links for the tools I use in the lab: www.curiousmarc.com/amazon-links
- Channel merch on Fourthwall: curiousmarc-shop.fourthwall.com
- Legacy channel merch on Teespring (I don't have everything transferred to Fourthwall yet): teespring.com/stores/curiousm...
- Support us on Patreon: / curiousmarc
"Elevator Music" Credit: Crinoline Dreams by Kevin MacLeod
Our lovely sponsors
- PCBWay: fast turn PCBs, www.pcbway.com
- Electro-Rent: www.electrorent.com
- Keysight: test instruments: www.keysight.com
- Samtec: connectors: www.samtec.com
- R&D Microwaves: rdmicrowaves.com
Get more technical info on the companion site: www.curiousmarc.com
Contact info: kzhead.infoa...
Master Ken's reverse engineering skills are awesome!
He's got to be one of the best in the world. His documentation is also clear, concise and detailed, an underrated art in and of itself. Ken is truly a master
@@tezinho81 Master Ken makes me feel inadequate.. A very special mind to be able to focus on a task, analysis and map system functions so methodically. Incredible work by all in the team who are recording for posterity what was almost lost. Best
"Yeah...... Obvious......" 🤣 I'd say "Master" Ken has truly earned that title!
That chip tour intermission was fantastic!
Always a pleasant surprise when the chip you bought on eBay actually works and meets the advertised spec!
Master Ken has truly gone off the deep end where you have learnt so much that you can't remember what a new person to the topic would and wouldn't recognise anymore :) Always awesome seeing such deep dives into engineering, while the chips we're looking at are still visible in optical light :D
Back in the 80s, my dad, a materials science engineer who worked in the valley, was doing a ton of silicon on sapphire work for DoD. They were really interested to see if chips using it would be hardened against EMPs since the sapphire substrate would be a fantastic insulator. Many weekends I would go into work with him and get all bunny suited up so I could watch him check on some run he had going in the fab. Fun times and great memories!
And Silicon-on-insulator is still a thing today!
It is amazing to see a group of REALLY SMART guys tackle these engineering challenges!
"10 beginner programmers' mistakes: syntax error, undeclared variable, off-by-one error." Nice work! Always happy to see you recombobulate vintage tech to a good working order. That chip looks extremely complex, I couldn't make anything meaningful out of it. Ken is a true Sheriff of Electronics, ha! And again, I hope I learned something.
Spent a fair amount of time programming instruments with HPIB, usually with Fortran. That was at Univac on a large HP test system with many programmable instruments. Two companies later, at Iomega, we were using custom disk drive controller chips. The first iteration of a new chi[ design came from fab with a bug. The brilliant chip designer with me 'helping' debugged the chip with micro-manipulators and was actually able to repair is by scraping metal and creating a short enabling us to make further progress on the new design debug. Good days.
The (almost) full team! One of those videos you watch multiple times! :D
You guys just astound me, I so enjoy watching you get stuff on its feet again.
I hope the next episode is repairing the faulty chip. That would be an impressive chalange.😊
I repaired prototype IDC (Integrated Disc Controller) boards for the HP 7906/20/25 in 1979. I graduated college a few months earlier so it was a great opportunity. The IDC also used the PHI chip. Fortunately I had lots of spare Silicon on Sapphire PHI chips.
You would have loved the 7900A 😁 Many pcbs
@@andrewallen9993 The 7900 drive was a tank. You could drop it from a roof and it wouldn’t get an error. They were built on a production line next to the 7906. I think 7900 production shut down in late 79 or 80. I did some programming with the 7900 and a 2100 computer but not much else. I maintained the software that tested the disc media. Punched in boot loader code from the front panel back in the good old days. 🙂
@@77leelg Both 1mb platters of it 😁
@@77leelg the IBM 360 guys had to change the oil and the air filters. We just had to change the air filters and position graticle light bulb and the nicad head retraction batteries every 2 months.
@@77leelg we had 2 2100 CPUs with 4 7900 disks set up as a tabulator for horse racing courses and they would move the system around in a truck built to suit it.
Bringing out the logic analyzer. -- meh Bringing out Ken and Eric. -- HERE WE GO!
You got it right! It's the people that count!
I watch these episodes in total awe. I can understand what you're explaining -- no problem following along -- but certainly cannot match the ingenuity it took to get there. You guys are really amazing and inspiring.
großartig - Thank you, CuriousMarc for showing us such a beautiful view of a chip layout under the microscope on KZhead. It would be great to see a modern chip for comparison sometime.
