ZählerSchau is a small, battery-powered unit that clips onto an electricity meter in Germany (water meters via a clamp adapter), takes a photo of the display once at a set date, and reads the digits itself. It sends the value as one short LoRa message to a server; when it is unsure, it adds a tiny grey picture of the digits as proof. Reading dates are adjustable, for example once or four times a year, and the goal is a battery life of 10 years or more, with a safe lithium cell and no soldering for the user. It reuses the camera and LoRa module the team has already got working, and it can be set up from a phone. Status: concept, research, the electronics board (corrected after a check found short circuits in its first production files), the mechanical design, a prototype print package, a presentation and a making-of film are done. Nothing has been built, ordered or measured yet. Open: who fits the radio module at the board assembler, real tests on meters, radio from meter cabinets, and the buying of the cell, capacitor and camera modules.
v1.7 (2 Nov 2026): check of the question in PCBWay's bill of materials. The 46 lines of PCBWay's copy match our board v1.3 BOM. The RTC part PCBWay asks us to confirm is the same part (manufacturer ordering code TA QC, −40 to +85 °C). One point is open: PCBWay's copy marks the LoRa module as not placed by PCBWay and says the team would fit it, which conflicts with the requirement to solder nothing. Ben did not answer within 120 minutes, so we assumed PCBWay places the module as in our quote, and nothing was sent to any service. The board itself is unchanged and stays orderable.
PCBWay's BOM copy matches our board BOM in all 46 lines; only two notes differ
The RTC part PCBWay asks about is the same part; the confirmation is ready and waits for approval
Open: PCBWay's copy says the LoRa module is not placed by PCBWay, which would mean hand soldering
PCBWay's copy totals 367.91 USD for 5 boards without shipping and without the module
An 11-slide overview of ZählerSchau: a small battery unit that clips onto electricity meters (water meters via a clamp), photographs the display on set dates, reads the digits itself and sends one short LoRa message. The deck explains the problem (only 2.1 % of German household meters report by themselves), the idea, the reading cycle, the fit on three meter types, the four setup steps, the battery budget and the honest project status: designed and checked on paper, not yet built. Every picture is a new studio render of the current design, and every number names its source. A portrait …
11 slides as PDF and PowerPoint, plus a phone version of every slide
New studio renders: unit on an electronic meter, a Ferraris meter and a water meter
Setup in four steps: stick the frame, click the unit on, tap the phone, confirm
Battery chart: about 24 % of one AA cell used in 15 years at 4 readings a year (estimate, 20 °C)
Small change v1.3: the main board is now the new board revision v1.3 from the electronics station. It replaces board v1.2, which was withdrawn because of short circuits in its production files. The real board model was compared part by part. Nothing the housing depends on moved; only a tiny resistor and capacitor (0.55 mm high) and a flat test pad were added on the side facing the back wall. They sit 2.0 mm below the back wall, 5.4 mm from the nearest pocket wall of the programming port and more than 3 mm from the chamber insert. The housing shell G1 and the chamber insert G2 do not change: …
Board revision v1.3 integrated: two new 0402 parts and a flat test pad on the back-wall side, nothing else moved
Housing shell G1 and chamber insert G2 unchanged: 0 mm³ geometry difference, print files and drawings byte-identical to v1.2
New parts 2.0 mm below the back wall, 5.4 mm from the programming-port pocket wall, at least 3 mm from the chamber insert
0 collisions in 5 states with the new board; smallest housing clearance still 0.35 mm
v1.6 (30.10.2026): full re-check of the main board after an external review. The most serious claim is confirmed: the production files of board v1.2 contain real short circuits, because the ground areas were not recalculated before the export, so v1.2 is withdrawn. The new board v1.3 fixes them, makes the wake-up reliable (power held at least 50 ms instead of 7.6 ms in the worst case, and no switch-on by leakage in sleep) and corrects the parts list; it is orderable because its exported files passed an independent check (0 shorts, 0 opens, wiring identical to the design). Outline, part …
