I wanted to thank everybody for help and advice! We finally got
everything working well between the MicroMag3 and LPC2119 using
SPI0/SPI1. The problem was with the sensor itself and PNI Corp has
replaced it with working one.

As to the multiple byte transfers on SPI in LPC2119, it works as
expected and demonstrated in code samples in this thread.

--- In l..., wrote:
>
> I was never able to get the SPI engine to send 16 bits. I ended up
writing my own code to do it by GPIO.
> I don't believe the LPC SPI engines can do anything but 8 bit data
word size.
>
> Chris.

An Engineer's Guide to the LPC2100 Series

irafalovsky wrote:

> In our code, we make chipselect using a separate pin, LPC2119
> requires SPI SSEL pin to be high if MCU is in master mode, therefore
> we connected it to VDD, as per user manual.
>
> I will ask the engineer who used && operand there for his reasoning,
> I think it wasn't a typo...
>

"&&" is a logical AND, while "&" is a boolean AND.

TomW
>
> --- In lpc2000@yahoogroups .com ,
> "tirfec" wrote:
> >
> > Perhaps it was a typo but the line where you wait for the sensor
> to
> > signal that data is ready...
> >
> > while(!(IOPIN0 && (1 << 22))){;;} // Wait for Data Ready
> >
> > has an extra & in the masking of IOPIN0. Don't you really mean...
> >
> > while(!(IOPIN0 & (1 << 22))){;;} // Wait for Data Ready
> >
> > The first line will only take the while() path once because
> (1<<22)
> > will always evaluate (and thus the ! of the whole statement will
> make
> > it leave the loop) to true and it won't ever matter what the read
> > of IOPIN0 is. This seems like it may be consistent with your
> failure
> > since the sensor is probably never ready when you start asking it
> for
> > the data (i.e doing you SPI reads).
> >
> > Try Alan's example code with this correction and maybe things will
> > start working..assuming you've configured the peripheral properly
> etc.
> >
> > Also, someone made the point about the SSEL1 toggling after every
> > byte. This may cause problems with the sensor which probably
> expects
> > the chip select to stay low through all 16 clocks for the data. So
> > unless you are manually controlling another GPIO for the chip
> > select...be aware of that behaviour on SSEL1.
> >
> > Hope this helps.
> >
> > Best regards,
> > Ryan.
> >
> > --- In lpc2000@yahoogroups .com ,
> "Alan Strickland"
> > wrote:
> > >
> > > To clear the SPIF bit in S1SPSR you have to access the S1SPSR
> then
> > > access the S1SPDR (ie read or write to it). You also have to
> shift
> > > out a byte to receive one, and wait for the SPI status to
> indicate
> > > SPIF before the received byte can be read.
> > >
> > > I use interrupts, so I can't speak to how well polling the SPI
> > status
> > > register works. This code is also ignoring any errors that the
> SPI
> > > may be setting as well.
> > >
> > > Try something along these lines:
> > >
> > > S1SPDR = 0x41; // Send Measure X command
> > > while((S1SPSR & 0x80)!= 0x80){;;} // Wait for SPI status
> > > temp = S1SPDR; // Clears SPIF in S1SPSR.
> > >
> > > while(!(IOPIN0 && (1 << 22))){;;} // Wait for Data Ready
> > >
> > > S1SPDR = 0;.
> > > while((S1SPSR & 0x80)!= 0x80){;;} // Wait for SPI status
> > > MM3_HB = S1SPDR; // Clears SPIF in S1SPSR.
> > >
> > > S1SPDR = 0;
> > > while((S1SPSR & 0x80)!= 0x80){;;} // Wait for SPI status
> > > MM3_LB = S1SPDR; // Clears SPIF in S1SPSR.
> > >
> > > wMM3_X = MM3_HB;
> > > wMM3_X = (wMM3_X << 8) | MM3_LB;
> > >
> > > --- In lpc2000@yahoogroups .com
> , "irafalovsky" wrote:
> > > >
> > > > If we attempt to send 2 bytes after peripheral indicates
> > available
> > > > data, we simply get zeroes in return...
> > > >
> > > > Here's what code looks like:
> > > >
> > > > IOSET0 |= 1 << 23; // Set RESET HIGH, this is to reset the
> MM3
> > > > delay(10);
> > > > IOCLR0 |= 1 << 23; // Set RESET LOW, pulse formed
> > > > delay(10);
> > > > S1SPDR = 0x41; // Send Measure X command
> > > > while((S1SPSR & 0x80)!= 0x80){;;} // Wait for SPI status
> > > > while(!(IOPIN0 && (1 << 22))){;;} // Wait for Data Ready
> > > > S1SPDR = 0;
> > > > S1SPDR = 0;
> > > > MM3_HB = S1SPDR;
> > > > MM3_LB = S1SPDR;
> > > > wMM3_X = MM3_HB;
> > > > wMM3_X = (wMM3_X << 8) | MM3_LB;
> > > >
> > > >
> > > > >
> > > > > Do you send 2 bytes out from LPC2119 to allow the SPI clock
> to
> > go so
> > > > > you can get back the 2 bytes that you are expecting?
> > > > >
> > > > > We have an application with LPC2138 running in slave mode and
> > > > > receiving packets of 9 bits characters from an host
> processor
> > and
> > > > > responding with packets of 9 bits characters.
> > > > > Naturally, in order to be able to send back the answer the
> LPC
> > must
> > > > > receive the spi clock from the master so the host processor
> > must pump
> > > > > out dummy characters to allow the answer to come back.
> > > > >
> > > > > Regards,
> > > > >
> > > > > Carlo Barbaro
> > > > >
> > > >
> > >
> >

