Where the outside numbers come from

A2P SMS

You read it as a text message. The industry calls it A2P SMS — application-to-person — and the distinction is not cosmetic. Twilio defines it as any kind of message traffic in which a person is receiving messages from an application rather than another individual, and which is not expected to receive a reply. A message from a friend is P2P, person-to-person, and the carriers treat the two differently.

That is why an alert from a controller is not the same job as a text from a phone. A2P traffic in the United States has to be registered with the mobile carriers before they will carry it, under a scheme called 10DLC — a ten-digit long code registered to a named business for a named purpose. It is a queue rather than a build, and it is the thing the check-in on Designed, not built is waiting on.

Is it really the same carrier path your bank uses? Close enough to say so, and worth being precise about. CTIA, the US wireless industry body, puts a bank's alert and a tank controller's alert in the same category under the same rules. Its Messaging Principles list large-to-small businesses, financial institutions, schools, medical practices, customer service entities, non-profit organizations together as one kind of sender, and name service alerts or notifications (e.g., fraud, airline) as the kind of message they send. Where a large bank is likely to differ is the number type. It may send from a five or six digit short code rather than a registered ten-digit number. Same category, same rules, a different doorway.

One more thing worth knowing, because it makes the term harder to look up. CTIA has stopped using the word. Its current edition notes that The prior version of these Principles and Best Practices used the terms ‘Consumer (P2P)’ and ‘Non-Consumer (A2P).’ The industry now writes Non-Consumer where it used to write A2P. The carriers, the registration and the rules are the same.

Twilio glossary, Application-to-Person Messaging — twilio.com. CTIA, Messaging Principles and Best Practices, May 2023 — ctia.org. Both checked 29 September 2026.

The $290 to $360 reference

Have your say shows you a price before it asks what you would pay. That number is $290 to $360, and this is exactly what is behind it.

How it was built. The entry controller from each of five brands sold in the United States, priced off the seller’s own page in September 2026. Not a review site, not a marketplace listing, not a figure I remembered. What is counted is the controller, its power supply, a temperature probe and a pH probe. What is not counted is switching equipment on and off, and adding that costs another 30 to 100 percent on top.

Where the range comes from. The five entry controllers run from $289.80 to $361.19, every price read twice. The brands are named and compared further down this page.

One line sits outside it. The brand that sets the $289.80 floor also sells a full-size controller whose equivalent basket is $657.80. The range is what it costs to get in, so that one is left out of it, and the comparison carries both.

Five entry controllers, priced from the seller’s own page, checked 29 September 2026 — aquariumcomputer.com, bulkreefsupply.com, topshelfaquatics.com, coralvue.com and bulkreefsupply.com. Due to be re-priced March 2027.

What six controller lines cost to do five jobs

Every comparison I could find on a manufacturer's site compares that manufacturer's own models. The ones that do cross brands are feature grids with ticks in them, and a tick does not tell you what the thing costs or whether the cheap unit can still do the job once you have added the second box. So here is the plain version: Six product lines from five brands, five jobs, cheapest first.

I am not in it, and that is deliberate. Nothing I have built has run on a tank yet, so putting Pharos Pods in this table would turn a reference into a sales argument, and you would be right to stop reading. When there are test results to stand on I will add a row and this paragraph will change. Until then this page is only useful to you if it is useless to me.

The five jobs. Each one includes everything above it, because that is how these systems are actually bought — nobody starts again at the second purchase.

The job	What it means
Watch	Reads temperature and pH and puts them on your phone. Switches nothing.
Watch and switch	Adds control of mains outlets, so it can turn a heater or a pump off.
Watch, switch and dose	Adds dosing pump control. Read the dosing table underneath before you read this column, because three different things are being called dosing.
Watch, switch, dose and notification	Adds an alert to your phone when something goes wrong.
Watch, switch, dose, notification and battery backup	Adds a battery backup the manufacturer sells or names, so the system survives a power cut.

What each one costs. Street prices in the United States on 29 September 2026, read off the seller's own page. Where a manufacturer and a retailer disagree the lower one is used and the gap is noted underneath. Sorted by the price of getting in.

