simon_e_hall wrote on 2025-11-24, 16:22:

Quick question, has anyone put a later version BIOS in this?

Rolling a new BIOS with sane defaults based on normal use would probably be a good idea. It had occured to me but I haven't taken a look yet.

Anyone know the toolchain handy for that offhand? I used to have GUI tools to do the work but don't think I have them archived.

A new BIOS, or any BIOS that actually takes advantage of the SoC features would be great to have. I'm not sure if there's a way to get the battery percentage reliably... And some way to shadow the VBIOS!

Also, what is the very last version of UniVBE that works with this? I am currently on 5.1a, as the latest release requires a 486 to run.

I've been looking for a way to get the original datasheets for the SoC with no missing pages, but it has proven difficult. If anyone can find it or knows someone who has it, please share it - what we have is missing essential pages concerning the AT bus clock timing issues, and some examples on overclocking the bus via software. We could increase the throughput of the video module without compromising the stability of the system if the PIT is accessed too often.

Yeah it does seem a shame that the video hardware is designed to work at higher ISA clock rates & the m6117 has some capability to do that, but we cant take advantage of it...
Id really like the datasheet for this as well, & the one for the RTC crystal (mine is marked 'K316CA' if that means anything to anyone?) so that I can verify the loading capacitor values- When I removed R38 on mine the clock stopped totally! I tried a few different caps & verified the clock frequency using a 1.8pF cap in series with a x10 scope probe (to minimize loading changing the freq) connected to a frequency counter with an oven temp stabilized crystal ref (and my spectrum analyzer for good measure...) It started up fine with 15pF (initially I guessed at 22pF for a 12.5pF crystal, the most common type, but it ran too slow) & seemed to give mine a clock that was about 90ppm slow - which, after i fixed the RTC clock/date and BIOS issue many have seen (ill do another post on that), it losses about 90ppm per day..I don't know the spec for the crystal so its hard to say this is OK - some are only 100ppm..

I should take my own advice on the parallel port; I blew mine up connecting a parallel to Ethernet adapter. After I replaced all the 74HC series chips I found that mine no longer had the problem I mentioned in a previous post where it couldn't pull the control lines low - it seems the 74HC04 originally fitted to mine was a bit 'weedy' and couldn't sink enough current.... Strictly speaking a fully working 74HC04 isn't actually specified to sink enough current to meet the type1 (let alone type2) parallel standards, however in practice its pretty likely to work OK - I shall try and go back and edit my previous posts so it doesn't mislead anyone...

Its still worth noting that these parallel ports are susceptible to damage connecting/disconnecting powering up/down things - as were the original IBM PC XT and AT parallel ports, they also used 74 (the higher power 74 & 74LS though) series chips directly connected to the port pins, so the pocket386 is really just inheriting this issue! Adding some resistors to the lines would offer some protection, 330R would limit currents to well under 20mA (max clamp current for 74HC) with 5V.. however it would further hamper its ability to drive cables and peripherals.... I haven't tested this yet, YMMV, Id try putting them at the peripheral end of the cable to help a bit...

If anyone does a new bios (Ive never done one... so Id have to learn..) there are some other default settings Id go for;
I think the memory refresh should be set to 15us (unfortunately!) its hard to be 100% sure on this as I cant find the exact datasheet for the DRAM chips, the attached is the closest I could find, but mine is fitted with A420616AS, the datasheet doesn't mention the 'AS', just the 'S', its perfectly possible the 'AS' is a different refresh variant, or it may just be something to do with packaging??? Mine seems stable with any of the three settings from brief tests - but hard to be sure from testing on this one, the refresh time is v temp dependent etc etc...
I also set:
GPCS to Disable
Parallel port to Bi-directional

I'm not 100% sure on this one either; the four control lines on the parallel port aren't open collector/tristate, so these four lines cant be used as inputs like on the original IBM bi-directional design... however the data lines should work fine for bi-directional use.

BIOS, date & time resetting and clock loosing time issue
Ive found the problem with the BIOS resetting, the power on reset/power good circuit to the M6117 is a bit too basic.
It seems when switched on the processor is running before the power is stable and corrupting the time/date/BIOS (though I guess its possible that it could be doing this from not powering down cleanly, i.e processor running as the +Vcc collapses).

At the moment it just uses R1 & C9 to set a time constant. The M6117 datasheet suggests either a ADM709MAR (a power supply monitor ic with a 4.4v threshold), or a LM339 based power monitor with only about a 3v5 threshold, but a substantial delay after that is reached.

