MicroMod Beta Testing

Sorry been screwing around today...

screenshot.jpg

Quick and dirty so far, need to add labels...
 
Sorry been screwing around today...

Quick and dirty so far, need to add labels...
nice board Kurt.

After testing been messing around myself. Think going to play with tensorflow until I get the parts.

Tim. Had a thought that you can test. Think spec says at 4bit camera power is about half, on phone now so can’t verify. If you run you board with 4bit you should see a power drop. If test runs for then another clue to the puzzle.

Hopefully will get parts next week
 
PJRC.BB sat unplugged ~1.5 hours
Repowered and it ran this long - i.e. 2 minutes:
LP#=34 Pisr#=111490 lastPR=107583967 ms=119645 ipCyc%=59.55 S5#=0 48C

Coming back I see it running this long:
LP#=36 Pisr#=114127 lastPR=107589263 ms=416176 ipCyc%=59.57 S5#=0 51C


But in the middle - it did this unattended - Each line a restart with the rest of the spew deleted - a self provoked restart where nothing runs but SSD1306 Box cycles and the LEDs:
Code:
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
SENSOR DETECTED :-) MODEL HM01B0
 
nice board Kurt.

After testing been messing around myself. Think going to play with tensorflow until I get the parts.

Tim. Had a thought that you can test. Think spec says at 4bit camera power is about half, on phone now so can’t verify. If you run you board with 4bit you should see a power drop. If test runs for then another clue to the puzzle.

Hopefully will get parts next week

The board does look interesting - just need a connector.

Mike - looking at the spec the worst 8 bit camera current is 1746 micro amps - not sure it would even be visible on my meter among the 4 mA drift that shows - not sure of the mode but it could also be under 300 uA in 8 bit.

This is puzzling ... the VReg never gets hot that I've felt. And it bumped up to cover the extra mA's fine - though it did repro sooner - and if a simple thing no apparent reason sitting 1.5 hours to 'cool' returned to quick failure. Problematic perhaps the camera running over voltage - but other than the init it isn't doing anything in the 30 spontaneous restarts inp#853 - though maybe the camera is not healthy and doing the i2c init is enough to wake it up and attack the voltage - but odd how the device meant to run under 2 mA could have the ability to take down the 1062 ...

So given that I'll jump to 4 bits and also not init the camera if it shows failure - and it can be reprogrammed ... it has been off 2+ hours now
 
Mike - doing 4 bit mode after long power down - made code change.

DMM showing range 208 mA to 211 mA - so maybe it did drop ... though now 208-212 after 11 mins

Code:
Using SDCARD - Initializing
initialization done.
HM01B0 Camera Test
[B]HM01B0_FLEXIO_CUSTOM_LIKE_4_BIT[/B]
------------------
START Fill Cache :: 3454
END Fill Cache :: 	Last is 82813	 @ 3501

T:\tCode\libraries\HM01B0\examples\HM01B0_example2\HM01B0_example2.ino May  1 2021 22:03:48
Send the 'f' character to read a frame using FlexIO (changes hardware setup!)
// HELP HERE

[B]SENSOR DETECTED :-) MODEL HM01B0[/B]
OSC_CLK_DIV: 0x29
ImageSize (w,h): 324, 244
Reading frame

Just to see what happens with AUTO restart - I moved the SENSOR DETECT/setup to after if (Serial.Available())

So if/when it restarts - it won't activate the camera until SerMon chars arrive - though if like Audio board it stays activated until power off ...

Opps - power dropped shutting off DMM after
LP#=34 Pisr#=112828 lastPR=107586641 ms=846489 ipCyc%=59.46 S5#=0 52C

Running 2.25 hours since posting [after board was off some 2 hours] - all good so far: No Restarts, VID Good, Boxes cycling, LEDs lit and sermon dumping 't' seconds data
LP#=35 Pisr#=112816 lastPR=107586641 ms=8707896 ipCyc%=59.32 S5#=0 53C
> Have to assume current value as it should be - some meter Power changes On/Off cut the line

And if it does restart this sketch will not power the camera until the first user input calling camStart():
Code:
[ATTACH]24686._xfImport[/ATTACH]
- if it restarts after that only LEDs and SSD1306 were active.

