TASK · FULL COSTS · COOLING
PHONE SENSING EXAMPLE
Reduce the heat where the work changes.
At a 20% sensing share, the strongest saved 23.0416667% task saving reduces total phone input power by 4.6083333%. At equal energy leaving in forms other than heat and equal stored-energy change, the heat reduction equals the input watts saved. For baseline device power P, ΔQ̇ = 0.046083333 × P W. The other 80% stays fixed, and the calculation includes the saved model’s planning cost.
The Wi-Fi arithmetic reference below has a separate rate, interface and complete model budget.
Whole-phone sensing calculation →DEFINED TASK / COMPLETE MODEL
D027 · Low-power Wi-Fi client electronics
Both designs return the same required output fields, encoded response bytes and final state within this defined model. Each job contains 4 useful frames and repeats every 10 ms.
Input power: 0.353272008 → 0.352366726 W at 100 jobs/s. ΔQ̇ = (Eordinary − Eselected) / job period, for equal energy leaving in forms other than heat and equal stored-energy change. The input-power model includes the unchanged reference load and shared costs.
Ordinary → selected completion: 720.1 → 1,090.89 µs; required deadline 8,000 µs.
The changes included in this design.
- Serial lookup arithmetic
- Compact state layout
- Streaming boundary
- Clock gating with its cost charged
- Schema DMA interface with serialized traffic charged
The model compares the selected design with the best eligible conventional design in the study. Both include compatible improvements. The energy calculation counts logic, memory, clocks, connections, setup, idle time, resets and recovery once.
Compared designs and full energy breakdown
Selected: H_G7_COMPACT_D5_STREAM_CG::W_SCHEMA_DMA_1B_CG::BODY_POLICY
Strongest ordinary: O_BALANCED_COMPACT_D5_STREAM_CG::W_SCHEMA_DMA_1B_CG::BODY_POLICY
Prior R6 selected policy: H_G7_COMPACT_D5_STREAM_CG::W_SCHEMA_DMA_1B_CG::BODY_POLICY. Extra saving against that policy: 0 µJ/job.
| Input-referred term | Strong ordinary (µJ) | Selected (µJ) |
|---|---|---|
| common common host control W | 222.222222 | 222.222222 |
| common common memory W | 55.555556 | 55.555556 |
| common common clock sensor W | 111.111111 | 111.111111 |
| common plant or external load W | 2,222.222222 | 2,222.222222 |
| installed FPGA shell | 888.888889 | 888.888889 |
| generic accelerator clock distribution | 6.916439 | 10.191751 |
| arithmetic resource leak | 12.021111 | 1.997578 |
| cache leak | 0 | 0 |
| allocated program data memory leak | 2.497744 | 2.497389 |
| additional array port capacity tariff | 2.497744 | 0 |
| occupied CPU units | 5.086667 | 5.088 |
| DSP products | 0.016489 | 0 |
| adder bits | 0.005895 | 0.016302 |
| nonadder LUT bits | 0.000322 | 0.013167 |
| divider sequence | 0 | 0 |
| register accesses | 0.0895 | 0.0895 |
| accelerator control | 0.002917 | 0.002917 |
| internal data instruction cache traffic | 0.615147 | 0.615164 |
| cache valid initialization | 0 | 0 |
| conditional original B core dynamic | 0 | 0 |
| conditional original B clock standby | 0 | 0 |
| external canonical wrapper CPU | 0.010667 | 0.010667 |
| external request commit readback | 0.758667 | 0.758667 |
| cold host preparation | 0 | 0 |
| cold logic program cache preparation | 0 | 0 |
| R6 wrapper logic | 0.451471 | 0.451471 |
| R6 wrapper internal fixed transfers | 0.037169 | 0.037169 |
| R6 wrapper resource leak | 1.107822 | 1.107822 |
| R6 wrapper memory leak | 0.187589 | 0.187589 |
| R6 wrapper extra port leak | 0 | 0 |
| R6 wrapper clock | 0.416717 | 0.602112 |
| Total | 3,532.720077 | 3,523.667264 |
Exact totals: 317944806911/90000000000000 J versus 317130053743/90000000000000 J. Source models use declared physical coefficients.
Operating conditions, uncertainty and fitting the design into a product
One reference controller/processor, shared memory, sensor/clock rails and the stated unchanged external load; compute-specific increments and added resources are priced by the R5 common cost-law artifact.
Supply efficiency in the model: 90%. Shared loads and loads that stay unchanged: common host control W 0.02 W; common memory W 0.005 W; common clock sensor W 0.01 W; plant or external load W 0.2 W.
- Input waveform and accepted EVM/ACLR/spectral thresholds at exact reference plane
- PA/regulator transfer and loss maps across voltage, load, temperature and frequency
- Causal forward/inverse model coefficients, lag/order/state history, numeric precision and observation chain
- Regulator replacement: retained protection, headroom and stability bounds
The portfolio has 13 savings at the main modeled point, 100 ties using the conventional fallback, and 7 unmatched comparisons across 120 references. Positive advantages passing both the strict and combined tighter uncertainty tests: 0.
APPLICATION COVERAGE
The comparison tasks.
| Comparison job | Input energy saved per job | Job rate | Heat reduction under the stated conditions | Outcome at the main modeled point |
|---|---|---|---|---|
| D027 · Low-power Wi-Fi client electronics | 9.052813 µJ | 100/s | 0.905281 mW | Selected changes costed together |
| D031 · Smartphone application CPU kernel | 0 µJ | 100/s | 0 mW | Conventional fallback: same energy |
| D032 · Smartphone modem PA electronics | A matching baseline and eligible candidate are still needed | 1,000/s | Open comparison | Unmatched |
| D033 · Phone camera image pipeline | 0 µJ | 30.30303/s | 0 mW | Conventional fallback: same energy |
| D034 · Mobile GPU frame scheduling | 0 µJ | 62.5/s | 0 mW | Conventional fallback: same energy |
| D035 · Mobile neural-network inference | 0 µJ | 100/s | 0 mW | Conventional fallback: same energy |
| D036 · GNSS receiver-side computation | 0 µJ | 50/s | 0 mW | Conventional fallback: same energy |
| D037 · Wearable sensor hub | 17.396051 µJ | 10/s | 0.173961 mW | Selected changes costed together |
| D038 · Bluetooth audio DSP | 0 µJ | 1,000/s | 0 mW | Conventional fallback: same energy |
| D039 · Low-power IoT sensor node | 0 µJ | 1/s | 0 mW | Conventional fallback: same energy |
| D040 · Phone battery-health estimator | 0 µJ | 1/s | 0 mW | Conventional fallback: same energy |
These are computational models of defined tasks relevant to this application. Each selected design includes the compatible changes listed in its comparison.
YOUR PRODUCT’S THERMAL INPUTS
Calculate the cooling mass.
Use the heat carried by this cooling path and the weight of its installed parts, including added hardware. The comparison task’s average power change contributes to that total device calculation.
Chosen model: effective area follows Q = hAΔT; the fixed mass stays, the area-scaled mass changes, and added parts are charged.
Enter the product’s heat-path and part-weight values to calculate the cooling change.
Explore the calculation.
Heat energy = input energy − energy leaving in forms other than heat − change in stored energy.
All 120 reference rows, exact values, complete ledgers and source pins →All product thermal budgets and FI tools →