Setup/enable I2C internal pull-ups

[Torello]

Hi,
How do I enable/setup the internal pull-ups for I2C? Preferably with a CCS build-in function.
I'm using the PIC18F26Q83 that does have them.
I would only like pull-ups on those two pins (C3/C4)
_________________
Regards, Edwin. PCWHD v5.114

[Ttelmah]

OK.
Several things. The internal pullups are _only_ acceptable for slow speed
I2C, with a very short bus (low capacitance). If you have a bus that is
more than perhaps 150mm long, or uses more than perhaps 2 devices,
or clocks over perhaps 200KHz, you will have to provide external pullups.
Technically you can enable fast mode on the pins with this, and turn the
current up to 20* the default level, but reviews from users are suggesting
this can be unreliable. So keep things slow if you want this to work.
100KHz max.
Now the good thing is you are on the standard I2C pins (you must still
set PPS to these pins). Only the pull ups on these pins support the
current needed for I2C.

[Torello]

Thanx Ttelmah,

I've found more a less the same and was also struggling to get to 400Khz. The use_i2C is quite fuzzy and not always working with given parameters as described in the manual

[Ttelmah]

Remember you have to do both pins. What you show only sets C3.....
The pins are individual.
You are selecting fast slew for 400K, but not setting a baud rate for the
#use I2C. Use a baud rate here, rather than writing to I2C baud.
However don't try to think of the internal pullups as being equivalent
to a resistor value. They are not, they are programmed current sources.
So draw more when the signal is low and decline as the signal goes up.
this is why people are finding them not reliable at higher baud rates....
Don't put the pins in the #use I2C. Using I2C1 is all you need. The compiler
(at least on current versions), automatically sets bot the 'in' and 'out', if
you just specify the signal as I show.
You do need to enable the weak pull up though.

[Torello]

