logo

Live Production Software Forums


Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

Options
Go to last post Go to first unread
madido  
#1 Posted : Thursday, February 21, 2019 6:38:13 PM(UTC)
madido

Rank: Advanced Member

Groups: Registered
Joined: 1/13/2016(UTC)
Posts: 32
Location: Italy

Hello,
I'm doing some tests to evaluate vMix as cheaper alternative to Tricaster for an all NDI system.
Guided by other users I discovered that vMix has some issues with NDI; in particular regarding latency.
Compared to other solutions (I tried OBS and VidblasterX) there is a couple of frames of delay in input.
And there is also an additional delay on output, even when project and input have the same resolution/fps (no scaling involved).
I do a lot of IMAG and this can be a show-stopper for me thinking about vMix as core of transition to video over IP that I will do this year.

Regards

Stefano
admin  
#2 Posted : Thursday, February 21, 2019 9:10:08 PM(UTC)
admin

Rank: Administration

Groups: Administrators
Joined: 1/13/2010(UTC)
Posts: 5,139
Man
Location: Gold Coast, Australia

Was thanked: 4136 time(s) in 1487 post(s)
Hi Stefano,

vMix uses a small buffer to ensure all live video inputs come in smoothly.
Without this, the sources would have lower latency, but would drop frames a lot due to network jitter or PC system load.

That said, the difference is only 1 to 1.5 frames at the most compared to other solutions that don't use a buffer.

In cases where vMix has a significantly higher latency than this, it is likely due to a configuration issue and would encourage you to
drop us an email to investigate further.

Regards,

Martin
vMix
admin  
#3 Posted : Thursday, February 21, 2019 9:22:17 PM(UTC)
admin

Rank: Administration

Groups: Administrators
Joined: 1/13/2010(UTC)
Posts: 5,139
Man
Location: Gold Coast, Australia

Was thanked: 4136 time(s) in 1487 post(s)
Just for fun I ran a test comparing the latest vMix and OBS releases.
(coincidentally both version 22!)
This is a simple loopback test, NDI out to NDI in using a video clip with printed timecode.

I didn't expect vMix to have a whole 2 frames better latency actually, but there you have it..

vMix (3 frames)
vmix-ndi-test.jpg (126kb) downloaded 5 time(s).

OBS (5 frames)
obs-ndi-test.jpg (132kb) downloaded 1 time(s).

zenvideo  
#4 Posted : Thursday, February 21, 2019 11:53:48 PM(UTC)
zenvideo

Rank: Advanced Member

Groups: Registered
Joined: 5/13/2014(UTC)
Posts: 514
Man
United Kingdom
Location: Manchester, UK

Thanks: 2 times
Was thanked: 183 time(s) in 130 post(s)
For me, it's a bit hard to imagine what that's actually measuring, given that the displays appear to be created within the applications under test. I would rather see a test where the the applications (OBS & vMix) are used to process an NDI stream (from NDI in to NDI out) but with the latency measuring being entirely external. That way you can start with a reference setup and then selectively add the application under test to see what changes.
madido  
#5 Posted : Friday, February 22, 2019 12:53:38 AM(UTC)
madido

Rank: Advanced Member

Groups: Registered
Joined: 1/13/2016(UTC)
Posts: 32
Location: Italy

My test was made with vMix, OBS and VidblasterX all receiving an NDI stream from VLC on the same PC.
I'm not in front of my PC, but later I will post a picture of the setup

Stefano
admin  
#6 Posted : Friday, February 22, 2019 1:22:51 AM(UTC)
admin

Rank: Administration

Groups: Administrators
Joined: 1/13/2010(UTC)
Posts: 5,139
Man
Location: Gold Coast, Australia

Was thanked: 4136 time(s) in 1487 post(s)
Originally Posted by: zenvideo Go to Quoted Post
For me, it's a bit hard to imagine what that's actually measuring, given that the displays appear to be created within the applications under test. I would rather see a test where the the applications (OBS & vMix) are used to process an NDI stream (from NDI in to NDI out) but with the latency measuring being entirely external. That way you can start with a reference setup and then selectively add the application under test to see what changes.


It's a full NDI send/receive loopback test.

1. Play a video of Timecode through the apps NDI output (send)
2. Bring the NDI source back in via an NDI input (receive) and compare.

