Dear NXP ,
Now I am using KL43Z I2C as Slave Device , and I need to read data from KL43Z ( I2C Slave ) , and i found a delay ( about 16us between ACK & DATA ( Slave ) . as below CLOCK setting and LA Picture , Why this happen ? how to modify to solve this , i want to slave response quickly . The Situation as below ,
IDE : MCUXpresso
SDK Project : reference "frdmkl43z_i2c_interrupt_b2b_transfer_slave"
I2C Slave : FRDM-KL43Z
Setting : At 125KHz, SCK should be Hi 4us and low 4us
and there are some delay (about 2 clocks = 16us) before slave mode out to master.
Thanks a lot !!
Hi eLin,
Thank you for your updated information.
Customer can use the optimize, actually optimize will decrease the code size, just decrease the code execute time.
On my side, I modify the SDK interrupt handler, and change it to the register control mode, then the ACK&Data time is also decreased.
From my attached code, you can find if don't modify, the executed time is about 12us, but if I use the register control, the time is about 6.8us.
Actually, it is relate to the Handler execute time.
I modify the SDK code in folder: SDK_2.4.1_FRDM-KL43Z\boards\frdmkl43z\driver_examples\i2c\interrupt_b2b_transfer\slave\iar
fsl_i2c.c void I2C_SlaveTransferHandleIRQ(I2C_Type *base, void *i2cHandle) like this:
void I2C_SlaveTransferHandleIRQ(I2C_Type *base, void *i2cHandle)
{
volatile unsigned char Dummy;
if (I2C0->FLT & (I2C_FLT_STOPF_MASK)) //stop
{
I2C0->FLT |= (I2C_FLT_STOPF_MASK);
I2C0->S |= (I2C_S_IICIF_MASK);
return;
}
if(I2C0->S & I2C_S_IICIF_MASK)
{
I2C0->S |= I2C_S_IICIF_MASK; //clear interrupt flag
if (I2C0->S & I2C_S_ARBL_MASK) //clear ARBL
{
I2C0->S |= (I2C_S_ARBL_MASK);
if (!(I2C0->S & (I2C_S_IAAS_MASK)))
{
return ;
}
}
}
if (I2C0->S & I2C_S_IAAS_MASK) //ADDRESS
{
I2C0->C1 &= ~I2C_C1_TXAK_MASK;
if (I2C0->S & I2C_S_SRW_MASK) //send mode
{
I2C0->C1 |= I2C_C1_TX_MASK;
I2C0->D = 0x55;
}
else //receive mode
{
I2C0->C1 &= ~I2C_C1_TX_MASK;
Dummy = I2C0->D ;
}
}
else //date range
{
if (I2C0->S & I2C_S_SRW_MASK) //send mode
{
if(I2C0->S & I2C_S_RXAK_MASK) //nack
{
I2C0->C1 &= ~I2C_C1_TX_MASK; //change to receive mode
Dummy = I2C0->D;
}
else
{
I2C0->D = 0x55;
}
}
else
{
I2C0->C1 &= ~I2C_C1_TX_MASK; //receive mode
//Rec_data[I2C_S_Recindex++] = I2C0_D ;
Dummy = I2C0->D;
}
}
}
This time should meet the customer's demand.
If the customer still want to minimize the time, they need to modify the .s file, just use I2C0_IRQHandler as the Interrupt handler directly.
Don't copy the to I2C0_DriverIRQHandler, then I2C_TransferCommonIRQHandler, then I2C_SlaveTransferHandleIRQ
As you know, enter deep function also need the code execute time, just simplify the code method.
But if the time already meet the customer's demand, it's better use the SDK directly, becuase that code structure is more better, just a little complicated to the code execute time.
I also attached my modified code for your reference.
This picture is SDK code without modification:
This picture is using the register controlled interrupt handler:
Wish it helps you!
Have a great day,
Kerry
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------
Hi e Lin,
From my own view point, I don't think it is the slave's problem. Because the slave already give the ACK to master, this delay should caused by the master, when the master receive the ACK, it is not send the data clock quickly.
Just from your picture:
Bit 9 is the ACK point, you can find the slave already give ACK, it is low during 9th clock.
So, maybe you also need to check the master side.
Could you also tell me what the I2C master you are using? You can check that master code, when the master enter read mode, and get the ACK, the wait time between receive ACK and the next Clock sending. As you know, all the clock in the I2C bus is sent by the I2C master.
Wish it helps you!
If you still have question about it, please kindly let me know.
Have a great day,
Kerry
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------
Hi John Wang,
Thank you for your updated information.
Today, I have tested the sdk code directly, this is the I2C bus wave:
You can find each byte have a little delay, actually, this also relate to the code, as you know, SDK code have called a lot of API.
This small delay won't influence the I2C usage.
But if you still want to shorten it, I think, you can use use the register control directly, instead of the API code, register control will save time.
Have a great day,
Kerry
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------
Dear Kerry Zhou 
Thanks for your checking ~
Could you help to provide code about " register control directly " Method to test ?
Thanks a lot !!
Dear Kerry Zhou 
ok , i' ll test example code with 2 board , One is I2C Interrupt Master , another one is I2C Interrupt Slave . To check the issue will happen on FSL_I2C_Master sample code or not , Update i ' ll let you know , thanks a lot !!!!
Dear Kerry Zhou 
As John Wang Says , Could you provide some advice to help him ?
I 'll check Example code the same time ~ thanks a lot !!
