The title says it all - is it okay to feed the 3v3 line via the qwiic connector?
In other words: no other power would be supplied to the camera, I would only connect the data and power lines of the qwiic connector and expect the camera to run (and my script to communicate via the uart1 tx/rx pins)
Best regards
Yes, this works, we need to do some more hardening on the I2CTarget mode driver. But, master mode is generally fine. If you use Target modeā¦ do not deinit during a transfer. This causes unrecoverable hardware lockups on the Alif chip.
OK, so if I understood it correctly:
Thanks for your answer!
Basically, the QWIIC cable is connected to the main 3.3V rail and we inject 3.3V onto that rail using an ideal diode from a USB 5V ā 3.3V switching regulator.
And here it is! OpenMV AE3 feeding its data into a LEGO Spike Prime hub 
The adapter is quite big right now, but target PCBs are being manufactured already and shall arrive after Easter.
I do have a āfunnyā though: if I connect this setup to my PC over USB-C, the connection keeps dropping and restarting. It works if I disconnect power from the LEGO hub. Not sure whatās the deal here, but I didnāt get to the bottom of the schematics yet (just printed them big, this is why I was pestering you for the schematics over at Kickstarterā¦ 
)
Thanks for the help! Really appreciated!
Just tested with switching the AE3 back and forth between 3.3V externally and also being powered via USB. Works perfectly. When USB is applied there is no load on the 3.3V PSU. When USB is removed the load switches over, and then switches back when USB is applied again.
If your 3.3V rail needs to be charged up though, this can result in a brownout if thereās a lot of big caps the AE3 has to power all of a sudden.
I will investigate it further, I have a suspicion itās not the AE3 thatās playing tricks but my laptopās usb. I donāt have issues if I connect via USBA to USBC cable, but it gets unstable if I go USB-C to USB-C, like if the laptop couldnāt decide if it should supply the usb or get supplied by the usb port.
I need to test it properly, but today I ran out of time and will not be able to get back to it for a couple days now - Easter is coming
So, I looked into it a bit more, hereās a summary for those who might land here via a search:
This condition can (but not always will) get triggered on my Lenovo T14Gen3 Intel laptop. It seems not to be related specifically to the AE3, but rather to the implementation of the USB system on the laptop.
Ah, weird. The AE3 just accepts 5V USB, and we have the two 5.1k pull-downs on the device. This is the correct implementation. It appears the laptop design isnāt quite right.
But, I have a Lenovo too, and all sorts of stuff isnāt quite right also. These laptops are buggy. My camera fails to work randomly along with many other issues.
Yeah, I blame the laptop here tooā¦ I posted it just as a warning/info that one should not get startled when observing the behaviour, but just connect to an USB-A and check if it persists.
I have connection problems too: AE3 <> OpenMV IDE.
Running on a Macbook Pro M3 with USB-C.
I have narrowed it down to this:
a main.py that is actively writing to UART(1) or I2CTarget(1) on the qwiic in a tight loop.
As soon as such program is running, USB and the IDE balk, doesnāt matter if it was booted via 3v3 on SPIKE or via USB. The only thing I can do is flash the AE3 again.
My first workaround was to put a 5s sleep before the program, so I could quickly connect after the green light stopped blinking.
My second workaround was to use the switch on the side. Now I have this in my main loop:
from machine import Pin, UART, I2CTarget
sw = Pin("SW", Pin.IN)
while True:
if not sw.value():
break
This stops the loop and the writing and enables the IDE to connect again.
Nonetheless, it would be much more userfriendly if USB always worked, regardless of what UART or I2C is doing.
Hi Aton,
Can you give me a demo program that crashes things for both? The issue is that the driver isnāt polling USB events internally when blocking for data. I can send Damien your code samples so he can fix these bugs. However, you can also avoid the issue by doing non-blocking calls for the UART or I2C too.
Here are the scripts. Both crash the camera and the IDE hard if I connect without pressing the switch and exiting the loop:
UART
```
import time
import math
time.sleep(5)
thresholds = [(0,20,-6,6,-6,6)] #Black
import sensor
sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.QVGA)
sensor.skip_frames(time=1000)
sensor.set_auto_gain(False) # must be turned off for color tracking
sensor.set_auto_whitebal(False) # must be turned off for color tracking
clock = time.clock()
# Start uart
import machine
import struct
sw = machine.Pin("SW", machine.Pin.IN)
u = machine.UART(1, 115200)
while True:
clock.tick()
img = sensor.snapshot()
blobs = img.find_blobs(
[thresholds[threshold_index]],
pixels_threshold=200,
area_threshold=200,
merge=True,
)
if len(blobs) > 0:
blob = blobs[-1]
