A feeder with a reader tells you which animal approached. A feeder with a load cell tells you how much food left the hopper. Separately, neither number means very much. Together they produce the only metric that actually answers a owner's question: how many grams did each specific cat eat, and when. The combination is where the engineering gets interesting, and it is also where most of the honest questions about a quotation live.
What each sensor knows alone
The reader produces events: tag detected, lid opened, lid closed. It does not know whether the cat ate, how much, or whether the second cat pushed in behind the first. The scale produces mass: grams dispensed, grams remaining, and after some arithmetic, grams consumed per interval. It does not know who. A single-pet household can live with the scale alone. The moment there are two animals, or one animal on a prescription diet, the scale's number becomes ambiguous, and the reader without the scale is a log of intentions.
The useful number is grams per tag per interval
Fusion is simple in principle: every weigh event gets attributed to the tag that was present. From that you get the curve that matters in practice, which is not the daily total but the pattern: grams per visit, visits per day, and the trend across days. Appetite change is one of the earliest visible signs of illness in cats, and it shows up as a change in this curve days before it shows up in behavior an owner would notice. A feeder that can produce a per-animal trend is a different product class from one that produces a daily total, and the difference is exactly the pairing of reader and scale.
Tag misses are the real design problem
Any multi-pet feeding program lives or dies on what happens when the read fails. A second cat follows the first through the opened lid. A cat stands with its head just outside the read field. A collar tag rotates away from the antenna. The reader sees nothing while the scale sees fifty grams disappear. Good firmware has an explicit policy for this: count the event as unattributed mass, or assign it by exclusion if only one tag could possibly be present, or hold the data for a confirmation window. Bad firmware silently attributes it to whoever was last seen, and the per-cat curve becomes fiction. Ask the vendor what happens on a tag miss, and ask to see a raw log with a miss in it, because the answer to that question is a better quality signal than any accuracy figure on the datasheet.
Chip reading and collar tags fail differently
The two tag technologies have opposite failure modes. Implant chips at 134.2 kilohertz are permanent and impossible to lose, but the animal must bring the implant into the read field and hold it there, which constrains bowl and antenna geometry and adds a fraction of a second the impatient cat may not grant. Collar tags read at a distance and do not care where the head is, but they depend on the collar staying on, and in multi-cat homes a removed tag is a re-pairing exercise. Some households need both paths supported. That is a firmware and hardware decision the factory makes early, not a setting.
The scale needs tare discipline to stay honest
A load cell drifts with temperature and with the residue that sticks to the bowl between meals. A feeder that tares only at startup will be wrong by evening, and by the end of a fortnight the remaining-grams figure is decorative. The credible designs tare before every dispensing cycle and again after each visit is attributed, and they expose the raw readings in diagnostics so a strange curve can be distinguished from a failed cell. Also worth checking: the bowl must be mechanically isolated from the hopper and the housing, because any stiff connection turns the whole body into part of the weighing platform and the grams stop meaning anything.
The other fusion question: what the app does with it
Data fusion is only useful if the product surfaces it. The minimum viable presentation is a per-animal view: grams today, trend across the week, and the unattributed remainder shown honestly rather than hidden. Features that add value: alerts when a tagged animal has not eaten within its usual window, and a lockout that enforces the daily ration for an animal on a weight-loss program. Features that add cost without much value are the usual suspects: calorie estimates from algorithms, social sharing, and anything that promises health conclusions from two weeks of data.
What to ask the factory
Five questions cover the ground. Which chip standard the reader supports, and whether collar tags are an option. What the firmware does on a tag miss. How often the scale tares, and whether raw readings are inspectable. Whether the bowl is mechanically isolated from the frame. And how the app attributes weigh events, with a sample log as evidence. A supplier who answers those five without hesitation has built this product before.
At XingLiao's pet feeder line, including the dual-bowl and weighed platforms, the tag-miss policy and the tare schedule are part of the specification conversation rather than buried in a manual, because those two items decide whether the per-cat numbers mean anything. On the B2B side, when a feeder is built for a distribution brand, the reader standard and the weighing architecture are settled before the enclosure is tooled, since the bowl isolation and antenna geometry cannot be retrofitted.
The same five questions work as a first email to any supplier, including through XingLiao or our contact page, and a seller who answers them with sample logs rather than adjectives has already told you which tier of feeder you are looking at.
Top comments (0)