Confirmed: the v1.2 production files had short circuits (8 signals tied to ground); v1.2 is withdrawn
New board v1.3 is orderable: its exported files passed an independent check (0 shorts, 0 opens, 0 clearance faults)
New release rule: export first, then check the exported files themselves; no check from before the export counts
Wake-up made reliable: power held at least 50 ms (v1.2: 7.6 ms worst case), leakage can no longer switch it on
v1.5 (30.10.2026, complete except what needs Ben's decision): PCBWay accepted the production Gerber files and quotes 5 assembled boards to Singapore at 168.90 USD (board 62.04, assembly 112.08 including the through-hole header J5 at +0.76); the parts price is not shown, because the BOM step needs the Special Notes agreed, which we do not do. JLCPCB has no price: the upload of the v1.2 files was refused by the platform (Upload not approved). The pin length is 6.0 mm. Recommendation: PCBWay. Ben decides on the Special Notes and on the JLCPCB upload. v1.4 (quote comparison, board unchanged): …
v1.2 finished: header moved 8.5 mm inboard, pin ends inside the board outline, no overhang
Cell connector moved to (57.0, 62.0) mm on the meter side; VTREF pull-up moved to (46.0, 52.0) mm on the user side
80 of 94 nets identical to v1.1 in exact nanometres; 14 nets re-routed; DRC 0 errors, 0 unconnected
Ben's requirement 1 kept: fitted header, Dupont plugs, no soldering
v1.4 (quote comparison, board unchanged): new instant quotes for 5 assembled boards, shipped to Singapore. PCBWay: 168.90 USD in total, with the through-hole programming header assembled and included (assembly 112.08, header +0.76). JLCPCB: board 41.97 USD, but no assembly, header or parts price without a file upload, and it does not offer the specified TG150 material or tented vias. Recommendation: PCBWay. The parts price for both still needs Ben's decision on a file upload. v1.3 (small change of v1.2): the main board is finished and orderable. The programming header stays fitted, so Ben …
v1.2 finished: header moved 8.5 mm inboard, pin ends inside the board outline, no overhang
Cell connector moved to (57.0, 62.0) mm on the meter side; VTREF pull-up moved to (46.0, 52.0) mm on the user side
80 of 94 nets identical to v1.1 in exact nanometres; 14 nets re-routed; DRC 0 errors, 0 unconnected
Ben's requirement 1 kept: fitted header, Dupont plugs, no soldering
Change v1.1: the housing parts G1 and G2 were rebuilt from Mechanical CAD v1.2, which adds a side window and plug channel for the programming cable in G1 and a notch in G2 over the moved cell connector. The print package is updated for these two parts and the plate: G1 now prints lying on its back wall (26 mm tall, 36 g, 0.7 h, about 0.05 cm³ of supports), where v1.0 needed 68 mm and 1.2 cm³ of supports. G2 is unchanged in print values. The plate is about 176 g and 3.3 h. Two small test pieces, a window section of G1 and a snap-hook coupon, are added for a check before the full print. The …
G1 rebuilt for the programming port: side window, plug channel and pocket walls, printed lying on its back wall.
G1 values: 26 mm tall (was 68), about 0.05 cm³ of supports (was 1.2 cm³), 36 g and 0.7 h.
G2 notch over the moved cell connector: no change in print values.
Plate 1 total about 176 g and 3.3 h; all other parts unchanged.
Small change v1.2: the finished main board v1.2 replaces the v1.1 board, whose programming header did not fit. The header J5 stays fitted, but now sits 8.5 mm further inboard, so its pins end inside the board outline. The housing got a rebuilt programming port: an 18.9 × 3.2 mm side window in the left wall, a plug channel over the board edge to the header, pocket walls that rest on the board, and a 0.35 mm recess in the back wall. Ben pushes ordinary Dupont plugs about 9 mm deep through the window onto the pins without opening the unit and without soldering. The housing now goes over the …
Board v1.2 integrated: programming header 8.5 mm inboard, pins end 0.5 mm inside the board edge