# These values depend on the blob not being circular - otherwise they will be shaky.
if blob.elongation() > 0.5:
img.draw_edges(blob.min_corners(), color=(255, 0, 0))
img.draw_line(blob.major_axis_line(), color=(0, 255, 0))
img.draw_line(blob.minor_axis_line(), color=(0, 0, 255))
# These values are stable all the time.
img.draw_rectangle(blob.rect())
img.draw_cross(blob.cx(), blob.cy())
# Note - the blob rotation is unique to 0-180 only.
img.draw_keypoints(
[(blob.cx(), blob.cy(), int(math.degrees(blob.rotation())))], size=20
)
u.write("{} {}\n".format(blob.cx(), blob.cy()))
else:
u.write("00 00\n")
print(clock.fps())
if not sw.value():
break
I2C target
# This work is licensed under the MIT license.
# Copyright (c) 2013-2023 OpenMV LLC. All rights reserved.
# https://github.com/openmv/openmv/blob/master/LICENSE
#
# Single Color RGB565 Blob Tracking Example
#
# This example shows off single color RGB565 tracking using the OpenMV Cam.
import time
import math
from machine import Pin, I2CTarget, LED
led = LED("LED_BLUE")
sw = Pin("SW", Pin.IN)
print(sw.value())
threshold_index = 0 # 0 for red, 1 for green, 2 for blue
# Color Tracking Thresholds (L Min, L Max, A Min, A Max, B Min, B Max)
# The below thresholds track in general red/green/blue things. You may wish to tune them...
# thresholds = [
# (30, 100, 15, 127, 15, 127), # generic_red_thresholds
# (30, 100, -64, -8, -32, 32), # generic_green_thresholds
# (0, 30, 0, 64, -128, 0),
# ] # generic_blue_thresholds
thresholds = [(0,20,-6,6,-6,6)] #Black
import sensor
sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.QVGA)
sensor.set_windowing([0,sensor.height()//2,sensor.width(),sensor.height()])
sensor.skip_frames(time=1000)
sensor.set_auto_gain(False) # must be turned off for color tracking
sensor.set_auto_whitebal(False) # must be turned off for color tracking
clock = time.clock()
# Only blobs that with more pixels than "pixel_threshold" and more area than "area_threshold" are
# returned by "find_blobs" below. Change "pixels_threshold" and "area_threshold" if you change the
# camera resolution. "merge=True" merges all overlapping blobs in the image.
# Start i2c
import struct
# Initialiseer de buffer met een grootte van exact 4 bytes
buf = bytearray(4)
# Start de I2C Target (Slave) met de gevulde 4-byte buffer
target = I2CTarget(1, addr=42, mem=buf)
n = 0
while True:
n = (n + 1) % 60
if n == 0:
led.on()
if n == 30:
led.off()
clock.tick()
img = sensor.snapshot()
blobs = img.find_blobs(
[thresholds[threshold_index]],
pixels_threshold=200,
area_threshold=200,
merge=True,
)
if len(blobs) > 0:
blob = blobs[-1]
# These values depend on the blob not being circular - otherwise they will be shaky.
if blob.elongation() > 0.5:
img.draw_edges(blob.min_corners(), color=(255, 0, 0))
img.draw_line(blob.major_axis_line(), color=(0, 255, 0))
img.draw_line(blob.minor_axis_line(), color=(0, 0, 255))
# These values are stable all the time.
img.draw_rectangle(blob.rect())
img.draw_cross(blob.cx(), blob.cy())
# Note - the blob rotation is unique to 0-180 only.
img.draw_keypoints(
[(blob.cx(), blob.cy(), int(math.degrees(blob.rotation())))], size=20
)
struct.pack_into('<hh', buf, 0, blob.cx(), blob.cy())
else:
struct.pack_into('<hh', buf, 0, 0, 0)
print(clock.fps())
if not sw.value():
break
Thanks for the post. Will get it debugged.
Okay, so, you need to update to the latest pre-release firmware via Tools->Install the Latest Dev release.
Note that the API has changed significantly for using python objects. Weāve removed a lot of older stuff that wasnāt pythonic and was also costing us valuable flash space.
