[Idea] Intel NUC boards in Thinkpads?

[axur-delmeria]

NVMe is (probably) useless because the bottleneck would be the RPi4's processor.

Maybe a PCIe multiplexer chip could work in order to have both SATA and USB 3.0? Problem is, the 500MB/s bandwidth would be split.

[atagunov]

I've ordered VAIO-SA/MBX-237/Foxconn V030 power board. I expect it to work: 20V Lenovo PSU to input (use a hybrid of T60 and Vaio cable), T60 battery to battery connector (attach Lenovo battery connector to Vaio board), possibly +5V or maybe +3V to some of the inputs and ground to some others. I expect it to supply power to outputs and possibly to charge the battery. There will be protection against battery overvoltage but sadly thermistor from T60 battery will remain unconnected..

I'm a lot more worried about EC now. Suppose I managed to build some board with IT8587E in the middle. Suppose I was running something like modified https://github.com/system76/ec on it. How will it communicate with the main system? Main system here is either a Rasberi PI or NUC or some other single board computer. That's the next hurdle, and I don't know if I'll manage to negotiate it. Whatever home-brew EC it is it will need to be able to tell the main system about battery power levels, it would need to be able to tell it that it's about to be shut-down because of lack of power etc, tell it to go to sleep because monitor lid has been closed..

Update: apparently the classic way for CPU to communicate with EC is via LPC bus created in 1998. CPU reads/writes data from/to registers on this bus. It seems in classic setup keyboard is also managed by EC and CPU gets keyboard strokes the same way over LPC. On SBC-s LPC may be absent or not exposed. If exposed it's still not guaranteed that registers wouldn't be taken by some SBC component. Even if not the OS would still know it needs to communicate to those registers.. I'm less certain of how things work in ARM systems.

In summary it seems that the most straight-forward way for a DIY EC to communicate to the main system would be over USB. That means that drivers/kernel modules/some other kind of coding would be needed on OS side. Guess there is no way around it. One small hope I had was that there may already be some drivers written for UPS-es, which also attach to a PC via USB. Maybe diy EC could mimic several USB devices: keyboard, mouse and UPS. I wonder if when on UPS computers display remaining charge similar to laptops on battery and if they apply power saving in the same way..

Update 2: for what it's worth there is a driver for BQ24735 based charger in Linux kernel sources. I understand VAIO-SA/MBX-237/Foxconn V030 power board has got BQ24725 which is very similar. The difference apparently is that 35 has got a "boost" mode e.g. it would have been able to power system from both battery and AC. 25 and 35 are said to be "software compatible" so 35 driver can probably be re-used for 25. Not sure if Linus will accept a change that makes the same driver be used in both cases For that driver to be usable BQ24725 would need to be connected to the main system via i2c. It appears this is easy with Raspberi PI though I'm not sure if this driver will compile/run on an ARM system. Attaching BQ24725 via I2C to NUC-s is a lot more interesting as I described earlier.. The hope is it can be done on those NUC-s that have a 2nd separate M.2 slot for a WiFi card. And then I'm not sure how far getting BQ24725 connected and serviced by a driver gets us.. But it's interesting.. It might give us some useful battery features..

Hmm.. it seems this driver would give us power_supply info in /sys fs for Linux but possibly not much more than that..

I'm once again thinking that a super low power Intel 8051 compatible microcontroller (8051 since that is what https://github.com/system76/ec is written for) with USB (CDC ACM probably) ability is what would be needed. Is “That cypress chip” fx2 series suitable?.. Back of the envelope calculation suggests it may last 160 years in sleeping state on 50Wh battery, so it might be okay. Or possibly something that https://chromium.googlesource.com/chrom ... latform/ec runs on - again extra low power, with USB. Generally this EC runs on a lot of controllers: https://chromium.googlesource.com/chrom ... ster/chip/ lists It83xx, lm4, mec1322, npcx, nrf51, stm32. Not sure which are extra low-power with USB. And there will still need to be some code on OS side to treat this USB device as a secondary EC (NUC for example does have normal primary EC). 3rd readily available open source embedded controller firmware is the one for One Laptop Per Child, for XO-1.75 and XO-4. They run on ENE KB3700 but that being another Intel 8051 instruction set chip would probably be portable to other chips with the same instruction set. Either of this would need to be adapted to communicate via USB. And in either case OS-side code would need to be adapted.

