Philips Semiconductors

Product data sheet

PCA9513; PCA9514

2004 Oct 05

6

OPERATION

Start-up

An under voltage/initialization circuit holds the parts in a

disconnected state which presents high-impedance to all SDA and

SCL pins during power-up. A LOW on the enable pin also forces the

essentially zero. As the power supply is brought up and the enable

is HIGH or the part is powered and the enable is taken from LOW to

HIGH it enters an initialization state where the internal references

are stabilized. The 92

µA input pull-ups on PCA9513 are also

enabled in the initialization state. At the end of the initialization state

the “Stop Bit And Bus Idle” detect circuit is enabled. With the

ENABLE pin HIGH long enough to complete the initialization state

and remaining HIGH when all the SDA and SCl pins have been

HIGH for the bus idle time or when all pins are HIGH and a stop

condition is seen on the SDAIN and SCLIN pins, SDAIN is

connected to SDAOUT and SCLIN is connected to SCLOUT.

A 92

µA pull-up current source on SDAIN and SCLIN of PCA9513 is

activated during the initialization state and remain active until the

power is removed or the enable is taken LOW. When the 92

µA

pull-up is active it will become high-impedance any time the pin

Connect Circuitry

Once the connection circuitry is activated, the behavior of SDAIN

and SDAOUT as well as SCLIN and SCLOUT become identical with

each acting as a bidirectional buffer that isolates the input

capacitance from the output bus capacitance while communicating

the logic levels. A LOW forced on either SDAIN or SDAOUT will

cause the other pin to be driven to a LOW by the part. The same is

generally ignored because a falling edge is only recognized when it

falling edge is seen on one pin the other pin in the pair turns on a

pull down driver that is referenced to a small voltage above the

falling pin. The driver will pull the pin down at a slew rate determined

by the driver and the load initially, because it does not start until the

or slow slew rate, if it is faster than the pull down slew rate then the

initial pull down rate will continue. If the first falling pin has a slow

slew rate then the second pin will be pulled down at its initial slew

rate only until it is just above the first pin voltage the they will both

continue down at the slew rate of the first.

Once both sides are LOW they will remain LOW until all the external

drivers have stopped driving LOWs. If both sides are being driven

LOW to the same value for instance, 10 mV by external drivers,

which is the case for clock stretching and is typically the case for

acknowledge, and one side external driver stops driving that pin will

rise and rise above the nominal offset voltage until the internal driver

catches up and pulls it back down to the offset voltage. This bounce

is worst for low capacitances and low resistances, and may become

excessive. When the last external driver stops driving a LOW, that

pin will bounce up and settle out just above the other pin as both rise

together with a slew rate determined by the internal slew rate control

and the RC time constant. As long as the slew rate is at least

1.25 V/

µs, when the pin voltage exceeds 0.8 V for PCA9513 and

PCA9514, rise time accelerators circuits are turned on and the pull

down driver is turned off.

Maximum number of devices in series

Each buffer adds about 0.065 V dynamic level offset at 25

°C with

0.150 V. The LOW level at the signal origination end (master) is

dependent upon the load and the only specification point is the

about 0.1 V below the threshold of the rising edge accelerator (about

0.6 V). With great care a system with four buffers may work, but as

in firing the rising edge accelerator thus introducing false clock

edges. Generally it is recommended to limit the number of buffers in

series to two.

The PCA9510 (rise time accelerator is permanently disabled) and

the PCA9512 (rise time accelerator can be turned off) are a little

different with the rise time accelerator turned off because the rise

time accelerator will not pull the node up, but the same logic that

∼0.6 V and a rising edge is detected, the pull-down will turn off and

will not turn back on until a falling edge is detected; so if the noise is

small enough it may be possible to use more than two PCA9510 or

PCA9512 parts in series but is not recommended.

MASTER

buffer A

SLAVE B

buffer B

SLAVE C

buffer C

SW02353

common

node

Figure 6.

Consider a system with three buffers connected to a common node

and communication between the Master and Slave B that are

connected at either end of Buffer A and Buffer B in series as shown

changing from Master to Slave B and then from Slave B to Master.

Buffer A of 0.3 V and its output, the common node, is

∼0.4 V. The

output of Buffer B and Buffer C would be

∼0.5 V, but Slave B is

driving 0.4 V, so the voltage at Slave B is 0.4 V. The output of

Buffer C is

∼0.5 V. When the Master pull-down turns off, the input of

Buffer A rises and so does its output, the common node, because it

is the only part driving the node. The common node will rise to 0.5 V

before Buffer B’s output turns on, if the pull-up is strong the node will

bounce. If the bounce goes above the threshold for the rising edge

accelerator

∼0.6 V the accelerators on both Buffer A and Buffer C

will fire contending with the output of Buffer B. The node on the input

of Buffer A will go HIGH as will the input node of Buffer C. After the

common node voltage is stable for a while the rising edge

accelerators will turn off and the common node will return to

∼0.5 V

because the Buffer B is still on. The voltage at both the Master and

Slave C nodes would then fall to

∼0.6 V until Slave B turned off. This

would not cause a failure on the data line as long as the return to

0.5 V on the common node (

∼0.6 V at the Master and Slave C)

occurred before the data setup time. If this were the SCL line, the

parts on Buffer A and Buffer C would see a false clock rather than a

stretched clock, which would cause a system error.