Still going properly - no restart ... ~4 hours
LP#=35 Pisr#=112811 lastPR=107586629 ms=14156981 ipCyc%=59.63 S5#=0 53C

Quick note : HM01B0_FLEXIO_CUSTOM_LIKE_4_BIT
LP#=35 Pisr#=112826 lastPR=107586641 ms=38917407 ipCyc%=59.69 S5#=0 51C
LP#=35 Pisr#=112816 lastPR=107586641 ms=38918407 ipCyc%=59.61 S5#=0 51C

Still running this time with only altered code? ? ? - a record 10.8 hours and counting ???

All systems powered and looking right
 
Last edited:
Morning all,

@defragster - I have not run any tests yet today, but it looks like you have been busy.

When mine was resetting I put my finger on Teensy and VR... and did not feel anything overly warm. That is why I started to think maybe it is the camera setup...


So I started playing with a quick and dirty adapter for converting the HIMAX camera ribbon to something I could probably plug into some test boards setup for the OV7670 camera
which I would prefer to use so far anyway...

As I mentioned yesterday started to do a quick and dirty layout, which is maybe close enough for now to maybe order ones and see what they do...

screenshot.jpgscreenshot2.jpg

There are some simple questions, like the first 2x8 pins are the same... The RST/Shutdn don't exist on the ribbon, but the two VRs have an enable pin, so far I have
wired both up to RST pin with PU resistor... Could have gone to Shutdown instead... Added extra 2 pins for TRG/INT In case ever wish to experiment... My current boards don't
have matching pins.

But assuming I do this, would be able to test camera on CSI pins on T4.1...
 
Quick note for post #855: HM01B0_FLEXIO_CUSTOM_LIKE_4_BIT

LP#=35 Pisr#=112826 lastPR=107586641 ms=38917407 ipCyc%=59.69 S5#=0 51C
LP#=35 Pisr#=112816 lastPR=107586641 ms=38918407 ipCyc%=59.61 S5#=0 51C

Still running this time with only altered code? ? ? - a record 10.8 hours and counting ???

All systems powered and looking right
 
Quick note for post #855: HM01B0_FLEXIO_CUSTOM_LIKE_4_BIT

LP#=35 Pisr#=112826 lastPR=107586641 ms=38917407 ipCyc%=59.69 S5#=0 51C
LP#=35 Pisr#=112816 lastPR=107586641 ms=38918407 ipCyc%=59.61 S5#=0 51C

Still running this time with only altered code? ? ? - a record 10.8 hours and counting ???

All systems powered and looking right

Well that's good news. Now wondering if your camera is seated good in the 8-bit port of the breakout that might be causing you some of the issues. Remember when I was having problems with the ML carrier, it would work in 4bit mode the giving me headaches in 8bit - problem was that the camera just wasn't seated correctly. That's one of reasons I like that board Kurt is designing or premade ones :)

Still might be power related though if 4bit actually draw less current.
 
Morning all,

@defragster - I have not run any tests yet today, but it looks like you have been busy.

When mine was resetting I put my finger on Teensy and VR... and did not feel anything overly warm. That is why I started to think maybe it is the camera setup...


So I started playing with a quick and dirty adapter for converting the HIMAX camera ribbon to something I could probably plug into some test boards setup for the OV7670 camera
which I would prefer to use so far anyway...

As I mentioned yesterday started to do a quick and dirty layout, which is maybe close enough for now to maybe order ones and see what they do...

View attachment 24691View attachment 24692

There are some simple questions, like the first 2x8 pins are the same... The RST/Shutdn don't exist on the ribbon, but the two VRs have an enable pin, so far I have
wired both up to RST pin with PU resistor... Could have gone to Shutdown instead... Added extra 2 pins for TRG/INT In case ever wish to experiment... My current boards don't
have matching pins.

But assuming I do this, would be able to test camera on CSI pins on T4.1...

Yeah - was thinking that with that kind of adapter you would be able to do a bit more with on the T4.x
 
Well that's good news. Now wondering if your camera is seated good in the 8-bit port of the breakout that might be causing you some of the issues. Remember when I was having problems with the ML carrier, it would work in 4bit mode the giving me headaches in 8bit - problem was that the camera just wasn't seated correctly. That's one of reasons I like that board Kurt is designing or premade ones :)

Still might be power related though if 4bit actually draw less current.

Last I reseated the Camera into 8bit connector it went in deep and straight and latched down well - and still sitting square. It would work well for those 3 or 9 hours periods in 8 bit. Though between high V and running shifted in socket once ??? Maybe 8 bit interface is compromised?

