M-740 Reverse-Engineering And Hacking Adventures

Project Logs

Collapse

Keyboard and Options (maybe)
ziggurat29 • 01/19/2026 at 19:10 • 0 comments
The board has some unknown 'options jumpers' labeled S, P1-5 which are selected by a solder short and go into IC4. We don't know definitively what these are, but speculate that they might be keyboard/symbol set options. There are a few reasons for this:
- there is a video Eric previously posted of a guy who has one of these units and showing it off. Scrutinizing that video, it can be seen that it is jumpered as 'P3', whereas Eric's unit is unjumpered. Looking further at the video it can also be seen that the keyboard has a slightly different layout. The YouTuber's website has a Netherlands top-level domain, so I infer that he is in the Netherlands, and that the unit is a European model. This is an ancient unit from thrift, so who knows the original location it was sold. But I am going to hypothesize that it likely somewhere on the continent.
- the 'service manual' that Eric found on Sears Parts Direct shows a keyboard layout but, but labels the keys in groups 'A', 'F', and 'K' with sequence numbers therein. My retro-designation is that these are 'Accessory', 'Function', and 'Key' based on where they are placed on the keyboard. So that suggests the unit was intended to serve a more general market.
By pausing and scrutinizing the YouTube video in various places, I can make out the alternative key layout with some confidence. (E.g. Shift-1 is hard to discern, but reason suggests it is an asterisk.) The key matrix for this device seems to be:
```
P3 option (Europe; probably UK)
        Column
Row     p5.0    p5.1        p5.2    p5.3    p5.4        p5.5    p5.6    p5.7
        --------------------------------------------------------------------
p0.0            MargRel     Line    TabSet  Marg        CrsR    CrsL    Bksp
p0.1    Up      Dn          Right   Left    RET         TAB     Space   Mode
p0.2    CE      Size        <A>     GRAPH   Color       =       +       -
p0.3    ×       ÷           ???!    CODE    ShL/ShR     CAPS    REPEAT  
p0.4    0)      1*          2"      3/      4@          5£      6_      7&
p0.5    8'      9(          ;:      ¢$      ,,          -?      ..      ½%
p0.6    =+      A           B       C       D           E       F       G
p0.7    H       I           J       K       L           M       N       O
p1.0    P       Q           R       S       T           U       V       W
p1.1    X       Y           Z       !¼                                  
```
And since I've gone to the trouble, I might as well make one for the 'neutral' mapping that uses the nomenclature shown in Sears Parts Direct:
```
Neutral (Sears Parts Direct nomenclature)
        Column
Row     p5.0    p5.1        p5.2    p5.3    p5.4        p5.5    p5.6    p5.7
        --------------------------------------------------------------------
p0.0            F5          A10     F9      F7          F19     F18     F6
p0.1    F3      F4          F2      F1      F11         F8      F17     A12
p0.2    A6      A9          A7      F15     A8          A5      A1      A2
p0.3    A3      A4          A11     F12     F13/F14     F10     F16     
p0.4    K10     K1          K2      K3      K4          K5      K6      K7
p0.5    K8      K9          K33     K23     K42         K11     K43     K44
p0.6    K34     K24         K39     K37     K26         K15     K27     K28
p0.7    K29     K20         K30     K31     K32         K41     K40     K21
p1.0    K22     K13         K16     K25     K17         K19     K38     K14
p1.1    K36     K18         K35     K12                                  
```
I'm not sure how useful that is, but the nomenclature is interesting to me because the 'A' and 'F' groups seem to be likely to not change by market, whereas the 'K' group has printable symbols, and likely does.

Also, as previously noted, the sequencing of the keys for the alphabetic and numeric characters seemed oriented towards simplifying ASCII translation. So I suspect the bulk of those keys to remain constant across models (else that benefit would be lost), and that the model customization would be mostly confined to the non-alphanumeric keys (which included the shift of the numerics).

So the current hypothesis is that the P-jumpers select some alternative symbols for that subset of the 'K' designated keys.

It would be interesting to assess this by trying out the various P jumpers and seeing what comes out. The unit has a 'correction print' mode which shows stuff on the LCD before printing. That could be used as a way to quickly explore the various key mappings.

In the case of the known European unit, the jumper was set at P3 and the only new symbol was pound Stirling shift-5 £. So shorting P3 should cause that to show that new symbol on the LCD when in 'CP' mode. And the other symbols show in their new locations.

