PowerTool
A device's power budget as a schematic you can compute: the power tree, the operating modes and the battery life.
The tool is internal, so everything is measured on its own screenshots.
IBM Plex Sans
HeadingsAa
- PowerToolRegular 400
- PowerToolSemiBold 600
АБВГҐДЕЄЖЗИІЇЙ abcdefghijklm 0123456789
IBM Plex Mono
Body textAa
- PowerToolRegular 400
- PowerToolSemiBold 600
АБВГҐДЕЄЖЗИІЇЙ abcdefghijklm 0123456789
Corner radii
- 6 px
- 8 px
- pill
Buttons: 6 px
Type scale
- AaReadout, 28 px
- AaH1, 15 px
- AaBody, 14 px
Contrast
- 14.6 : 1text on background, AAA
- 2.3 : 1button label
Context
An internal tool by Renat, the studio's co-founder, used by a team that designs autonomous devices. It's closed and doesn't run on the open web, so we show the live version on a call. For this case we built a complete demo model: a LoRaWAN soil and air sensor on a single 18650 cell.
The Challenge
A power budget usually lives in Excel: a column of currents, a column of efficiencies and a sum at the bottom. That sheet knows nothing about the device's modes. It treats converter efficiency as a constant, while in reality it depends on current and voltage. And it will never show that a mode taking half a percent of the time eats four fifths of the charge.
The constraint
The answer has to hold down to the microamp. So every part is described the way its datasheet describes it: voltage and current limits, quiescent current, efficiency curves for several input voltages. Numbers the datasheet doesn't give, the model doesn't make up.
The decisions, pinned where they live
Why we made it this way.
The mode switch at the top: sleep, measure, GNSS fix, uplink. The boxes stay put and every number changes, so you can see exactly what separates one mode from another.
R1 in the battery line works as a measurement shunt. Next to the calculated 59. sits what the meter read: 6 .
Every converter shows input, output, efficiency and loss in the current mode. Efficiency comes from the curve at the real current and voltage, so the 88.9% here is calculated.
GNSS power runs through the SW1 switch. In this mode it's open, the rail turns dashed and the module shows . The switch's 2 of leakage still counts.
The sensor head is drawn on its own sheet. The 5V and 3V0 rails reach it through a named port, the way a schematic crosses a connector. The two-meter cable has its own loss model.
The mode switch at the top: sleep, measure, GNSS fix, uplink. The boxes stay put and every number changes, so you can see exactly what separates one mode from another.
R1 in the battery line works as a measurement shunt. Next to the calculated 59. sits what the meter read: 6 .
Every converter shows input, output, efficiency and loss in the current mode. Efficiency comes from the curve at the real current and voltage, so the 88.9% here is calculated.
GNSS power runs through the SW1 switch. In this mode it's open, the rail turns dashed and the module shows . The switch's 2 of leakage still counts.
The sensor head is drawn on its own sheet. The 5V and 3V0 rails reach it through a named port, the way a schematic crosses a connector. The two-meter cable has its own loss model.
What we did, and why
Mission profile and battery life
The period, the length of each mode, and one mode that takes whatever time is left. Out come the average current, the charge per period and the cell's life. Nobody typed the uplink length in: the radio module's sheet computes it.
A calculation sheet inside the part
For loads whose current comes out at the end of a calculation. The formula engine knows 46 functions, plus four that Excel doesn't have: the sheet knows which mode it's being asked about. Here it computes LoRa airtime from spreading factor and packet length, and the result becomes the mode's length.
plotgrab: curves out of a datasheet
Efficiency curves exist in a datasheet only as a picture. Two reference points per axis, an eyedropper on each curve, a box over the legend, and the table for the part is ready. Across a set of 69 curves the median error is 0.023% of frame height.
Model against the bench
What the meter read in each mode sits next to what the model says. During the GNSS fix the model expects 96.6 and the meter shows 99. , a +3.25% gap. A model nobody checks stays a drawing.
Where it landed
The tool shows what a sheet of averages hides. In the demo model, the GNSS fix takes 0.56% of the time and 81.5% of the charge. The model is checked on the bench: in the same mode the calculation gives 96.6 , the meter reads 99. , a 3.25% gap.
A closer look
Click any frame to open it full screen.