Update 3: I am wondering if for Linux it would be sufficient to write a kernel module similar to http://www.thinkwiki.org/wiki/Tp_smapi It is possible tp_smapi could be used as a template to write new code. Also we would break out the keyboard controller into a separate unit. So there would be the power board borrowed from Sony, some rudimentary EC to handle the battery and communicate with main system over USB and a secondary controller powered via USB from the main system to handle the keyboard and possibly the screen backlight brightness. One unsolved question here is how the main system would learn that it has transitioned from AC to battery and back..

[TheForgottenKing]

How did I not see this thread until now?

I've been slowly working on a project called the 701R - it's a 701C with a custom motherboard and frame and an OEM keyboard. At the moment I'm using a GPD Pocket 1 motherboard, but other boards could be used as well. Still working on wiring up the keyboard.

It's far easier to use a laptop motherboard as you don't need to handle powering it with a battery.

[TheForgottenKing]

I've actually been using a T420s that had a HP Probook 10510U+MX250 board inside for a while. As for the screen any 16:9 EDP screen will work fine using the HP screen cable. Ports are accessible through the CD-ROM port (empty now). Keyboard it used a daughterboard like the one in SK8855 and is wired to an internal USB. Touchpad and fingerprint should be possible but I've got no plans cutting the case to fit the HP ones.

wtf, pics?

[xytech]

A homebrew solution can be written to take care of the charging problem. Modern batteries communciate over the Smart Battery System protocol and most/if not all TP batteries uses this (with the notable exception of 2010 - 2013 batteries because Lenovo decides to do weird things with their batteries). Over at the /r/tp discord, there's apparently someone who's trying to come up with his own charging design based on a attiny85. However, he's designing the battery management of the battery, not the battery management of the EC. Integrating this design into 1 could fully fix the power management problem.

I've been talking with tagunov actually about a PCB shroud design with slotted design for SBC. I'm pretty for FOSS, and a core board like this: http://en.t-firefly.com/ enables unlimited upgrades with a standardised pinout for a slotted PCB (motherboard) shroud designed for an older laptop. Integrate the EC (kb, touchpad, power management, fan etc) onto the motherboard while using the main core board for the 'brains' of the machine would enable near unlimited upgrades for an older machine.

In fact, with the advent of RISC computing, an upgradeable machine with separate slots for CPU, RAM, storage and wireless would definitely make sense for the next decade. ThinkPads and other good industrial design don't have to go to the landfill simply because they're dated.

[TheForgottenKing]

Over at the /r/tp discord, there's apparently someone who's trying to come up with his own charging design based on a attiny85

https://hackaday.com/2020/09/02/buildin ... ad-battery

Not sure if he's one of us.

Still easier to just rip apart other laptops though. Dell Inspirons have a good motherboard layout for it, especially the 5368 motherboard (and other similar boards) which can be used quite easily and at a reasonable cost.

[atagunov]

https://hackaday.com/2020/09/02/buildin ... ad-battery

Hi, I was also interested in cobbling together a DIY battery. I wanted to place extra 18650 or bigger cells into ultrabay on R60. That'd be a DIY 2nd battery for a Frankenpad..

