I would like to add my experiences with this device. My father and I bought one of these at a garage sale in the early 2000's for a few bucks. It had a bad CMOS battery at that point, so POST would always land in the two beep "System battery is bad, replace and run setup / F1 to resume, F2 for setup" state. Easy to get past that, F2 to enter setup, navigate through the "checksum bad" warning, load defaults, save, and exit. The machine worked fine, although the battery had corroded and eaten some of the conductive plating off the inside of the case as well as eaten one of the battery contacts. I soldered in some random vanadium pentoxide rechargeable coin cell and that fixed the system battery warning.
Fast forward ~20 years to today, and I found it in a box, wondered if it would boot. Nope, it POSTs with 2 beeps, but the screen was all black (backlight on, but contrast setting too high). In the course of blindly screwing around with power switch/F2/F10/Esc I somehow managed to corrupt CMOS (?) and it stopped POST'ing. That was scary, until I popped it open to look for damage and decided maybe I should remove that coin cell. Back to POSTing, but screen still dead. Good.
I wondered if something was wrong with the contrast voltage generator so I took it apart and noticed that the contrast adjustment pot had the worst case of tin whiskers I've ever seen. Some of these whiskers were almost 5mm long.
The tin whiskers had shorted out the pot somehow: I removed them all with sticky tape, and after that the machine booted fine again. I guess maybe it will need this treatment again in another 15 years. Maybe there is some permanent remedy for tin whiskers (reflow the tin with a soldering iron?) but I'm not going to mess with it more. Having all those microscopic hairs of metal floating around right above the motherboard was scary. I blew it out as well as I could after that.
xjas said:
I was able to get it to POST by applying 12V to the battery contacts but the screen didn't come on.
The 12V to the battery contacts and successful POST beeps match my symptoms before fixing the potentiometer. 12V on the battery contacts should work fine. It seems to need about ~0.6-1A to successfully boot. It idles around 0.65A in DOS and 0.85A in Windows, if I'm remembering right. Booted just fine with a bench supply at 12V on the battery contacts.
I also reverse engineered a little bit of the power circuitry and quite a lot of the circuitry in the charger, since I wanted to build a replacement lithium ion battery pack, just for fun. The charger circuit is rated 13V, but it puts out the rated 1.3A up to around 15V, and then it continues to put out juice at reduced currents up to 16V. This is apparently due to the diode drop of the extra (somewhat mysterious) diode which follows the power supply's buck circuit that does the battery charging.
This schematic might not be too useful, most of this is the charger/PSU brick. Basically it's a standard flyback converter, with ~16.3V output for the laptop, and then powered from that same output is a standard buck converter with high side current sense (the 0.15 ohm resistor) followed by a series diode of unknown function, and this is what supplies the CHG/battery pin.
Then, within the laptop, this charge circuit (CHG pin, in my diagram) is connected to the battery positive terminal through a P channel MOSFET in parallel with a 68 ohm resistor. The laptop apparently controls the charging function with the FET, although the 68 ohm resistor always allows a tiny trickle even if charging has been terminated.
Interestingly, for whatever reason, the actual power output of the brick comes through an 0.1 ohm resistor that was added as an afterthought. It's soldered to the PCB on one end and heat-shrinked and then the conductor of the power brick's cable is soldered to the other end. All of the resistors (the 0.1 ohm, the 0.15 ohm, and the 68 ohm) are about ~2W power resistors. The 68 ohm 2W resistor looks hilariously out of place on a laptop motherboard.
After all of this the power brick's positive pin and the battery positive terminal are diode OR'd and the resulting power is sent off to the rest of the laptop. Not shown is that apparently there is a separate diode OR system for the backlight driver. Not clear on why that's the case.
I built a pack for the battery compartment using 8 Vapcell F12 14500 cells in AA holders in a 4S-2P arrangement, a slightly modified 4S protection IC (Amazon, "ANMBEST" brand, based on S-8254A chip), plus hot glue and a strip of FR4 PCB material to hold it all together and to form PCB conductors to mount the contact pads. I used the tab weld compatible pads (which had to be desoldered from the protection board before soldering wires) in order to make the new battery pack's contact pads. The missing battery contact spring was replaced with a strip of phosphor bronze sheet.
This works quite well, gets 2:40 of run time before low battery warning, at least 10 minutes of run time after low battery, and the charge/warn thresholds match up well enough. The cells end up charged to ~4V per cell (so pretty full). Laptop low battery beep is at 11.8V (2.95V/cell) and low battery triggered sleep is at 10.7V (2.67V/cell), which is a slight bummer because the protection trips at 2.7V/cell so sadly sleep doesn't ever happen.
I haven't reverse engineered the Ambra's charge algorithm yet. I expect it depends on the pack's thermistor to detect end of charge, with a time-out. Whatever the algorithm it doesn't actually charge this pack all the way. It seemed to stop fairly short at 15.3V (80%, ~1.8 Ah), and then only the tiny trickle from the 68 ohm resistor caused the pack to creep up towards 16V over many, many hours. I don't think I'll try to change this. As a note the system definitely does charge without any thermistor attached (I don't have anything wired to the third, middle contact).
If I was to do it again, I would pay more than ~$1 per AA holder (oh my god the contact spring resistance, had to jump over the springs with little bits of wire wrap wire), and I would find a protection IC that properly matched the F12's 2.5V cutoff voltage so that the Ambra's sleep function had a hope of actually working before protection trip. But overall super fun.
One cautionary note, the plastic snap tabs visible in the battery compartment in the image with the pack removed are plated with metal. This metal is VERY conductive (probably for ESD/EMI shielding of the machine). These tabs are actually touching the cells when the pack is installed, and only the cell sleeve is preventing a short and possible battery fire. Before I put this thing back in the box I'm going to shave the metal plating off of those tabs so they can't short the pack even if the cell sleeves are damaged.
Anyway, hopefully this is helpful to someone. Pretty sweet to have the thing running untethered.
