aboutsummaryrefslogtreecommitdiffstats
path: root/notes/i2c
diff options
context:
space:
mode:
authorbnewbold <bnewbold@robocracy.org>2010-04-07 00:59:35 -0400
committerbnewbold <bnewbold@robocracy.org>2010-04-07 00:59:35 -0400
commit2596f1409a328c6e321054a9017826ef64f1b657 (patch)
tree7f1b9a20f38f411a3c6048dde73c01bbdafbdd89 /notes/i2c
parent32feb1c68ba389c90b1002576d1c3dfb17d0df93 (diff)
downloadlibrambutan-2596f1409a328c6e321054a9017826ef64f1b657.tar.gz
librambutan-2596f1409a328c6e321054a9017826ef64f1b657.zip
updated README, LICENSE, notes
Diffstat (limited to 'notes/i2c')
-rw-r--r--notes/i2c134
1 files changed, 134 insertions, 0 deletions
diff --git a/notes/i2c b/notes/i2c
new file mode 100644
index 0000000..c324bb2
--- /dev/null
+++ b/notes/i2c
@@ -0,0 +1,134 @@
+
+Maple -------------------------------------------------------------------------
+
+We'll probably want both HardwareI2C and SoftI2C implementations, with the
+soft being bitbanged on arbitrary pins (and possibly only master-mode; almost
+certainly not multi-master).
+
+The first simple implementation will support up to 400khz operation, only
+7-bit addresses, no clock stretching (on our part), only blocking master
+read/write (with timeout), and only master n-write,n-read combined messages.
+
+A more advanced implementation would be event-driven, support DMA, higher
+clock speeds, handle error interrupts, etc.
+
+Should all of these return a status code? or just fail? or fail silently?
+
+Ring buffers etc are not needed because the length of the read/write is always
+known (or handled by the application code for slave mode).
+
+// port = {1,2}
+i2c_init(port, freq)
+i2c_disable(port)
+
+// blocking/polling
+i2c_master_read(port, slave_addr, *data, len) // wrapper around writeread
+i2c_master_write(port, slave_addr, *data, len) // wrapper around writeread
+i2c_master_writeread(port, slave_addr, *wdata, wlen, *rdata, rlen)
+
+// callbacks/interrupts
+i2c_slave_set_addr(port, addr) // enables callbacks etc if they aren't already;
+ // set ACK high
+i2c_slave_set_rx_callback(port, *function) // takes a byte?
+i2c_slave_set_tx_callback(port, *function) // gives a byte?
+
+How to init:
+ - setup pins
+ - setup peripheral input clock: 4MHz for fast mode
+ - configure clock control registers
+ - configure rise time register
+ - configure interrupts
+ - configure I2C_CR1 to enable the peripheral
+ - set ACK flag low: won't be a slave until addr is set
+
+How to master_write:
+ - START, addr[0:7], W, ACK, DATA[0:7], ACK, DATA[8:15], ..., STOP
+ - software sets START high, which toggles M/SL to master mode
+ - when START actually happens, SB is set and interrupt happens; hardware
+ waits until address is written to DR
+ - address shifts out and an interrupt is thrown with ADDR high; if LSB of
+ address was low, in transmitter mode. TRA reflects this
+ - software writes to the first byte to DR and clears ADDR
+ - first byte shifts out and when there's an ACK an interrupt is thrown
+ with TxE high; if no new byte was written to DR since the previous
+ transmission BTF goes high
+ - software writes next byte to DR and clears BTF, or sets STOP bit to end
+ data transmission, or sets START to begin next part of combined session
+ - after STOP is set, hardware goes back to slave mode
+
+How to master_read:
+ - START, addr[0:7], R, ACK, DATA[0:7], ACK, ..., NACK, DATA[-8:-1], STOP
+ - software sets START high, which toggles M/SL to master mode
+ - when START actually happens, SB is set and interrupt happens; hardware
+ waits until address is written to DR
+ - address shifts out and an interrupt is thrown with ADDR high; if LSB of
+ address was high, in receiver mode. TRA reflects this.
+ - software clears ADDR; hardware shifts in first byte from slave and sends ACK
+ if ACK is set. interrupt is thrown with RxNE high, with BTF set
+ - if only reading one byte, ACK should be set low and STOP high; hardware will
+ then NACK and STOP after reading
+ - software reads from DR and clears BTF bit. next byte is shifted in
+ - software sets NACK low after second-to-last byte is read by setting ACK low
+ after reading from DR, and either setting STOP or START high
+ - after STOP is set, hardware goes back to slave mode
+
+How to master_writeread:
+ - [START, addr[0:7], W, ACK, WDATA[0:7], ...],
+ [START, addr[0:7], R, ACK, RDATA[0:7], ACK, ..., NACK, RDATA[-8:-1]]
+ STOP
+ - implement the above read/write but set START high instead of STOP high at
+ the end
+
+How to slave tx (master read):
+ - if address is matched, ACK is send (if ACK flag is high) and ADDR goes high;
+ event interrupt is triggered if ITEVFEN flag is set. TRA indicates whether
+ this is a read or write request
+ - program sets DR degister and clears ADDR flag; hardware will clock stretch
+ until ADDR is low
+ - if ITVENFEN and ITBUFEN are set, the TxE interrupt is triggered after the
+ byte is sent and the master ACKs. now the hardware stretches until BTF is
+ cleared before sending again.
+ - after the STOP is sent, hardware sets STOPF, throws an event if ITEVFEN is
+ high, and waits for a read of SR1 and write CR1
+
+How to slave rx (master write):
+ - if address is matched, ACK is send (if ACK flag is high) and ADDR goes high;
+ event interrupt is triggered if ITEVFEN flag is set. TRA indicates whether
+ this is a read or write request
+ - the hardware shifts in the byte to DR and sends an ACK if ACK flag is high
+ - if ITEVFEN and ITBUFEN are set, an RxNE interrupt is sent. hardware waits
+ until the BTF flag is reset
+ - after the STOP is sent, hardware sets STOPF, throws an event if ITEVFEN is
+ high, and waits for a read of SR1 and write CR1
+
+i2c General Reference ---------------------------------------------------------
+
+http://en.wikipedia.org/wiki/I²C
+
+4 modes/roles for a given device. These designations are transaction specific;
+a device can switch roles from one transaction to the next:
+ master transmit
+ master receive
+ slave transmit
+ slave receive
+
+
+STM32 Specific Reference ------------------------------------------------------
+
+see stm32 docs: reference manual (doc 13902 rev10) pg637/1003
+
+see STM32F10x_StdPeriph_Lib_V3.2.0/Project/STM32F10x_StdPeriph_Examples/I2C/Interrupt for an example project (reading/writing between the two i2c buses) based on the stm32library
+
+>100khz is "fast mode", which goes up to 400khz. stm32 supports up to 400khz.
+ things are different in fast mode: 4mhz input clock
+
+interrupts see pg652/1003
+registers see pg654/1003
+
+PEC = packet error checking. don't think this is i2c standard
+
+each i2c port has an error and an event interrupt
+
+the stm32 is in slave mode by default; it enters master only when executing
+a request
+