Help, where is this latency coming from ?

[isophase]

Hi all,
not sure if this is the right place to post this?
So the setup is taking place in an art gallery for an exhibition with sound and video. There are 6 rooms in which exactly the same program is being played and so there is only one source (video file)
Here's the deal.
The program is played with a minimac through HDMI output, then feeds an HDMI active splitter (x8 but only 6 outputs used), then feeds x6  little HDMI over Ethernet/RJ45 converter, then the  signal goes to an Ethernet switcher that distributes the signal "in wall" through RJ45 cables to the different rooms (25 meters is the longest length ) in each room there is a pair of powered audio monitors and a projector for the diffusion of the audio and video program. In each room the program is converted back to HDMI using another "cheap" little converter, then that feeds yet another "cheap" little convertor that will covert HDMI to HDMI and extract the audio from that and convert it to analog 
To my great surprise, this setup seemed to be working good when i started to do some listening tests yesterday however, as one might imagine i soon realized there where some latency problems.
The latency i am trying to describe here is more about the audio latency that develops between each recievers/rooms rather than the latency between the image and the sound of the video, even if that may also be a concern later on.. (the films are rather short and abstract and the music and sounds that plays along does not have to be perfectly in sync)
The rooms are not totally closed and in fact quite "open" so there is some sound spill all over the gallery. In simple words, the intention here is to fulfill the whole gallery with some sound...
I did not have so much time to do some listening tests yesterday but this is what i discovered. after about 20 minutes of playing i can hear an echo (which should be around 50ms, haas effect) and the only "quick" fix that i found is to unplug the main HDMI cable feeding the splitter and re-plug it hot, this is putting the sound back in phase again. i guess until some latency develops again but i didn't have enough time to test more...
So can one of the experts here tell me where in the chain the latency is happening?
Can someone confirm that this setup is not good and will not work even if we test with less "cheap" converters?
thank you

 

[joaquins]

50ms is about 17m at the speed of sound, I don't understand well but if you are hearing from the other room those 50ms could be easily from this, I don't know the distance between each room, but at least a figure to have into consideration.

If you want to get the sound at the same time in each module is usually the best way to go to avoid different path, you make a distribution and each path goes exactly through the same amount of converters, usually only one. Daisy chaining converters may introduce data corruption and unpredictable delay, depending on their design. If you have all identical converters and all identical signal path (except the wire length that connects them which won't be a problem in audio timing) you shouldn't have latency problems, they all should match pretty closely. If you have different converters and daisy chaining between them you can't know what's going to happen.

JS

 

[totoxraymond]

Hi,

You're problem may come from clock errors. There is no perfect clock and over time you can get this kind of delay. Especially when working 'over ethernet'.

Try to drive your audio systems from one source (but you may lose synchronization with video). Or, better, use SDI rather than ethernet to feed your video and audio data (can support long cable lengths). Problem is HDSDI converters are not cheap.

Thomas

 

[isophase]

Hi,
thanks for your replies.
So i did some more testing today and apparently this setup is really sh%#ting some bubbles...
The latency is random, i had made a quick audio file loop with snare and bass drum in order to make the latency stand out more clearly, the latency is random in the sense that after only 3 minutes of listening this morning the latency was well over 50ms already, hot unplugging and replugging the master HDMI cable still puts the system back in phase for a while.
Also there is NO ethernet switcher in the signal path as mentioned in my previous post (mistake) and a decision was made to not use all the rooms/spaces so that makes it 4 systems in total comprising of 4 projectors and 4 pairs of speakers.
Signal path is:
Minimac HDMI output  > HDMI splitter  > x4 HDMI/RJ45 converters > x4 RJ45 cables each into > RJ45/HDMI converters into> HDMI/HDMI converter with analog audio out > stereo powered speakers.

There clearly is some jittering/wow and flutter happening because i can hear the latency also reduce at some point (no wonder why it was a bit weird the other day when i was trying to fine tune the levels, sound appeared too loud on one side and after adjusting.. still appeared to loud on the same side haha!)
Yes the speed of sound, i forgot to say that all critical listening was done while standing exactly between two speaker sets fed from two discrete devices/receivers. in fact, two of the rooms are quite "open" in front of each other and the speakers have been installed in the corners so its easy to hear the effect excluding the extra speed of sound parameter from the equation.

I am not personally responsible about this setup, my role in all this was to prepare all the audio files for the exhibition thats all.
I did ask the gallery about the sound system and sound diffusion setup and the first question i asked when they started to mention HDMI over Ethernet was about latency.
I was guaranteed  by them that this setup worked fine and the video tech also confirmed it.
Funny thing is that all the gear is brand new and freshly installed,  all the small converters are all the same and no brand is written on the units..

Anyway,
I know for sure now that this setup with multiple receivers doesn't work for critical sound diffusion, but it got me curious if the quality of the converters would make a difference if they were better grade like some people suggested or if its just bullsh#t?
I don't see how any latency could be induced from the HDMI splitters so I suspect that the latency is happening the most in the small HDMI converters. I don't know much about HDMI and Ethernet protocols and i should read about them (i know) .
I can't help but think that a similar setup with AES signal instead of RJ45 would work properly even if using some cheap DAC at the other end. i'm sure it wouldn't be as much catastrophic as this.

 

[joaquins]

Ok, I like the clock drift idea. I don't know about the protocol with HDMI over ethernet but if there's no word  clock distributed there you have a problem, better converters could and should but may not have a better clock, which means you have lower drift. Let's say you pay twice as much and you get twice as better clocks so they drift half as much, instead of working fine for 3 minutes they will work fine for 6 when you will need to unplug and plug them again, as you used to do... Not much a solution I guess.

How scattered is all this setup? I mean, as many converters as used here may not be cheaper than analog installation wire, which with you'll never loss sync with the distribution, for the video you still have the issue, but that won't be a problem for your case since you don't need audio/video sync as much as you need audio/audio sync.

A clock distribution between all the converters may be used but probably they don't have that capability built in since cheap stuff is usually intended to use from your computer to a media center or something like that, not designed for sync'ed distribution...

With VGA and 100V speakers lines this things never happened!

JS

 

[isophase]

Oh yes, since the beginning i said and suggested the easiest/safest way to do it was an all analog setup. And as of yesterday, after three days that i keep telling them that their setup doesn't work for the sound, they decided to change the setup to full analog... its funny how many people actually think the analog setup would not work because of the different cable lengths and they think the latency will be worse, haha!
So now the "tech" from the gallery has ordered an active splitter and some cable lenght, and apparently they're planning to install it the day before the exhibition opens 
i asked about the splitter but they don't know the brand or model and don't have any info about it.. great!
So now lets see if there will be some terrible ground loops!
i don't like to be mean, but really this situation a quite funny