I removed C9 and fitted a BD5245-2M power supply monitor chip - I had one to hand & its small easy to graft on upside down where C9 is (attached pic ). I fitted it with a 220N capacitor to give a switch on delay of around 1 sec - very long, but I wanted to counter issues where i struggle to press the power button cleanly. These ICs are available in 0.1V steps, this one is a 4.5V one.

This a downside & so isn't a complete solution yet. However it completely stops all problems with the BIOS getting corrupted & resetting to default & it also keeps date and time without issue - mine just looses several seconds per day, which is in line with what Id expect after measuring the RTC 32.768kHz clock with a frequency counter (also entirely reasonable for a 100ppm crystal if that's what it is!!).

The problem is that this monitor chip will reset the processor with even the slightest drop below 4.5V. It does this when its running from battery and you plug in the 12V external supply - as you might when the battery is running low... whats happening is the 12V input immediately switches off the battery, but it take about 0.5mS for the 12V to 5V regulator to charge up C21 470uF. During this time the voltage across C24 (Vcc) drops from 5V to about 4V.....
there are solutions to this - increasing capacitors to increase hold up time etc- though there is little space for larger capacitors, only about 9mm height (maybe put an 8mm dia one on its side..)... also could cause problems with battery supplying inrush at start up.
C21 could be made smaller to reduce the time it takes to charge up, but I think it doubtful you'd find a smaller one with adequate ripple current rating....
Another option might be to use the voltage on C21 to switch off the battery instead of the 12V input - this ought to work well.
Ive also noticed that with the BD2545 mine seems to reset occasionally when running from battery, though my pocket386 does seem a bit poorly atm, so hard to be sure if its another fault or the supply dropping and the BD5245-2M IC being so 'keen' that's doing it. -Edit- I've verified that this was a fault on my unit, Ive fixed this and it seems fine now!

having the power monitor with something like a 4.5V threshold does seem a good idea in many respects... But these issues would need resolving....
OTOH these drops in the +5V supply voltage don't seem to have caused any other issues; without the BD2545 its seems to run fine as you plug in the 12V supply, so another avenue might be to ignore them! and either choose a BD25**-2M with a lower threshold, or use a simpler circuit. Its possible that the existing R1 &C9 circuit with a longer RC time constant might be enough, but I'd be tempted to add a diode across the resistor to discharge the cap immediately when the supply drops (classic, basic logic reset circuit). Another option is to add a resistor across C9 to help change the threshold a bit.

When Ive more time Ill test some of these solutions..... so long as my pocket386 doesn't get more poorly!!

I appreciate the work you are putting in here to make this less novelty/toy like and a more solid system.

NP, id like it to be a bit more solid! (though my wife seems to want to see more of me!!)
Ive done some more tests today...but it maybe beer o clock here in the UK now!
In short I'm leaving the BD5245-2M power supply monitor chip in place, its the best solution I have so far...it helps it behave solidly & the only downside is that it re-sets if you plug in the 12V supply in when running on battery...
Increasing the supply capacitors just caused other issues.... from the 12V to 5v regulator overshooting on startup(eek!) to the battery driven 5V reg taking too long to power up, etc etc..
Using the +5V (on C21) to switch off the battery works great when you plug in the 12V supply, but causes an issue when you unplug it (DOHH!)...
Increasing the resistor on the existing RC power good circuit and adding the usual discharge diode I think improved things a bit from standard, but was far from solid - the BIOS reset to default after the fourth try switching off/on...

I could try a BD52** with a lower threshold, but TBH I'm not sure Id guarantee stable operation like that. I think the best plan is to delay the switch off of the battery when the 12V is plugged in by 1mS or so, but of course it would need to switch the battery back on as quick as it does now when the 12V is unplugged, no great difficulty to design a circuit to do that, the trick is to design one that's easy to graft on the the existing PCB, in a way that's likely to be reliable.....
Another option might be to use a second BD52** maybe a 5V one, to monitor the voltage on c21(output of the 12v to 5V reg, actually I think its 5.2V output or so to take into account the diode drop to Vcc) and use this to switch on/ off the battery- the output of the BD5245 would need inverting, but actually as I type I reckon that's the most promising option...

Hello all!

I’m looking for some advice from people who have owned the Pocket 386 for enough time to encounter some of the issues and annoyances I’ve heard about.

Now that you’re over the “honeymoon” period and the novelty has worn off, do people find themselves still using this device with any regularity, or do the issues mean it mostly collects dust?