--
Tom Walsh - WN3L - Embedded Systems Consultant
http://openhardware.net, http://cyberiansoftware.com
"Windows? No thanks, I have work to do..."
----------------

>Original Message:
>-----------------
>From: derbaier
>>--- In l..., "subscriptions@..."
wrote:
>>>
>>>
>>>
>>>From: derbaier
>>>
>>>But the LPC2119 does toggle slave select every 8 bits, which indicates
>>>to me that it's internal state machine is resetting regardless of what
>>>the peripheral is doing?
>>
>> If that's the case it's broken. Don't use the automatic slave select.
If
>> it toggles every word it will work with only a small number of slave
>> devices.
>
>Isn't that what this thread is all about???

I didn't think so. I thought there was an injunction to check the chip
select early on. We seem to have wandered off into the weeds on it but I
thought the main thread had already dealt with it, as in chipselct was oon
and stayed on through the transaction.

Robert
--------------------------------
mail2web - Check your email from the web at
http://mail2web.com/ .

The code that Alan presented should be correct...however, you
absolutely MUST change the double ampersand (&&) to a single one (&)
in your wait loop.

For example,

while(!(IOPIN0 && (1 << 22))){;;} // Wait for Data Ready

must be replaced with...

while(!(IOPIN0 & (1 << 22))){;;} // Wait for Data Ready

In the top case you will never wait since && means a logical 'AND'.
Therefore (1<<22) will always evaluate to non-zero (therefore
TRUE)...and its very likely that at least one of the bits in the read
of IOPIN0 will be high thus making that part TRUE and the whole
expression will evaluate to TRUE...which after you negate with the !
out in front will cause you to leave the loop immediately (which is
not what you intended for it to do).

What you mean to do with your expression is to MASK the bit of
interest (in this case bit 22) with a read of IOPIN0...that's what
the single & is for...take a moment in any C reference to review the
meaning of the operators and you'll be back on track.

I took a quick glance at the datasheet for the sensor you are using
and the Data Ready signal may not be asserted for 500 microSeconds in
the best case and could take as long as 60 milliSeconds in the worst
case (based on the Period Select setting...which I'm not sure which
one you are using). Since your wait loop never actually waits, you
will NEVER expect to see valid data! Even if you have the SPI and
chip select configured properly you will NEVER see valid data because
you are not actually waiting for it.

I'm sure you've checked the direction of Pin 22 but just for sanity
make sure you have it configured as an input also. You may want to
pay fairly close attention to this pin and whether or not there is an
internal pull-up enabled on it. It's very unlikely, but perhaps the
sensor's Data Ready output isn't strong enough to pull the line low
when data is not yet ready. Maybe just double check the levels to
which the pin is toggling and see that they are at expected levels.

I hope this helps you get things working...I know what it feels like
to be in pain sometimes.

Best regards,
Ryan.
--- In l..., "irafalovsky" wrote:
>
> Would you be willing to share a code example with the group or send
> it to me? We're banging our heads on the wall here... :-)
>
> --- In l..., "subscriptions@"
> wrote:
> >
> >
> >
> > >Original Message:
> > >-----------------
> > >From: derbaier
> > >--- In l..., "subscriptions@"
> > > wrote:
> > >
> > >> There is no way for the slave HW to know how many bits the
> master
> > >>uses per word so it cannot have any dependancies on word size.
> > >
> > >I thought that the transition of the slave select signal
> identified
> > >the end of the current word, and reset the slave state machine to
> > >begin the next word???????????????
> >
> > Nope, Chip select is done on a frame basis. Of the three devices
> I've got
> > on one board at the moment, one has a frame size of 12 bits, one
> of 24 bits
> > and one is an indeterminate multiple of 8 bits (the frame size
> depends on
> > how many 8 bit bytes I want to read/write at once). All
> controlled via the
> > Philips inital SPI with a fixed 8 bit word size. If you toggle
> the slave
> > select between words you will definitely fail.
> >
> > Robert
> >
> >
> > ------------------------------
-
> -
> > mail2web - Check your email from the web at
> > http://mail2web.com/ .
>