Since the only new symbol on the YouTube keyboard is £, this suggests...
Read more »
Random URL dump +photos
Eric Hertz • 01/09/2026 at 06:58 • 2 comments

I've too many tabs open!

The blog that started this addiction...

https://biosrhythm.com/?p=2143

Plausibly-compatible pinion-gear and very similar mechanism that may be the same brand/parts:

https://www.ebay.com/itm/133559488667

Mitsubishi Single Chip Microcomputers 1989 guide with datasheets for our two chips:

https://archive.org/details/bitsavers_mitsubishiishiSingleChip8BitMicrocomputers_54200624/page/n1/mode/1up

Here's the same databook from 1984.

The 1989 databook talks of devices it does not contain data for that refers back to the 1984 edition (specifically: I/O Expander chips made for the 740, which I thought *might* help us understand the LCD wiring weirdness, but no more... as the I/O expander port is definitely unused on our 740). But, apparently they removed *some* info, so who knows what/how-much by then?

(Thank you, Dave, for sending this *much less* scary link so I could delete the other I had!)

https://archive.org/details/bitsavers_mitsubishiishiSingleChipMicrocontroller_11129380

Another plausibly-compatible mechanism, source for pens, gears, etc... "The german guy" is linked here:

http://tandy.wiki/CGP-115

M54539P : 6-UNIT 700mA TRANSISTOR ARRAY WITH CLAMP DIODE --Stepper-Drivers, yeah?

https://www.alldatasheet.com/datasheet-pdf/download/99078/MITSUBISHI/M54539P.html

Renesas (Mitsubishi) 740 disassembler

https://github.com/mnaberez/m740dasm

Brother Type-O-Graph BP-30 Manual

https://archive.org/details/brother-type-o-graph-bp-30-manual

LOTS of useful info:

1984 Brother Type-O-Graph #H41530729

Status: Sightings

Hunter: Ted Munk (munk)

Created: 11-18-2016 at 06:49PM

Last Edit: 06-19-2025 at 08:49PM

https://typewriterdatabase.com/1984-brother-typeograph.7064.typewriter

Also from Ted Munk:

The Handwriting, Drawing Typewriter: Brother Type-O-Graph BP-30, with User Manual!

https://typecast.munk.org/2016/11/19/the-handwriting-drawing-typewriter-brother-type-o-graph-bp-30-with-user-manual/

Refilling old plotter pens — the Brother BP-30 (Type-o-graph)

https://darrengoossens.wordpress.com/2021/07/15/refilling-old-plotter-pens-the-brother-bp-30-type-o-graph/comment-page-1/
Here's the instruction-set/programming manual for MELPS740...
1989_Series_740_Software_Users_Manual
https://ia904505.us.archive.org/35/items/bitsavers_mitsubishioftwareUsersManual_11158124/1989_Series_740_Software_Users_Manual.pdf

This might be the same, the name sounds even more appropriate, but...
This keeps crashing Chrome before it finishes loading... but looks to be promising.

FOREWORD This software manual is for users of the MELPS 740 series. Register structure, addressing mode and instructions are introduced in each section. The enhanced instruction set with enhanced data

https://www.datasheetarchive.com/datasheet/7150000d4757ffe5?term=melps%2520740%2520programming%2520manual

MELPS740 vs 6502

09.24.98 Mitsubishi MELPS740/WDC W65CO2S: EDN’s 25th Annual Microprocessor/Microcontroller Director

https://www.edn.com/09-24-98-mitsubishi-melps740-wdc-w65co2s-edns-25th-annual-microprocessor-microcontroller-director/

My previous/first brief encounter with MELPS740... how to potentially replace the firmware with an external ROM (and maybe even leading to how to dump the internal rom for e.g. fonts, etc? Thoughts on that to come later...)

https://hackaday.io/project/188508-repurposing-otpmask-rom-microcontrollers/log/214010-vcr-board-m50950-melps-740-w65c02

Random image dump within random URL dump:

This is what we've traced so-far along with potential for Eva-mode hackery

This might be up Dave's alley, it's a bit beyond me... Someone's trying to add MELPS740 to ghidra...

