I have similar experience with DeckLink Mini Recorder inputs: Video stuttering was very bad and was continous after vMix was run for several hours with External Output to DeckLink Extreme 3D output enabled.

I restarted vMix and the stuttering was sill going on. I did exit vMix again and turned power off from the SSD recorder and turned it on again and restarted vMix and all was smooth again.. for a while.



64 bit Windows 7 Pro SP1 + latest updates

Motherboard: Asus Gryphon Z87
Intel Core i3-4130,3.40 GHz
8Gb RAM

Installed cards:
- NVidia GeForce GTX 750 Ti based gfx card with latest drivers

For input:
- 2 x DeckLink Mini Recorder (Using HDMI in)

For output:
- DeckLink HD Extreme 3D (Using SDI out)
- HyperDeck Shuttle 2 -SSD recorder.

DeckLink Driver: 10.6.5

Can't remember exatly due screen tearing on vMix small preview windows but the stuttering was definetly seen on External output after enabled it.

I have been silently working away at a very similar set of problems with DeckLink dropped frames and thought it a good time to share my experiences. I have a DeckLink Duo and a DeckLink SDI and a bars and tone MP4 I made in Premiere with a clapper in the screen I can very easily see when the video stream is dropping/rebuffering/jumping around/whatever... First off, I can see the first half second of a clip stutters when played in vMix when Loop is enabled, it does not cleanly jump back to frame 1.

Moving on to external output, I started with the Duo, sending Program out port 1, preview out port 2, within 24 hours on port 1 the video would still look OK, but audio would play for 5 seconds or so, then silence for 5 seconds, then sound again. The sound itself would be out if sync with the video by a second or two. Interestingly, port 2 didn't have this dropping audio issue, but A/V sync would still be a problem. Having genlock hooked up to an AJA sync generator didn't seem to make a difference one way or the other. I would also occasionally hear minor stuttering of the tone at times when doing stuff in the vMix interface. I also noticed setting the renderer to VMR9 had less frame dropping than EVR, which was counter intuitive to me, but still some frame drops were still occurring. Fullscreen display rarely drops anything, but on occasion will. I should also mention i was running these tests with audio output in Windows disabled altogether to eliminate, in my mind, the idea of audio hardware/drivers contributing.

Next, I replaced the Duo with an SDI and played the clip loop. Audio/video sync issues and dropped frames on the bars/tone are noticeable at 24 hours, but no audio dropping outright.

Next on my list was to downgrade to Windows 8.1 or maybe 7 and see if conditions were different. I sent a private message to Martin following up the Windows 7 fullscreen bug about which OS seemed more stable, if such a statement could even be made, but never got a response. But so far i'm not super impressed with Windows 10 as a whole. Forced updates and a lack of simple means to control that are a big part of that.

I can give you system specs upon request later if you want to follow up on this writing.

Since you can't recreate these issues in your lab, I would be willing to ship this computer to you at my expense if it was possible to get it to be rock solid and not drop frames/stutter/lipflap. At this point, all the great new features of NDI in v17 aren't useful yet if I put this on a production machine because I I can't have this DeckLink issue happening. I have purchased several licenses in the past just to do this type of testing on.

I also wanted to mention that vMix 17 does change the way it outputs to the decklink card with the goal of making the output smoother as we received a number of reports
of dropped frames when using the decklink output in vMix 16.

If everybody here says it works perfectly fine in 16 then I can easily switch back to that method, but I suspect there is a bit more to it than that....

I am a new vMix user and have held back from contributing to this as I fear that I might have done something silly that contributed to my woes! However after reading this thread its seems that it might be sensible to pipe up so here goes..

I have a vMix setup with the BMD quad 2 on an i7 system with 32Gb RAM, GTX 750Ti with 2GB (the machine also has a monitor connected to the Intel HD2000 port so I can leverage quick sync in Edius), 2X Samsung 840 pro SSD's in RAID 0 for media use and Windows 10. My vMix version is the 4K one which (even though I am just working HD) I purchased specifically so I could use the Multicorder facility for ISO recordings.

The first time I used the system for recording in church, I had installed the first upgrade to vMix 17 (from 16) around NAB time. The system started stuttering and the controls in vmix became unresponsive after about half an hour of recording (system had been running (but not recording) for about 2 hours by this point. At the time I was trying to use multicorder to capture 2 of the camera's with avc mp4 codec at 17Mbps as well as recording the mixed output in MP4 (at 17Mbps) and streaming via Youtube 1080p25 at 3Mbps.

I also had SD card and SXS card (in camera) recordings and also an Atomos Samurai Blade SDI recorder on the output of one of the BMD quad 2 SDI ports.

After noticing the unresponsive behaviour, a couple minutes went by and then the system became responsive again and all seemed well until I stopped recording. At this point I noticed that the multicorder recordings were totally missing, but the MP4 recording of vMix was present. The Atomos recorder had created several files - probably at each discontinuity in the SDI output during the stuttering.

The MP4 recording from vMix was interesting because around the time of the unresponsive behaviour, the vision appeared to slow down and lose sync with the sound.

A fortnight later, I had updated to the 17.0.0.85 release when we next attempted recording at church. This time I didn't bother with multicorder as I wondered if this had been implicated. I simply streamed to Youtube (108025p @ 3Mbps) and recorded the output of vMix as an mp4 file at only 8mbps. I did my main recording on the Atomos Samurai unit via BMD Quad card output and ISO recordings in camera.

This time there were no problems at all and everything worked fine.

I would like to know if it is likely that my issues were because I was trying to do too much at the same time or if something got fixed in the .85 release?

On both occasions I was working with three SDI camera's and an NDI source of Open LP from another PC. All camera's and vMix output to youtube and the Atomos recorder were at 1080 25p

Hope this info is useful!

Andrew

Hi all,

I think I have made some good progress in solving this issue.

1. Download http://www.vmix.com/down.../decklinkplugin27may.zip
2. Extract the DLL inside the zip file
3. Replace the DLL with the same name in C:\Program Files (x86)\vMix\avplugins\decklinkplugin with this new file
4. Run vMix, start External Output and report back your results!

For those curious, the problem might be related to differences between the system clock and the decklink clock.
If one clock fell behind the other, frames may drop rapidly in the decklink driver, even if vMix is showing 0 dropped.
This update should hopefully prevent that once and for all!

Regards,

Martin
vMix

Now after hours and hours of testing the new decklinkplugin27may beta -plugin with vMix 17..85 I have not caught any stuttering with live cameras and decklink output!

Kind regards,
Jan

Yes, the audio was audible and in sync. I used clapper board and examined the footage sync frame-by frame and it was perfect. I will repeat this test today once more.

Kind regards,
Jan

Hi Martin,

I have tested your new dll file for about 2 and a half days now, and I can report some progress. The audio dropouts are no longer happening. The video stuttering also appears to be much reduced. Only thing I am noticing is that right now the audio is about 2 seconds behind the video when checking with my SDI monitor.

I have a single Decklink SDI card with external output sending to this device in Windows 10.

Also, I noticed the External Output duration counter seems to loop back to 0 after some time, maybe this variable should be an Int64 if it isn't already.

This experiment with the timestamps, what am I looking to test from it? I have Premiere CC on one of my systems here I could bring a recorded file into, am I looking to quantify the number of frames the video and audio are separated by? Or is just reading the time code, assuming there is embedded time code,
going to reveal something interesting?

Also, what MultiCorder container would I need to record with?