How to locate a union at a specific location in memory?

[matrixofdynamism]

I have used bitfields in the past but only on rare occasions. I have not tried to fully understand how to use them as I was programming in C++ and VHDL. In my present situation though I am trying to figure out every single detail that I consider relevant to bitfields. I have habit of taking notes. When I come across a problem, I try to understand all aspects of it, answer all questions by asking real people or forums or reading books or application notes that then write it down in my notebook.

I wanted to write a program where I can access SFRs like variables. That is solved by using struct, getenv() and #byte. Then I wanted to be able to access fields in an SFR like field in struct, that is solved by using bitfields. However, there are a few more questions that must be answered which I have posted later in this thread, and for them I am going to do some experiments today as another person has advised.

I have an 8-bit ADC. It takes a 16 bit word as input. The first 4 bits are programming bits then come 8 data bits, these 8 bits are clearly split between the upper and lower byte. Then we have 4 bits of don't care. This can be created as a bitfield, or it can be implemented using bit-shift and logic operation.

I shall also have to write more code which can be simplified with bit fields and for all this I am trying to fully understand how to use bitfields in CCS PIC C compiler so I know what I am doing.

[temtronic]

re:
I have an 8-bit ADC. It takes a 16 bit word as input. The first 4 bits are programming bits then come 8 data bits, these 8 bits are clearly split between the upper and lower byte. Then we have 4 bits of don't care. This can be created as a bitfield, or it can be implemented using bit-shift and logic operation.


The easy solution is to rearrange the 16 bit word as
4 bits of programming bits
4 bits of don't care
8 bits of ADC data

this way the ADC data is 'whole',can be accessed as a byte, faster.

This same regrouping helps with 'bit' data as well. Group 8 similar bits into a byte, then 8 bit integers,then 16 bit words. Usually bits are used as 'flags' so grabbing a 'byte' of flags for testing or configuring and tends to be easier on the eyes to read them.

As the programmer you have( or should have) control over the format of the data.

Jay

[matrixofdynamism]

The DAC is MCP4801. The don't care bits must be after the data so I guess I shall have to put them this way in the struct also. See Figure 5-3 in http://ww1.microchip.com/downloads/en/DeviceDoc/22244B.pdf#page=23

Basically, the data containing struct which has bitfields shall be directly written into the SSPBUF for SPI write, one byte at a time.

By the way, should the bit field members be declared as int8 or unsigned char or something else? I am sure that a convention exists for the CCS PIC C Compiler that is perhaps not portable.[/img]

[RF_Developer]

It's a DAC, not an ADC, of course.

That sort of arrangement is common. Its done so that you can treat all three versions, 8, 10, and 12 bit the same way, i.e. as if they were 12 bit, and deal with the data you send as an int16. That simplifies writing a driver as you can have one common driver for all versions of the family. The 10 and 8 bit chips simply ignore the two or four "don't care" bits.

Of course the compiler has to do much the same bit shift and logic operations in bitfields as you would do if you constructed the two bytes yourself. It'd do it less efficently - giving bigger and slower code - than we humans can. This is a pretty trival example that would be much simpler, smaller and faster the shift way. That would be my first and probably only approach to this problem.

EDIT: Indeed, it is how I approach the problem for the very similar MCP4822, which is essentially two MCP4812s in a single package.

[matrixofdynamism]

I got the DAC to work finally. However, I was held for a while. This is my first time using SPI on the PIC in actuality. The DAC would not do anything and after looking at the waveforms by sending 0x01 and 0x80, 0x20 and 0x40, I realized that the data was changing at the rising edge of clock and thus expected to be latched at the falling edge. This was opposite of what was given in the datasheet for the DAC.

The DAC mentioned mode 00 and mode 11 as valid SPI modes. I did not know what that even means but I came across description of these modes by chance in the PIC datasheet I have printed out. It gave the values for CKP and CKE that correspond to each mode. I used the CCS PIC C debugger through the ICD-U64 to read the SFRs. It was clear that the values of these bits in the SSP1CON1 and SSP1STAT as existed in the PIC do not give me mode 00 or mode 11. Thus, I eventually used assembly instructions to set and clear the CKP and CKE bits accordingly.

How am I supposed to set a specific mode of SPI operation using the setup_spi command? What does SPI_H_to_L and SPI_L_to_H mean?

The SPI bus modes are described here:

http://ww1.microchip.com/downloads/en/DeviceDoc/39960d.pdf#page=289

[PCM programmer]

[Ttelmah]

As a comment though, if you are new to CCS SPI, then really don't use setup_spi....

This is the 'old' SPI interface command. The new (and current) way to do SPI configuration, is #use spi. #use spi, 'knows' SPI mode numbers directly, and is a much more powerful and flexible way to work. For instance it allows you to directly send 8bit, 16bit, and 32bit values in a single command. More importantly though it allows you to have multiple SPI interfaces accessed by name. It's a much tidier solution all round.

[RF_Developer]

[matrixofdynamism]

hmm, I thought that the #spi was for "software SPI" i.e bit-banging as it is called and setup_spi() was for hardware spi. So that is not true?

RF_Developer, if the bit_clear and I supposed by that extension that bit_set are redundant or outdated, what is the newer method to do this? The reason I went for assembler is that it will enable understanding the PIC hardware internally. Yes, it is true that I have written the first 2 assembly instructions of my life last night which are BCF and BSF.

[RF_Developer]

Read the CCS manual. What does it say?

[RF_Developer]