Line	Watch	and switch	and dose	and notification	and battery
GHL ProfiLux Mini	$289.80	$409.90	$966.80metered, apart	$966.80	not sold
Neptune Systems Apex	$319.99	$644.98	$1,019.97metered	$1,019.97	$1,019.97plus a battery you buy
Red Sea ReefControl	$329.00	$648.00	$1,097.00metered, apart	$1,097.00	not sold
Reef Factory	$349.98	$449.97	$576.96metered, apart	$576.96	not sold
CoralVue HYDROS	$361.19	$608.99	$699.07timer only	$699.07	$1,432.47plus a battery you buy
GHL ProfiLux 4	$657.80	$952.60	$1,367.40metered	$1,367.40	$1,393.30plus a battery you buy

Read the dosing column with the note in it. Three different things are being sold as dosing, and a price on its own does not say which one you are buying. The next table is what the money actually buys.

What the money buys. Same order, same lines.

Line	How it doses	A screen at the tank	Text message	If the power goes
GHL ProfiLux Mini	Metered, but on its own. GHL sells a stand-alone doser, the Doser 3 SA, with no additional GHL master device required. It runs beside this controller rather than from it, because the Mini has no bus for a doser to join. The cheaper route is one of the four sockets on a timer, running a pump you supply, and GHL's own guide says for most dosing setups, configuring a dedicated dosing pump is recommended instead.	On the controller itself, no extra cost. An illuminated blue graphic display with six keys, on the controller itself. GHL publishes no size and no resolution for it. It is not a touchscreen, and this line has no port for the one GHL sells.	Not attachable. GHL's SMS modules connect over RS232 or over the PAB bus. This controller has neither.	Not documented for this line, and the hardware says why. There is no power cut input to detect the cut with. An alarm sounds and blinks at the tank.
Neptune Systems Apex	Metered, driven by the controller. Two heads, calibrated with a measuring cylinder that comes in the box.	Apex Display Module, $149.99. A backlit LCD with mechanical buttons. Neptune describes it in those words. It is not a touchscreen, and Neptune does not sell one.	Discontinued 5 November 2025. Neptune's own page says SMS messages are no longer being sent, and that no new SMS methods could be added for US and Canadian carriers from 20 August 2024 because of the carrier registration rules. Under those rules, in Neptune's words, an alert by text is at best a gray area. Email and app push remain.	Documented: if its cloud goes down the controller keeps running locally, and there is an audible alarm at the tank. For a power cut, Neptune's reference manual has you put the base unit on a UPS you supply, and a line in an alarm's program sends the alert when the mains goes. Your router needs power as well, or the alert cannot leave the house.
Red Sea ReefControl	Metered, but on its own. Bulk Reef Supply's guide to the ReefBeat app lists the doser among the products that use their own ReefBeat-compatible hardware or dedicated controller and appear independently inside the app. So the four hundred and forty nine dollars buys real metering in the same app as the controller, as a device of its own rather than one the controller drives.	ReefControl Power 8 Control Panel, $99.00. Eight backlit on/off buttons and a master. It shows no readings at all. To see a number you open the phone.	Discontinued 18 November 2024. The text sent when a device had been offline more than six hours was retired. Push and email remain. Its only published offline threshold is six hours.	Its cloud notices a device has stopped checking in and pushes an alert, which covers a power cut and a router failure by the same mechanism. That is the best architecture of the six. Its published threshold is six hours, and whether alerts fire faster is not documented.
Reef Factory	Metered, but on its own. One head, metered, and it takes its instructions from the company's app rather than from anything in your house.	None. No system display. The pH meter has a small screen showing its own reading and the temperature monitor has another. Nothing shows the tank as a whole.	Cannot be verified. SMS was historically offered as the one paid extra. No current documentation says whether it still works. Treat it as push only until the company says otherwise.	Not documented, either way, for a power cut or a router failure. Its devices are mains powered WiFi clients with no battery.
CoralVue HYDROS	A socket on a timer. Its own listing for this pump says a 120V timer or controller is required for automatic on/off operation, sold separately. The controller is the timer, and the pump runs at its fixed rate for as long as the socket is on.	None. The whole catalog was checked. The interface is the phone, plus a web page on a computer.	Discontinued 1 October 2021. CoralVue's stated reason was to make HYDROS a more sustainable service. Push replaced it, and it noted push carries no carrier charges.	Documented. Its cloud notices the controller has stopped checking in and tells you the tank is unreachable. That alert says the tank went quiet rather than why, and other alerts wait until the link returns. With the internet down and the power on, the controller keeps running its own logic, and a Bluetooth override works from beside the tank.
GHL ProfiLux 4	Metered, driven by the controller. Four stepper-driven pumps on the controller's bus. The doser cannot run without a controller and the controller cannot reach it without the cable, which is why the cable is in the price twice.	GHL Control Pad, $249.90. A real 4.3 inch capacitive touchscreen, glass fronted, powered down one cable from the controller. Of the six, this is the only first-party touchscreen that exists.	Not purchasable. GHL documents an SMS module in its notification guide, and the older outdoor version even advertised working through a blackout on its own battery. The module for current controllers appears in neither of its store catalogs, and the only SMS product listed is for a discontinued controller line.	Documented in the most detail of the six. With the controller and the router both on a UPS, the power cut adapter flips the state, a maintenance program sheds load to stretch the battery, and an alert goes out. Without a UPS there is no alert. Its app also reaches the controller over the local network with the internet down, which none of the others document.

