Thread by @Ascii211

Got one of those "camlink" USB converters because I was curious about em, and WHOAH I was shocked. UVC output, 20ms delay and signal looks to be properly grabbing 1080P@30. Single chip solution under a small sink and costs under $20. What in the heck is this.

I'll prob dump the eeprom that's on board, and take a look at the part number of the chip, bet there's gonna be no datasheets though.

The plot thickens. As requested by @esden the main chip picture. Unfortunately not a FANTASTIC FPGA deal, rather an ASIC, not listed on MacroSilicon's page.

Quick capture from my DSLR plugged into it, no color washout, nice and clean signals. Shockingly great for the money, I really don't think you're gonna beat this for bang/buck.

Also is DEFINITELY capturing at 1080P.
All reported resolutions from UVC class are listed here

The quick capture metadata. Interestingly it seems to auto-synch to the input framerate(up to 30hz)

All up, a damn good option for adding HDMI input to your streams or whatever for a budget and not dealing with camlink scalpers.

Anyway I'm naming this thing the "Can't link" for joke potential.
Just like the harbor freight hardcases are "Pelican'ts"

Bought this because the several hundred MS delay of the OG Elgato HD was annoying for my microscope camera

Anyway: link time
ebay.com/itm/1742801499…

Uh it appears to have HDCP handshake too.... WOW

People wanted to know if you could use it to grab video output of a raspberry pi, the answer is yes.

Anyway, I'm dumping the EEPROM later today and will post the binary for interested people to peruse.

Weirdchamp news bois.
Whole dataflash is ZEROES! I am MAXIMUM CONFUSE

The plot thickens, it shows up as a strange display on my windows machine.... with no resolution sink options.... Interesting

Works fine with the switch tho

Theory is that it doesn't have EDID but the GPU I2C query definitely showed it up as "HDMI to USB" which is an EDID string.... HMMMMMMMM

more formats and framerates

Real bad news, my "USB3" 1080P 60FPS unit arrived and yep, it's the EXACT SAME thing as this user has. Just a blue USB 2.0 version of the aluminum case with a higher price tag lol.

https://twitter.com/_Ninji/status/1271610223522963456?s=19

At this point the cheapie 2.0 aluminum case is the FAR FAR better buy

Further updates: on some USB 2.0 controllers it can have pixel loss and juddering. Mostly seen on older 2/3rd gen Intel LGA1155 era boards. Works fine on modern AMD2.0/3.0 and older USB3 controllers

Btw pic related to the "USB 3.0" capture card. It's on the left, real 3.0 on right.

well good news and bad news, seller partially refunded. bad news is partial refund. Eh whatever

Later this week I'm making a return to this thread thanks to getting new flash dumps of different variants that advertise over UVC as being able to do 1080/60, will be trying to crossflash to see if it's actually a firmware or just a silicon mask rev.

I should also send some to @johndmcmaster for decapping and imaging of die.

Sent in by @rombik_su, an EEPROM image from one that enumerates 1080@60 vs one of my 1080@30. Diff is highlighted.
VERY INTERESTING, both have the Macrosilicon string but there's a LOT of difference between them.
Crossflash next!

Well, a crossflash did not change the available frame rates... interesting.

I wonder if it is a different mode strap ngl.
I'll go buy some of the 60FPS units and do a full traceout of the schematics on a stream.

oof this is gonna get expensive.

Ya know, looking at this, I think the mode strap resistors COULD be R1 and R5.
@rombik_su can you trace out these pins on yours and see what they're strapped as?

You know, these things are so hilariously cheap that if you get them in bulk to hit scale discounts, you could 1000% be THAT house on halloween that hands out capture cards
😂

Well now things get even MORE interesting. I have the firmware from my 1080@30, Rombic's 1080@60 and now @EternoJax's submission that has a TOTALLY different ID string. I wonder if the second difference in the top line is the firmware date revision?

I thought it could be Year Month Day, which holds for the first two, but the last one does not as there is no 20th month. Side note, the more recent images seem to get bigger, middle is oldest, then right and left. More crossflashing to do

Ok wacky mystery that piqued my interest. @boichipIsaccept DM'd me about a card that wouldn't work normally and looks like some weird wish tier knockoff of a shanzhai product which is 🤔already. So of course I ask em to crack that open to see what kinda guts we got.

It gets WEIRDER, the screws reluctantly came out aside from ONE that spins. attempting to extract board normally failed, so gentle tapping was tried and failed. Eventually slammed it on a desk so hard it dented the surface multiple times, also to no avail.

Side note; the video I was sent of them slamming it into the carpet was so unintentionally hilarious that I couldn't help but laugh.

Update, back cover is OFF and the board is still extremely reluctant to come out.... hammertime?

Hooboy, this one gets interesting.... component size dropped to 0402 mostly, and there's SUPERGLUE outgassing residue ALL over it. Waiting for chip ID numbers, no heatsinking makes me concerned.

And here's where it gets REALLY interesting; ITS A WHOLE DIFFERENT ID SCHEME ON THE MS2109!
No 20XX id string below the second string?

I wonder... this could be a new cost down rev, combined with it might even be a mask rev on the main ASIC, dropping a process node and increasing efficiency removing the need for a relatively expensive heatsink?

Point of note from user is that it will capture the xbox just fine but colorbars on a switch video output which makes me wonder if the switch has some sort of atypical HDMI that the ASIC doesn't like

Anyway, that's another hmmm tier update but the RE work on figuring out the firmware differences in the 30 v 60 is

is continuing in the background. Still waiting on hardware deliveries.