Oak-D-Lite Happyness

[Pito Salas]

Sergei

After getting wound around the axle by version incompatibilities, spurious error messages, examples from Luxonis that didn’t work, and misdirection from Claude, I was about to give up until I upgraded the power supply to the Oak-D-Lite. 

I just tried your depthai_rospi and it just worked! And the current draw on the whole robot including motors, lidar and oak-d-lite is 0.7A, so well within the range. No brownouts no problems.

I am sure I am cheering prematurely. These things always go 

While true:

solve-one-problem

find-next-problem 

I am going to scour your repo to see what is different. But give us a clue. If there is a single thing you did to make it all work. Was it the FPS? Or something else?

Anyway, I answered my own question from the previous post. Yes, it is possible to run a oak-d-lite directly from the raspi 5 without an extra Y-connector and power supply. Yes there is a Santa Claude!

Thanks!

Pito

Boston Robot Hackers && 

Comp. Sci Faculty, Brandeis University (Emeritus)

[Marco Walther]

Pito & all,

Just keep an eye on it.

The problem is, all those chips/boards/... are smart with their power
usage. They *do not* consume a constant amount of power and even worse,
they might create some power spikes for [for the user] random reasons!

Your Pi + Oak-D might work or it might freeze/reboot/... once in a while:-(

Yes, I'm powering my Oak-D through a Pi5 as well. But I feed the Pi with
an 8A(15A max) power supply and all the other USB devices are connected
to a powered hub.

So, there are essentially two ways to go:
* Account for the spec'ed max power for each device and make sure, they
can get that.
* Live on the edge, watch things, and make changes as needed.

We as hobbyists mostly take 'door two', but we have to stay conscious
about it.

Have fun;-)
-- Marco

On 1/26/26 07:48, Pito Salas wrote:
> Sergei
>
> After getting wound around the axle by version incompatibilities,
> spurious error messages, examples from Luxonis that didn’t work, and
> misdirection from Claude, I was about to give up until I upgraded the
> power supply to the Oak-D-Lite.
>

> I just tried your depthai_rospi <https://github.com/slgrobotics/
> depthai_rospi?tab=readme-ov-file> and it just worked! And the current

> draw on the whole robot including motors, lidar and oak-d-lite is 0.7A,
> so well within the range. No brownouts no problems.
>
> I am sure I am cheering prematurely. These things always go
>
> While true:
> solve-one-problem
> find-next-problem
>

> *I am going to scour your repo to see what is different. But give us a

> clue. If there is a single thing you did to make it all work. Was it the

> FPS? Or something else?*

>
> Anyway, I answered my own question from the previous post. Yes, it is
> possible to run a oak-d-lite directly from the raspi 5 without an extra
> Y-connector and power supply. Yes there is a Santa Claude!
>
> Thanks!
>
> Pito
>
> Boston Robot Hackers &&
> Comp. Sci Faculty, Brandeis University (Emeritus)
>

> --
> You received this message because you are subscribed to the Google
> Groups "HomeBrew Robotics Club" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to hbrobotics+...@googlegroups.com
> <mailto:hbrobotics+...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/
> hbrobotics/E58A626F-5699-4362-8DFB-E92D3CCEDD9A%40gmail.com <https://
> groups.google.com/d/msgid/hbrobotics/E58A626F-5699-4362-8DFB-
> E92D3CCEDD9A%40gmail.com?utm_medium=email&utm_source=footer>.

[Sergei Grichine]

Pito,

Thank you for trying my code and confirming that it works.

"...I am going to scour your repo to see what is different. But give us a clue..."

- sure,  here it is

"If there is a single thing you did to make it all work. Was it the FPS? Or something else?"

- I'd guess you might find the answer by looking at the code (git history might help) - but I really can't remember now any one thing. I am not actively using OAK-D Lite at the moment.

Marco rightfully noted that most issues are related to powering the device (1 A is not a joke). He feeds the camera through Pi's USB - I am not fan of it and explained my point here. I'd guess, if you read it - you'll find that Luxonis agrees.

OAK-D camera can adjust to available USB link speed and most scripts and launch files will report it ("LOW", "FULL", "HIGH", "SUPER", "SUPER_PLUS"). For USB2 speed will be "HIGH", for USB3 - "SUPER". Make sure you know what speed is your camera using.

If you (or anybody else) would find any issues or want to submit a pull request - please use GitHub. Contributions are highly appreciated. "Starring" a repo makes it more visible to other struggling folks.

Best Regards,

-- Sergei

[Steve " 'dillo" Okay]

On Monday, January 26, 2026 at 6:02:35 PM UTC+1 Marco Walther wrote:

Pito & all,

Just keep an eye on it.

The problem is, all those chips/boards/... are smart with their power
usage. They *do not* consume a constant amount of power and even worse,
they might create some power spikes for [for the user] random reasons!

I also have a good bit of experience with the OAK-D Lite, using it on Tenacity and for a number of client projects. 
IME, I've found these "random power spikes" to be correlated with the camera working in a "target rich environment". In other words, when the model loaded on the camera has a lot of "hits", that's more the onboard vector processor has to do and so it starts pulling more current. Depending on what else you have attached to that particular USB bus, it might trigger a power-draw interrupt from the USB controller, which disconnects the offending device off that bus. This is not a great thing to have when you're depending on that device for navigation & collision-avoidance with your robot. The D400-series Realsense devices have this problem too. 

'dillo

[A J]

My older optical disc drive had 2 USB 3.0 plugs for power. Maybe there is a power hack for the cameras.