BBC FM latency experiment

[CambridgeWorks]

I remember years ago visiting a microwave link repeater station that carried the Eastern Radar (Watton) "pictures" to Midland Radar at North Luffenham and being told that the digital network was so slow 2 jet aircraft would have probably collided before they saw the danger on screen and had time to warn them!
Given what was said earlier in this thread and how slow digital was decades ago, it was probably true?
Rob

[cmjones01]

Originally Posted by paulsherwin

Yes, that's what you'd expect. There is much more buffering and processing involved with consumer digital video and audio codecs, which are effectively streamed rather than 'broadcast' in the traditional sense. You can expect a delay of several seconds, though the professional encoders used by the broadcasters are much faster than the decoders in domestic equipment.

There's also the issue that digital compression of audio and video signals are very different processes. Video compression of a live TV signal requires, at a basic level, looking for changes in the picture and detecting similar elements in successive frames and how they've moved. This means that it has to introduce a delay - in order to detect and analyse changes in the picture, it necessarily has to look at a lot of frames. Because the frames are of the order of 20ms apart, that takes time. The more frames you do this process across, the better the quality of the resulting compressed video. The high compression levels used in broadcast TV may do this process across tens of frames, so it's easy to end up with a second or more of delay. That's before you've added any buffering to allow for temporary loss of signal and so on.

Audio compression doesn't generally use this sort of technique, concentrating more on what's audible to the human ear within a frame of audio data which may be as little as 1ms long (in NICAM). So there's little or nothing to be gained in terms of compression or transmission quality by taking more time over it.

Chris

[m0cemdave]

I have noticed how the situation of audio and video latency has changed over the years, having spent a lot of time working on conferences and large-scale entertainment events that used video displays and audio reinforcement or amplification.

Back in the late 1980s / early 90s at pop festivals the video was all analogue and almost instantaneous. Although it is natural to see things before we hear them, 40 or 50m from the stage the big screens used to make the performers look as though they were miming, due to the propagation delay of the sound. I remember once suggesting to a video production crew that a delay would be useful, to sync things up in the middle of the audience. They laughed, and told me what a video framestore cost - which in those days was more than enough to make the idea economically unviable.

These days the situation is reversed. Digital audio has low latency but digital video is very slow, the delay being in multiples of frames. The video delay is often much more than that which would usefully synchronise for the audience who are far enough away to rely on the screens for a good view of the show. Hearing sound in advance of picture is not natural, and consequently far more noticeable than the other way around.

This can cause difficulties at conference events where the presenters are all on radio (or top table) mics and appear on screens along with the videos, powerpoints, etc. We would get asked to delay the sound to match the picture, which of course is technically simple now that the audio desks are all digital. But the required delay can be surprisingly large, resulting in the presenters hearing an echo of themselves via the sound reinforcement system (and the room) which can make it very difficult to keep speaking coherently. Being corporate execs, rather than musicians, they don't want to be wearing in-ear monitors. In the worst cases there has to be a compromise between having them look silly on the screens, and having them become hesitant or confused when speaking. And it gets even worse when mics get handed around the audience for the Q&A session...

[G6Tanuki]

Audio latency can be fascinating: it is a big issue in 'communications' applications too; if the audio sidetone you hear in your headset is significantly delayed it can lead to a degree of additional mental-processing overhead you really don't want in a combat situation.

OTOH it can be a useful diagnostic tool; one of my relatives worked as an assessor for claims of hearing-damage in the mining industry; the test kit was a tape-based delay loop, you got the claimant to wear headphones and, talking into a microphone, read a page from a newspaper while the degree of audio-delay was varied.

If you've got good hearing a small amount of delay is tolerable, but lots will soon cause you to go to pieces. If you've got compensation-worthy hearing damage you don't notice the delay even when it's half a second or more. He found lots of 'supposed' claims of hearing damage were spurious.

[mhennessy]

This is a big deal for presenters wearing earpieces or headphones. They are generally given a "clean feed", which should not include their voice, but sometimes that isn't the case, often by accident, sometimes by design.

Hearing yourself can be distracting, but with the right delay it can be disastrous. Rarely a problem in the analogue era, but with digital circuits and data rate reduction codecs, delay is inevitable today. A real problem for systems based on AoIP over WANs, which is why the ViLoR (local radi) project uses a locally generated analogue mix for the presenters.

With this in mind, we test every student who comes through our training. Of course, it's an informal test - just a bit of fun, really, but the learning point is to demonstrate that latency can be a real problem for the presenters and guests that we have to look after.

The test is really simple - we put them in a studio with headphones on and ask them to read something. Meanwhile, we gradually increase the delay to their headphones and listen carefully to how they get on.

We find that people start to notice delay at 10-20ms, but aren't bothered by it. It just sounds like echo. As humans we're used to that - think about the speed of sound and the size of rooms we tend to inhabit.

