Status of DUE&RADDS support

dc42

dcnewman has two sets of RRF repositories. The older ones are RepRapFirmware-RADDS and CoreNG-RADDS and the readme files say that development of these was stopped on 21 April. The more recent ones are RepRapFirmware and CoreNG and they were last updated on 16 July. I don't know whether he has any plans to continue development of that fork.

My policy is that although I do not have time to maintain RRF for RADDS, I will gladly incorporate changes for RADDS into my fork on request provided they comply with the coding guidelines. That is what I have done for dcnewman. Also, when I make changes that I know will need additional work for RADDS, I flag the code with "#ifdef RADDS … #error ... #endif" blocks to draw attention to them.

What would be ideal is if either the manufacturer of RADDS or a committed RADDS user took over maintenance of the RADDS build.

mb0

i tried to contact a RADDS supplier but did not get a response so far. i've seen him posting about RRF on another forum so i thought he might be the right person to ask.

for now i can try to get it working on one of my spare boards. since i never used eclipse for anything serious i'm wondering how i can change the build configuration to set the RADDS preproc define. setting "SAM3x_CoreNG" as active build config doesn't seem to define RADDS. i could set it manually but that can't be the right way to do it .

another thing i'm facing. eclipse (mars) manages to build the firmware but the ui part is not able to resolve the headers and everything is underlined red this are the times i miss good'ol'makefiles and vi.

dc42

@mb0:

for now i can try to get it working on one of my spare boards. since i never used eclipse for anything serious i'm wondering how i can change the build configuration to set the RADDS preproc define. setting "SAM3x_CoreNG" as active build config doesn't seem to define RADDS. i could set it manually but that can't be the right way to do it .

I suggest you create new configurations of the CoreNG and RepRapFirmware projects (which is what I expect dcnewman did), copied from the SAM3X configurations bur with RADDS added to the preprocessor defines in the C and C++ tool settings. I will do this in my next release too.

@mb0:

another thing i'm facing. eclipse (mars) manages to build the firmware but the ui part is not able to resolve the headers and everything is underlined red this are the times i miss good'ol'makefiles and vi.

Right click on the project, then select Index->Rebuild from the menu. That should fix it.

I am aware of two changes that will need to be made to build a RADDS version from the latest source code:

1. The SD card interface has changed. In fact the Duet builds of RRF now support SD cards connected via SPI as well as via HSMCI, and that should make it much easier to support RADDS.

2. The logical pin numbering and inverted/not inverted status for M42 and M280 commands needs to be defined for RADDS.

mb0

thanks! re-indexing fixed the red.

i grepped through dcnewmans coreng and rrf forks. he provides pin configurations in the /RADDS subdirectory.
there are some changes to coreng as well. one is related to spi dual use or something. rest is more or less just jitter.
there is one #ifdef which defines DUET if SAM3X is defined. this logic needs to be revised as well, but this is of
course trivial.

actually he introduced SCons as build system so i guess he didn't bother using eclipse as build platform. this is
unnecessary.

best is i make a fork on github and send a pull request. i think i can start working on it as soon as my dyzend stuff
is working on the duetwifi (it is heating but auto-tune does not work so the pid is very slowly approaching temp,
but then rock solid). i hope i manage to finish the printer by the end of next week.

how is the esp connected to the duetwifi? would be awesome to bolt that onto radds as well.

dc42

@mb0:

how is the esp connected to the duetwifi? would be awesome to bolt that onto radds as well.

On the Duet WiFi the high speed SPI interface on the SAM4E mcu is dedicated to communicating with the ESP so that we can achieve high data transfer speeds.

mb0

i tried to make sense of eclipse project management and now i am wondering what the canonical way is to maintain the .cproject files. if noticed you added the RADDS build configuration to the firmware part. of course it is also needed in coreng. but unfortunately eclipse maintains paths to arduino gcc stuff in the same .cproject as it maintains build targets. unbelievable i now. thus i can't make clean git patches.

how can we colaborate without these stupid config conflicts? isn't there some kind of user profile overlay for cross toolchains?

dc42

It's possible to declare build variables in Eclipse, so it might be possible to declare the path to gcc in a build variable and refer to that build variable in the .cproject file. I suspect that this may just move the problem to the .project file, but the .project file is rarely changed anyway.