Something about an unusual jump instruction?

https://github.com/NationalSecurityAgency/ghidra/issues/5002

I'm sure I lost several in the copy-paste-close-shuffle :/

Feel free to add any of your own!
brief stepper analysis
Eric Hertz • 01/09/2026 at 00:44 • 4 comments

Could it really be so simple?

This is the motor-winding-phases as sent to the driver-board when "arrowing-right".

When the button is first pressed, it seems to go exactly one character-width, then if it's held for longer than half a second, or so, it starts moving continuously.

My logic-analyzer isn't set up to capture the continuous motion, but looking at this, it looks rather clear to me that there is no acceleration/deceleration used.

Seems crazy, considering all the massively-varying loads.

It also looks like it times-out after several miliseconds of non-motion, cutting power to the windings.

This, I think, explains why the initial step *almost* looks like it's accelerating. But, really, it's because the first step turns on only one winding. If you look at the hex-value represented by the four wires at the start it goes 000000111333222

that same 1-3-2 pattern repeats, later, at sample 26.

But, it doesn't start with 0, instead it continues from 9. The "last" winding (D3) is still on when the pattern begins again.

This is [all] intriguing to me because I've done quite a bit of stepper-control over the years, and, frankly, I still don't really "get" it.

Most the steppers I've worked with are of the "NEMA" sort, whereas these are "pancake". I don't really think that, specifically, should make a difference in how they're driven except that, in a case like this, the motors are geared-down quite a bit. The pancake motors usually have a much lower resolution, e.g. 32 steps per revolution vs 200... Which, frankly, I'd've expected would make them *weaker*, since the electromagnets have to rotate the rotor that much further, across a greater air-gap.

But, maybe I'm mistaken... because this system has "wowed" me from the start at its speed and its ability to switch-direction so often without, apparently, losing steps.

So, then, we have another weird-to-me observation with these phases' powering-down when unused... In my experience doing similar, the motors often snap into a new position.

So, then, it comes down to "full-stepping" vs. "half-stepping", which I've gathered are actually not-well-defined. (whereas, maybe, "half-stepS" and "full-stepS" might be, confusingly, well-ish defined).

So, I'll take a moment to re-explain as I see it:

One way unipolar steppers can be driven is to only power one phase at a time, in sequence. Another way is to power two at a time. The photo shows both cases. Alternating between powering a single winding and powering two.
That very first step is only *one* winding powered.

Why is this important?

Well, most of the info I've seen is rather vague about what "half-stepping" vs. "full-stepping" really means, and-yet, talk about it like it's fundamental.

So, if you imagine the rotor as being made of a ring of four electromagnets, you might wonder what the difference would be if powering *one* vs. powering *two*...

It really depends on *how* the rotor and stator are designed to work together. It may well be that there's a de-facto standard, or it may be that there are numerous designs out there and yet the vast-majority of documents out there about controlling steppers just happen to make no mention of these differing designs!

OK, so imagine ONE simplified design:

The rotor is like a compass, with a bar-magnet attached to the shaft, and the stator has four electromagnets at the N, W, S, and E markings.

Imagine only one winding is ever powered, and it's "unipolar" in the sense that: when the electromagnets are powered they *attract* the North pole of the bar magnet/compass-needle... When they are *unpowered* they do not attract. And, being "unipolar," they are never reverse-powered to repel.

THEN: the needle will [what *I* would call] "full-step" between the NWSE positions.

OTOH, you could also power it with *two* adjacent windings: N&E, S&E, S&W, N&W... and the needle should point at those, what I'd call "half-steps," NE, SE, ... Yet even though it's...
Read more »
Sears Parts Direct
Eric Hertz • 01/06/2026 at 23:45 • 4 comments

Dave tells story of a similar plotting-unit he had long ago that succumbed to a broken nylon gear, and, well, it'd be a shame if that happened with this...

So, I had a weird moment-of-insight remembering that as-of just a few years ago (after the closure of the Sears nearby) I discovered that Sears Parts Direct still existed, and I thought it'd be worth a try to see if this machine was even in their database...

YEP!

https://www.searspartsdirect.com/model/mxnaod3pgi-000934/sears-26853700-office-parts

Unfortunately, but not surprisingly, none of the parts are available...

But there is some useful info, there, including maybe part-numbers that might be searchable elsewhere... but also, some diagrams!