However what we see on the photo is a small battery management board (BMS) that you can purchase cheaply on Aliexpress or expensively on Amazon. I have actually rejected the idea of using this board because it looked too simplistic to me. Actual boards in original Lenovo batteries are a lot more elaborate. The boards in actual batteries have about 3 levels of fail-safe protection and incorporate a fuse. There is a controller running there and a couple of ways to stop the battery charging/discharging that are used if the controller fails and conditions get dangerous (overvoltage and possibly over-current). Pls. don't think I know this because I reverse engineered the boards I know this because I looked at "standard application" circuitry suggested by TI the maker of popular BQ8030 series of battery controllers and of other battery controllers often used in Thinkpad batteries and they have this. I feel confident the design of actual original batteries is no worse than that.

There is a fuse that cuts the battery off either if there is over-current flowing through it (like a normal fuse) or if the controller on the board decides it's time to cut cut it (there are 3 or 4 inputs to the fuse - 2 for passing through battery current and 1 or 2 extra to heat up an extra wire to blow the fuse).

I actually would feel a lot safer using a battery management board from a real battery. I wouldn't mind connecting new cells and re-setting the controller, I wouldn't mind a board from some 10-year old Apple battery, I could consider using the board from a KingSener battery - if it looks sensibly engineered, but I will not use this board from AliExpress/Amazon.

[flyingfishfinger]

I did this for an X61 battery:

viewtopic.php?f=30&t=124273&p=817895#p803911

R

EDIT: Just realized you already saw it / commented! Did you ever find something to charge your batteries with?

[atagunov]

I did this for an X61 battery ... Did you ever find something to charge your batteries with?

Rafael, my deepest respect for pulling off such a fabulous stunt I might try a free version of http://be2works.com/ with CP2112 at some point and give up if it fails.. Re charging I've ordered a VAIO-SA/MBX-237/Foxconn V030 power board from Aliexpress, so another month before it arrives. I'm a serial procrastinator so may never get around to actually working on it, but that board is my best plan. It's got some BQ chip but sadly it can't talk to thermistor in Thinkpad batts. I linked to its schematics a few posts above. The board will need to have some of the inputs pulled high or low, probably to 3v or 5v. In a finished solution an ultra low power microcontroller aka "an external EC" would need to do that. Cyprus FX2 might be suitable..

BTW in that hypothetical finished solution.. If that "external EC" senses battery has run low it would (1) send signal over USB hoping drivers in OS will shut it down or suspend (2) briefly "press" the power button on SBC hoping that it goes to suspend or switches off (3) but if all that fails it will need to just cut power. Questions:

- what kind of a switch would I use to cut power? It would need to be some power MOSFET right? Which one would I take?.. Need something easy to order.. I expect there'd be 20V there and the maximum current allowed would be several amps; ideally I'd use something that under no circumstances would need a radiator or would get too hot; it would be open almost always and only closed when battery is really drained
- what kind of a switch would I use to "press" power button on the SBC from this microcontroller? it would be nice if here the switch would work regardless of which polarity we have on the on/off button because you typically have just a 2-pole header that is very easy to plug a wrong way into... would I use some optocoupler? Okay, I guess I could use optocoupler, and and if user has plugged it the wrong way it will just fail to switch on the SBC and user will be instructed to try the other way then.. So which optocoupler?.. Or shall I use two optocouplers in reverse polarity so that one of them will work regardless? One thing I totally want to avoid is any possibility to leak current here.. Suppose that hybrid computer is put into storage for a year with battery in.. I'd still want to drain as little power as possible. It almost sounds like I want a mechanical relay

I'm such a noob unfortunately Please tell me if I'm being cryptic here

Update: I'm reading up on MOSFET-s a little

[TheForgottenKing]

Still waiting on my GPD Pocket to arrive, going to use the board with the Atom CPU in my custom 701C. It's been stuck in customs hell for over two weeks now with no end in sight. Soon I shall have my reward.