I love when the music starts early! Means it's going to get interesting! Great view of the chip as well!
"It's a bad bit. A bad, bad bit..." "And here are the Castle Marc, we have but one punishment for data curruption!"
"The Input Buffer shall be equal to or greater than zero in length." - IEEE 488.2 apparently feeling the need to explicitly exclude the use of negative-length buffers
We don't like these either.
Apparently SOS (Silicon on Sapphire) technology is making a comeback, only this time it won't be Silicon that's the functional transistor material anymore - transistors have gotten far too small for Silicon to function properly, so it's potentially GaNOS (Gallium Nitride on Sapphire) mostly for thermal management, along with Gallium Nitride material providing the best efficiency trapping electrons within transistors as it should in order to function correctly, as Sapphire is obviously very good at heatsinking (it got decent thermal conductivity which could / would help with getting rid of heat from sub-2 nanometers gate-all-around transistors, getting heat to the heatsink or water block). Silicon on Sapphire technology is quite interesting because apparently it made some classic chips that used this technology radiation-proof, that even if you got powerful X-ray tube shining at the chip, it wouldn't miss a beat, thus was also popular for usage in satellites (RCA 1802 CPU was used in a few early digital satellites).
Master Ken - you are famous even in Switzerland. I hope we meet one time in person! Maybe I can take a short trip during my next US visit to have a coffee and some words.
I would have never expected to gear Marc say "i tought i was cooking" but hey. Anyways as always a great video. Always a pleasure to watch you'r content.
You know you've hit a big problem when you need to assemble the brain trust :-P - Amazing work as always, thanks for taking us along with you for the ride!
Great video, again. Many thanks for the information about Hi-speed data transfer. Anyhow, about the GPIB @ 3:38, it is a 8 bit parallel bus, capable of 1 MByte/s (not 1MBit/s) data rate. I as well still use a lot of GPIB based instruments, especially (again) the HP3458A, where I really needed this high speed transfer for data acquisition, 200kByte/sec for a 16Bit, 100KSa A/D conversion. That was in 1989, for a real-time FFT analyzer, programmed on a 20MHz AT PC, in assembler language, with a lot of tricks, like DMA, SRQ handshaking, low level PIO and GPIB commands.
I did lots of programming with HPIB with many different HP computers. Seeing the Parallel Polling reference brought back memories. One of the funnest thing I did was write a boot loader in assembly for several HP 21MX computers connected together. One was a server running RTE 1000. The boot loader used HPIB serial polling to interrupt the RTE machine and run a load program that transferred assembly language code to the requesting 21MX. The assembly language code was a program that performed several tests on HP disc media for 7906/20/25 drives. Operators no longer had to insert discs with the program code to boot the MX machines. The boot loader and HPIB solution saved a lot of time. Seems simple today but was a blast to create for a novice programmer starting what turned out to be a 28 year career at HP.
You guys are fantastic. Giving the "shoulders of giants" a big leg-up!
Master Ken is… a Master indeed! Silly makeshift Phi “repair”, could it be emulated with FPGA, or very cheaply with code and an ordinary microcontroller? Edit; Well, there’s E-Bay… 😅
amazing.. I know a lot.. But this is amazing. You guys are geniuses. Trouble shooting is a lost art...
Did you try resetting it? 😏 Fascinating equipment available today, but we have also created a plethora of protocols and modes. Thank you for sharing, this reminded me of troubleshooting a GESPAC stepper card. Talking about obsolete systems - I sincerely hope you‘ll someday see a GESPAC G96 system for an acceptable price..
As soon as Marc mentioned "disassembly", i knew that Master Ken would sooner or later show up in this episode 🙂
Congratulations ! We can see the old technology working. Not nowadays, as everything is encapsulated or done by software.
Master Ken , the Guru!
Why did the PHY chip go bad? Did Ken find any damaged transistors?
We still don’t know. Ken could not find a fault visually.
Beer, chicken wings, and 40 minutes of @CuriousMark Man, what a night!
This is soooo impressive!
I love this stuff. Thanks Marc and Ken.
Wow! Thank you!
Excellent episode! Thanks so much Team
Interesting that this AMIGO protocol seems to be a block-level disk access protocol, similar to SCSI. I always thought those drives communicated at a FILE level, with commands like "open file", "read from file", etc, like the Commodore PET/C64 drives.
