4K Cameras – Design of the new eye bulb – Stint 2

[pattacini]

Follow-up of #7.

[salvi-mattia]

I think we can remove the two dragonfly board on the side of the head [red arrows], isn't it?

and what about RFE and PS-CPU arethey both going to be there?

then we have to add the Jetson Nano and Quark carrier and the two little adapter between this and the cameras themselves.

is there any other board to add/remove? Should the Jetson Nano and Quark carrier be placed in any particular place?

@maggia80 @MrAndrea @fiorisi

cc @marcellosavoldi

[pattacini]

I think we can remove the two dragonfly board on the side of the head [red arrows], isn't it?

Yes 👍🏻

is there any other board to add/remove?

We have to replace the present COM-EXP Type 6 with the latest type 10. See https://github.com/icub-tech-iit/tickets/issues/2032#issuecomment-1117191139.

[pattacini]

Should the Jetson Nano and Quark carrier be placed in any particular place?

We should consider for that some aspects related to heat dissipation and dynamical behavior of the head while moving (i.e., close to some major rotational axes).

For the second point, we may then keep @mfussi66 in the loop.

[salvi-mattia]

@MrAndrea @maggia80

trying to create the cad model for Quark Carrier for NVIDIA® Jetson Nano™ with @fiorisi he would like to know wich of this code are we going to use

does it already exist in wingst?

[maggia80]

NGX004-02, but I have ordered also the passive heatsink

[MrAndrea]

Reference for the part numbers ordered:

• https://github.com/icub-tech-iit/purchases/issues/1768
• https://github.com/icub-tech-iit/purchases/issues/1692
• https://github.com/icub-tech-iit/tickets/issues/2121

[salvi-mattia]

Yesterday, talking with @maggia80 we saw we have to remove all the boards with red arrows

the main boards in the neck are going to be replaced by 16446 - Quark Carrier + NVIDIA® Jetson Nano™ Module + Active Thermal NGX004-02 and the group comE10+carrier+heatsink alerady present in the ergocub chest [as @MrAndrea said]

is it correct @maggia80 ?

it wuold be nice that your new head frame would fit this mounting, @marcellosavoldi

[maggia80]

yep!

[pattacini]

As agreed w/ @maggia80 and @salvi-mattia, we're going to design two alternatives:

1. The "standard" approach with the o-rings to fine-tune the CCD position.
2. A second solution w/o the o-rings and one unique sensor support.

It's fundamental to make a 3D-printed copy of the standard-wise support to allow for testing the SW pipeline.

[salvi-mattia]

define the lens is quite blocking at this point we have to understand volumes and lenght in order to nicely fit the bulb/eyelid sistem we have ordered to MWS a les holder that fits basler 4K sensor board in order to run some tests

this is a good candidate and we have it to test @icub-tech-iit/fix is going to run this test

this is the MWS job we r waiting for

[salvi-mattia]

decisions made reviewing with @maggia80

• we are going to use M2.5 screws for the camera board [ M3 are too big and the positionig is made by pins]
• we are going to minimize the fine camera tuning using only the compliance of 4 Orings between camera and pan frame - red arrow in pic [this because we trust new prismatic coupling for the tilt shaft ]
• the screws fot this tuning are in the back of the eye, in consequence of this we no longer need the bulb to be made in two parts
• pan axis will not be coincident with image plane [it will be moved slightly frwd or bckwd depending on the lens choice]

@fiorisi is the historical problem we had with eyes' pan harmonic resolved? is this fix implemented also in the motor/reducer cases in the neuro head? [we are going to use them]

[fiorisi]

@fiorisi is the historical problem we had with eyes' pan harmonic resolved? is this fix implemented also in the motor/reducer cases in the neuro head? [we are going to use them]

@salvi-mattia if you refer to the tolerances of the housing, this was the outcome of the meeting with HD and @Mick3Lozzo:

• point 7: https://github.com/icub-tech-iit/cad-mechanics/issues/127#issuecomment-840577826
• conclusion: https://github.com/icub-tech-iit/cad-mechanics/issues/127#issuecomment-842939392

Concerning the "neuro head" I think that we should investigate with @Mick3Lozzo.

[salvi-mattia]

from this two comments it's not clear to me which measures and toleraneces do I have to check I'll ask @Mick3Lozzo for being sure

[salvi-mattia]

A short list for me to remember

• Create 2.5mm ORing in cad and wingst [waiting fo feedback from supplier]
• Create the lens in cad and wingst [waiting for test]
• Modify the bulb to fit both new mech and new lenses
• Create connection bulb-mechanics
• Model the “adapter” board and fit them in the head
• Add new frame from @marcellosavoldi
• Check with @Mick3Lozzo for the pan HD tolerancies
• Cross the fingers

For everyone who wold like to give a look at the design it’s in this repo [mech branch] Open IC_011_G_001_4K Left eye: new eye Right eye: neuromorphic eye

[salvi-mattia]

talking with @Mick3Lozzo we found these differencies

I'm going to modify measures and tolerancies in the right images in order to match left ones

cc @fiorisi @Mick3Lozzo

[salvi-mattia]

I found also these parts are different

also in this case I'm going to matche the left side image, the one without tolerances

is there anything else to be modifyed?

@fiorisi @Mick3Lozzo

[salvi-mattia]

workshop job is done

[salvi-mattia]

being the distance between the image plane and the tip of the candidate lens 23mm

and being the distance between the image plane and the support face 2.3mm

the lens is supposed to be 20.7mm from the support face.

being the Smount [M12x0.5] in the plastic support really poor and badly realised it is not possible I'm going to mout it as next as I can

we are going to take some images

we have also test the camera with its original support and we took some images.

is there any particular test you want us to run to test the camera?

@pattacini @maggia80

[pattacini]

is there any particular test you want us to run to test the camera?

I would limit this preliminary analysis to the FOV of the lens.

[salvi-mattia]

something new after today's meeting with @DanielePucci and @S-Dafarra

tha basler sensor we r going to mount has different dimensions WRT dragonfly:

it is 1/1.8" while dragonfly has 1/3" CCD

it means we r going to obtain a wider FOV even with the same focal lenght furthermore the lens now on iCub is a f=2.9 while the candidate lens is f=2.8 [slightly shorter, slightly wider]

here some results from an online FOV calculator

we are going to provide the two mountig in order to compare them but in the mean while we r quite confident that the increment in FOV is significant and I think we can proceed with the mechanical design considering this mounting.

any consideration?

cc @maggia80 @pattacini @mfussi66

[salvi-mattia]

in case we r going to use 2K cameras

[mfussi66]

Response from Basler!

Dear Mattia, Lets answer this in two stages:

1. What is the difference between mc and mci cameras? The difference is the ISP (Imaging Signal Processor) wich is included on the board in mci cameras and not mc cameras. One could say the mci cameras are more targeted to the Jetson boards because we can not use the ISP that is included in the Jetson platform yet. And we created the mc cameras to be used with the NXP boards that include an ISP that we can use.

2. What kind of embedded platform (e.g. Single-Board Computer similar to the Jetson Nano) can be used to take advantage of the daA3840?

You can use the daA4200-30mci-MX8MM-VAR - Embedded Vision Kits

With best regards, Johannes-Lars

Basler Customer Service

daA4200-30mci-MX8MM-VAR in question.

The embedded vision kit mentioned by them does not provide a Jetson Nano, but a Variscite SoM (System on Module), aka one of these: https://www.variscite.com/products/system-on-module-som/?cpu_name=NXP%20i.MX8M%20Mini