I couldn’t help but notice a pretty decent community devoted to correcting issues and adding features, which is awesome. To give context I have no issues performing any community discovered hardware mods which improve devices like this as long as they are reasonably well documented.

Some initial posts and videos noted a lottery on the video chip which comes with the device. Is this still the case and are the consequences of receiving a 386 Pocket with an unfavourable chip that consequential?

For those of you who most put the device away after the novelty wore off, what were the main reasons?

I look forward to your opinions and thanks in advance!

Hi,
Ive had mine just over a year now, well over the honeymoon, still using several hours a week! quite a few hardware issues - some obviously just one off faults with my particular unit (weedy 74HC04 on the parallel port, bad solder joint on D8, iffy connection on the serial port to main PCB connector...)
I may have a go at documenting the more general issues Ive found in one place, together with my fixes for them, along with any other fixes from around the web...
There is one more issue Ive found - Ive had to replace the battery charge IC 3 times now ( in a lot of use over the year), its generally failed presenting a short circuit to the battery (which fortunately has built in protection) & in 2 of the 3 times also presented a low load to the +5V rail. Ill say more on that in another post I think...
Only yesterday the ribbon cable on my VGA/PS2 dongle failed - from flexing near the connector - it had tons of use though & I use mine with a mouse plugged in all the time. Fixing it was just a matter of cutting it short and re-doing the connectors..
The vendor does seem helpful when Ive contacted them.

In terms of things that would stop me using it - I think if i were to use the keyboard for writing a lot - coding or whatever Id find its small size frustrating, its 'good enough' for games and a bit of command line/editing batch files etc.
So for serious writing you'd prob have to be OK with an external keyboard, having said that I used to use the original Eeepc 701 a lot, its keyboard cant be far off a similar size...

I guess you need to be realistic about whats doable with a 386 - Id forgotten that back in the day, by the time i was doing any serious web browsing, that had Jpegs in etc or looking at anything but very small .PDFs (handy for game maps & walkthroughs) Id moved on to a 486 dx2 66. This machine is pretty painfully slow for that kind of thing...
So really its had most use playing old games, here its downsides are; 1. no digitized sound playback (surprising how much is just done wit the Adlib though), no joystick interface (flight sims, maybe x-wing - though its prob a bit slow for that anyway) and 2. its speed limits you to kinda pre-Doom ish games.
So if you have enough games that you'd enjoy on it you should be OK. Ive spent hours on elite & wolfenstein etc

I may look at addressing some of the gaming shortcomings with a docking station, Ive done a schematic for soundblaster playback , joystick & one ISA slot, I could be tempted to add 10baseT Ethernet (though realistically pulling the compact flash out to plug it into another machine or using a USB stick is prob all you REALLY need for file transfer).
If i do a docking station I do have another, very niche, use - running old electronics test equip that has ISA interface cards & DOS/early windows software... be a nice compact solution for that...

Ive seen a couple of these in my hands, both had the Cirrus Logic CL5420 Graphics IC, which seems pretty decent for this era machine, whilst bottom of the range in that series it seems OK for everything Ive used it for, there were some higher ones in the range with some 2D acceleration (aimed at windows?), slightly higher clocks/memory speed, but TBH not sure you'd really notice that much - perhaps on Doom?? Of course some of the improvements from the later chips came from when used with a VLB... I suspect CL5420 is a decent 'match' for the processor capabilities....

Maybe if someone figures out how to get it to use the higher ISA clock speeds that might give a noticeable diff with any of these chips, even the CL5420 supports it. I seem to remember a vague mention in the datasheet that the M6117 can be set to use higher clock speeds for particular address ranges, so maybe its possible....

I have an embarrassing amount of retro devices so I don't give the Pocket 386 as much time as it might deserve, but even though I don't game, I don't regret the purchase. Since I got it at the beginning sales, mine came with Win95, which was a non-starter, literally--the boot took forever, and while programs installed, opening them was glacial.

But with Win311, and a serial or gpio wifi adapter, one can ftp to the mother ship for file exchange and telnet (there's a telnet lynx site to text-browse the web), and along with Word 6, an uptime monitor, and a couple of other small desktop apps, actual work can be done.

The key-ed mouse is meh and my real rub is the lousy battery life. I use an external battery, but the interior battery won't give more than 2 hours.

What do you use for an external battery? I could be tempted...
If i do a 'dock' i was planning on it still being portable, maybe leaving space for a 10000mAh battery in there...

