F3 and X-Vert support #2113

kevindehecker · 2017-09-28T08:31:16Z

Adds support for:

STM32F373x
The Horizon Hobbies Eflite X-Vert VTOL (based on the STM32F373xC)
STM32F303x (preliminary)
The F3 Discovery board (based on the STM32F303)

All are based on Chibios. In order to fit the RAM memory footprint on the 32K RAM of the STM32F37, the thread working area sizes have been reduced. It all seems to work without fault, so I don't think it is necessary to increase the size for larger chips such as the F7.
Another thing is that the PPRZ_TRIG_CONST needed to be defined as CONST. This placed the sin integer lookup table on flash. The other defines such as PPRZ_TRIG_INT_COMPR_FLASH do not seem to have any effect. Of course, the new define PPRZ_TRIG_INT_USE_FLOAT also works.

The x-vert has been flight tested. It can hover manually. Further steps (autonomous, forward flight) are planned. A wiki page about the x-vert is in the works.

Known problems:
The X-vert ESCs provide motor feedback, but this is not yet read out
The X-vert may have a current sensor, but it is not yet read out
The X-vert board has an CYRF chip for radio control, but this is not supported yet in this PR.
The Chibios stm32F37 implementation has a problem with the hard FPU. When enabled, totally weird results happen.

podhrmic · 2017-09-29T16:53:31Z

I ll look over it over the weekend.

podhrmic · 2017-10-01T21:43:58Z

conf/boards/xvert_1.0.makefile

+## FPU on F3
+
+##TODO: there is some bug when using the hard FPU unit!!
+USE_FPU=no


can you file a bug report at ChibiOS forum about it? Giovanni is typically fairly responsive.

It's probably something with wrong settings for the minimal float number. Was planning to look into it a bit further myself later.

podhrmic · 2017-10-01T21:45:33Z

sw/airborne/arch/chibios/mcu_periph/adc_arch.c

 #endif

  adcgrpcfg.circular = TRUE;
  adcgrpcfg.num_channels = ADC_NUM_CHANNELS;
  adcgrpcfg.end_cb = adc1callback;
  adcgrpcfg.error_cb = adcerrorcallback;
+#if defined(__STM32F373xC_H)


it would be nice to have a better way of specifying the driver than adding ifdefs, perhaps separate files for each chibios arch - but that is indeed out of scope of this PR

podhrmic · 2017-10-01T21:46:06Z

sw/airborne/arch/chibios/mcu_periph/i2c_arch.c

-        (uint8_t*)i->dma_buf, (size_t)(t->len_r),
-        tmo);
-    memcpy((void*)t->buf, i->dma_buf, (size_t)(t->len_r));
+               (I2CDriver *)p->reg_addr,


stylecheck pls?

Fix_code_style is actually what caused this...

I wrote this part of the code. The code_style script should be correct, it is probably my VIM settings for aligning multi-line parameters that are a bit different.

podhrmic · 2017-10-01T21:48:06Z

sw/airborne/arch/chibios/mcu_periph/i2c_arch.c

@@ -198,7 +198,7 @@ static struct i2c_init i2c1_init_s = {
 struct i2c_errors i2c1_errors;
 // Thread
 static __attribute__((noreturn)) void thd_i2c1(void *arg);
-static THD_WORKING_AREA(wa_thd_i2c1, 1024);
+static THD_WORKING_AREA(wa_thd_i2c1, 128);


I would be worried that this might overflow for larger i2c buffers - could you add ifndef guard and specify this number in the board file to override the default?

I don't think that's how it works. The data goes in different buffers. E.g. the ublox I2C gps (used in this PR) actually is a rather large buffer user compared to most other sensors, it does not seem to be a problem. Also, the chibios examples themselves use this number. I think they were oversized in pprz. Of course could add the ifdef overrides if still required, but think these numbers here should be the default.

I'm not sure though what happens when the buffer is to small, I may try that when I get the chance,
It would also be nice if someone with e.g. a Chimera can test this.

(same for all the other THD_WORKING_AREA's)

OK, sounds reasonable. I think ChibiStudio plugin for Eclipse can tell you the used memory for each thread.

i2c buffers are global variables so not counted for the stack size. Only automatic variables (more or less the local variables) created by the thread will use the stack memory, a few pointers in the case of this driver. So I guess 128 is fine.

podhrmic · 2017-10-01T21:49:10Z

sw/airborne/arch/chibios/mcu_periph/spi_arch.c

@@ -347,15 +346,15 @@ static __attribute__((noreturn)) void thd_spi1(void *arg)
  }
 }

-static THD_WORKING_AREA(wa_thd_spi1, 1024);
+static THD_WORKING_AREA(wa_thd_spi1, 256);


again - would be better to have this board-dependent with larger defaults.

podhrmic · 2017-10-01T21:54:53Z

sw/airborne/subsystems/ahrs/ahrs_float_cmpl.c

-  } else {
-    ahrs_fc.weight = 1.0;
-  }
+    static struct FloatVect3 filtered_gravity_measurement = {0., 0., 0.};


isn't pprz style with 2 spaces? Pls run the stylecheck script on this.