I'm sorry I forgot to check in the changes to CoreNG. I'll check them in shortly.

mb0

nice thanks! when you're at it dcnewman defines the following in SConstruct build file:

[[python]]
if platform == 'radds':
   env.Append( CCFLAGS = [
         '-D__RADDS__=1',
         '-DSD_MMC_SPI_MODE',
         '-DSPI_PIN=77',
         '-DSPI_CHAN=0',
         '-DSD_SS=4',
         '-DSD_DETECT_PIN=14',
         '-DSD_DETECT_VAL=0',
         '-DSD_DETECT_PIO_ID=ID_PIOD',
         '-DUSE_SAM3X_DMAC=1',
         '-DDMA_TIMEOUT_COMPUTE' ] )
   if use_usb_combo:
      env.Append( CCFLAGS = [
         '-DUSE_USB_COMBO',
	 '-DUSB_DEVICE_VENDOR_ID=0x2341',
	 '-DUSB_DEVICE_PRODUCT_ID=0x003e',
	 '-DUSB_DEVICE_PRODUCT_NAME=\\"Arduino Due\\"',
	 '-DUSB_DEVICE_MANUFACTURE_NAME=\\"Arduino LLC\\"' ] )

you might want to add these flags as well to the RADDS build configuration. then i can keep a local copy of the .cproject and ignore it (hoping so).

dc42

When Dan did the RADDS port, my fork didn't have code for accessing SD cards over SPI, but it does now. So it makes more sense to define those parameters (except for RADDS) in the Pins_RADDS file, just as they for the Duet and Duet WiFi. I already put most of them in Pins_RADDS,h except for the ones I didn't know the values for.

I think you should be able to use the CoreNG SAM3X build from my githib repo as-is for RADDS, except that you will need to change variant.cpp and variant.h to set up the pins appropriately.

dc42

PS - I've updated my copy of CoreNG with Dan's copy of variant.cpp and variant.h, ma de a few other changes, and now I can actually build a RADDS version of CoreNG. I'll commit it to GitHub tomorrow when I have tested that I haven't broken anything in the Duet builds.

mb0

thanks! i'll have a look at it when i'm back from very short vacation!

you are right after writing my post i realized that these defines should go into a header instead.

there are only a few other mentions of RADDS in coreng. One is in AnalogIn.cpp line 94:

[[c]]
#if SAM3XA
#ifdef __RADDS__
			if (ADC_TEMPERATURE_SENSOR != GetAdcChannel(channel))
#endif
				adc_enable_channel(ADC, GetAdcChannel(channel));
			if (GetAdcChannel(channel) == ADC_TEMPERATURE_SENSOR)
			{
				adc_enable_ts(ADC);
			}
#endif

1st i disklike nested #ifdef whatever, this makes everything unreadable. 2nd yoda style (ADC_TEMPERATURE_SENSOR != GetAdcChannel(channel) is a no no.
however i'm asking myself if this is actually needed because in any case it would call adc_enable_ts if needed. in the RADDS case not calling adc_enable_channel
before then. so either there is a bug when doing this when using the radds board or it is not neccessary and can be forgotten.

if your time permits you might want to look into SharedSpi.cpp as well. in dcnewmans fork are a 2 more #ifdef's for the RADDS case in lines 186 and 386

[[c]]
#if defined(USE_SAM3X_DMAC) && !defined(__RADDS__)

you know better than me if these are necessary or not facing the current implementation. Afaics, these were the only changes to the actual source which depend
on anything radds related. if so coreng would be radds safe after your changes.

i looked at the radds schematics and think that using a small adapter pcb the esp could also be connected to spi and the sdcard rewired to the hsmci port. thus
enabling high speed uploads there as well (preventing however the use of an lcd but not serial panel or whatever).

dc42

I've checked everything in. The RADDS version builds but I haven't tested it.

