SPI Modes and Timing.


In a lot of cases, when using SPI, we do need to use "SPI_Init_Advanced". It has a number of parameters. Here the parameters regarding the SPI "mode" are described. The "mode" consists of in
procedure SPIx_Init_Advanced(..., data_sample, clock_idle, edge: word);

The "data_sample" parameter does not belong to the actual "SPI mode", it is an extra feature of the MCU, see here.

The SPI "Modes"

SPI knows 4 "standard" modes, reflecting the SCK's polarity (CPOL) and the SCK's phase (CPHA).

The definition is:
0 (or 0,0) 0 0
1 (or 0,1) 0 1
2 (or 1,0) 1 0
3 (or 1,1) 1 1

The meaning is:

(*): the transmit edge is the clock edge at which the SDO level changes

In a timing diagram this looks like(only one clock pulse shown here):

The Transmit edge is the clock edge at which the SPI output data changes,
the Sampling edge is the clock edge at which the sampling of the SPI input data takes place.
The sampling edge is normally the opposite one of the transmit edge, but see also here.

SPI and PIC/ dsPIC

The PIC MCU supports a 4 SPI modes, but the MCU registers involved are named differently and behave differently.
The 2 MCU registers are CKP (Clock Polarity) and CKE (Clock edge). The relation between those 2 registers and the SPI modes is:

0 (or 0,0) 0 1
1 (or 0,1) 0 0
2 (or 1,0) 1 1
3 (or 1,1) 1 0

As you can see, CKP behaves the same as CPOL, CKE is the inverse of CPHA.

This means that a simple "SPI_Set_mode" looks like this:
procedure Set_SPI_mode(CPOL_, CPHA_: byte);
// The 2 paremeters are:
// CPOL: SPI clock polarity: 0 = Clock Idle LOW;               1 = Clock Idle HIGH
// CPHA: SPI clock Phase:    0 = Transmit edge active to idle; 1 = Transmit edge Idle to active
  CKP_bit := CPOL_;
  CKE_bit := (CPHA_ xor 1) and 1; // invert bit zero
Above routine can e.g. used after a usage of an mE SPIx_Init... routine to (re)set the SPI mode.

The mE SPI library usage

In the mE SPI library the "SPI mode" is set by 2 parameters of the "SPIx_Init_Advanced" function.

The the parameters for PIC are : "clock_idle" and "transmit_edge"
The the parameters for dsPIC are: "clock_idle" and "edge"

Important: in both cases the last parameter ("transmit edge" or "edge" parameter) is NOT the transmit edge, but the other one, called the "Sampling edge", see here. As you can see the name of the PIC parameter is wrong (it says "transmit_edge").

Furthermore the constant values used to give the parameters their values are:

For PIC:

For dsPIC:

(**) Again: the "edge" is the "Sampling Edge", not the "Transmit Edge", see here.

Deriving the parameters from the device's "SPI mode"

The easiest way is using the "SPI mode" if known, the hardest is deriving the parameters from the device's timing diagram.

The table below gives the relationship between an SPI device's "SPI mode" and the parameter values to use with "SPIx_Init_Advanced":

SPI MODE clock idle parameter Edge parameter
= the "Sampling" edge

Important: In mikroPascal the parameter "edge" represents the clock edge where in input data is sampled, not the "transmit edge" (the clock edge at which the SDO data is changed).

The parameter "data sample" is not included in the "SPI mode", but usually "_SPI_DATA_SAMPLE_MIDDLE" will do fine, see however also here.

Defining parameters from the device's timing diagram

This manner has to be applied if the "SPI mode" of the device at hand is not (explicitely) defined.

To find out the necessary values of the parameters we have to look to the device's SPI timing diagram we want to handle:

Keep in mind that in the "device" diagram the "device" timing is shown. For the PIC timing input and output should be interchanged (which was already done in the above diagram, it shows the timing from the PIC's perspective).

1. The "clock_idle" parameter

See 1 in the diagram. The level wanted is the SCK level at the moment notCS changes. In the diagram this level is zero, so: the clock idle is low.
This means also that (again in our case) the clock active level is high.

2. The "edge" parameter

See 2 in the diagram. The edge wanted is the one in the middle between 2 PIC output (PIC SO) datachanges (two "transmit" edges, represented by "x" in the timing diagram).
In our case we see that it is the edge from low to high, or from idle to active.

The "data_sample" parameter

The data_sample parameter is not defined in the SPI mode, it is an extra feature of the MCU.
Normally the "Sampling edge" (see here) is the opposite one of the "transmit edge". This is achieved by setting the "data_sample" parameter to the value "_SPI_DATA_SAMPLE_MIDDLE".
The sampling moment can however be shifted to the same moment as the transmit edge by setting the "data_sample" parameter to the value "_SPI_DATA_SAMPLE_END" value, see here).

How to define which one to use:
See 3 in the above diagram.
First define the "data output time", represented by an x in the timing diagram. It is the time from one PIC output data (PIC SO) change to the next.
Now compare the device data out (SO = PIC SI) with the "data output time". The data_sample value we have is the best moment in the "data output time" the PIC should take a sample of its input data (the device output data). In our case, the middle of the "data output time" seems to be the most appropriate (at the end would be not bad either).


a. Timing diagram (derived and modified from Mchp datasheet)

This is the (master) timing diagram with the terms explained: