It is currently Sat Jul 22, 2017 12:38 pm Advanced search

Video Delays QOD+ output vs input

Report unexpected behavior of QTAKE

Video Delays QOD+ output vs input

Postby bbvcbob » Tue May 09, 2017 1:47 pm

I’ve been upgrading my rigs to v1.3 and running some tests to try and quantify the video delays throughout the systems. Everything was much as I expected until I compared the outputs of the QOD+ device on one of the rigs with the old Matrox DH2G/Geffen combo on another.

With two identical monitors (Sony PVM-A250 OLED’s) I viewed an image with burnt in timecode direct from the source on one monitor and the output of my rig on the other. I recorded the monitors using an iPhone set to record slo-mo at 1080 120fps - this allowed me to detect delays to an accuracy of 0.2 of a frame.

With the old Matrox/Geffen rig I was seeing a 1.2 frame delay between direct and the GPU out. This was with a 1080P 24fps input signal

With the rig using the QOD+ device, the delay was 4 frames with 1080P 24fps signal.

Changing the input signal to PsF, the delays were 1.6 frames and 4.6 frames respectively.

I reset the QOD+ and reinstalled the firmware (01.03.86 DPv1.2) but the results were the same.

I also tried setting the output format of the QOD+ away from the default 'Auto' to match the format of the incoming signal but there was no change to the delay.

My questions are;
1. is there a problem with my QOD+,
2. is it possible that I have it configured incorrectly
3. is this all correct and I am expecting too much of the QOD+

Both of the tested systems are built around the Apple Mac Pro Tower Desktop Xeon E5 3.5GHz, 16GB RAM, 256GB SSD, FirePro D500 Dual GPU OS Yosemite 10.10.5 with 2x Kona 3G cards. Distribution is via BMD Smart Video Hubs. Both systems running the latest v1.3 Qtake software. The rig with the QOD+ device fitted was run in both HDx2 and HDx4 (the results were the same in both modes) - the other rig is HDx2 only.
bbvcbob
 
Posts: 151
Joined: Tue Apr 19, 2011 8:07 am

Re: Video Delays QOD+ output vs input

Postby switcherdude » Tue May 09, 2017 6:41 pm

I am not sure how you had a 1.2 fr input-output delay on the gefen as it takes 2 frames to just get it into qtake even with the kona cards
switcherdude
 
Posts: 43
Joined: Mon Feb 09, 2015 12:08 am

Re: Video Delays QOD+ output vs input

Postby switcherdude » Tue May 09, 2017 8:06 pm

One thing to try is to make sure the settings are the same one both monitors.
Make sure you can see the incoming frame rate displayed n the monitors.
Start with a 60 frame source signal so there are more frames to work with(60P or 60i)
Use a 60 frame timecode burn in so you can calculate subframes of 24 frames after
The Gefen will always output 60i anyways hence one of the reasons QOD was developed
Set the QOD to output 60i or whatever the Gefen is outputting.
Now the systems should have the same input output parameters.
If you shoot 120f on the iphone keep in mind every frame will just be repeated as you are only supplying a 60 frame source.
Try to convert frame delay to milliseconds as it saves the confusion of knowing which frame rate we are talking about.
(1 frame @ 60 fps = 16.66 ms, 1 frame @ 24fps = 41.66)
If after doing this you still get the 1.2 fr delay ,then I will BUY your Gefen system.
My biggest beef with Qtake has always been the input output delay . It makes it more difficult to do Live mixing /keying with precise cues and dialogue, GPI input triggering , and Color Grading within Qtake when the DOP is used to minimal delay from the camera.
My Qtake systems with the QOD + at best are 4 fr @ 24 (166.64 ms) in to out.(using the pulldown which adds a frame)
switcherdude
 
Posts: 43
Joined: Mon Feb 09, 2015 12:08 am

Re: Video Delays QOD+ output vs input

Postby bbvcbob » Wed May 10, 2017 1:58 pm

Hi switcherdude, thanks for your reply.

"I am not sure how you had a 1.2 fr input-output delay on the gefen as it takes 2 frames to just get it into qtake even with the kona cards"


I re-ran my tests to check this, just in case I had misread the results or was comparing the wrong streams but, no, the results were the virtually the same (new tests showed a 1.4 frame (58.33mSec) delay.) Are you thinking of the added delay through the camera? I was feeding the output of another Qtake system with a file I generated in After Effects that has a 24fps TC ‘burn in’ so there are no added delays.

One thing to try is to make sure the settings are the same one both monitors.
Make sure you can see the incoming frame rate displayed n the monitors.


Yes, I did that - and used two monitors of the same make and model. Different monitors have varying reaction times dependent on how they process the incoming signal. This can be particularly noticeable when comparing processing times between P and PsF signals. Having said that, when testing the GPU outputs I had ‘live’ monitor at 1080P 24fps and the GPU out monitor (obviously) at 1080 60i as this is how it works in real world situations.

Use a 60 frame timecode burn in so you can calculate subframes of 24 frames after


As above, I’m trying to judge real world delays at the frame rates I use on set, rather than under false test conditions. Although I’m only quoting the 24P figures here I also tested at 23.976, 29.97 and all speeds at PsF. All the tests had roughly the same results.

If you shoot 120f on the iphone keep in mind every frame will just be repeated as you are only supplying a 60 frame source.


As I'm showing the two images on adjacent monitors I can see the point at which each monitor feed switches between, say, 19:23 and 20:00. So when the 'live' monitor makes that transition, I can count the number of frames on my slo-mo playback (5 for each 'realtime' frame) until the output feed makes the transition.

Try to convert frame delay to milliseconds as it saves the confusion of knowing which frame rate we are talking about.


