uALFat Baud Rate

Kit Morton

Hi all,
In my latest project I needed FAT32 support so I got a uALFat chip, and I am trying to interface it to the propeller.
I am using the serial interface with the FullDuplexSerial object. And I can talk to the the chip but I want to use a baud rate over 9600. But when I try to change the baud rate of the chip, then change the serial objects baud rate, I can't communicate with the chip any more .
Has anyone tried to change of this chip, and if so what did you have to do to make it work?
Or should I just bag the Serial and go for SPI?
Thanks
-Kit

tekochip

I had the same trouble, everything was fine at the lower baud rates, but it couldn't communicate at the higher rates.· All the timing was spot on, I just assumed it was some trouble with the serial driver not being able to handle full duplex at the higher rates.· I think I may have gotten it up to 115K without trouble, and then decided that I really needed to go I2C.· The SPI can go faster, but I was disappointed by the pin count required for SPI.

For the I2C you can get up to the 400KHz spec without trouble, and since you already have to dedicate pins for SDA and SCL, the only penalty is the handshaking line.· I only use one handshaking line to see if the uALFAT still has data.· The only precaution I can give is that you'll need a START, COMMAND, DATA, STOP for each character, so it's twice as chatty as it needs to be.· Even still,·you can move 4K images in 400mS.

Oh yea, the uALFAT also has a 512 byte buffer, so you'll get 512 bytes then the uALFAT will go silent for·5 mS while it fetches more data.· I was streaming audio and needed a buffer to get over the dead period.

·

Kit Morton

Thanks for the tip.
I tried to do this, but I couldn't figure out how to make the I2C object work in the rite way.
Could you post some code snippets for how you implemented it.

Thank you
-Kit

tekochip

Sure, I started with the Hydra I2C driver, but it got so gutted that very little of the original remains.· It's sort of like George Washington's axe; I had to replace the handle, and I had to replace the blade.

{*******************************************ReadAlfat*********************************************
It's a pain, but the Alfat can only read one byte at a time, so all transfers are twice larger
than the really need to be because you have to address the chip all over again
**************************************************************************************************}
ReadAlfat
              '--- Setup Normal Condition ---
              andn     dira, HandShake          'Make sure it's an input
 
:Loop
              'Wait until ready
              mov      dataByte, ina                    'Grab the handshaking line
              test     dataByte, HandShake wz           'Is it low?
        if_z  jmp      #:Loop                  '        Yes, cycle until ready
                                                         
                                                         
              call    #StartCondition                   'generate start condition
              mov      dataByte, #$A5                   'The Alfat's read address
              call    #SendByte                         'send device address byte
              call    #CheckAck                         'check for ACK from eeprom
                                                 
                    '--- Read Byte ---
              call      #ReceiveByte                    'receive byte
              wrbyte    dataByte,BufferAddress          'store received byte into buffer
              add       BufferAddress,#1                'increment buffer address                                                
              sub       paramByteCount,#1    wz         'decrement number of bytes left
              wrlong    paramByteCount,byteCountAddr    'let other cogs know the progress so far
                                                         
                                                         
                                                         
              test    LongMax,#1    wc                  'set C to 1 for "No ACK" (meaning "done")
              call    #SendBit                          'send "ACK"/"No ACK"
              call    #StopCondition                    'generate stop condition
        if_nz jmp     #:Loop                            'loop                                                                                  
              
ReadAlfat_ret
              ret
 
 
{*****************************************StartCondition*****************************************
I2C is supposed to idle SCL and SDA high, and I believe this has caused a number
of timing problems, so everything will be changed to make it more robust and faster
*************************************************************************************************}
StartCondition
              or       outa,datMask             'Make sure data and clock are high
              or       outa,clkMask             'pull clock high
              or       dira,datMask             'set data pin to output
              or       dira,clkMask             'Verify clock is too
              
              mov      timer, #T_HIGH           'Get the high period
              add      timer, cnt               'Add the time
              waitcnt  timer, #T_HIGH           'Wait and get ready for next high period
              
              andn      outa,datMask            'pull data low
              waitcnt  timer, #T_HIGH           'Wait and get ready for next high period
              andn      outa,clkMask            'pull clock low
              waitcnt  timer, #T_HIGH           'Wait and get ready for next high period              
StartCondition_ret
              ret                            

There is no error checking if the uALFAT doesn't respond because in my application it wouldn't matter, the machine would be broken, so you may want to add some sort of abort to the handshaking loop.· The main issue with the Hydra driver is the way waitcnt is used for bit timing.· It·gets a little flakey when you·reach full speed·and that's why I·wanted to rewrite it.· You'll want to make sure that each transfer starts with reloading your timer, or else waitcnt will roll over and you'll be stuck waiting·50 seconds for the next bit to shift out.
·