I don't know if I'da started clicking-around more, then maybe I'd've found higher-res shots... but here's what I mined:
Keyboard Matrix
ziggurat29 • 01/05/2026 at 21:08 • 5 comments
I transcribed the traces into a kind of netlist of the suspected row and column elements. For every wire there is a list of keys on that line. And for every key there is exactly two wires. Well, there should be. I made a couple booboos with a couple keys seemingly having 3 wires but that was easy enough to correct against the board layout since I have all the wires on different photoshop layers that I can flick on and off to hone in on a specific set.

I mostly used macros made of muscle memory to pivot and merge data, and just used Excel for sorting in various ways. Turns out my prior thinking of rows/columns was not what I intended, but also my 50/50 guess of which group are which failed, so this was a happy case of fate negating my ignorance and it turned out that I labeled the networks correctly. So I avoided a re-label effort.

Short story is that Port 5 is the 'column' port, meaning that it reads a row of bits from each column. Port 0 and the first two bits of Port 1 are the 'row' ports, meaning that they drive the row lines in a 'one-hot' manner to scan the keyboard. An interesting thing is that column p5.4 is also wired to IC4's (the second MCU) /INT pin. This column contains keys such as 'return' and 'shift', so maybe something to do with that (especially 'shift'). And since the /INT is active low, that suggest that P5 is probably configured to have internal pullups.

The keyboard matrix is: (hopefully ascii art will come through, and unicode chars)
```
Unoptioned (US)
        Column
Row     p5.0    p5.1        p5.2    p5.3    p5.4        p5.5    p5.6    p5.7
        --------------------------------------------------------------------
p0.0            MargRel     Line    TabSet  Marg        CrsR    CrsL    Bksp
p0.1    Up      Dn          Right   Left    RET         TAB     Space   Mode
p0.2    CE      Size        <A>     GRAPH   Color       =       +       -
p0.3    ×       ÷           ???!    CODE    ShL/ShR     CAPS    REPEAT  
p0.4    0)      1!          2@      3#      4$          5%      6¢      7&
p0.5    8*      9(          ;:      ½¼      ,,          -_      ..      /?
p0.6    '"      A           B       C       D           E       F       G
p0.7    H       I           J       K       L           M       N       O
p1.0    P       Q           R       S       T           U       V       W
p1.1    X       Y           Z       =+                     ...
```
Read more »
Tracing Keyboard Matrix Finished
ziggurat29 • 01/05/2026 at 16:33 • 1 comment

After a couple days of tracing (and correcting booboos), I finally mapped out the keyboard. As mentioned before, there are 18 gpio lines, and in this case they map to IC1 Ports 5 (all), 1 (two bits), and 0 (all).

Now I need to collate these results into an actual keyboard matrix. I'm planning to transcribe this net data into Excel and do a little data massaging there to get the rows/columns worked out. I hypothesized that there are 8 bit columns which are read at once, and 10 row lines that are sequentially activated during scanning. There's no basis for this hypothesis other than it would be sane and likely simplify code in this constrained device. I semi-arbitrarily named P0 as the column port, but this could be wrong. (At least I know it's not port 1, haha.) The shape of the resulting matrix should make things apparent.

My mapping was fraught by 3 nets that are hidden under IC2. I have a guess that I feel is pretty good because of 2-D topology of the component side of the PCB (if there were more vias then this would be more ambiguous). So Eric will need to buzz it out to verify. This should be a quicky because two of those nets have easily accessibly test points, so it would just be testing continuity between a testpoint and a chip pin. I'll request that separately with an indication of the test points involved and my guess where they go.

There's 36 trace layers for the keyboard (18 lines, two sides), so I'm not going to post those. Instead I'll just post this one that is a merging of all of keyboard-side nets against a solid background (it's too noisy against the board shot). I also made a keyboard legend to help identify what button is what.

Oh, and there's a 'secret' key I labelled "???!" that exists on the board but does not exist on the physical unit case. What's even more trippy is that the elastomeric sheet does have a key for that button. And it has keys for buttons that are not on the board! This board has 5 buttons on each side, but the elastomeric sheet has 7! Another curiousity is that the board has two buttons for 'Return' (which are wired in parallel, so not really two buttons), but the elastomeric sheet only has a key for one of those buttons. Hmm! Mysteries of manufacturing.