Four things the table does not say out loud

1. Five of the six cannot send you a text message today, and the sixth cannot be confirmed. Two discontinued it outright and published the date. One replaced it with app push and said so. One still documents the hardware and does not sell a module that fits a current controller. One cannot attach a module at all. One cannot be verified either way. The dates run from 2021 to 2025 and the reasons differ. The earliest gave sustainability. The most recent names the United States carrier registration rules for application-to-person traffic, which is the same scheme anything new has to go through, and says that under them an alert by text is at best a gray area.

2. Hearing about a power cut takes a battery or a check-in. One line sells a power cut detector and has you supply your own uninterruptible power supply. One documents the same route with no detector to buy, the base unit on a UPS you provide and a line in its alarm program. One sells a power center that can hold the controller up, with guidance to keep heaters and return pumps off it. In every case the router needs power as well, or the alert has no way out of the house. The other three name no battery option. The second route is a check-in. A cloud notices the tank has stopped reporting and tells you, which covers a power cut and a router failure the same way. Two lines work this way, and one publishes its threshold, six hours.

3. The cheapest way in stops short of the rest of its own range. The entry controller at the top of the table reaches the first two jobs for less than anything else here. It has no expansion bus, so it cannot take the touchscreen or the power cut detector that the same manufacturer sells, and its own specification table prints a dash against the bus, both ports and email. Dosing is still open to it, through a doser that runs beside it rather than from it. Getting the screen or the power cut alert means replacing the controller, not adding to it. That is not visible from any price on its own.

4. One first-party touchscreen exists across all five brands. One 4.3 inch capacitive panel, $249.90, on the most expensive line. Everything else is a backlit readout with mechanical buttons, a row of lit on/off buttons that shows no readings at all, a small display on the controller that the manufacturer publishes no size for, or nothing. Three of the six lines answer a screen at the tank with the phone in your pocket.

One of the six has no head unit

Reef Factory is priced here, but it is built differently from the other five, and the comparison would mislead if that were not said. There is no head unit. Each box is an independent WiFi appliance with its own power supply, and the logic that ties them together runs in the company's cloud rather than on anything in your house. The one closed loop it documents, alkalinity, is computed by the app. Two further facts a buyer should have: the company entered bankruptcy in May 2025, and in September 2026 it was reported to have a new owner, a private individual who had paid to keep its servers running through the bankruptcy. The cloud stayed up throughout.