Brilliant stuff, many thanks team
This was another excellent episode as always gentlemen! Marc, you have one of the best channels on KZhead! Thank you.
Really amazing work, Marc and the Marc-o-nauts.
Really good episode; loved the gate level chip explanation
24:35 an open three-ring binder is actually a great anti-static surface
And here I was thinking that Ken was going to reverse engineer and then build a PLA emulation of the chip! 🙂
Great analytics and repair. Maybe try to insert a smart buffer to that bad output line, Could be slightly logic voltage levels output is bit off, so Z80 see a 1 instead of a zero. I think if you probe it, you see its out of logic level specs. @ "Master" Ken, thanks for the chip inside tour.
Great that you were able to obtain a working Phi chip.
I love seeing the HP 16C on the desk, hopefully helping out!
Master Ken on the job again!
"So this chip is giving bad data to this chip.." "No, that's the good board." "Listen here you little...."
Never understood the "Silicone on Sapphire" meaning of the clash song till now. Thanks!
glad to see you use a generac generator!
As a 6502 guy I’m obliged to blame the Z80. It’s always the Z80.
Astounding. As usual. 👍👍
Better stock up on the Z80 spares, Zilog announced they're EOL recently.
From doggeral in the log of the Stanford AI Lab PDP-6/10 combo, circa 1970: The memory with the margins has the bits that are busted While the words that are working are in the core by the door. -- Apologies to Danny Kaye's classic movie The Court Jester
Ken is a legend !!!!
so much storage, what would one man ever need so much storage for? 😆 Nice to see the effort to keep these working. And sad to see another Z80 chip go bad now that they have gone out of production and become obsolete.
Don’t worry, FPGA will save the world… again
I have had to replace many dead Z80's..... Don't know why they die, but they just do. I keep a stash of them.
I think you can place last time orders for Z80 until end of may...
Also don't forget that someone is attempting to put a Z80 clone back into production at the end of the year. So not all hope is lost!
@@RikkiCattermole Maybe one day, my son will travel back in time (like John Titor did) to retrieve a few Z80 to save humanity.
incredible....!!
It's all normal nerd bull geekdom... then Ugwei Ken shows up and frickin' decaps and r-engineers the PHI chip. Mind blown.
"If you follow the channel" at the start is so much classier than "like share and subscribe my dudez"
Ken's fun comments are my absolute favorite part of these videos. 24:53 had me rolling... Marc: Ok, absolutely bad board with absolutely bad chip... Ken: what if it works? 🤣🤣🤣🤣
I noticed some microphone crackle. You can drop your video MP4 files into any if the spectral demix websites and cleanly extract voices from noises
Wow. You probably have the most HP9895A's that are working anywhere on the planet! BTW Marc, can you tell us what your microscope/camera setup is for examining the PHI chip? Thanks.
It’s an Olympus BH-2 with the metallurgic illuminator (can’t recall how this is called) with the original lamp replaced by a LED, long working distance dark-field capable objectives, and a custom adapter I made in my lathe for an old Sony Alpha camera body. I should make a video on it!
@@CuriousMarc Please do! It will certainly inspire others to make their own. The images it gave are first rate!
that was kool to see :) excelent team :)
Neat, neat !!
most entertaining!
What's with the map of Australia on the service manuals
This mark is on all documents shared by the HP computer museum in Australia
It’s because Australia is the center of the world. Everyone knows that.
As an Aussie, I agree 😂
@@LongnoseRob We still use those machines in Australia. It's all the latest technology.
So late in the design process management decided to switch from low active signaling to high active signaling. Since the documentation was already written, to save money HP subcontracted out the publication to an Australian firm. As everyone knows, due to the Coriolis Effect and Quantum Superposition of Spooky States the Australian printing presses (being upside down) inverted the signals at no cost! The only side effect being that the image of Australia was superimposed on the documentation. This cost saving effort saved HP from almost certain failure as a company, and allowed them to take the company-saving profits and reinvest them into executive compensation and stock dividends.
28:01 - I can see the bug in the Z80 - the chip actually jumped a few cm on its own! Looks like a bug!!
@28:52 - This very thing happened to me at the Staind/Seether concert Wednesday night. I went to kneel down and my knee overextended and collapsed underneath me. I ended up doing a small roll across the carpeted club level. By the time I got back to the hotel at midnight, i had a balloon for a knee and walking was hard. It sucks getting old.