This eliminates any other potential causes of latency and determines with good accuracy the internal processing delay of the application + NDI SDK.
Any other latency observed can then be determined to be caused by something external to the application.
thanks 1 user thanked admin for this useful post.
sinc747 on 2/22/2019(UTC)
RichDanby  
#7 Posted : Saturday, February 23, 2019 4:46:21 AM(UTC)
RichDanby

Rank: Advanced Member

Groups: Registered
Joined: 2/23/2019(UTC)
Posts: 129
United Kingdom

Thanks: 14 times
Was thanked: 26 time(s) in 24 post(s)
I just ran the some tests here and got broadly the same results
Code:
+---------------------+-------+------+
| ROUTE               | 29.97 |  60  |
+---------------------+-------+------+
| vMix -> vMix        |   4   |   4  |
| OBS -> OBS          |   5   |   6  |
| vMix -> OBS -> vMix |   9   |  10  |
| OBS -> vMix -> OBS  |   9   |   9  |
| vMix -> Studio Mon  |   5   |   5  |
| OBS -> Studio Mon   |   9   |  10  |
| VLC -> vMix -> SM   |  4(6) | 4( 5)|
| VLC -> OBS -> SM    |  7(9) | 6(10)|
+---------------------+-------+------+

For the "VLC -> ?? -> Studio Monitor" tests the first figure is the difference compared to another instance of Studio Monitor connected straight to VLC. The figure in brackets is the difference between the OBS/vMix program view and the Studio Monitor picking up that output (so VLC -> SM has more of a delay than VLC to OBS or vMIX, but I also think the SM delay might be more variable so probably not that reliable a test).

There is a bit of variance but I think just down to when the windows update and the screenshot captured, which is more noticeable at 29.97

It was all looking like vMix having less delay than OBS for NDI input and output but then I thought I should try vMix -> OBS and OBS -> vMix. Testing at 60p vMix -> OBS was 1 frame delay and OBS -> vMix was 9 or 10. Combining that with the other results it seems like OBS might have a fast input and slow output and the opposite for vMix. It would be interesting if someone else could do test along those lines.

It could just be running both apps on one machine/GPU is confusing things - or it could be my error. I will try to find time to redo it on another machine with a better GPU and between the two machines.

I think the machine may have been suffering a bit as on one screen shot of vMix -> OBS it shows these frame numbers
vMix input : 18
vMix preview : 14
vMix program : 13
OBS preview : 16
OB program : 14

Back to the original question, I believe the other benefit of Tricasters is they can handle more sources on a network as they use the low bandwidth streams and switch instantly to the full stream when a source goes live. Other NDI users face a delay switching streams so have to always consume the high bandwidth streams.
JoseL  
#8 Posted : Saturday, February 23, 2019 6:20:22 AM(UTC)
JoseL

Rank: Advanced Member

Groups: Registered
Joined: 4/15/2018(UTC)
Posts: 64
Man
Location: Spain

Thanks: 6 times
Was thanked: 19 time(s) in 13 post(s)
Why not an option on preferences for “low latency ndi”. Like capture cards.

Is a fact that ndi is not ok actually for imag, if we can have a choice to down 1.5frame ndi latency capture, it will be fantastic.

OBS has less latency on ndi capture. And maybe more latency on ndi output. But for imag we use hdmi output, not ndi.

Please give us the choice to use “smooth ndi” or not.

Thanks!

Jose.
ask  
#9 Posted : Saturday, February 23, 2019 2:23:28 PM(UTC)
ask

Rank: Advanced Member

Groups: Registered
Joined: 10/13/2012(UTC)
Posts: 1,161
Australia
Location: Melbourne

Thanks: 220 times
Was thanked: 199 time(s) in 181 post(s)
What you are describing is all ready there. It is the Low Bandwidth option that you can select when choosing an NDI input. NDI itself does not have a low-latency mode. It is designed to have the smallest latency possible, which is around a frame. Further latency will be added by you network and the number of hops the data signal goes through, the efficiency of the switches and routers you are using, your NIC, the CPU, and how your software processes the NDI signal on your PC.

Martin (Admin) has given a good guide as to how to test in a closed loop on your PC. For more complete testing you need to simulate the entire network you intend to use in the field.
JoseL  
#10 Posted : Saturday, February 23, 2019 5:26:15 PM(UTC)
JoseL

Rank: Advanced Member

Groups: Registered
Joined: 4/15/2018(UTC)
Posts: 64
Man
Location: Spain

Thanks: 6 times
Was thanked: 19 time(s) in 13 post(s)
Better to share a pic. Format is 50P, from camera to decklink quad and to proconvert ndi.

Vmix decklink input: tc: 00:01:12
Vmix ndi proconvert: tc: 00:01:10
Obs ndi proconvert: tc: 00:01:12

Both obs and vmix opened at same time side by side on same computer and monitor. Obs has 2 frames less latency from proconvert ndi input on the monitor than vmix.

FF8CD422-3D77-4DF9-9651-C780047BF9F0.png (3,774kb) downloaded 43 time(s).
Users browsing this topic
Forum Jump  
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.