Programming port rebuilt: side window 18.9 × 3.2 mm, plug channel to the header, pocket walls on the board, back wall 1.65 mm locally
Assembly path free: the housing goes over the board stack with 1.0 mm to the header and 0.8 mm past the antenna
7 ordinary Dupont plugs fit with 0.24–0.33 mm clearance; 6.0 mm of free pin per the part drawing; 4.75 mm of each plug stays outside to grip
v1.3 (small change of v1.2): the main board is finished and orderable. The programming header stays fitted, so Ben plugs in his programmer with Dupont cables and solders nothing. Its pads moved 8.5 mm inboard (x 59.9 → 51.4 mm), so the pin ends stay inside the board outline and nothing hits the housing wall on the assembly path. The cell connector and the VTREF pull-up moved to make room. Exact comparison with v1.1: 80 of 94 nets are identical; the 14 changed nets were re-routed, all checks pass (DRC 0 errors, 0 unconnected). The housing still has to provide a side window for the plugs, and …
v1.2 finished: header moved 8.5 mm inboard, pin ends inside the board outline, no overhang
Cell connector moved to (57.0, 62.0) mm on the meter side; VTREF pull-up moved to (46.0, 52.0) mm on the user side
80 of 94 nets identical to v1.1 in exact nanometres; 14 nets re-routed; DRC 0 errors, 0 unconnected
Ben's requirement 1 kept: fitted header, Dupont plugs, no soldering
v1.1 (small change of v1.0): the programming connector is now a fitted 1 × 7 pin header (2.54 mm, right angle, placed by the assembler) instead of seven unfitted pads. Ben plugs in his STM32 programmer with ordinary Dupont cables and does not have to solder. The board, the copper and all other parts stay as in v1.0 (copper compared with the v1.0 files: identical tracks and vias, DRC 0 errors). The housing has to make room: the 2.5 mm header housing fits the 3–4 mm gap behind the board, but the mating pin sits 4.3 mm above the board and the header reaches about 8.5 mm past the right edge, so …
v1.1: programming header fitted, Dupont cables, no soldering; pin order documented (reversed vs MicBoard)
Housing: header housing 2.5 mm fits the gap; pin axis 4.3 mm needs a 1.5 mm lid recess, body overhangs the edge 8.5 mm (decision open)
Copper unchanged: 1324 tracks and 306 vias identical to v1.0; DRC 0 errors, 0 unconnected
62 × 88 mm, 4 layers, 114 parts; 10 × 10 mm window for the camera
Small change v1.1: the main board is now PCB v1.1 with a fitted 1 × 7 right-angle programming header J5, and the housing got a programming port for it: an 18.9 × 3.2 mm opening in the left side wall, a pocket whose walls rest on the board, and a 0.35 mm recess in the back wall. In the finished state 7 ordinary Dupont plugs fit onto the pins without touching anything, and no straight light path reaches the dark chamber. The header drawing shows the header is only 2.5 mm high (the 4.3 mm value is a length), so a deep lid recess is not needed. But the unit cannot be assembled with this board: …
PCB v1.1 integrated: the J5 header sits on the user side at the housing's left wall (the PCB notes call it the right edge)
Programming port designed: opening 18.9 × 3.2 mm, pocket walls on the board, back wall 1.65 mm above the header
Finished state checked: 7 Dupont plugs fit with 0.24–0.33 mm clearance, 5.4 mm of pin free outside, 0 straight light paths into the chamber
Blocker: with the fitted header the unit cannot be assembled (header must pass the wall; 887 mm³ collision on the assembly path)
The housing and adapter parts of ZählerSchau v1.0 are ready as a print package: 12 parts on one plate, about 177 g and 3.3 h in Creality Hyper PLA. These are prototypes for fit and function only. Hyper PLA is not flame-retardant and softens at high temperature, so the parts must not be used in the field at the 55 °C design case. The housing parts are printed without the header pocket, because the programming header of the board is not yet in the housing design.
One plate for the Creality K1 Max, 12 parts, about 177 g and 3.3 h (estimate ±30 %).
Prototypes only: Hyper PLA is not V-0 and softens at high temperature; not for 55 °C field use.
Housing parts G1 and G2 are printed without the header pocket; reprint after the pending CAD change.
The 8 heat-set inserts stay, pressed in by an assembler with a hot tool; Ben solders nothing.
