Asgeir Hustad wrote:I wouldn't use NDI for sources, because I very often need to project back to screens in the venue, and as such need as little latency as possible. But for transporting the (low-priority) program or an AUX to the other side of a location, or from location to a MCR? Where there already is a network with sufficient capacity?
Sure, any time (At least unless it's projected on a huuuge screen where you really need the uncompressed image).
If I have to lay down cable anyway to get a signal (cause I don't trust the local networks and network-guys with my sources), I'll either use SDI or fiber. It ensures uncompressed quality until the master recording at least, and has longer range than a network cable, requires fewer switches and less resources of them, and is generally a known technology.
This might be a great opportunity to get some wider feedback.
In my earlier post about migration from SDI to IP Video, I made a comparison with transitioning from CDs to MP3 - essentially accepting some small compromises, in return for the massive boost in flexibility (workflow).
It seems that the 'compromises' IP Video might present could be latency, and image quality. In the case of SMPTE2110, at present it could be increased cost, although in theory 2110 wont introduce the other 2 compromises.
Those who have tried IP video will already appreciate the benefits, and most likely have already accepted any compromises.
Given that most modern processing devices already introduce latency, as do projectors, screens, even cameras. I guess really the question is - how much latency *is* acceptable as a compromise in IP Video ?
It might depend on the application - as you mention MAGI is an area where a large latency is noticeable, but as far as I have observed, existing MAGI systems with baseband video already have quite a delay. My guess is that an extra frame or 2 may not make any practical difference. Remember that you are already delaying the audio for the back speakers.....
Another area where latency is critical is for 2-way communications where one talent interacts with another via a video link. Here, I am certain that a couple of frames is perfectly fine - in fact up to half a second appears to work OK.
Remember that we long ago embraced satellite transit of live video, which has massive latency and compression, for the very same reasons - convenience - and unlocking global real time video workflows - so clearly the industry is perfectly happy to exchange compromises for flexibility.
On the topic of quality - again - how much compression is acceptable ? once again it may depend on the application. Things like very high quality green-screen compositing require very clean signals and many green screen pros would even reject your beloved 4:2:2 SDI video for this application anyway, so maybe that one is golden in terms of quality.
In other areas, again - we have already accepted compression for convenience - almost all professional camera recorders use compression, and its pretty rare to see uncompressed video files outside of a D.I or Grading environment. All our final TV and Internet delivery is massively compressed - so the end consumer almost never sees uncompressed pictures. And when was the last time you saw an uncompressed still photo?
It seems that the professional production industry has accepted that really good quality compression like ProRes, or DNxHD is perfectly acceptable - even for really high quality productions like Movies. If we dig into the compression systems proposed for IP video, we find the same sorts of quality, so maybe it should not really be a concern ?
I dont know much about TICO, but I suspect its pretty much undetectable in terms of quality loss on a typical signal. Unfortunately, Tico isn't going to get the same take-up in the market as other standards due to its proprietary commercial model.
We are already seeing the AV1 open compression standard casting massive waves across H.265 and others since its promised to be license-free. Codec licensing has turned out to be a huge issue for video manufacturers.
JPEG 2000 has already taken quite a hold in IP Video, and quality/latency is good, although a lot of the applications are with hardware devices, since J2K is pretty heavy processing.
In the case of NDI, it uses a codec which is technically and quality comparable with ProRes and DNxHD, and is very low latency (technically minimum 8 video lines, typically 1 frame). The NDI codec also benefits from zero generation loss - you can compress and recompress the same image 1000 times and it never gets any worse beyond pass 1.
For most people in anything other than a scientific image-analysis environment will probably find that the compression in NDI (and TICO, and J2K) will be no-compromise at all in practice. You certainly wont notice *any* compression compromise on a typical big projection screen in a live environment.
So, ultimately I think I have convinced myself that compression is a non-issue, which leaves latency as the possible compromise and it is something you can notice in some cases. Perhaps this single issue will determine where IP Video gets to spread its wings (and unlock incredible workflows) and where the industry will insist on zero(ish) latency connectivity. Frankly the difference between 0 and 1 or 2 frames is not a lot for almost every application, so maybe this is not such a big deal.
Anyone operating in professional video has a responsibility to themselves and to their customers to fully explore this new world.
** IP Video is now inevitable, in the same way that SDI once was compared to analog. **
There are lots of free IP video tools out there which will let you draw your own conclusions. Sitting on the fence scowling at IP Video isn't going to save your business.
Dont be the last guy standing doing 4:3 Analog video productions.....
Sun Oct 29, 2017 5:54 pm
We are both hardware / software developers in our field.. I understand the concept of NDI and IP..