So, now a lot of net transcribing into the warm fuzzy world of machine readable text and on to Excel collation....
More photos for clarity
Eric Hertz • 01/04/2026 at 21:26 • 1 comment

LOL, OK, so it would seem that simply cropping the photos on my phone causes them to be saved in a higher compression, as well... Sorry... No longer cropping.

Oh the irony that of the 13 photos I uploaded, this, the blurriest, was the only one that appeared. "for clarity".

Let's try again!

Holy Moly, that took *minutes*... I guess that's what one gets for hiding out from the rain in a faraday-cage.

Uhm... 11 more...?!

gotta stop there... many were duplicates anyhow. hope these help! (see Two photos with C5 lifted)
Main Board Photo Reverse Engineering 002; Keeb and Fiends
ziggurat29 • 01/04/2026 at 15:44 • 2 comments

The coup de grace will be mapping out the keyboard matrix. So boring! And 76 buttons! That implies a

⌈76/⌊√76⌋⌉ + ⌊√76⌋ = 18

lines for a minimal rectangular array.

It's going to be a lot of wires, and I don't know what are rows and what are columns, and I probably need to formulate a game plan before rushing into it. So to procrastinate creatively, I decided to map out the wires connecting the two MCUs. As mentioned before, the LCD data lines go to a smaller M50740 (3 Kib ROM, 96 by RAM), but the control lines are on the larger M50745 (6 Kib ROM, 192 by RAM -- how did we get anything done in those days?) So there's wiring between the two devices to coordinate.

Tracing:
So there's 5 lines between the two. The thing that is striking about the M50740 is that pretty much everything except for the 8 LCD data lines and these 5 'coordination' lines is tied high or low. So surely it must be doing something more that providing another 8 bit port because we used more than 8 bits to get that! EDIT: Eric clarified via buzzing out that all the LCD lines are brought to IC1, so 6 of these 11 are those.

Oh, there is a little more -- there are 5 'option' solder bridge pads that tie directly to the M50740. I have no idea what they are. None are bridged in this unit.

OK, having done that, I couldn't resist doing at least one wire of keyboard just to see what it looks like. So I picked an arbitrary one to kick the tires with the process:

The thing that is striking to me is that this has 8 buttons in it. Coincidence? And it runs to a gpio on IC1. So I think IC1 does the keyboard scanning, too.

As said before, I think the keyboard is going to be some work, so I'll try to get a game plan. I might wind up doing one layer for each line, row and column, but that's going to be 18 * 2 layers, hahahaha! I wonder if my ancient Photoshop can handle it? Yeah, I probably should have done this tracing in a vector graphics tool. Hindsight!

Anyway, one bit of reasoning is that since I have traced out a row (or is it a column?), then all the other lines on those buttons must then be columns (or rows -- I guess it just depends on the orientation of the matrix). So I'll next proceed doing columns of those buttons.

Because it is likely a 8x10 matrix, it seems sane that the side with 8 lines would go to the same port. That would simplify code, and this things only got 6 KiB rom and 192 bytes ram. So that can be an intuition that might help in puzzling things out at some point.

Eric's addition, since photos can't go in comments:

How 'bout that groovy key-layout... there's a cents symbol!
Beeping Results...
Eric Hertz • 01/04/2026 at 05:17 • 2 comments

