I've heard mixed reviews on the TBC10, but I am a fan of SDI capturing whenever possible for the reasons already mentioned. Basically, all TBCs convert the incoming video to a digital buffer and then output the frames at a digitally-precise rate, so when using SDI output from one, you skip the conversion back to analog and there are fewer analog to digital and digital to analog conversions in the chain, each of which has some opportunity for image loss.
Uncompressed and lossless are confusing terms though, a zip file that is compressed 5:1 (takes 1/5th of the space of the files in it) but then unzips to 1:1 is still "compressed" and ACTUALLY lossless. This is only possible because the starting files are digital. It technically isn't possible to have truly lossless digital to analog or analog to digital reproduction as there are infinite possibilities of analog values. In digital, everything is a 1 or a zero, but when you try to have that go to analog, there's always a transition period between the 1 and zero, hence no perfectly square waves from zero to 1 and one to zero. You could divide the horizontal resolution into however many pixels you want for even more resolution, but they settled on 720 as the standard, so that's what we use. Vertical resolution is fixed though since there's a fixed number of scan lines per frame/field.
When most people refer to lossless captures, they are doing it at 8 bit color which has far fewer possible pixel color assignments than 10 bit does. Humans can EASILY tell the difference between 8 and 10 bit color pallets in certain types of scenes, mainly ones with low dynamic range and with mostly one color like sky, sunsets, or dark scenes, you'll often notice "blocks" of solid colors without any variation where the values in that area all rounded to the same value in an 8 bit system, whereas in a 10 bit system, they wouldn't appear blocky.
This is an over-exaggeration (similar colors get rounded to the same value), but you get the idea:
Screenshot 2024-12-22 at 11.31.47 AM.jpg
For that reason, you might find that ProRes422 or ProRes422HQ while not technically lossless actually looks better and may take up the same or less space depending. It gets away with that because it is compressed, but has more actual information present.
To me, uncompressed 8 bit 4:2:2 isn't as desirable as partially compressed and visually lossless 10 bit color (ProRes422/HQ), but to each their own. It used to be an issue where programs like AVIsynth only worked in 8 bit color spaces, but I think there are now 10 bit implementations.
All of this of course depends on the TBC you are using though. If the TBC is using an 8 bit DAC (like the recommended TBC1000/AVT8710), that limits the color variability to 8 bit going forward in the chain, so there'd be no benefit to capturing in 10 bit. This may be why the ES10/15 passthrough is said to be posterized as it may have an effective conversion rate of a bit less than 8 bit which would cause that effect. Looks like the TBC10 is actually a 10 bit and wouldn't surprise me if that is what the "10" in the model number is trying to tell you.
In most standard definition scenes, you probably won't be able to tell the difference for 8 bit versus 10 bit, but in other scenes, you definitely will be able to tell. Sunset is the classic one where the different shades of yellow/orange will appear as bands around the sun as opposed to a smooth transition between them all.