The change to AnalogIn shouldn't be needed, because the RADDS version of RRF shouldn't ask to use the temperature sensor channel.

The changes to SharedSpi are because he got it working under DMA, but my build doesn't use DMA for the shared SPI. I think I left the sspi_start_transmit and sspi_start_receive functions there so that he could use them, but he seems to be calling spi_start_transmit and spi_start_receive instead, and I don't know where he defines those. You may find that it works OK without using DMA anyway, although using DMA would be preferable because on the RADDS all SD card access goes over SPI.

mb0

managed to build the RADDS flavour as well. There were some minor quirks (arm-*-objcopy was unable to locate the .elf because the path to it is wrong objcopy -binary /RepRapFirmware.elf /RepRapFirmware.bin) i checked the eclipse settings but they seamed correct. Anyway, did it by hand. Ahh and i switched from external make
to internal build tool in eclipse. Maybe you can change this setting as well in the project file to reduce one more error one can have. I copied the gcc from my arduino install
to the directory you have in the eclipse config, so in the future i don't have to touch this settings (at least under windows).

to use bossa (under Tools/Windows) you need to set the corresponding COM port to 1200 baud. This is done using the command
[c]mode comN:1200[/c] where N is the port number of the due. under linux you do [c]stty -F /dev/ttyACM0 1200[/c].

The target binary of RRF for RADDS should be renamed to RepRapFirmware-RADDS.bin for clarity.

After uploading the firmware to the due i could connect to it using pronterface and issue some commands. It seems that only the external SDCard
is working, maybe some configuration issue. I couldn't test the full deal because i ran out of time. Will do tomorrow. Just wanted to drop a quick note…

dc42

@mb0:

…. It seems that only the external SDCard is working, maybe some configuration issue....

Does RADDS support 2 SD card sockets then? I only allowed for one in the Pins.h file.

StephenRC

The RADDS has a microsd card slot and a SD extern connector that's used by the lcd sold for it.

dc42

Ok so you need to change the SD cards section of thea RADDS pins.h file to define both cards. If the external card is working, then I suspect that the CS pin in the existing definition is for the external card however the CD pin may be for the internal card. You can use the M122 command to see the state of the CD pin, from that you can work out which one it is.

mb0

thanks, there are two chip selects (CS0 and CS1), the CD (card detect) is shared between both.

http://www.dr-henschke.de/LCD_Enc_SD_12.pdf

i think the default config for RADDS should use the onboard SDCARD slot because one can not expect the lcd (which is not supported) to be connected.

dc42

Do you know what the pin assignments are for CS0 and CS1?

bruce356

Hi, I am not a technical person, wondering if I could use the firmware that you are working on here for my dual x carriage Cartesian (Mendel) printer with Due/RADDS. I would probably require a bin file as I know nothing about compiling this type of firmware.
Thanks and regards - bruce

mb0

CS0 -> Arduino Pin 10 -> SAM3X PA28/PC29 -> Physical 102
CS1 -> Arduino Pin 4 -> PA29/PC26 -> Physical 137

btw. the printer is running now for >4h without any issues (related to the firmware).
the current setup is RADDS&RAPS128@1/128 0.9 stepper for x/y. z@1/32, extruder@1/16.
24V. It can't go full speed at this microstepping rate. i noticed some "pauses" when printing
large rounded corners. it seems that braking after travel movement is quite aggressive, i get
strange sounds as if the hotend would have bumped into the print.

edit: after setting the raps to 1/32 for all steppers the bumping is gone. is suspect planner
errors because out-of-time situations.