It is still running 14.1 hours!- except now the shading is the TOPO outlines with added BLUE - much different than the monochrome Pin Art. But active Video and SSD1306 and 5 LEDs. TEMP is lower with 4bit Flexio:
Code:
LP#=35	Pisr#=112816  lastPR=107586641  ms=50789616  ipCyc%=59.48  S5#=0  52C
LP#=35	Pisr#=112750  lastPR=107586503  ms=50790643  ipCyc%=59.65  S5#=0  51C

Did TyComm Reset and restart is fine - not any sign of trouble - stopping or restarting.

The other net effect of the edit made it the Sketch starts and the camera is ignored until first user USB input? Let me rebuild to 8 bit with that same startup ... and let it run to see if it make 4 hours or more ... >> _hmConfig == 2 : 12:32pm w/"fFFtV" - proper start @208-212 mA {same as p#855 w/4bit 212 mA is rare}
Code:
LP#=34	Pisr#=111448  lastPR=107583841  ms=204745  ipCyc%=59.34  S5#=0  51C
LP#=34	Pisr#=111419  lastPR=107583841  ms=205731  ipCyc%=59.42  S5#=0  52C
... Dang ... DMM Off restarted sketch ... Restart fine @12:37:
Code:
HM01B0_FLEXIO_CUSTOM_LIKE_8_BIT
...
LP#=34	Pisr#=111436  lastPR=107583841  ms=123147  ipCyc%=59.40  S5#=0  51C
LP#=34	Pisr#=111514  lastPR=107584021  ms=124159  ipCyc%=59.23  S5#=0  52C

Running fine 3 hours into 8 bit ... opps ... unit shifted on desk and restarted, worn USB cable likely shifted : All systems go - good images ... before and after
Code:
LP#=33	Pisr#=111506  lastPR=107584021  ms=10443269  ipCyc%=59.23  S5#=0  53C
LP#=35	Pisr#=111433  lastPR=107583841  ms=10444281  ipCyc%=59.50  S5#=0  52C

Restart fine so far: {seems about 15 hours powered and running since 2 hours off? 3 hours back on 8 bit:
LP#=34 Pisr#=111420 lastPR=107583841 ms=441929 ipCyc%=59.43 S5#=0 52C
LP#=35 Pisr#=111506 lastPR=107584021 ms=442942 ipCyc%=59.79 S5#=0 53C
 
Last edited:
@defragster - sounds like you are having fun testing.

Anyway, kind of going to be done testing the camera for until new parts arrive. So in the meantime spent the day getting Tensorflow Lite working with the TMM.

TENSORFLOW LITE:
Started with the Arduino_tensorflow_lite that can be downloaded from using the library manager. Boils down to having to do 2 things:
1. Copy and Paste a copy of the cmsis_gcc.h file into the library - @Paul doesn't look like it made it into the core.
2. Edit one file to add include <string.h>

Ran and compiled the hello_world example making 1 minor change as noted on the other thread regarding tensorflow - adding a delay(1) to the output_handler function and it started working. Want to play with some of the other examples and then will post some more detailed info on making it work.
 
Yeah - was thinking that with that kind of adapter you would be able to do a bit more with on the T4.x

I will probably order a set tomorrow... Was playing around with another layout of it... Not cleaned up...

screenshot2.jpg

Lots of wasted space on top, but idea is with that camera could put a little double faced tape behind it and stick it to the board... Again not sure it is worth the extra space...

Playing around than with MMOD version of display board, removed ribbon...
screenshot.jpg
 
@defragster - sounds like you are having fun testing.

...

I should ask about your hearing if this sounds fun :) Just hoping to see what is what - oddly now with the only real change being delayed Camera .init() steps - it seems to be not reproducing prior behavior for a long time running.

All is well at this point from power up last night.
Code:
LP#=34	Pisr#=111417  lastPR=107583841  ms=6176002  ipCyc%=59.71  S5#=0  52C
LP#=35	Pisr#=111439  lastPR=107583841  ms=6177015  ipCyc%=59.79  S5#=0  53C

Unpowering for a minute to make sure the Camera resets - and starts up just as current init dictates.
So 5:16pm here and came up proper. as "HM01B0_FLEXIO_CUSTOM_LIKE_8_BIT". Will go back to ignoring and see if it runs until ...
Code:
LP#=34	Pisr#=111518  lastPR=107584021  ms=1187233  ipCyc%=59.62  S5#=0  53C
LP#=34	Pisr#=111421  lastPR=107583841  ms=1188245  ipCyc%=59.20  S5#=0  53C
> temps showing steady 53C now not 52/53 abd current 210-214mA just now?