Here's the total pinout of IC4... Dave was spot-on about everything he could see in the photos, and even the guesses he had made where it couldn't be seen.
I turns out the LCD's data lines are actually all connected to both chips. My guess is that IC2 actually drives the LCD, and that IC2 is also using that same data-bus to communicate with IC4... But there's still some weirdness about it all... because what, then, does IC4 do?!
Its INT pin does seem to go all over the place, and I was able (I think) to track it down to a keyboard key in the board photos (I'd really rather *not* take that board off its mounts, since I feel I was lucky getting that LCD back into the case without having to remove it from the zebra strips and risk total misalignment).
In the past couple years (was it the AlphaSmart?) I finally learned that there *is* a way to use a key-matrix without *constantly* scanning. I can't recall the details, something about setting all rows low, and ORing all the columns? I suppose that could be done with a resistor network... (seeing as how few diodes and gates are on the board). So, then, when a Low is detected, only *then* run a scan for which row/col... So obvious, once I saw it... So, here, maybe that's what IC4's INT pin is doing...? Just check if *any* column goes low, then trigger IC2 to scan which one...? Dunno.
(I just realized there's no diodes in the matrix!).
Again, Dave's results were spot-on... Here I've also figured out the NOR logic used for the Solenoid and the left-side-detect switch.
Interestingly, I wiggled the solenoid's shaft by-hand, and I think Dave's theory of its latching is correct, it seems to hold in either position. It turns out the solenoid shaft is also magnetic! So, then, one polarity would suck it in, and the reverse polarity would spit it out! Never seen the likes. I've seen similar where *two* windings were used, but this solenoid has only two pins! Cool! And, kinda surprising that its holding-force is enough to overcome the spring in the pen, and movement along the pen-pusher thinggy.
So, I think Dave's right, the weird blue chip is an H-Bridge, and the NOR logic assures that it doesn't try to go both forward and backward at the same time...? Maybe converting a "pulse/direction" uC output to actual polarities for the H-Bridge... or vice-versa.
The Low-Batt signal does go to two pins on IC2... one is an interrupt, the other I think is just a GPIO for reading its value...(?)
The "jumpers" (S, P1-P5, all unpopulated) all go through a pull-up resistor-network... (poorly-drawn).
I think that's it.
Yep, *most* of IC4's GPIO pins are either grounded or tied high. Crazy. What's that thing *do*?
Main Board Photo Reverse Engineering; Peripherals
ziggurat29 • 01/03/2026 at 16:20 • 6 comments

Having figured out the signals on the interboard connector of the driver/power board, I felt impelled to propagate that knowledge onto the main board circuitry. The main board is two sided, so the existing photos are helpful.

As with the 'driver board', I deformed and color tweaked the images to where they were the same size and shape and had adequate visible contrast. The big difference with this board is it's much larger and has a lot more traces that bounce back and forth through vias. So it's easy to get lost.

To help with this I used Photoshop's 'layers' feature and put each side on a separate layer. Then I mirrored one of the sides so that the holes would line up. This required more finessing of the distortion, so I made the two layers transparent and colorized to make the tweaking easier. Since most of the stuff was on the 'component' side, I mirrored the keyboard side to match it. This will confuse me later, I'm sure, since the keyboard will be mirror of what the manual depicts, but making work easier here is worth it, I think. Also, I'm sure you can do all this stuff in gimp -- I just happen to have an ancient (2005?) Photoshop installation and I'm familiar with the tool. And I'm sure there's better tool choices even than using a raster editor.

The layers then looked like this:

Color looks like crap, but that was deliberate to help while I was tweaking the distortion to get the vias to line up perfectly since I made these low opacity.

Work then proceeded by copying those two layers and manually tracing out circuit paths on those new copies.

To keep myself sane through all the traces, I decided to do one peripheral at a time. The first one I did was the carriage stepper phase control. I used the same colors as the phase wires to make it more obvious what was going on. Result:

So that device goes to four lines on IC2. I haven't looked at the datasheet, but it's a microcontroller, and surely that is one of its GPIO ports.

Now I need to add more stuff. The platen motor makes sense, but I don't want to confuse myself since it has the same color wires. So I make a new set of layers focused just on that peripheral. I also decided that it would be good to know what traces have already been mapped, so I developed a process where when I start a new peripheral, I use the prior peripheral mapping and color all its traces grey. Then I map out the new peripheral on top of that. Doing it this way means that progressively I will accumulate mapped traces separate from unmapped traces, and still have the focused visibility on the peripheral I'm working on. My photoshop file gets really big as I add loads of layers, but disk is cheap these days. (It wound up being over 500 MiB, so most of a CD-ROM. And I've still got more work to do.).

The platen motor traces result:

A challenge was that one of the lines disappears under one of the parts, so I don't know where it goes. But since these are in a logical group, my guess is IC2.52.

The pen up/down solenoid was colored according to the jack. The colors don't really mean anything because I don't know what IC7 does on the driver board, however there is a control line from the main board that is physically near the phase line of the solenoid, so I just propagate color that way. But it does not have intrinsic meaning.

Unfortunately the solenoid wiring goes to and under IC1, so I can't trace it back to the MCU from the photo. But this is as far as I got. It will probably need some buzzing out of stuff to finish this part.

There is a limit switch that indicates the pen revolver has reached the beginning of the line. I thought it was a magnetic reed switch, but Eric says it's a leaf switch. It just has a single line going to IC1.12:

The low battery detect similarly is a single line,...
Read more »