Things to check before you buy

GHL ProfiLux Mini. GHL's own product page for this controller advertises an email client. Its specification table on the same site prints a dash against it, and its notification guide lists the controllers that can send email without this one among them. Two of the three agree, and the two that agree are the specific ones.

Neptune Systems Apex. The entry controller has no outlets of its own, and the dosing pump's cable is made for the outlet bar, which powers the pump and links it to the controller in one run. Without the bar the pump can join the controller's own bus with a power supply of its own, which Bulk Reef Supply confirms for this controller. The table prices the bar route because the switching tier buys the bar anyway.

Red Sea ReefControl. Bulk Reef Supply lists every ReefControl item as out of stock as this is written. Three others hold all of it at these prices, so this is one shop rather than the line, and it is worth knowing which shop you are looking at.

Reef Factory. The two boxes that make up the entry tier are in stock. The outlet bar above them is not, and that is the tier the rest of the table is built on. It is out of stock at the one United States retailer that carries the brand, it has come off that retailer’s own brand listing while its page stays up, and four other United States shops were checked and none of them sells a Reef Factory outlet bar at all. So every figure in this row from the second column rightwards is a price rather than a purchase.

CoralVue HYDROS. The controller is sold without a power supply, and the power supply is out of stock on its own store as this is written. Buy the pack rather than the parts.

GHL ProfiLux 4. Neither the powerbar nor the doser includes the cable that connects it, and the cable is not obvious on either product page.

Every one of them charges nothing for the cloud. Six lines, five brands, no subscription on any of them. I looked for one because a recurring charge would be the obvious thing to find and it is not there.

Documents quoted above, read again on 1 October 2026: Neptune's SMS notice, Neptune's reference manual on power failure (a mirror of Neptune's own copy), Neptune's help page on power monitoring, the DOS page at Bulk Reef Supply, Red Sea's ReefBeat page, Bulk Reef Supply's ReefBeat guide, ReefControl stock at Bulk Reef Supply, GHL's Doser 3 SA, GHL's ProfiLux Mini specification, Danireef on Reef Factory's bankruptcy, Reefkeeping Today on its new owner.

Part numbers

These are two different kinds of identifier and printing them as one list would be a small lie. Each group says whose number it is. One manufacturer prints a catalog number ending in -VAR on anything sold in more than one color or country fitting, and the orderable item underneath it carries a number of its own, so that is the number its page shows and the one repeated here.

GHL ProfiLux Mini — GHL SKU, as GHL's own store prints it. ProfiLux Mini WiFi, PL-1614-VAR, $209.90. pH electrode, BNC, straight into the controller's own BNC input, PL-0071, $79.90. ProfiLux Mini WiFi-Set, PL-1814-VAR, $409.90. GHL Doser 3 SA, PL-2344-VAR, $556.90.

Neptune Systems Apex — Bulk Reef Supply item number, the retailer's, not Neptune's. A3 Apex Jr System, 251247, $319.99. EB832 Energy Bar, 210221, $324.99. DOS QuietDrive, two heads, 254973, $374.99. Apex Display Module, 210220, $149.99.

Red Sea ReefControl — Red Sea SKU, US retailers append their own suffix to it. ReefControl Lite, R35812, $329.00. ReefControl Power 8, R35830, $319.00. ReefDose 4, R35320, $449.00. ReefControl Power 8 Control Panel, R35832, $99.00.

Reef Factory — Bulk Reef Supply item number, the retailer's, not Reef Factory's. pH meter, 253234, $242.99. Thermo view temperature monitor, 253223, $106.99. Power switcher X4, four outlets, 255204, $99.99. Dosing pump, one head, 253228, $126.99.

CoralVue HYDROS — CoralVue SKU, as CoralVue's own store prints it. Control X3, HDRS-CX3, $248.31. Power supply, which the controller does not come with, HDRS-C24-PSU, $28.86. Temperature sensor, HDRS-TS-100H, $30.02. Milwaukee pH probe, CV-MA911B-2, $54.00. Control X3 and XP8 PRO Pack, HDRS-X3PRO, $608.99. 2 mL/min dosing pump, one head, HDRS-DSP2ML, $90.08. Kraken 24 V Power Center, HDRS-KRKN, $692.99. Kraken battery backup cable, HDRS-KRKN-CBL, $40.41.