Thanx for your extensive help!
I think am doing both pins: RC3I2C=RC4I2C=0b01100001;
~~
Maybe a complier bug (i'm stuck at 5.114)
Fast=xxxxxx in the #use_i2C(..) does not work. It compiles ok but the device is not working on the correct freq.
If I comment out the I2CBaud=2 and FME=1
#use i2c(stream=MCP9808_Stream, Master, I2C1, CLOCK_SOURCE=HFINTOSC, FAST) compiles ok but F-SCL=~800Khz...
#use i2c(stream=MCP9808_Stream, Master, I2C1, CLOCK_SOURCE=HFINTOSC, FAST=100000) same as above
In both cases the read back of the I2Cbaud=0 and FME=0. Thus the clock source (4Mhz) only gets div'd by FME=0 (=div 5). Which makes ~800khz
And the fast=xxxxxx hz parameter doesn't do anything. I know the exact frequency can't be made. But then I would expect the closest.
The parameter CLOCK_DIVISOR does work. But that only sets the bit FME.
I2C_Speed() function raises a compiler error.
So looks like using the I2C1Baud register is the only thing that works for now.
~~~
What you describe is a resistor pull-up behaviour, I would say. As the voltage across the resistor declines the current does also.
A good current source would keep the current the same independand of Vpin. The slope of the signals would then be more a straight line.
But they are not. Maybe caused by the slew-rate control?
~~~
Indeed I can leave the SDA SCL pins out. I2C1 is enough.
~~~
The PU bits in the RC3I2C/RC4I2C doesn't need the weak to be enabled, for what I see on the device.
I tested it by setting the bits explicitly low, readback and verify they are/stay low
I also watched the slope of the SCL impove at higher currents and still keeping the weak pull-ups disabled
_________________
Regards, Edwin. PCWHD v5.114

[Woody]

Hi,

I don't want to start a new thread as this one exactly describes the problem I run into.

My question: did anyone ever find out why #use I2C() does not work for setting the I2C clock speed?

I keep struggling with it and can only come to the same conclusion: best to do the I2C settings manually as #use I2C() only gets the clock source right.
Setting FAST=nnnnn and CLOCK_DIVISOR=x don't do what I expect them to (at least, in my compiler, 5.115 and with my chip, 18F14Q41)

Doing this 'by hand' works better but even then I find that the chip (18F14Q41) does not follow its datasheet.

When I set hfintosc (system clock is 16MHz internal) as the source in the #use i2c(), set I2C1BAUD=31 and FME to 1 I get a nice 125kHz i2c clock. As advertised in the datasheet, page 647. But if I then select FME=0 I expect a 100kHz clock (DS page 646) but get 83.5kHz.

I do not see any i2c clock related issues in the errata, so why this happens is unclear to me. I really hate it when I don't understand things like this

I'll renew my compiler today to see if that makes any difference to the use i2c() settings; there was a change that might influence this (although for S/W i2c):

-> 5.118 Updated S/W #use i2c() so that slower baud rates can be used from a fast clock speed on PCM/PCH.

[Woody]

Man, this is complicated.

I renewed my compiler. This seems to partly solve my problem. I now can use the FAST=nnnnnn parameter in #use i2c() and that gives me the desired baud rate.

But I find I have to take into account how the compiler decides to make the baud rate, with FME=0 or FME=1. Because this FME bit does also influence the SCL timing, which has implications for the peripheral that is used. For instance:

In my case, where I have a 16MHz FOSC, if I use

[Ttelmah]

I would raise this with CCS.
The problem here is that the #use I2C command was written originally for
the MSSP peripheral. With this there were none of the extra settings such
as FME. Now I'm assuming that you are using the internal I2C pull-ups?.
If so, FME has the extra function, of changing the current this peripheral
produces. Hence why the bus works when the FME bit gets set, but not
when it isn't. The SCL line is also held high for one clock cycle longer
when FME is 0.
I must admit I agree with you that selecting the clock_divisor as 4 ought
really to enable FME, and to 5, disable this. Seems that whoever wrote
this did not quite understand what this should be doing.

[temtronic]

OK, I'm a 'dinosaur' and prefer to use REAL resistors as pullups, even when I have to hand solder them onto the PCB cause, well as a 'dinosaur' I tend to um, ah, er, forget things.....
I don't trust 'computers'.....

[Woody]

[temtronic]

GREAT but please....be careful soldering !! Don't need to see 4 pages about 'it was working' fine until .....oopsy...soldered both pins together, grr....WHY did I listen to 'dinosaur'...he's supposed to be extinct !!!!

[Woody]

I hope you did not lose too much sleep over it

Being close to extinction myself I like a challenge so I heated up the welding torch and got to it. I'm not too miffed about the result:

https://box.postkamer.eu/cloud/index.php/s/wNzB2BynfSHkbye
(Sorry, no idea if and how this [img][/img] is supposed to work)

I switched off all internal pullups (set RxyI2C to 0x01, I2C-specific input thresholds).

Now risetimes for both SDA and SCL are ~130ns, and fall times are ~15ns. These used to be more 'in line' with each other. This might be a ifference between using the internal current source and external resistors.

Anyway, the resistors are in and functioning. Note to self: do not forget them. Then again, I'll probably mislay that note....

[temtronic]

re: I'll probably mislay that note....

yes !!! reminds me that I need to make a 'hidden hidden driveway sign ahead'.
sigh, I'm the hidden driveway on a corner and city trees are blocking the big warning sign yellow sign.

Naturally when you said you were adding resistors, I assumed REAL resistors ! You know ,multicouloured cylinders with wires on both ends NOT some silly, itty, bitty, blob with '2431' on it !!

sigh, I feel OLD now......

BTW does 2431 mean red-yellow-orange-brown ??

[Ttelmah]

The really small ones don't even have printing on them. Sizes like 0201
and 0402. Hand soldering these is something that should be used as the
basis for a silly TV program. 'Resistor wars', or something. Only way
really is hot air and solder paste with a mini finger holder and a microscope.
Classic "don't try this at home" exercise...

[Woody]