Not sure if you opened prior posted code - but broke setup() call to move CAM init to first USB input, i.e. no ref to "hm01b0." until camStart() is called ... has 8 hrs to run until Zzzz's ::
Code:
 setup(){
// ...
 while (!Serial) ;
  Serial.println("HM01B0 Camera Test");
  Serial.println( hmConfig[_hmConfig] );
  Serial.println("------------------");

  //hm01b0.init();
  delay(500);

  tft.fillScreen(TFT_BLACK);
  setupPR();
  showCommandList();

}

void camStart () {
  static int runOnce = 1;
  if ( runOnce != 1 ) return;
  runOnce = 0;

  uint16_t ModelID;
  ModelID = hm01b0.get_modelid();

// ...
}

gone TOPO again - room dark to light on? - restarted: all good 8:32p
LP#=35 Pisr#=111409 lastPR=107583793 ms=10409432 ipCyc%=59.25 S5#=0 53C
LP#=34 Pisr#=111376 lastPR=107583743 ms=10410418 ipCyc%=59.56 S5#=0 52C

odd vid again - but otherwise fine - no restart or stopping:: Reset and restart fine
LP#=35 Pisr#=111493 lastPR=107583967 ms=10683440 ipCyc%=59.32 S5#=0 52C
LP#=33 Pisr#=111416 lastPR=107583841 ms=10684398 ipCyc%=59.59 S5#=0 52C

Odd again : 1:11a - good otherwise and fine on restart
LP#=35 Pisr#=111510 lastPR=107584021 ms=6022010 ipCyc%=58.84 S5#=0 53C

Odd again - more Pin Art w/TOPO - not restarting - will see in the morning ... 2:42a
LP#=34 Pisr#=111392 lastPR=107583793 ms=5425409 ipCyc%=59.67 S5#=0 53C
 
Last edited:
Let me just send you all some appreciation.
Me and my friend are so excited over this that we check the forum every day. We have designed our own PCB's that are ready to order once this launches.

I just feel like saying THANKS for putting your time into this. Truly epic work and it'll help the community so much.
Hands down guys, you all rock! :cool:
 
TENSORFLOWLITE

Ok all just pushed my updated copy of the Arduino_tensorflowlite library and 3 working examples:
1. micro_speech - need the audio shield and microphone: https://forum.pjrc.com/threads/57441-Tensorflow-on-Teensy?highlight=tensorflow , tested on the PJRC.BB.
2. hello_world - fading LED
3. magic_wand - works sort up. Probably accel settings need changing - testing on the ML Carrier

There are 2 other examples I haven't worked on though

Created a new repository: https://github.com/mjs513/Arduino_TensorFlowLite_t4.
 
@mjs513 - Sounds like you have been busy! At some point may take a look, but math makes my head hurt ;)

Side note: I did order a set of 3 MMOD boards from OSHPark as well as 3 of the Himax to OV7670 adapters...
Will be interesting to see if I can solder any of them :D
 
@mjs513 - Sounds like you have been busy! At some point may take a look, but math makes my head hurt ;)

Side note: I did order a set of 3 MMOD boards from OSHPark as well as 3 of the Himax to OV7670 adapters...
Will be interesting to see if I can solder any of them :D

Makes may head hurts as well - thats why I just converted the 3 examples - have no clue how to set it up from scratch but am sure some folks are interested :)

Cool that you ordered the adapters - should be interesting to see how it works out.
 
Well p#864 long run never stopped and never restarted! Video going Damascus/TOPO - left it that way last night the same this morning - replugged a few secs and it quickly went back to that:
LP#=34 Pisr#=111423 lastPR=107583841 ms=29955812 ipCyc%=59.73 S5#=0 52C
LP#=35 Pisr#=111521 lastPR=107584051 ms=29956824 ipCyc%=59.62 S5#=0 51C
LP#=35 Pisr#=111417 lastPR=107583841 ms=29957837 ipCyc%=59.10 S5#=0 52C
LP#=33 Pisr#=111447 lastPR=107583841 ms=29958822 ipCyc%=59.57 S5#=0 52C
LP#=35 Pisr#=111514 lastPR=107584021 ms=29959834 ipCyc%=59.46 S5#=0 51C

Seems the temp may be 51to 53 based on room temp?

Will restore the git setup() and see if it then goes problematic without camStart()?

9:54a - called camStart (); at the end of setup() so won't be called in loop()

New code upload started fine - noticed one diff - in recent runs ?