Great video and nice success too. Loved it all but the chip tour was awesome.... Q: What Nano Meter architecture was the PHI chip ?....(Probably Micro Meter thinking about it....)
That pulled chip does make me wonder if it was pulled from a board that could otherwise have been repaired if it made its way into the hands of your team. Or worse, it was working, but parted out because they thought they could make more by selling the chips individually.
This begs the question: Floppy music from the drive stack?
I was surprised at 27:04 Marc didnt check /INT /NMI, /WAIT, /RESET and /BUSRQ lines on the Z80. If any of these lines were stuck high, the Z80 wouldnt behave normally. if /BUSRQ was stuck high, that could have accounted for /RD floating.
17:02 I found the use of '8' and '48' here very confusing. I assumed they were talking about decimal numbers. Whereas later it becomes clear the '48' they are talking about is 0x48, a hexadecimal number. 0x08 (hexadecimal) = 8 decimal = 0B00001000 binary 0x48 (hexadecimal) = 72 decimal = 0B01001000 binary Therefore the dead PHI chip was wrongly pulling the Z80's D6 pin high while the Z80 was executing its IN command.
👍👍
I wonder if the chip analysis would show the cause of the error from the phy?
Master Ken could not spot anything wrong visually.
16:02 the radiation sounding static really freaked me out! My iPad made it sound like it was coming from the window beside me! Yikes.
What is the sailplane model in the background hanging on the wall?
Eight drives.. time to set up RAID 60 on them! :D
👏👏👏👏👏
LD A,byte LD HL,table LD BC,table_length CPIR
Hello, Would a HP 2737a original paper reader interest you at all?
I have almost identical to this Agilent 1670g logic analyzer in work, but older iteration with HP branding. Unfortunately it doesnt have any use :(
Time to get out the IR camera and see what’s heating up!
Now that Master Ken reversed engineered the chip, if you run into another bad phi chip, maybe you could replace it with an FPGA on a breakout board that can fit in that socket? That would be a cool little project :)
Master Ken is actually Master Yoda disguised .
Eric! Where did you get these glasses!?!? Awesome!!!
Some day someone is going to have to start making new chips in their garage to get this old equipment running again! Is there software that can use image processing to automatically reverse engineer some old chips just with a picture of it?
I loved the vidoe, it's so intresting!! I am curious though, Is somebody paying you for fixing them? Do you sell them after they are repaired? or is it more like a hobby ??
You haven't figured it out yet? Mark is quite an accomplished guy, so I guess he can choose what he does nowadays. ;)
It’s just for fun. But we sort of became pros at it…
@@CuriousMarc That's fantastic! its a good hobby
I'm not a dumb guy, but I'm also not the brightest bulb in certain rooms. This room simply makes me feel inadequate in my technical knowledge; completely and utterly. Master Ken is the kind of guy that is just so smart that he's socially awkward. I revel in the time when you feature Master Ken. It's just fun to listen to him explain things. A curious question I have - are common life items as easy for you guys? Like if you had to replace a clutch or change oil or rotate tires? Do any of you cook well? I just imagine you guys being perfect at nearly everything you touch. I'm a former (recovering) perfectionist. Once I learned what I was, I realized that it was very debilitating. I always leave an imperfect mark on my projects as penance for abusing myself all those years.
I am *shocked* that they didn’t use their favorite MC6800 8-bit CPU.
How hard would it be to simulate the controller board hardware? It seems that you have completely characterized, or could completely characterize, the behavior of it. It seems at some point that it might be worth doing as their is not an infinite number of old phi chips lying around
If the bad Phi was consistent with its error, I wonder if it would have been possible to mod the rom code to work around the issue and fix the buffer overrun but as you were able to find a good chip you didn't need to try that
I was hoping master Ken is going to point the failure on the chip.
you can now make a 8 ways RAND storage : Redundant Array of Noisy Disk
Can you do a 8inch "floppotron" with all the drives?
Can you use the buffer overrun for arbitrary code execution?
I would bet almost certainly. 😅
Could I please get the HP85 test code for the drive? I wanted to run it on mine.
Watching this makes me wonder how hard it would be to reproduce a Fi chip given the fact that they've been successfully reverse engineered.
The next step is to make new chips from the reverse engineering process. AND fix the software error while at it.