GHL ProfiLux 4 — GHL SKU, as GHL's own store prints it. ProfiLux 4, PL-1407-VAR, $577.90. pH electrode, PL-0071, $79.90. Powerbar 6E-PAB, PL-1725, $279.90. PAB cable, 1 m, PL-0682, $14.90. GHL Doser 2.2 EXT, four pumps, PL-1938, $399.90. Power cut monitoring adapter, PL-1607, $25.90. GHL Control Pad, PL-2402, $249.90.

Six product lines from five brands, priced in the United States. Every figure read off the seller’s own page on 29 September 2026 — GHL's own US store, bulkreefsupply.com, topshelfaquatics.com, saltwateraquarium.com, coralvue.com, bulkreefsupply.com. Specifications, compatibility and discontinuation dates from each manufacturer’s own documentation and knowledge base, read the same day. No marketplace listings were used. Due to be re-priced March 2027, and a price checked once is a price that will drift.

What a router actually draws

Every runtime figure on this site rests on one number: how many watts your internet gear pulls. There is a good source for it. The US Voluntary Agreement for Ongoing Improvement to the Energy Efficiency of Small Network Equipment publishes an annual report, and the 2024 edition gives weighted averages for equipment actually sold and deployed in the United States.

A router or access point averages 8.45 W. A standalone broadband modem averages 9.33 W. An all-in-one gateway, which the report calls an integrated access device, averages 14.69 W. Across every category it is 11.37 W.

An average hides a wide spread, and yours may not be average. The report lists hundreds of individual models. They run from 1.85 W for an ASUS RT-N300 B1 up to 26.10 W for an ASUS GT-BE98 PRO, a factor of fourteen between the two ends. That is why the runtime on Lighthouse Mode is quoted against a wattage rather than as a flat number of hours.

Why this source and not a blog. The signatories are the internet providers and the equipment makers themselves. The report says they served approximately 91.3 million U.S. residential customers, representing 85% of that market in 2024, and the figures come from confidential procurement and sales data submitted by the signatories together with independent verification testing of models of each reporting signatory. It is the industry reporting on itself, which is a reason to read it carefully rather than a reason to dismiss it, and it is the only dataset of this size that exists.

2024 Annual Report, Voluntary Agreement for Ongoing Improvement to the Energy Efficiency of Small Network Equipment — energy-efficiency.us. Checked 29 September 2026.

What a router backup costs, and why there is no price on the page

Lighthouse Mode tells you to buy a router backup and gives no price for one. This is why.

What I found. Two units with the same stated capacity, 38.48 watt-hours, one at $38.35 and one at $105.99. A third at $17.25 that states 1800 mAh and no watt-hour figure at all, which means you cannot work out its runtime from anything the seller tells you. The one name-brand product in the category, the APC Back-UPS Connect 50, comes back as no longer available, and it turns out to be an AC unit with wall sockets and a simulated sine wave rather than the DC-straight-from-cells design this page is describing.

Watch the units, because this is where the money gets lost. A pack sold as 10400 mAh is 38.48 Wh, because the cells are lithium at about 3.7 V. Assume those milliamp-hours are at 12 V and you will calculate three times the runtime you are actually going to get.

So price is not a proxy for capacity here. The category is mostly marketplace sellers under a dozen different names, and saying that plainly is worth more than a dollar figure that would be wrong in six months. Buy on the watt-hour number. If a listing does not print one, that is the listing telling you something.

XISAOK 12V 3A 38.48WH Mini UPS, $38.35 — walmart.com item 5479641757. BFKK DC10400 Pro, 38.48 Wh, $105.99 — newegg.com item 9SIB7VEK918171. TOP5EX 12V 2A Mini UPS, $17.25 — walmart.com item 1101470650. APC BGE50ML — bhphotovideo.com. All checked 29 September 2026, and a price checked once is a price that will drift.