Using the camStart (); bypass the first 'f' was not apparently AutoExposure corrected? It was DARK, but "F" then corrected for good balance.
> Putting camStart (); into setup() resulted in exposure corrected on first 'f'

Ran fine this long:
LP#=35 Pisr#=111413 lastPR=107583793 ms=7754567 ipCyc%=59.51 S5#=0 51C
LP#=33 Pisr#=111424 lastPR=107583841 ms=7755553 ipCyc%=59.49 S5#=0 52C

Did 10 sec unplug to see about resetting the camera for the old style startup init ordering ...
>> That's all that changed since it would go 'bad' before in 3-9 hours - now up for 36 hrs?
Note: camera seating was not adjusted - now 8 bit again and as noted with prior startup order mostly in place - except I did put the camStart (); at the end of setup() ... it was right after while(!Serial ...

No problem with half return to last uploaded in 1.45 hrs + prior restart:
LP#=34 Pisr#=111416 lastPR=107583841 ms=5228137 ipCyc%=59.59 S5#=0 52C
LP#=34 Pisr#=111518 lastPR=107584021 ms=5229122 ipCyc%=59.69 S5#=0 52C

Just set that aside and went to the last GIT PR version, reprogrammed okay to start - then 10 sec unplug restart okay

Now 4+ hours again and no problems - not even odd display ???? - running last PR to Github that repro'd in under 4 hours - and 42 hours? since back to the PJRC.BB?
LP#=35 Pisr#=111405 lastPR=107583793 ms=15423922 ipCyc%=59.54 S5#=0 53C
LP#=34 Pisr#=111495 lastPR=107583967 ms=15424908 ipCyc%=59.64 S5#=0 53C

Well running that code it Started the restart process about 20 minutes ago

Tried getting the prior PR Version uploaded to run and it dies - and restarts - again quickly { with CamStart() delayed) }and repeatedly now that it is in that state - even with some unplug power off restarts
 
Last edited:
@defragster - @KurtE
If you want a little distraction attached is a working copy of the tensorflow lite person detection using a HB01B0 on the ML board. Not sure I did the image crop code correctly but seems to work ok - think needs some more work like add a display. But getting tired now and almost time for diner. So if you all want to play :)
 

Attachments

  • person_detection.zip
    352.6 KB · Views: 45
Good Morning all,

@mjs513 - Looks like fun... Downloaded and builds :D Will have to play later.

The last few days started to do some more needed outside work, and did some mowing, and some hand brush clearing along edge of trails. Lots more to go, as the berry vines want to completely take over, and there are areas where some of the vines appear to grow something like a foot a day.

Plus spending time trying to better understand some Lynxmotion Smart Servos...

Back to playing and infusion of coffee
 
Good Morning all,

@mjs513 - Looks like fun... Downloaded and builds :D Will have to play later.

The last few days started to do some more needed outside work, and did some mowing, and some hand brush clearing along edge of trails. Lots more to go, as the berry vines want to completely take over, and there are areas where some of the vines appear to grow something like a foot a day.

Plus spending time trying to better understand some Lynxmotion Smart Servos...

Back to playing and infusion of coffee

Morning @KurtE
Was kind of fun tying the camera stuff in with TensorFlowLite. Sounds like you are having a lot of fun as well.

A couple of notes:
1. Just realized that the library.properties has a couple of things set I may have to remove or figure out how it works:
Code:
ldflags=-lm
includes=TensorFlowLite.h
precompiled=full
basicaly ldflags and precompiled?
2. While it is set up to use the HM01B0 camera I left the code in to use the OV2640 2mp mini Arducam camera (requires Arducam). With some modifications can probably get that to work.
3. Still not 100% sure I did the cropping correctly down to 96x96 pixels correctly.

Any have fun playing. I did push it to the library so I wouldn't loose it.

Now as you said - more coffee.
 
Also downloaded new PersonDetect here - but not touched.

I restarted the VID test on pjrc.bb last night - still running this AM ????? The image was good first thing this morning - but now gone TOPO

Wanted to test the restart fail case if the camera is pulled after it gets in that state - using the code that doesn't init the camera until USB is received.
 
@KurtE - @defragster

This morning I finally got around to replacing the regulator on the PJRC BB, i.e., new 1a, regulator.

Reran the Zelda sketch to see if we would still get that popping/distortion when running in video mode and guess what NO Distortions/Popping sounds heard.

Just thought you all would like to know :)
 
Back
Top