The meter reader is now a complete detail design (stage 3): one camera unit, 68 × 94 mm and about 101 g, that clicks onto a thin frame glued to the meter. Frame E is for electronic meters, frame F for old Ferraris meters, and a clamp adapter fits water meters (register housings Ø 62–100 mm). The unit stands about 37 mm proud of an electronic meter, 31 mm of a Ferraris meter and 33 mm above a water-meter glass, and the camera sees a 69 × 52 mm field 25 mm away. After an independent review, the water adapter got a cut-to-fit foam seal that keeps the camera chamber dark on every register size, …
Stage 3 detail design: unit, frame E, frame F and water adapter complete with all screws, inserts, foam and tape
Depth in front of the meter: 37.3 mm (electronic), 31.3 mm (Ferraris), 33.3 mm (water); all within the agreed limits
Lens 25 mm from the digits (24.9 on its new stop), 69 × 52 mm field; nothing in the view except the meter itself
Water meters: the foam seal sits on the register rim and is light-tight for Ø 62, 80 and 100 mm (checked)
v1.1 (small change of v1.0): the programming connector is now a fitted 1 × 7 pin header (2.54 mm, right angle, placed by the assembler) instead of seven unfitted pads. Ben plugs in his STM32 programmer with ordinary Dupont cables and does not have to solder. The board, the copper and all other parts stay as in v1.0 (copper compared with the v1.0 files: identical tracks and vias, DRC 0 errors). The housing has to make room: the 2.5 mm header housing fits the 3–4 mm gap behind the board, but the mating pin sits 4.3 mm above the board and the header reaches about 8.5 mm past the right edge, so …
v1.1: programming header fitted, Dupont cables, no soldering; pin order documented (reversed vs MicBoard)
Housing: header housing 2.5 mm fits the gap; pin axis 4.3 mm needs a 1.5 mm lid recess, body overhangs the edge 8.5 mm (decision open)
Copper unchanged: 1324 tracks and 306 vias identical to v1.0; DRC 0 errors, 0 unconnected
62 × 88 mm, 4 layers, 114 parts; 10 × 10 mm window for the camera
The main board for ZählerSchau is designed, routed and checked: 62 × 88 mm, four layers, 113 assembled parts, with a window in the middle through which the camera looks at the meter. Between photos only the wake-up timer runs (about 0.2 µA, roughly 26 mAh in 15 years), and the board wakes by timer, phone tap or button. As Ben asked, the brightness of each of the two lights is adjustable in about 350 steady steps, so glare on the glass and too-dark pictures can both be avoided. All seven checks pass and the files for a turnkey order at PCBWay are ready. First quote: USD 163.90 for 5 boards …
62 × 88 mm, 4 layers, 113 parts on both sides; the lens looks through a window in the board
Sleep current about 0.2 µA: only the wake-up timer stays on
Two lights with adjustable brightness (about 350 steps, no flicker)
Wakes by timer, phone tap (NFC) or button; settings come from the phone
The research pass checked the ZählerSchau concept (v1.1) against its sources. The core choice holds: one board with a camera, a LoRa radio and a lithium-thionyl chloride cell with a pulse capacitor. The 15-year life is limited by the capacitor's stated life (10 years at room temperature, 7 years at 60 °C, to 80 % capacity) and by the cell's self-discharge in a hot cabinet (67 % of one cell at typical rates, 85 % at the maximum). An older study suggests the 55 °C rate could be much higher for some electrolytes, so the maker must confirm it. The stretch-release adhesive frame is not supported …
Battery: the capacitor's stated life (10 years at room temperature, 7 years at 60 °C, to 80 % capacity) is shorter than the 15-year target. This is the biggest new risk.
Hot cabinet at 55 °C: 67 % of one AA cell at typical self-discharge, 85 % at the maximum. A 1993 study predicts a far higher rate for its electrolyte, and cell makers recommend storage at 30 °C or …
Currents: the pulse capacitor gives about 2 A for 1 s and 0.5 A continuous; the bare AA cell 40 mA continuous. Fuse and layout follow from these numbers.
Mounting: no stretch-release product fits. Plan VHB 5952 on plastic and painted steel (93 °C long-term) and a removal method. A patent has a close mechanical concept; a patent attorney should review …
Recommendation: Option 1 "Chip reads digits" – a 70 × 96 × 32 mm unit clipped onto a frame around the meter display. The microcontroller and camera that already work together read the digits and send one short LoRa message on the set dates; when unsure, a tiny grey photo of the digits follows as proof (Option 2 runs on the same board). Everything is switched off by a nano-power timer between readings, so one AA lithium cell with a pulse capacitor is estimated to need only about a quarter of its charge in 15 years at four readings a year – the cell's own losses, not photos or radio, set the …
Recommended: Option 1 – the chip reads the digits, one ≈ 20-byte LoRa message per reading, proof photo strip only when unsure
One board for Options 1 and 2: the tested camera + microcontroller pair and the tested LoRa module
Timer switches everything off between readings: ≈ 0.25 µA asleep; 1–12 readings a year barely change the battery life
Battery: AA lithium-thionyl chloride bobbin cell + hybrid pulse capacitor, fuse, no charge path – the meter-industry standard