Hardware to Build a Low Latency VT4 or VT5 machine.

[Chris.]

Hello All,

I am looking for details (ie. exact system specs and Windows settings) for building an extrememly low latency live production system.

We currently have a VT4 system with several frames (5.5 to be exact) of delay between our cameras and live output on a video projector, and we find this quite noticable. Though some of this latency we can attribute to other hardware (ie. camera, upscaler and projector) we are looking for ways to reduce latency in the the VT4 system as well, possibly even upgrading to the VT5 if necessary. The Newtek tech support have noted some people are reporting latency of only half a frame to a frame and we were wondering if some of those people would care to share there setups with us as well as their method of testing the latency, that could help us in our decision.

Basically:
- Hardware components
- Windows Settings (Config files etc.)
- Method you used to Test Latency

Thank-you in advance to all that respond.

[billmi]

Are all of your cameras gen-locked? That more than anything will make the biggest difference.

[Chris.]

Unfortunately at this point they are not.

I do have a television broadcasting background so this was one of the first things I looked into when coming onboard his project. We currently have 4 non-genlocked robotic cameras and will be adding a 5th automatic tracking camera (already purchased and non-genlocked) for when we move into our new building. The new building seats close to 1000 people and we have been asked that the service now be live projected on the stage screen (both music and speaking).

To replace these with genlockable cameras is currently a cost we would like to avoid and is why we are looking at other means to reduce latency. One thing we are looking into is to remain digital from the VT4 through an SDI out card (instead of returning to analogue) this will prevent out video scaler and projector from having to reprocess an anologue to digital signal twice as the current setup is doing so we should be able to shave of 1-2 frames of latency there. However we are hoping to be able to shave some off at the VT4 (or maybe VT5) as well as we are currently losing a fair bit there.

Thanks for your response, I appreciate it.

[billmi]

Making less work for the scalers will definitely help.

However, if you are looking to shave off more latency at the VT end, that's not about what kind of hardware is in the VT's host computer, it's about the sources you are feeding it. If your sources are not gen-locked, then the VT will have to delay the sources long enough to synchronize them (as would TBCs.) That delay gets added on to the couple of frames delay that it's doing just processing the signals.

[Chris.]

You make a good point about the non-genlocked cameras introducing delay. We have done extensive testing on our system and have actually accounted for each part of the 5.5 frames of delay we are getting. We found 2.5 frames of the latency is introduced by the VT4 itself, this if after it having to sync all the cameras.

Newtek tech support has informed us that computer hardware CAN make a big difference, so we were wondering what others have found with their systems.

[bbeanan]

If the VT4 is only adding 2.5 frames the only way to get it better than that would be genlock. From what I have heard VT4 will need approx. 2 frames to sync the cameras.

So maybe by doing some hardware changes you may be able to get a .5 frame better but the real key would be gen-lock.

Maybe you could borrow or rent some higher end cameras to do a test for the boss types then make your case for pro cameras.

Keep in mind you will have this problem with ANY switcher by any company since they will all need a frame or two to sync.

[IMPERIAL]

I dont know... i used VT3 a lot for live switching (3-4 cameras + recording to HDD in DV)... never had any latency...any noticable anyway...and i used it on
Athlon 3200+
1GB ddr400
160GB HDD sata
MSI platinium booard with 939 slot. (Nvidia chipset)

Are you using analog inputs or firewire?

[bbeanan]

Imperial... 98% of people would never notice a 5 frame delay 50% of people will not even notice a 15 frame delay.
Your mind has a cool way of re-syncing after watching it for a few minutes... But then some of us get trained and we are screwed and can never watch an out of sync video again.

It is the technical types around here that notice it and drive us nuts! But even a 2 frame or even a 4 frame delay the tech types need a camera to actually see the delay by playing back the footage.


PS Here are a few other things that drive me crazy...
When you are at a movie theater next time keep your eye on the upper right hand corner. About every 20 min. you will see a black dot flash up telling the operator to switch reels (unless your local theatre has gone digital)

Also notice that the image is actually jumping up and down like crazy best noticed by watching the top edge (which is why they normally put flat black boarders around the screen to help make it disappear)

Never get trained on proper video calibration… you will find yourself walking into your friend’s houses fixing their TVs because even though you are not watching it and it is in the background you find it impossible to let it sit there with the Saturation way to high the Brightness way to high and just overall messed up (TV are shipped set so they look brighter, not better)

Oh there are so many more that if you never are told about you go along life without problems… then one day someone shows you and you are screwed… (never I mean NEVER learn what goes into a hot dog!!)

[IMPERIAL]

But even a 2 frame or even a 4 frame delay the tech types need a camera to actually see the delay by playing back the footage.


Delay for me means that, video output is shown later than pressing TAKE button. (how fast it switch channels). So how can you see the delay if you play back the tape? ... unless you map a sound to a button or something....wich sounded good idea.. and i maped sound to my "take" ENTER key wich is used for "take" (i dont have RS-8).

Then recorded the output ... there was at most 2 frames delay while switching cameras , not always.. most of the time it was in same frame.

however.. if you dont put next output in preview bus first and then switch...then you have jumping and delay... unless you have genlock cameras i guess.. i never tried that. For me.. if it dont jump..its around 14 frames delay....not always but most of the time.

[billmi]

Imperial, I think we are talking about two different kinds of delay here.

Video delay - the delay between video coming into the VT and going out.

And switching latency - the delay between the time you press take, or hot-punch a source on the program row and the time it switches on the output.

Those are very different issues.

[billmi]

We have done extensive testing on our system and have actually accounted for each part of the 5.5 frames of delay we are getting. We found 2.5 frames of the latency is introduced by the VT4 itself, this if after it having to sync all the cameras.

Newtek tech support has informed us that computer hardware CAN make a big difference, so we were wondering what others have found with their systems.

Based on the sum knowledge in the forum, I don't think you're going to be able tighten up much past 2 to 2.5 frames delay.

Something worth noting, though, if you are using this in a 1,000 seat venue depending on your audio arrangement, you will probably need to have some video delay to keep things in synch for the folks in the back rows, who are relying on the video the most. Every 36.7 feet that the listener is from the speakers, adds a frame of delay to the audio track due to the difference between the speed of light and the speed of sound.

A 5.5 frame delay may be noticeable when you are watching the screen and a gesturing speaker at the same time (this is VERY common in IMAG systems even with huge-high-dollar set-ups - latency is a fact of life with digital processing - just watch a high-end concert on TV and you'll see it) or it will seem like a lot if you are in the control room watching the output screen and listening on headphones.

To someone sitting 200 feet back from the prime speakers (in a proper sound speaker placements further back in the venue will be driven with delay boxes, otherwise they create an echo) everything will be in perfect synch. To someone sitting even further back, the video will be going ahead of the audio.

[Chris.]

Every 36.7 feet that the listener is from the speakers, adds a frame of delay to the audio track due to the difference between the speed of light and the speed of sound.

Thanks again billmi, we had a very brief discussion about the speed of sound traveling and how it may be to our benefit but weren't sure by how much (and I never did look it up online :) ), it's nice to see some actual working numbers with that, I'll have to see again what the dimensions of the sanctuary are going to be to see exactly how much this will affect our situation. Thank-you, I really appreciate your advice.

I was really hoping I could knock the total system down to about 2.5 to 3.5 frames of latency from camera to projector but I am really only seeing about 1 frame or so that we can try to knock off without a real cost investment unless of course someone knows of reasonably priced genlockable robotic camera? :)

[Blairness]

Maybe I'm missing something, but won't genlocking the sources only reduce each source's latency by 0-1 frame?

Having said that, 0-1 frame less latency is better than 0 frames less latency. I just think genlocking is getting pushed as a solution more than it should be, especially when there is 5.5 frames of latency.

[Jim_C]

What has your experience shown you with genlocked and non-genlocked cameras and a VT system? What solution do you have?

How do you push a proven solution 'more than it should be?'
Should we be counting how many times people suggest to genlock cameras and make sure we don't go over a given quota?

[Chris.]

We will not be pursuing the genlocked cameras at this time due to our research showing the cost of genlockable cameras to be quite costly (Unless someone knows of a relatively cheap good robotic camera with a genlock feature.).

We will be testing out our VT4 system with a SDI out card and a TVone scaler with SDI in and going DVI to our projector. We are hoping to shave some latency by not going analogue through this step like we are currently doing.

As soon as we get this tested I'll let you know of the results.

[jimhouston]

You may be able to use frame syncs to genlock each camera signal. I bought a Hotronic 41 on ebay for $100.00. I use it to sync up a JVC 500 to the CCU controlled and synced JVC 550s. You can adjust color and black and video levels with the frame sync too.
Jim

[Quiet1onTheSet]

[QUOTE=IMPERIAL;708750]I dont know... i used VT3 a lot for live switching (3-4 cameras + recording to HDD in DV)... never had any latency...any noticable anyway...[Quote]

As was stated earlier, given today's technology, any digital mixer induces some latency -- so yes, you've got latency, and there are ways you can detect the amount of it, as well, Imperial.

It would appear the original poster would do well to look into a cost-effective camera genlocking paradigm, that'll relieve the VT system of that responsibility, but he'll only gain about .5 frame advantage then, if the info he's supplied us is accurate.

Q1

[Quiet1onTheSet]

You may be able to use frame syncs to genlock each camera signal. I bought a Hotronic 41 on ebay for $100.00. Hmmmm. How much delay do you suspect that Hotronic 41 saved you?

Q1

[billmi]

You may be able to use frame syncs to genlock each camera signal. I bought a Hotronic 41 on ebay for $100.00. I use it to sync up a JVC 500 to the CCU controlled and synced JVC 550s. You can adjust color and black and video levels with the frame sync too.
Jim


Whether you have the signal from asynchronous cameras brought into synch by the VT's internal frame synchs, or by an external frame synch, you're still going to be introducing a delay to the camera's signal.

The point of using genlocked cameras is that they generate their signal already synchronized, so that it does not need to be delayed to get it in synch.

[johnlacko]

We will not be pursuing the genlocked cameras at this time due to our research showing the cost of genlockable cameras to be quite costly (Unless someone knows of a relatively cheap good robotic camera with a genlock feature.).

We will be testing out our VT4 system with a SDI out card and a TVone scaler with SDI in and going DVI to our projector. We are hoping to shave some latency by not going analogue through this step like we are currently doing.

As soon as we get this tested I'll let you know of the results.


Just my $.02...but why don't you skip the whole TVone scaler and use your output 2 (assuming you have a decent dual head graphics card) and go directly from your video card to projector.(also assuming you have an DVI graphics card)