Actually, all these changes were done by fix_code_style

If I run it again, it gives the same results. It does give some weird error warning though:
~/paparazzi$./fix_code_style.sh sw/airborne/subsystems/ahrs/ahrs_float_cmpl.c
Using astyle version 2.05.1
(standard_in) 1: syntax error
./fix_code_style.sh: line 18: [: -eq: unary operator expected
Unchanged sw/airborne/subsystems/ahrs/ahrs_float_cmpl.c

Don't know why the script does the warning, maybe @flixr knows more?

However, the file broke the style check - compare lines 206 const struct FloatVect3 vel_tangential_body = before and after. Clearly the correct style is before applying the style check.

Pls revert and check your changes manually.

Also there is a typo at line 212 - the end bracket should be } instead of )

podhrmic · 2017-10-01T21:56:09Z

sw/airborne/subsystems/ahrs/ahrs_float_cmpl.c


-}
+    const float cphi   = cosf(ltp_to_imu_euler.phi);


what is the reason for changes here (they aren't mentioned in the PR description)?

Weird, all this is done by fix_code_style. I only did the two minor changes in commit 8cdc6d7

Ok makes sense

podhrmic · 2017-10-01T21:57:04Z

sw/tools/qtc.creator.user_template

@@ -1,10 +1,10 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <!DOCTYPE QtCreatorProject>
-<!-- Written by QtCreator 4.1.0, 2017-01-23T23:23:44. -->
+<!-- Written by QtCreator 4.4.0, 2017-09-08T21:31:25. -->


this file should probably be a part of .gitignore

Actually, it should not. The template is used to generate a valid qtc project.user. Unfortunately it currently means that the template has to be updated when qtc changes. Looking for a better solution, but nothing as of yet.
(other than that, I can recommend qt over eclipse. It only shows you the files that are actually used)

podhrmic · 2017-10-01T21:57:31Z

sw/tools/find_vpaths.py

@@ -3,55 +3,71 @@
 #It finds the correct VPATH to each source file from make, and then feeds all info to gcc -MM.
 import sys
 import os
-from os import path, getenv
+from os import path, getenv, environ


how are these changes related to PR?

In order to get a good qt project with a chibios based board, some changes were needed. Since they are in a clear seperate commit I did not bother creating a separate PR

podhrmic · 2017-10-01T21:57:45Z

sw/tools/qt_project.py

-while line:
-    line=ap_srcs_list.readline()
+for line in ap_srcs_list:
+    #line=ap_srcs_list.readline()


remove the comment?

podhrmic · 2017-10-02T21:48:07Z

@kevindehecker the only remaining thing is the code style for ahrs.c file (see my comment above), the rest looks good (at least for the time being).

kevindehecker · 2017-10-05T08:09:58Z

Done!

podhrmic · 2017-10-05T17:23:22Z

LGTM. @gautierhattenberger or @flixr do you want to add anything or good to merge?

flixr · 2017-10-05T18:53:30Z

sw/airborne/firmwares/rotorcraft/stabilization/stabilization_indi_simple.c

@@ -340,9 +340,17 @@ static inline void stabilization_indi_calc_cmd(int32_t indi_commands[], struct I
  indi.u_in.r = indi.u[2].o[0] + indi.du.r;

  //bound the total control input
+#ifdef INDI_FULL_AUTHORITY


This new option should be documented in the module xml....

Also please use #if FLAG instead of #ifdef FLAG according to http://docs.paparazziuav.org/latest/stylec.html#stylecpp

gautierhattenberger · 2017-10-05T21:49:02Z

conf/airframes/TUDELFT/tudelft_xvert.xml

+    </module>
+  </firmware>
+
+	<modules main_freq="512">


would be nice to move this modules to the firmware section

gautierhattenberger · 2017-10-05T22:03:27Z

sw/airborne/arch/chibios/mcu_periph/i2c_arch.c

@@ -198,7 +198,7 @@ static struct i2c_init i2c1_init_s = {
 struct i2c_errors i2c1_errors;
 // Thread
 static __attribute__((noreturn)) void thd_i2c1(void *arg);
-static THD_WORKING_AREA(wa_thd_i2c1, 1024);
+static THD_WORKING_AREA(wa_thd_i2c1, 128);


i2c buffers are global variables so not counted for the stack size. Only automatic variables (more or less the local variables) created by the thread will use the stack memory, a few pointers in the case of this driver. So I guess 128 is fine.