See: v5.0.0 - OpenMV MicroPython 1.28 documentation
I am still working on the tutorial, but, once complete, Iāll be updating the IDE with things and getting this out the door.
import time
import math
time.sleep(5)
thresholds = [(0,20,-6,6,-6,6)] #Black
threshold_index =0
import sensor
sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.VGA)
sensor.skip_frames(time=1000)
sensor.set_auto_gain(False) # must be turned off for color tracking
sensor.set_auto_whitebal(False) # must be turned off for color tracking
clock = time.clock()
# Start uart
import machine
import struct
sw = machine.Pin("SW", machine.Pin.IN)
u = machine.UART(1, 115200)
while True:
clock.tick()
img = sensor.snapshot()
blobs = img.find_blobs(
[thresholds[threshold_index]],
pixels_threshold=200,
area_threshold=200,
merge=True,
)
if len(blobs) > 0:
blob = blobs[-1]
# These values depend on the blob not being circular - otherwise they will be shaky.
# if blob.elongation() > 0.5:
# img.draw_edges(blob.min_corners(), color=(255, 0, 0))
# img.draw_line(blob.major_axis_line(), color=(0, 255, 0))
# img.draw_line(blob.minor_axis_line(), color=(0, 0, 255))
# These values are stable all the time.
img.draw_rectangle(blob.rect)
img.draw_cross((blob.cx, blob.cy))
# Note - the blob rotation is unique to 0-180 only.
# img.draw_keypoints(
# [(blob.cx(), blob.cy(), int(math.degrees(blob.rotation())))], size=20
# )
u.write("{} {}\n".format(blob.cx, blob.cy))
else:
u.write("00 00\n")
print(clock.fps())
if not sw.value():
break
This works with V5.0.0 without crashing. I didnāt fix all the calls to the new API, but, everything is documented.
Doesnāt work on my side. Still hangs the camera. Where did I go wrong?
I have OpenMV IDE 4.5.0
When I install latest dev release on AE3 via tools it still says:
OpenMV v4.8.1; MicroPython v1.26.0-77; OpenMV-AE3 with AE302F80F55D5AE
Type āhelp()ā for more information.
(Expecting 5.0.0, MicroPython 1.28)
The framerates seem to be lower.
And I had to update the script you shared to this:
import time
import csi
sensor = csi.CSI()
sensor.reset()
sensor.pixformat(csi.RGB565)
sensor.framesize(csi.VGA)
sensor.snapshot(time=1000)
sensor.auto_gain(True) # must be turned off for color tracking
sensor.auto_whitebal(False) # must be turned off for color tracking
clock = time.clock()
# Start uart
import machine
import struct
sw = machine.Pin("SW", machine.Pin.IN)
u = machine.UART(1, 115200)
while True:
clock.tick()
img = sensor.snapshot()
blobs = img.find_blobs(
[(0,20,-6,6,-6,6)], #Black
pixels_threshold=200,
area_threshold=200,
merge=True,
)
if len(blobs) > 0:
blob = blobs[-1]
img.draw_rectangle(blob.rect())
img.draw_cross((blob.cx(), blob.cy()))
u.write(struct.pack( 'hh',blob.cx(), blob.cy() ))
else:
u.write(struct.pack('hh',0,0))
print(clock.fps())
if not sw.value():
break
OK. Updated the IDE to 4.8.11 manually. A whopping 5Gb. (I thought is was auto-updated) Then installed the latest dev release.
Now it says:
OpenMV fe74ac2725; MicroPython 0f831e89f0; OpenMV-AE3 with AE302F80F55D5AE
Type "help()" for more information.
But camera still hangs on this (adapted again) main.py. And framerates are in the low 20ās, while on the 4.81. firmware, framerates were in the 60ās!
import time
import csi
sensor = csi.CSI()
sensor.reset()
sensor.pixformat(csi.RGB565)
sensor.framesize(csi.VGA)
sensor.snapshot(time=1000)
sensor.auto_gain(True) # must be turned off for color tracking
sensor.auto_whitebal(False) # must be turned off for color tracking
clock = time.clock()
# Start uart
import machine
import struct
sw = machine.Pin("SW", machine.Pin.IN)
u = machine.UART(1, 115200)
while True:
clock.tick()
img = sensor.snapshot()
blobs = img.find_blobs(
[(0,20,-6,6,-6,6)], #Black
pixels_threshold=200,
area_threshold=200,
merge=True,
)
if len(blobs) > 0:
blob = blobs[-1]
img.draw_rectangle(blob.rect)
img.draw_cross((blob.cx, blob.cy))
u.write(struct.pack( 'hh',blob.cx, blob.cy ))
else:
u.write(struct.pack('hh',0,0))
print(clock.fps())
if not sw.value():
break
Yeah, the IDE is big because of the STEdge AI compiler. Ibrahim is working on a new, smaller IDE thatās coming along. The install size on disk is similar, but it downloads the tools after installation.
What OS are you on? Iām not getting a hang running the script. Itās working as expected.
As for the decrease in frame rate, mmm, we changed the memory architecture for allocations. This may be a one-liner to fix. Will check.
sensor.framesize(csi.VGA)
You were using QVGA before. Change it back to that, and you get 60 FPS.