Why the panel is 10 inches, and why that decides the board

The screen is the Raspberry Pi Touch Display 2. It comes in three sizes and they are not the same product underneath. The 5 inch and the 7 inch are both 720 × 1280 and both run on a Raspberry Pi B+ and later. The 10 inch is 1200 × 1920, and Raspberry Pi lists it as working with Raspberry Pi 5 and later only.

So the size and the board are one decision, not two. Picking the panel that does not look cramped with a tank's worth of readings on it is the same act as picking a Pi 5. That is what The system means by chosen by what a Pi 5 can actually drive.

It also settles something about the enclosure. The display is Powered directly by the host Raspberry Pi device, requiring no separate power supply, through a cable to the GPIO header, and the 10 inch carries video and touch over a 22-way to 22-way flat cable to the DSI port. Two cables, neither of them meant to be unplugged and replugged in normal use. That is why the panel and the controller are one object rather than a screen you could lift away.

Raspberry Pi, Touch Display 2 documentation — raspberrypi.com. Checked 29 September 2026.

Why a dry pump still draws its amps, and why the outlet watches watts

This is the claim on The system most worth checking, because it decides a part. If an outlet can catch a dry-running pump by watching current, the cheap measurement is enough. If it cannot, the outlet needs a real power measurement, and that is a different chip on a different board.

It cannot. A motor's current has two parts. The working part carries the mechanical load and the magnetizing part just excites the iron, and the magnetizing part does not care what the pump is doing. As one reference puts it, the magnetizing current does not depend on the loading on the motor, and, practically, the magnetizing current has a constant value. So when a pump loses its water, the working current falls away and the magnetizing current is still there, filling the space it left.

The numbers are the point. A dry-running write-up on exactly this problem states that the reactive current increase offsets the real current decrease, yielding only ~10% current change—insufficient for reliable detection, while True power (Watts) does drop 40-50% at zero flow. Power factor on a lightly loaded motor lands in the range of 0.1 to 0.3 against something near 0.8 working.

So a current-watching outlet proves the motor is energised and spinning. It does not prove water is moving. Ten percent is inside the noise of a real installation. Forty to fifty percent is not. That is why the twelve volt ports measure current, where the voltage is fixed and amps and watts are the same fact, and mains measures real power.

On the sources, honestly. The mechanism is textbook and two independent references agree on it. The specific percentages come from one vendor's engineering note rather than from a standard or a paper, so treat the mechanism as settled and the exact figures as one good measurement rather than as a specification. They will be measured on my own bench before anything ships.

Pump Dry-Running Detection: Why Current Monitoring Fails (and What Works) — industrialmonitordirect.com. No-Load Power Factor of Induction Motor — electricalvolt.com. Both checked 29 September 2026.

Why there is no continuous ammonia probe

For stores says a fish-only rack is wiped out by ammonia after a death, and then says there is no continuous ammonia probe worth trusting in saltwater. That second half is the reason this system does not offer something people ask for, so it had better be right.

The measurement is hard for one specific reason, and it is potassium. Continuous ammonia sensors are ion-selective electrodes, and they answer to ammonium rather than to free ammonia. A review of twenty years of that chemistry describes the remarkable cross-interference of potassium over the ammonium response of nonactin-based ISEs, and names the hard case exactly: detecting close-to-micromolar levels of ammonium in real samples with a significant potassium background level (i.e., millimolar concentration).

That is a description of a reef tank. Full-strength seawater carries 380 to 395 milligrams of potassium per liter at 34.5 to 35.2 parts per thousand salinity. Divide by potassium's atomic weight and that is about 10 millimoles. The ammonia a reef actually cares about is micromolar. The paper's worst case is this tank on an ordinary Tuesday.

Drift is not the problem. Upkeep is. A review of ISE fouling in aquaculture reports that conventional ammonia ISEs exhibit a 3% drift over six months, which is fine. The same review puts calibration intervals from daily (e.g., pH ISE Mettler Toledo InPro) to bi-weekly, and describes fouling running from reversible fouling (2–4 weeks) to irreversible fouling (3–5 months), with response time climbing exponentially after roughly the first week.

