I2C bus capacitance is a finite budget shared by every device, trace, connector and cable on the two wires. The bus specification sets a maximum because rise time depends on that capacitance and the pull-up resistance, and an edge that rises too slowly no longer meets the timing the protocol assumes. The commonly cited limit for standard and fast modes is four hundred picofarads, and it is easier to exceed than most designers expect.
Where the capacitance comes from
Each device connected to the bus contributes its pin capacitance, and its datasheet states the figure. Board traces contribute in proportion to their length and the stack-up geometry. Connectors add a small amount each. Cable is the big one: ribbon or multi-conductor cable contributes capacitance per unit length that is far higher than a board trace, so a run between boards can consume the whole budget on its own. Protection components, series resistors with their parasitic capacitance, and unused device footprints all add a little. The sensible practice is to total these explicitly during design rather than discovering the problem when a system is assembled with all its options fitted.
What happens as the budget is exceeded
The lines are open-drain, so the falling edge is driven hard by the device pulling down while the rising edge is left to the pull-up charging the capacitance. As capacitance grows, the rise slows, and the signal spends longer crossing the undefined region between the low and high thresholds. Setup and hold times measured from those threshold crossings shrink. The first symptoms are not a clean failure but occasional corrupted transactions, devices that fail to acknowledge, and a bus that works with a short cable and misbehaves with a long one. Because the failures are intermittent and load-dependent, they are frequently misattributed to software, to a particular device, or to electrical noise.
The pull-up cannot solve it alone
The obvious response is to reduce the pull-up resistance so the line charges faster, and within limits that works. But the lower bound is set by the maximum current the bus devices can sink while still holding a valid low level, and that limit is part of the bus specification. Once the pull-up is as strong as the weakest device on the bus permits, no further speed is available that way. At that point the options are to reduce the capacitance, reduce the bus speed so the slow edges still meet timing, split the bus into segments, or introduce active circuitry. Attempting to run an overloaded bus at a higher speed by strengthening the pull-up beyond specification produces marginal low levels and a different intermittent failure.
Ways to buy more headroom
Bus buffers and repeaters divide a bus into segments, each with its own capacitance budget, allowing more devices or longer cable than one segment could support. Some also provide level translation between different supply voltages, which is often needed at the same time. Multiplexers and switches let groups of devices be connected only when addressed, which additionally resolves address conflicts between identical parts. Bus accelerators drive the rising edge actively and then release to a passive pull-up, achieving fast edges without excessive static current. For connections leaving an enclosure, a differential extender converts the bus to a more robust form for the cable run and converts it back, which is far more reliable than simply using better cable.
Measuring the bus you actually built
Put a scope on both lines with a low-capacitance probe, since a standard probe adds enough capacitance to change what you are measuring. Look at the rising edge with the maximum number of devices fitted and the longest cable connected, and measure the time between the low and high threshold voltages rather than eyeballing the shape. Compare that against the specification for the speed you are running. Check the low level as well, confirming it stays below the input low threshold when the weakest device pulls down. Run the test at the highest operating temperature, since leakage and device characteristics both move. If the margin is thin, fix it before shipping rather than after the returns start.
What to check before you act
- Total the capacitance of every device, trace, connector and cable segment.
- Measure rise time between the actual threshold voltages with a low-capacitance probe.
- Keep the pull-up within the sink current the weakest device supports.
- Use a buffer, multiplexer or extender rather than overloading one segment.
Common questions
How long can an I2C cable be?
There is no fixed length, only a capacitance budget. A cable’s capacitance per unit length and the rest of the bus loading together determine what is possible at a given speed.
Will lowering the clock speed fix a slow rising edge?
It often does, because the timing requirements scale with the clock period. It does not help if the edge is so slow that the line never reaches a valid high before the next transition.
The practical takeaway
Budget bus capacitance during design, including every cable and connector the product might have fitted. Measure the real rising edge against the real thresholds on a fully populated system, and when the budget is spent, segment the bus with a buffer rather than pushing the pull-up past what the devices can sink.
References and further reading
- Department of Energy: how electronics losses turn into wasted energy and heat
- FCC: electromagnetic compatibility and radio frequency safety questions answered
- CPSC: safety guidance for electronics and electrical products indoors
Numerical scenarios are illustrative unless identified otherwise. Follow the exact product instructions; component ratings and local installation requirements can differ.