[atagunov]

Getting myself a couple of STM32F072-s to prototype external controller for SBC.
STM32L476 would have been a lot better - much lower power consumption and 128Kb or RAM instead of 16Kb, but it's more expensive I hope to make do with lower grade boards at least in the initial prototype

If I really want to be able to power down the SBC completely from the controller I may use VN5160S-E so called "high side integrated load switch".
For "pressing" down power button on SBC I'm getting some bog standard 4N25 optocouples.. these require 20mA of current to activate but it won't be often it's needed.

I'm now trying to figure out if these STM32 microcontrollers keep level of their output pins when put to sleep..
Keeping SBC powered up in off state (for example to help it keep the date) would require a "high" level on VN5160S-E control input and "external-ec" would be asleep to save power..

Or is it better to keep the SBC completely powered off when the whole notebook is off?

[TRS-80]

I had made a post over in What after classic Thinkpads? thread, but I can see you guys are quite further along here on some very similar ideas.

I don't have time to help out with many implementation specifics right now, but in general I agree strongly with things that xytech said about sticking strictly with F/LOSS/H.[0] The reasons for this are many, but essentially as independent hackers we simply don't have the resources to keep re-inventing all these different wheels when coming up with our solutions. Locked down hardware is essentially useless for our purposes, not to mention the fact that companies who engage in such dinosaur business practices should be left to die off as they don't deserve to be a part of the future. They don't want to share and share alike with everyone else who is participating in advancing things forward, so leave them out of it!

Unfortunately (from some people's perspective, not mine, lol) this precludes many of the familiar names you may have heard of like Broadcom / RPi as well as the Intel NUC mentioned in OP and even AMD. As I mentioned in my above linked post, I think the hardware we should focus on should come (roughly) from Armbian Supported Devices list as a rough proxy for "non locked down current ARM based hardware", along with perhaps some others in the future (although ARM is pretty much "there" right now already).

In addition to that, some truly modular architecture would be ideal as technology is always marching forward, as xytech also pointed out. And thus another key to all of this (IMHO) is sharing our progress, designs, etc. with one another. I am imagining for instance making detailed measurements of the interior of a 4:3 T60 frame, which could be shared, and then custom roll cages could easily be designed (using e.g., FreeCAD) to accomodate whatever SBC or board you like. Once one rollcage was designed and shared publicly under a F/LOSS/H license, others could be adapted easily from that. Soon (if we are not there already) we should be able to 3D print things like that with strong enough resins to make them viable. Or perhaps even traditional subtractive manufacturing.

I have a lot more ideas, but unfortunately for now they must remain as ideas as I have no time to implement. I am preparing this very moment[1] to go back into a work cycle to make some more money to continue to fund these sort of adventures. But the end goal is to buy some land out in the boonies where I can live cheaply enough to be able to afford to spend more time on my hacking endeavours and advancing the state of certain arts, many of which are directly related to a project such as this (3D printing, software, hardware, etc.). In the meantime, all I can do is shout my support and advice from the sidelines.

Cheers,
TRS-80

[0] Free/Libre Open Source Software/Hardware
[1] In fact this is what brings me back around these forums as I am preparing some of my ThinkPads to take with me when I go back out on the road for work.

[atagunov]

To report on this I'm not really making progress. I ordered a lot of relevant hardware from TaoBao and the parcel has just started its journey, but even once it arrives I'm not sure if I will set to work. Theoretically speaking I don't see any difference between using Raspberry Pi and "non locked down current ARM based hardware" - in either case you have to supply power to the board, connect to it via USB, toggle it's power button, and take video out of HDMI or DVI or eDP. No difference really - all my loyalty to open source notwithstanding.

Does it not seem like it might be possible to fit the smallest of the current boards like RPi compute module into the existing rollcage? Ultrabay is a particularly big space with a convenient point to get fan exhaust out of, HDD bay isn't that small either and of course there's the place where original CPU used to sit. In my understanding height is the biggest problem. Intel NUC-s for example tend to be quite tall and fitting into T60 isn't possible. They might fit into R60 just about if SATA connector is desoldered - straddled between HDD bay and ultrabay it seems. Low profile boards should generally fit more easily.