So it is upkeep that rules it out. A probe that wants recalibrating every week or two, inside a box whose whole promise is that it keeps working when nobody is looking at it, is not a sensor. It is a chore with a reading attached. That is why the store screen watches level and temperature and keeps a record, rather than chasing a number it cannot trust.

Why ammonium detection is particularly challenging but insightful with ionophore-based potentiometric sensors, Analyst, Royal Society of Chemistry, 2020 — pubs.rsc.org. Fouling Control of Ion-Selective Electrodes (ISEs) in Aquatic and Aquacultural Environments: A Comprehensive Review, Sensors, 2025 — mdpi.com. Seawater potassium from Typical chemical characteristics of full-strength seawater, Global Seafood Alliance — globalseafood.org. All checked 29 September 2026.

What a safety listing actually costs you, after the test

People hear safety listing and picture a test. The test is the cheap part. OSHA's rule for a nationally recognized testing laboratory requires that the lab Inspects the run of production of such items at factories for product evaluation purposes to assure conformance with the test standards, and separately that it Conducts field inspections to monitor and to assure the proper use of its identifying mark or labels on products.

So the factory is in scope, not just the sample. UL describes its own Follow-Up Services as designed to confirm that a manufacturer's UL Certified product continues to comply with the original certification requirements, carried out through factory inspections and audits at locations where UL Certified products are manufactured and the UL Mark applied. That is an ongoing obligation attached to a place, not a certificate you file.

This is why Pharos Power is a decision about what kind of company this is rather than a circuit-design problem. The electronics are the easy half.

29 CFR 1910.7(b)(2), Definition and requirements for a nationally recognized testing laboratory — osha.gov. UL Solutions, Follow-Up Services — ul.com. Both checked 29 September 2026.

Why my name on a battery makes it mine, and the law that actually does it

This is the reason there is no Pharos battery, so it is worth getting right rather than just sounding right.

It is not the product safety statute. The Consumer Product Safety Act defines a manufacturer as any person who manufactures or imports a consumer product, and puts a brand owner in a different box of its own: a private labeler, meaning an owner of a brand or trademark on the label of a consumer product which bears a private label. So under that statute, putting my name on somebody's cells does not literally make me the manufacturer. It makes me a private labeler, which carries its own duties.

The thing that does it is tort law, and it is older and blunter. The apparent manufacturer doctrine: One who puts out as his own product a chattel manufactured by another is subject to the same liability as though he were its manufacturer. The later Restatement turns on whether the seller puts it out under his name or affixes to it his trade name or trademark, and courts ask whether a reasonable buyer would have believed the brand owner made the thing and relied on that reputation.

Which is exactly the position a Pharos pack would put me in. A buyer seeing my name on a battery beside my controller would reasonably assume I made it, and that assumption is the whole mechanism. Same liability as though he were its manufacturer is not a paraphrase of my sentence. It is nearly the sentence.

Two honest limits. The doctrine is state law rather than federal, so it varies, and the Third Restatement has been read as questioning how far it still reaches. And I am not a lawyer. This is the reading that informed a decision not to build something, which is the safe direction to be wrong in.

15 U.S.C. §2052(11) and (12), Consumer Product Safety Act definitions — uscode.house.gov. Apparent manufacturer doctrine, Restatement of Torts, quoted in Christopher Carton and Jasmine Owens, When A Nonmanufacturer Is The ‘Apparent Manufacturer’, Law360, 23 June 2022 — bowmanandbrooke.com. Both checked 29 September 2026. Not legal advice.

Research compatibility

Research compatibility means a part whose specification I have read and believe will work, but which has not yet run on my bench. Everything in this section is in that state until it has run. A part leaves it in one of two directions. Either it gets tested and named as something that works, or it gets tested and this page says why it did not.

Nothing in this section is a recommendation to buy. It is what I am looking at and why, published while it is still a question rather than after it is settled.