--- In l..., "subscriptions@..."
wrote:
>
> >Original Message:
> >-----------------
> >From: derbaier
> > wrote:
> >>
> >> Nope, Chip select is done on a frame basis. Of the three devices
I've
> got
> >> on one board at the moment, one has a frame size of 12 bits, one
of 24
> bits
> >> and one is an indeterminate multiple of 8 bits (the frame size
depends on
> >> how many 8 bit bytes I want to read/write at once). All
controlled via
> the
> >> Philips inital SPI with a fixed 8 bit word size. If you toggle
the slave
> >> select between words you will definitely fail.
> >
> >But the LPC2119 does toggle slave select every 8 bits, which indicates
> >to me that it's internal state machine is resetting regardless of what
> >the peripheral is doing?
>
> If that's the case it's broken. Don't use the automatic slave
select. If
> it toggles every word it will work with only a small number of slave
> devices.
>
> Robert
>

Isn't that what this thread is all about???

-- Dave

>Original Message:
>-----------------
>From: derbaier
> wrote:
>>
>> Nope, Chip select is done on a frame basis. Of the three devices I've
got
>> on one board at the moment, one has a frame size of 12 bits, one of 24
bits
>> and one is an indeterminate multiple of 8 bits (the frame size depends on
>> how many 8 bit bytes I want to read/write at once). All controlled via
the
>> Philips inital SPI with a fixed 8 bit word size. If you toggle the slave
>> select between words you will definitely fail.
>
>But the LPC2119 does toggle slave select every 8 bits, which indicates
>to me that it's internal state machine is resetting regardless of what
>the peripheral is doing?

If that's the case it's broken. Don't use the automatic slave select. If
it toggles every word it will work with only a small number of slave
devices.

Robert
--------------------------------
mail2web - Check your email from the web at
http://mail2web.com/ .

>Original Message:
>-----------------
>From: irafalovsky
>
>Would you be willing to share a code example with the group or send
>it to me? We're banging our heads on the wall here... :-)

Check the newlib-lpc stuff at http://www.aeolusdevelopment.com

It's only for SPI0, developed originally for LPC2104/5/6 so it has none of
this fancy word size selection etc... ;)

There is an FRAM driver though and that does use the spi driver to to the
underlying work. Since the structure is meant to be built on to abstract
the SPI functionality away from the device driver it's not the clearest
exposition of the SPI itself. It is, however, commented.

Robert

--------------------------------
mail2web - Check your email from the web at
http://mail2web.com/ .

Would you be willing to share a code example with the group or send
it to me? We're banging our heads on the wall here... :-)

--- In l..., "subscriptions@..."
wrote:
>
> >Original Message:
> >-----------------
> >From: derbaier
> >--- In l..., "subscriptions@"
> > wrote:
> >
> >> There is no way for the slave HW to know how many bits the
master
> >>uses per word so it cannot have any dependancies on word size.
> >
> >I thought that the transition of the slave select signal
identified
> >the end of the current word, and reset the slave state machine to
> >begin the next word???????????????
>
> Nope, Chip select is done on a frame basis. Of the three devices
I've got
> on one board at the moment, one has a frame size of 12 bits, one
of 24 bits
> and one is an indeterminate multiple of 8 bits (the frame size
depends on
> how many 8 bit bytes I want to read/write at once). All
controlled via the
> Philips inital SPI with a fixed 8 bit word size. If you toggle
the slave
> select between words you will definitely fail.
>
> Robert
> -------------------------------
-
> mail2web - Check your email from the web at
> http://mail2web.com/ .
>

--- In l..., "subscriptions@..."
wrote:
>
> >Original Message:
> >-----------------
> >From: derbaier
> >--- In l..., "subscriptions@"
> > wrote:
> >
> >> There is no way for the slave HW to know how many bits the master
> >>uses per word so it cannot have any dependancies on word size.
> >
> >I thought that the transition of the slave select signal identified
> >the end of the current word, and reset the slave state machine to
> >begin the next word???????????????
>
> Nope, Chip select is done on a frame basis. Of the three devices
I've got
> on one board at the moment, one has a frame size of 12 bits, one of
24 bits
> and one is an indeterminate multiple of 8 bits (the frame size
depends on
> how many 8 bit bytes I want to read/write at once). All controlled
via the
> Philips inital SPI with a fixed 8 bit word size. If you toggle the
slave
> select between words you will definitely fail.
>
> Robert
>
But the LPC2119 does toggle slave select every 8 bits, which indicates
to me that it's internal state machine is resetting regardless of what
the peripheral is doing?

I hope that Mr. Irafalovsky posts his final outcome, because now my
curiosity is really arroused.

-- Dave

>Original Message:
>-----------------
>From: irafalovsky

>I wonder, what mechanism is used to "deselect the slave" if the pin
>for chipselect is not controlled by SPI but by a user program, is
>that some sort of internal workings of SPI implementation on LPC?

That's your job. You select the slave before starting the frame and
deselect it after frame end.

Robert
--------------------------------
mail2web - Check your email from the web at
http://mail2web.com/ .