[TRS-80]

To report on this I'm not really making progress. I ordered a lot of relevant hardware from TaoBao and the parcel has just started its journey, but even once it arrives I'm not sure if I will set to work.

Oh I expect a project like this to take months, years possibly of R&D / tinkering (well, depending on exactly what we are talking about, more below).

Theoretically speaking I don't see any difference between using Raspberry Pi and "non locked down current ARM based hardware" - in either case you have to supply power to the board, connect to it via USB, toggle it's power button, and take video out of HDMI or DVI or eDP. No difference really - all my loyalty to open source notwithstanding.

Does it not seem like it might be possible to fit the smallest of the current boards like RPi compute module into the existing rollcage? Ultrabay is a particularly big space with a convenient point to get fan exhaust out of, HDD bay isn't that small either and of course there's the place where original CPU used to sit. In my understanding height is the biggest problem. Intel NUC-s for example tend to be quite tall and fitting into T60 isn't possible. They might fit into R60 just about if SATA connector is desoldered - straddled between HDD bay and ultrabay it seems. Low profile boards should generally fit more easily.

If you are going for the "quick, simple" solution then of course this will be the way. Just find something that physically fits and chuck it in there. Biggest problem then would be interfacing the display with the board I guess (which depends a lot on what display you are using). Probably some other things I am not thinking about, too, which you have considered further up thread already as I can see.

I guess I was thinking of a more proper integration, perhaps even having a board made (or maybe some interface board, so the ports are around the edge in the proper places). Then the questions become what display to use? And what is long term availability of things like keyboards? 4:3 displays are getting very hard to find, to say nothing of the old plastics just keep getting older... So maybe it is a pipe dream on my part. Your way is certainly more attainable, so there is a lot to be said for taking that approach.

The deeper you get into the development/hacking, the more you will start to come across some of the problems I alluded to about locked down hardware. Of course if you are just taking an existing board and chucking it in, you may not have any trouble. But then it's a philosophical discussion: do we want to promote / use / add to the ecosystems of locked down platforms? Personally I am not going to. You are of course free to do as you choose. In addition, my time and energy are too valuable and scarce nowadays to invest in learning about / improving any platforms which are locked down. As any endeavour like this is going to involve a lot of tinkering / learning, these are the conclusions I have come to, for myself. You may see things differently, of course.

In any case, I wish you the best of luck and look forward to reading on your progress, no matter what hardware you end up going with. It's all valuable expereience towards the greater effort! I hope anything I have said have not put you off the idea too much.

[TRS-80]

I had come across this back in early February (according to my notes) and it reminded me of this thread, so I thought I would re-post it here:

https://www.reddit.com/r/diyelectronics ... top_build/

There may also be some technical bits in there worth studying?

It will be some months away yet, but at some point I will have at least most of the parts (and time) to at least begin playing around with something like combining PineBook, modern/powerful SBC, and/or ThinkPad parts into some sort of Franken-monster...

Probably in the ThinkPad frame, keep the screen, and keyboard of course, just replace the guts with something much more modern and powerful.

Just dreams, for now, but dreams are important, they keep us going in the meantime...

[axur-delmeria]

I had come across this back in early February (according to my notes) and it reminded me of this thread, so I thought I would re-post it here:

https://www.reddit.com/r/diyelectronics ... top_build/

There may also be some technical bits in there worth studying?

If I had money to burn, I'd use a LattePanda as a base:

https://www.lattepanda.com/products/lat ... vated.html

[TRS-80]

Ever since they pulled their IME nonsense, Intel (well, all of x86, really, as AMD have their equivalent, too) have been dead to me! I am mostly now interested in ARM (currently) and perhaps in future RISC-V or even POWER arch.

A few posts up you can find a link I posted to Armbian Supported Devices List, which will lead you to many more interesting options.