In my experience video delays are usually expressed in terms of 'Frames' and audio delays are expressed in milliseconds but I take your point and have quoted both. As I said, the figures I'm quoting here are for 24fps Progressive only.

My Qtake systems with the QOD + at best are 4 fr @ 24 (166.64 ms) in to out.(using the pulldown which adds a frame)


I hadn’t thought about the pulldown option - I’ll test again with the option disabled. As you are experiencing the same delay on your QOD+ output, I think you've answered my question, and that this is the normal delay through the QOD+.

Thanks again.
bbvcbob
 
Posts: 151
Joined: Tue Apr 19, 2011 8:07 am

Re: Video Delays QOD+ output vs input

Postby switcherdude » Wed May 10, 2017 7:48 pm

I re-ran my tests to check this, just in case I had misread the results or was comparing the wrong streams but, no, the results were the virtually the same (new tests showed a 1.4 frame (58.33mSec) delay.) Are you thinking of the added delay through the camera? I was feeding the output of another Qtake system with a file I generated in After Effects that has a 24fps TC ‘burn in’ so there are no added delays.

No , not thinking about camera delay , just input to output. I am envious that you are getting 1.4 frame input to output.
From Day 1 the lowest I have achieved was 3 frames @ 24P . Ingest through Aja IOXT ( 2 frame) and output through GPU (1 frame). Maybe the Kona 3G ingests faster.What kind of expansion box are you using? Since your QOD tests are on par with mine ,I see no problem with your testing.I have a couple Kona 3G cards so will try to duplicate your test in the next week or so.

Yes, I did that - and used two monitors of the same make and model. Different monitors have varying reaction times dependent on how they process the incoming signal. This can be particularly noticeable when comparing processing times between P and PsF signals. Having said that, when testing the GPU outputs I had ‘live’ monitor at 1080P 24fps and the GPU out monitor (obviously) at 1080 60i as this is how it works in real world situations.


So was the QOD ouputting at 60i or 24P ? It should be a 3 fps delay for 60i output and 4 fps delay for 24P output(pulldown enabled)


As above, I’m trying to judge real world delays at the frame rates I use on set, rather than under false test conditions. Although I’m only quoting the 24P figures here I also tested at 23.976, 29.97 and all speeds at PsF. All the tests had roughly the same results.


Your right. I was just trying to minimalize counting of duplicate frames.

Are you able to test the ingest portion of the delay by recording "live" monitor and GUI screen with View 1 or 2 in full screen?
I understand they are different monitors but would be interesting to know the initial delay coming into the system.
Are you adding any audio delay in the Project settings of the Gefen system ? I have always set at 2 frames to match the input delay that I am getting on my QOD system
switcherdude
 
Posts: 43
Joined: Mon Feb 09, 2015 12:08 am

Re: Video Delays QOD+ output vs input

Postby switcherdude » Wed May 10, 2017 8:42 pm

I know you are probably doing everything right but can I ask you to flip or zoom the image in DVE in Qtake and do the test again .
Specifically for the Gefen system.
So u should have Live feed on one monitor and zoomed or flipped feed passing through Gefen system on the other monitor.
Thank You
switcherdude
 
Posts: 43
Joined: Mon Feb 09, 2015 12:08 am

Re: Video Delays QOD+ output vs input

Postby bbvcbob » Fri May 12, 2017 12:05 pm

Maybe the Kona 3G ingests faster


My portable rig with IoXT’s is currently out working so I can’t test the throughput on that for the moment. BTW, I don’t know if it makes any difference but I have the latest firmware & drivers for the AJA’s - and it may be worth checking which firmware option (4K or not) is loaded - I think (but can’t check just now) the HDx2 system with the Geffen is set for non-4K and the HDx4 with QOD+ is 4K

So was the QOD ouputting at 60i or 24P?


It was set either on ‘Auto’ or I manually set it to output the same format as the input - there didn’t appear to be any difference in the delay.

Are you able to test the ingest portion of the delay by recording "live" monitor and GUI screen with View 1 or 2 in full screen?


but can I ask you to flip or zoom the image in DVE in Qtake and do the test again .


I won’t be back in my workshop until the end of next week but will setup these tests as soon as I can

Are you adding any audio delay in the Project settings of the Gefen system?

I use the Focusrite Scarlett 2i2 USB box for the audio in/out and I have found that, with all the Qtake settings at 0 delay, the throughput in Live/Record mode is 28.8ms (~2/3 of a frame) so, if you assume that the Camera is giving around a 1.5 Frame delay, your 2 frames is just about correct - the odd 8mS or so delay being unnoticeable.
bbvcbob
 
Posts: 151
Joined: Tue Apr 19, 2011 8:07 am

Re: Video Delays QOD+ output vs input

Postby bbvcbob » Wed May 24, 2017 6:12 pm

I have now had time to double check my results and to perform the additional tests as requested and have put together two documents:

The first document (Monitor Delays) compares the inherent delays in some of the monitor models I have in my hire fleet - this is interesting because it shows that some monitors are better at handling 'P', others PsF.

The second document (Qtake Delays) compares the input to output delays of the various SDI feeds available on my Qtake rigs.

These tests are not meant to be definitive, they were made under real-world conditions to reflect the way the kit is used and configured on set and all the results are subjective.

The two documents can be found here: 

https://www.dropbox.com/s/kq049rsrq62p2we/Monitor%20Delays.pdf?dl=0
https://www.dropbox.com/s/5qak9cl4layqrdp/Qtake%20Delays.pdf?dl=0
bbvcbob
 
Posts: 151
Joined: Tue Apr 19, 2011 8:07 am


Return to Bug reports

Who is online

Users browsing this forum: No registered users and 1 guest

cron