Which board for which job
Six things people actually measure, and which of the two boards suits each. Both reach about 3.2 amps, so range rarely decides it — channels, step size and the power register do, and neither board can see a sleeping microcontroller.
Pick the thing, read the board
Two boards, six jobs, and the answer is genuinely different for most of them. Step through the options: the pattern is that the TK119 wins on range and on sign, the INA3221 wins on channel count and on noticing things by itself, and neither wins on small currents.
The short version
One rail, and you are not sure how big the current is. The TK119. It is the only one of the two whose range you can move, moving it is one line of setup, and its steps are eight times finer than the INA3221's.
Three rails. The INA3221, and not only for the channel count. Its three conversions happen inside one chip on one schedule, so the three readings belong to the same moment — and its critical alert can add the three shunt voltages together and compare the total against a single limit, which no arrangement of separate boards can do.
A motor, or anything that stalls. Either. A stall is several times the running current, and both boards reach about 3.2 A — the TK119 on its widest setting, the INA3221 on every channel. The INA3221 can also flag the stall in hardware with its critical alert. Set a short conversion time or the peak gets averaged away.
A battery being charged and discharged. Either; the TK119 for one battery. Both boards read current in both directions over the same range, and the TK119 adds finer steps and a power register for the log.
A sleeping microcontroller. Neither. One count is 0.1 mA at best and the fixed error is worth about ten of those — twice that on the INA3221.
What "close to a limit" means
Some of those answers came back amber rather than green, and it is worth being clear that amber is not a warning about damage. It means the board will do the job with less margin than you would like: the TK119 on a waking ESP32 tells you that it woke and roughly how hard, but not what it drew while asleep.
The number that decides most of these is the error at the bottom of the scale, and that is the next page.
When it does not work
Yes, and it costs you three of the four addresses, three sets of wires, and the fact that three separate I²C reads do not happen at the same instant — which matters if the three rails feed one thing and you are looking at how the load moves between them. What you cannot copy is the INA3221's summed-current alert, which adds the channels in hardware and compares the total against one limit.
Not from these boards, whatever you do in software: the range is the chip's shunt input divided by the resistor, and the resistor is soldered down. A smaller shunt would widen the range and shrink the signal in the same proportion, which is a board design rather than a modification. For larger currents, use a monitor sold with a shunt sized for them.
Either works, and both read charge and discharge as positive and negative current up to about 3.2 A. The TK119 has the edge for one battery: 0.1 mA steps rather than 0.8, and a power register that makes an energy log a running sum. The INA3221 is the choice when the battery, the charger and the load are three rails you want to watch together.
Both do, and with nothing between them and the microcontroller: the chips run from 3 to 5.5 V and the board's pull-ups pull the bus to whatever is on the supply pin. The rail being measured is separate and may be anything up to 26 V, so a 3.3 V ESP32 reading a 24 V supply needs one shared ground and no level shifting.
Edit this page — content/books/inaset/which-board-for-which-job.mdx
Questions about this product
See what other owners have asked, and read their solutions.
This page covers several products. Choose yours to see the right discussions.
Discuss this article
Ask about this page. The answer stays here, on the page it belongs to, for whoever hits the same wall next.