Powering the Pod from a battery

The Pod is a Raspberry Pi 5 behind a 10 inch panel, and the Pi 5 documentation is specific about what it wants: 5 V at 5 A. Raspberry Pi’s USB Power Delivery white paper says that supply makes 1.6A available to the USB ports, and that on a 3 A supply the board will automatically limit the total power available to the USB ports to 600mA and show a warning message saying power is limited.

That rules out more power banks than you would expect. Most reach their headline wattage at 9 V or 20 V and still give 5 V at 3 A or less, which is the rail the Pod actually runs on.

The TalentCell NB7102 is the one I looked at closely. 3.7 V 17500 mAh / 64.75 Wh. Input 29.4 V/0.95 A. Four ports on the end face: a USB-C output rated 60 W that negotiates 5, 9, 12, 15 or 20 V at 3 A, a DC1 barrel at 29.4–21 V/3 A which is also the charge input, a DC2 barrel at 19.5 V/3 A, and USB-A at 5 V/2.4 A.

Sixty watts is not the number that matters. On the 5 V rail the pack gives 3 A, which is 15 W against the 25 W the Pod asks for. It reaches its headline figure at 20 V, which is the trap this section is about. The pack would run the Pod with its USB ports throttled to 600 mA and a warning on the screen.

Two limits come off the manual rather than the web page. Only one of the three main outputs should be discharging at a time. And the pack must not be charged through the USB-C port — charging is on the DC1 barrel. That second one is the more interesting half, because charging and discharging being on different ports is a prerequisite for holding a load up through a power cut, though nothing in the manual says whether it actually does.

Its capacity line is also a worked example of the milliamp-hour trap: the manufacturer writes 3.7 V 17500 mAh, and it is the 3.7 V that turns those milliamp-hours into 64.75 watt-hours rather than into a number three times larger.

Decided 29 September: whatever holds the Pod up stays outside the enclosure. There are UPS boards that sit on the Pi itself and meet the 5 A requirement cleanly, and they are out for a reason that is not electrical. A board like that charges cells inside the case, and Designed, not built says why that is a line worth holding — putting my name on a battery makes me its manufacturer in the eyes of the law, whoever built it. So the pack stays on a cable, outside, and it stays yours.

So a pack has to meet two requirements, not one. It has to hold 5 V at 5 A, and its output must not interrupt when the wall supply goes. A separate charge input is a prerequisite for the second, not the same thing as it — a pack can have one and still drop its output the moment it starts charging.

Raspberry Pi, USB Power Delivery on Raspberry Pi 5 — raspberrypi.com. TalentCell NB7102 specification table and product photographs, which contradict each other — talentcell.com. The instruction manual covering both the NB7101 and the NB7102, which settles it — grainajz.com. All checked 30 September 2026. Research compatibility — neither has run on my bench.

Why a store loses a tank, and the study that does not exist

For stores rests on a claim about the world: that a store tank lost overnight went to a leak, a stuck heater or a stopped pump rather than to chemistry drifting. That claim decides what the store screen watches, so it is worth saying exactly how much weight it can carry.

It carries my judgement and nothing else. I went looking for data. The published work on aquatic livestock loss is about the trade rather than the shop floor — transport and handling stress, disease treatment, microbiome collapse after purchase. I did not find anyone who has counted retail tank losses and split them between equipment failure and water chemistry. If that study exists, I have not found it, and I would rather say so than dress up a guess with a citation.

Why I still believe it. Chemistry drifts over days and a person walking past sees it. A heater sticking on, a return pump stopping and a sump overflowing all happen in hours, and they happen at night when nobody is walking past. The argument is about timing rather than about frequency, and timing is what an overnight alert is for.

What would change it. Numbers from stores. If you run one and your losses do not look like that, that is worth more to me than the reasoning is, and the contact page is right there.

No source. Searched the ornamental fish and aquaculture literature 29 September 2026 and found nothing that answers the question. Listed here because a claim with no source behind it should be as easy to find as one that has three.