H/W SPI on a pic24

geokonst · Joined: 04 Apr 2009 Posts: 27

Dear all,
Has anyone implemented a hardware spi on a pic24 before? Even better if using Peripheral Pin Select(Pin_select) ?

If yes could you please post the code because I am stuck for days (see my other post on SPI_READ) and about to give up.

Thank you in advance.

FvM · Joined: 27 Aug 2008 Posts: 2337 Location: Germany

I suppose it's working regularly. Of course, there may be some bugs as with other PCD built-in functions. Ask me a few days later.

Guest · Posted: Tue Apr 07, 2009 10:03 pm

geokonst-

Any luck with your SPI module? I'm working with the PIC24FJ256GB106 and am also having problems getting it to function. Are you using the #use spi functionality? I've tried both that and writing directly to the control registers without any success.

geokonst · Joined: 04 Apr 2009 Posts: 27

Dear Guest,
I am afraid I haven't done any progress at all. Yes I've ended up using #use spi because ccs support recommended that. But still I cant read the spi repsonse. see also :
http://www.ccsinfo.com/forum/viewtopic.php?t=38457

It seems to me there is some weird bug with h/w spi for 24f. What exactly is your problem with spi?

Wish you luck.

FvM did you have a chance to look at it?

Guest · Posted: Wed Apr 08, 2009 10:13 am

geokonst-

I noticed you other post after I responded to this one and it seems we are having nearly the same problems. i am unable to get so much as a clock out signal from my SPI bus, and the designated DO pin will not put out any data. I've tried different forms of configuration and checked the status and configuration registers to verify that they are being set correctly (which it appears they are). I've also viewed the the pin select configuration registers to check if the pins are being mapped correctly (which again they appear to be). At this point I'm at a loss as it seems to be a H/W problem like you mentioned. I've contacted Microchip's support center about it an am awaiting a reply. I'll post again if I find out anything useful.

geokonst · Joined: 04 Apr 2009 Posts: 27

Thank you.

In the meanwhile could you please post the version of your code that writes directly to the control registers to give it a go?

cheers

Guest · Posted: Wed Apr 08, 2009 12:20 pm

The register locations are defined as follows:

geokonst · Joined: 04 Apr 2009 Posts: 27

So are you using SPI1BUF then to send something or spi_write? What are you trying to communicate with?

Have you tried the second code I posted? That doesn't doesn't even generate a clock?

I don't believe that something in the errata has anything to do with our issue, but maybe you might want have a look at it. There are three articles on SPI

http://ww1.microchip.com/downloads/en/DeviceDoc/80369d.pdf

Please let me know if have had any progress, because I really need to make this work ASAP.

Thanks

Guest · Posted: Wed Apr 08, 2009 1:08 pm

I've tried both SPIBUF and spi_write without any success, I'll take a look at the second code you've posted and see if I can't make some changes on my own. I'm still awaiting a reply form Microchip as I am also needing to get this done as soon as possible.

geokonst · Joined: 04 Apr 2009 Posts: 27

Are you working with an accelerometer too? I actually need to implement two spis (like one was not enough); one between the accelerometer and the gb106 one between the gb106 and another pic.

The good news is that I've already done software spi for the accelerometer (in case I can't make HW work). Unfortunately software spi will be too slow for comms between the two pics.

Tomorrow morning (UK time)I will try the registers approach and see what I get on the scope.
Please update me if you find anything.
Cheers!

Guest · Posted: Wed Apr 08, 2009 1:31 pm

A few questions:

Is the spi.h file a user defined one or one that came with the compiler?

In your code you have an initial spi_read() and then another when you send 0xFF. Is the value based on your sensor's communication protocol or is there another reason for it?

I looked over the errata and agree, I don't think any of the listings are relevant to the problems we're having (although it may have to do with the 30 us delay you had to insert between read functions).

geokonst · Joined: 04 Apr 2009 Posts: 27

Dear Guest,
I just happened to name my c file spi so spi.h was the file generated by the wizard.

The reason I am used two spi_read is that if I tried with one write and one read (as it would be the normal thing to expect) I would get only 8 clocks. What I found out later is that the two clocks for the two commands overlapped. That's where the delay comes in.
I have no idea why using two reads works when two writes or one write and one read won't (without the delay) .

The FF on the second read is just dummy data to keep the clock going which is I believe common practice. I chose FF instead of 00 because the mode is 3 on the acc. 8F is the accelerometers who is register.

Any news from microchip? have you tried ccsinfo customer support? They tend to answer within hours. They still have answered me though in my last email :(

cheers

FvM · Joined: 27 Aug 2008 Posts: 2337 Location: Germany

As expected, the PIC24FJ128GA106 SPI is working regularly. As with other PIC24, it's recommended to read the datasheet thoroughly and to check, which PCD built-in functions are working correctly and which are apparently faulty. I used PCD V4.084. There may be SPI related changes towards V4.090, but rather unlikely.

I used this setup, with some additional definitions.

geokonst · Joined: 04 Apr 2009 Posts: 27

Dear Fvm,
Thank you for your reply.
Could you please clarify for me the following;
What exactly SPI_CPHA_0 is doing?
How did you find out that RPI45 was not working with pin_select? Did you get an error message? could this be causing my problem?
Finally why are you using four spi_reads? Do you need to send 3 bytes before the memory chip responds or is it some other reason?

Thanks for you help.

FvM · Joined: 27 Aug 2008 Posts: 2337 Location: Germany