gautierhattenberger · 2017-10-05T22:08:40Z

sw/airborne/firmwares/rotorcraft/main_chibios.c

@@ -42,7 +42,7 @@
 * Also include FBW on single MCU
 */
 static void thd_ap(void *arg);
-static THD_WORKING_AREA(wa_thd_ap, 8192);
+THD_WORKING_AREA(wa_thd_ap, 1024);


Here, I don't agree to reduce the stack size. We are running some estimation filters that requires some memory. So I suggest to keep the 8K by default and make it configurable.

gautierhattenberger · 2017-10-05T22:13:49Z

sw/airborne/arch/chibios/mcu_periph/i2c_arch.c

-        (uint8_t*)i->dma_buf, (size_t)(t->len_r),
-        tmo);
-    memcpy((void*)t->buf, i->dma_buf, (size_t)(t->len_r));
+               (I2CDriver *)p->reg_addr,


I wrote this part of the code. The code_style script should be correct, it is probably my VIM settings for aligning multi-line parameters that are a bit different.

Make mpu9250 compatible with mpu6500, by providing option to disable mag

8cdc6d7

podhrmic requested changes Oct 1, 2017

View reviewed changes

kevindehecker added 9 commits October 2, 2017 08:52

Upgrade QTC version, add more QTCreator compatibility with chibios

206d266

F3 support

8f086f5

Add priliminary support stm32f3 discovery board using chibios

cf71adb

Reduce RAM footprint, so that pprz + chibi fit on a 32k f37

c62c643

Add support for the stm32f37 xvert board by means of chibios

eb1dd4f

Motor driver for the vertx

40ee3f5

Give INDI simple full control authority

01a59be

Ublox over I2C needs extra time

5e87ed9

Cleanup

4720267

kevindehecker force-pushed the f3xvert branch from b65d9f4 to c969d49 Compare October 2, 2017 06:52

kevindehecker added 3 commits October 5, 2017 10:03

Fix code style

5044255

Fix ADC F3 support

4e35f30

Added battery levels

6c4599b

kevindehecker force-pushed the f3xvert branch from c969d49 to 6c4599b Compare October 5, 2017 08:06

podhrmic approved these changes Oct 5, 2017

View reviewed changes

flixr reviewed Oct 5, 2017

View reviewed changes

gautierhattenberger requested changes Oct 5, 2017

View reviewed changes

kevindehecker added 3 commits October 6, 2017 10:49

Refactoring indi full control authority define

9caca65

Clean up xvert airframe file

8174501

Create define out of required THD_WORKING_AREA space

46a36d4

gautierhattenberger approved these changes Oct 6, 2017

View reviewed changes

gautierhattenberger merged commit 2a6ad55 into paparazzi:master Oct 6, 2017

kirkscheper mentioned this pull request Oct 15, 2017

[sim] updates to gazebo so simulate pressure and acceleration #2120

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

F3 and X-Vert support #2113

F3 and X-Vert support #2113

kevindehecker commented Sep 28, 2017 •

edited

Loading

podhrmic commented Sep 29, 2017

podhrmic Oct 1, 2017

kevindehecker Oct 2, 2017

podhrmic Oct 1, 2017

podhrmic Oct 1, 2017

kevindehecker Oct 2, 2017

gautierhattenberger Oct 5, 2017

podhrmic Oct 1, 2017

kevindehecker Oct 2, 2017 •

edited

Loading

podhrmic Oct 2, 2017

gautierhattenberger Oct 5, 2017

podhrmic Oct 1, 2017

podhrmic Oct 1, 2017

kevindehecker Oct 2, 2017

kevindehecker Oct 2, 2017 •

edited

Loading

podhrmic Oct 2, 2017

podhrmic Oct 1, 2017

kevindehecker Oct 2, 2017

podhrmic Oct 2, 2017

podhrmic Oct 1, 2017

kevindehecker Oct 2, 2017 •

edited

Loading

podhrmic Oct 1, 2017

kevindehecker Oct 2, 2017

podhrmic Oct 1, 2017

podhrmic commented Oct 2, 2017

kevindehecker commented Oct 5, 2017

podhrmic commented Oct 5, 2017

flixr Oct 5, 2017

flixr Oct 5, 2017

gautierhattenberger Oct 5, 2017

gautierhattenberger Oct 5, 2017

gautierhattenberger Oct 5, 2017

gautierhattenberger Oct 5, 2017

F3 and X-Vert support #2113

F3 and X-Vert support #2113

Conversation

kevindehecker commented Sep 28, 2017 • edited Loading

podhrmic commented Sep 29, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kevindehecker Oct 2, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kevindehecker Oct 2, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kevindehecker Oct 2, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

podhrmic commented Oct 2, 2017

kevindehecker commented Oct 5, 2017

podhrmic commented Oct 5, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kevindehecker commented Sep 28, 2017 •

edited

Loading

kevindehecker Oct 2, 2017 •

edited

Loading

kevindehecker Oct 2, 2017 •

edited

Loading

kevindehecker Oct 2, 2017 •

edited

Loading