Long delays - 200ms or more - are also fine. Distracting, but most people can ignore it. That's because the brain is able to rationalise the situation - it's just a delay. There is quite a range in reported annoyance, but it rarely stops people from being able to talk.

However, there is a region in-between these where most people find it incredibly hard to speak. Some just have to stop - others can keep going, but their words are heavily slurred, as if they are seriously intoxicated. This can be anywhere between 80 and 160ms, but it does depend on the individual, because it's related to the way the brain processes hearing and speech particularly. As you increase the delay into this zone, you notice some minor slurring of words before it gets really bad.

But some people are totally unfazed by this, whatever delay you do, and no matter how carefully you repeat the experiment. It's a definite minority - perhaps only 1-2% in our tests. But they exist.

The test subjects are generally young, and have normal hearing. Yet despite this, there is an absolutely huge variation in results. Our test pool might not be massive, but it's more than 100 by now. The variation in our tests is caused by brains, not ears.

As a result of this direct experience, I seriously call into question the validity of the methodology described in #25, and would appreciate references containing more information about the rational behind it, and the weighting given to that test (presumably one of several tests forming the overall conclusion).

[red16v]

Originally Posted by GMB

I think it was interesting that it worked. I am glad that the FM distribution is not adding noticable delay to FM

And at new year I timed the delay between the bong on TV and my Rugby-synched clock as about 11.5secs.

Originally Posted by SiriusHardware

Sorry, I had never heard of that before. Nicam must be quite quick though, I think the 'N.I.' in Nicam stands for 'Near Instantaneous' and it really needs to be, given that it is transmitted along with real time analogue video and real time analogue sound for non-Nicam equipped receivers? If the delay was significant then both the analogue video and the analogue (non-Nicam) audio would need to be delayed by the same amount in order to keep everything in sync? (Maybe they were).

I think you’re referring to the analogue days of old. In those days the analogue tv signal leaving the broadcast centre had the audio encoded onto it in the sync interval using a technique referred to as DCSIS (Dual channel sound in syncs). So the signal arriving at the transmitter site always had vision and sound in sync. The dcsis signal was extracted from the analogue video signal and transcoded at the transmitter site for broadcast both as a conventional analogue signal and a dual channel digital (nicam) signal. The time taken to transcode dcsis to nicam and dcsis to analogue audio was negligible I believe.

[Stockden]

Originally Posted by mhennessy

As a result of this direct experience, I seriously call into question the validity of the methodology described in #25, and would appreciate references containing more information about the rational behind it, and the weighting given to that test (presumably one of several tests forming the overall conclusion).

Yes, I would as well.

This effect was demonstrated during my professional training, using exactly the same sort of tape based system described in post #25. Given that this was 1975 and we were in our early 20’s (with hearing that met or exceeded armed forces entry standards) it seems improbable that it’s anything other than a psychological (psycho acoustic?) effect.

The only way in which I can see it being used to assess hearing loss would be to induce the effect in the test subject and then reduce the volume fed to the headphones until normal speech returns. What advantage that might have over the usual methods of assessing hearing I don’t know.

Hugh

[G6Tanuki]

I can't provide any references about the logic of the delayed hearing test, it was back in the early 70s, my relative is long since dead but worked for what was then the National Coal Board in Nottinghamshire.

The effect definitely existed though, when tried on me. I am assuming it was in part because of the combination of through-bone sound conduction and that through the air. Get the respective levels of the two components right and add some delay, then resolve the needed amplification to cause the speech dysfunction at different degrees of delay..

[Julesomega]

I can't see any argument against this cunning test being used to evaluate hearing loss during employment. If only 1-2% of those tested can continue to read under that challenge, that means 1-2% of the claims might be fraudulent, while correctly assessing all those with impaired hearing.

[mhennessy]

Other than the fact it's deeply flawed.

If you've set the level in the headphones at a sufficient level to overcome the worst of the hearing impairment, then you're just testing brain processing, not hearing. They are two very different things.

I just feel sorry for those who were wrongly denied support and compensation as a result of this. Hearing loss is a terrible condition to live with.

[paulsherwin]

Back on topic please.

[stuarth]

Interesting numbers from Mark Hennessy in post #8 about the speed of sound.

Many years ago, working with fast logic devices, since light travels 30cm in 1ns, we sometimes called a nanosecond a “light-foot”.

I shall have to start calling a millisecond a “sound-foot”!

Stuart

[Radio Wrangler]

For PAL colour TV, on flesh tones, the human eye can spot one degree of phase error which is roughly 1ns which is roughly 6" of pcb track on FR4

Rough but useful

DAvid

[Graham G3ZVT]

Received wisdom (see what I did there?) has it that the timing of the BBC time signal is most accurate when received 100 miles from the Droitwich station.
It just so happens that I live
"Only, one hundred miles from Wychbold".
Could be the title of a song? Perhaps not.