The battery charge IC failure Ive seen I think is down to an instability, very occasionally Ive seen an oscillation of about 30kHz on the power rail input to it & other side of the input diode, D7 . It took peaks of the 5V supply beyond 6V when I saw it. Im guessing that it has, very occasionally got too high and blown the 2312 battery charge IC. Putting some larger package 10uF caps across C26 &C27 stopped the oscillation. I notice on the datasheet a 2R resistor and 10uF cap in series across the input to 'absorb high voltage pulse when the charger is inserted and protect the 2312', so I can see why the designer of the pocket386 left them off, it sounds like the IC manufacturer is talking about some kind of resonant behavior when a long (inductive) lead is plugged in? hard to say.... But it could also be to stabilize the IC under some circumstances, so it could well be worth a go if you find you have to replace this IC.
the way this has failed, a word of warning, if you do try a replacement internal battery , make sure you get one that has an internal short cct protection /PCB/IC, not all do...!!!
when mine failed it got hot enough to desolder resistor caps etc around it when the +12V was plugged in!...

If anyone is trying my BD5245-2M reset cct mod I'd hang fire a bit; after running mine on battery for less than an hour I got a reset.... so it needs more investigation... I couldnt say what the battery voltage was at the time - the STC8G2K32S2 on mine has a couple of damaged inputs (I think when the charger IC went..) & cant read the battery voltage correctly 🙁 a new chip is on the way from the vendor...
It may be that Ill have to up the input capacitor, or just choose a lower threshold BD52**, its not likely Ill get around to this for a month or so....
Of course there is also the workaround of a Bios programmed with better defaults that other have tried...

@Jules_nerd Offline a while ago, I had picked up two new-but-remaindered Tekkeon external batteries on ebay. They're deprecated now, but still put out 2 hrs per. They came with a 12v setting and a collection of tips, one which worked.

OTH, you seem so electronically adept, I can't imagine you couldn't build one yourself. As another aside, do you think there's a more powerful replacement battery readily available that may fit inside the P386?

OK, Thanks those Tekkeon type products look ideal. Yep I could design an external battery pack... I do have some parts kicking around (the fastest/cheapest way to get replacement 2312 battery charge ICs was to order 10 pre-made charging PCB modules), so its tempting.... But I am too good at finding projects to do, I have more than I can possibly get through already! 😉
So I'm trying to train myself to buy products rather than make them! having said that I'm enjoying playing around with the pocket386 hardware mods.

I did look at the internal battery size, particularly as the foam around the battery on mine did make me think there could be space for a slightly bigger one.... But the 'rectangular' box the battery fits in does have some features; ribs, lid catches etc that encroach on the space, reducing it to not much bigger than the existing battery, I guess you might be lucky and find a 5000mAh one, but I haven't so far & its a lot of hassle for a modest improvement...
All Ive found is a replacement (https://www.ebay.co.uk/itm/358179902458) that is exactly the same capacity and the same size (well 1mm shorter!) It also has a protection PCB built in, I tested it and it cuts out with anything less than about a 1Ohm load (with internal voltage drops at this load that corresponds to about 3A current limit), this is rather more keen than the protection on the original battery, but to seems to run mine fine - even though right now mine seems to have a fault that is loading the +5V rail very heavily.
It came fitted with the wrong connector so i had to fit one, its a JST PH2.0 connector. If you do this be careful, there is no standard for the polarity on these connectors, if you buy pre-made ones (likely your best option, they are crimp connectors that require special crimp tooling ), then often the red and black wires are the wrong way around.
You can swap them if you are careful, the crimps are held in the connector housing with a little barb, you can lift the small plastic tab that the barb hits & gently pull out the crimped connector.

Wow, what a hassle for the neophyte. Good thing you know your stuff.

I am not getting the play time at the moment. (edit: To go further with this myself at the moment, but... ) Something interesting came up on another thread, when Jan (ChkCPU) said...
Re: Can a Pentium Overdrive 83 be installed on a FIC 4386-VIO?

What further I could find about it was..
486 Chipset Register Definitions from 1993

Now I know this is a 386, but I believe that ali used a chipset internally in this SOC that is a close relative of standalone chipsets which were used on 486 boards, thus we may get somewhere with changing ISA dividers with this tool.... not sure it's plug and play quite yet though, but gives us something to mess with.

Edit2: "Stalking" Jan around the forums has found some interesting possibly applicable things about Ali chipsets Re: Overclocking a TI486DX2 this config file for the 1429 MIGHT be a starting point for messing with this.