room

Hi,

I myself got about 80% correct at around +30db S/N ratio (spoken), some of the guys who where at it longer and had probably better hearing than me got much better recognition with even more noise. It is a serious strain doing though.

Absolutely. With a Human we cannot seperate perception and which part of what we percieve is "reality" (back to Philosophy and Audio) and we have been unable to relaibly account for the impact of this in measurement and even in most Audio DB Tests....

Ciao T

 

Haha not so fast thorsten! There is a difference between the term perception as I mean it in terms of technical pattern recognition and what you are describing. Equipment is capable of measuring anything the human ear can measure - that the ear is somehow beyond test equipment is a myth. What the human brain does better is to take that information and recognise patterns such as instruments. That is entirely seperate from pure "analysis" which does not require this step. In simple terms as an example - speech recognition may not work below the noise floor (though in fact it can) but we can still measure below the noise floor as well as the ear can. Crucial difference.

Are you sure you dont mean 40 db SNR at RF rather than on the actual audio signal? If you were simulating static rise or jamming then you would compromise the RF SNR. If you did it as added noise on the audio band then you did it wrong .

 

I think you guys will probably lose poor old Johnny with this,as and when he gets back....he may even be forced to start another thread.

 

Sorry thorsten got me going with his audio mysticism!

 

Just be interesting to see how it will get routed back....

 

Hi,

But that is my point. The human auditory system, especially with training can still get the whole pattern with good reliability even though the patter is burried under loads of noise. Machines need more signal than noise to be able to recognise the pattern.

I did not "do it" as in conduct the "experiment", I was one of the subjects. The simulation was done on pure audio circuits, very primitive.

The aim was to train us to be able to still get HQ's commands after the yankees would have dropped the mothersucka on our troups, in other words under condition where you had practically no communication, all high tech stuff blown out and static at a few time of the output of your radio.

The Radios we used where also rather primitive and heavy, but definitly EMP Proof (all valve, batteries and mechanical transverter to get the HT), very basic AM Modulation, not all that much power (had to be portable and last long). My job was just to take the sheet the coder gave me and read it and hear what came from the other end accuratly, not knowing what I was hearing or saying really and having my ears full of noise, loads of it, not just jaming but literally meant to be in the ABC suit right next to the nuke dropzone.

All of that was rather pointless in my view but tell that the instructors.... ;-) Takl of military intelligence as oxymoron.

Ciao T

 

Not so fast biovizier !

As far as analysis goes, the ear works the same as a computer.
It performs an fft on the signal.

..''Equipment is capable of measuring anything the human ear can measure '' This is a myth in itself.

In any case we audiophiles aren't interested in mere signal analysis. What we really want to know is how music is affected by various distortions.

So you can quote all the facts and figures till the cows come home, but the bottom line is, it is irrelevant.

 

Given the current state of medical knowledge on the subject,I'm not sure that anyone is aware of the precise method that the human brain uses to process audio signals,given that no-one has defined the brain's operating system,let alone what software it runs.

 

Johnny - Actually the ear does not perform an fft physically speaking, neither does all the nonlinear signal processing that we call hearing. Hearing behave likes a neural network type of program not a conventional algorithmic piece of code like an fft. The fact that the cortex can register leading edges outside of the steady state bandwidth is proof that the ear and the brain are not as well understood as people think - this phenomena arises from processing of the differing firing times across the cilia. Audio is not magic it is engineering - therefore measurements are entirely relevent - you just have to make the right measurements. Music may be more than a signal to us in the emotional realm but it is still just a signal in the physical realm so dont make it more than it is. How distortion affects the signal is entirely a signal analysis issue. A spectrum analyser has greater phase and amplitude discrimination, wider bandwidth, greater dynamic range and a lower noise floor - its no myth. You just have to apply the kit we have in the right way and thats where audio engineers are currently falling short. FFT and THD dont tell the whole story.

Thorsten - a spectrum analyser does not perform pattern recognition. Show me which chip holds the algorithms for that! I beleive my example about speech recognition in noise holds the key. Recognising and measuring are not the same thing. The experiment does seem a little pointless and misleading - the radio would drop out long before a negative audio SNR was ever reached, especially if they are digital radios. For some reason I visualise you wearing a golden mask and waving a short hilted curved dagger whilst typing on the forum ;-). Unfortunatelty radios by definition can never be fully EMP proof and the front end is most at risk during a blast rather than circuits.

 

Hi,

My point. Place a signal below a suitable steady state noisefloor and it will not register.

Well, the radios where analog, valve and most likely AM, possible outside bet on SSB. This was 1950's style gear, almost WW2. It was still around in the 80's because it would survive where newlyfangled moden stuff would not shoud the actor on the throne of American King decide to push the button on us in the evil empire.... ;-)

I am not sure where the russians got their original data from, probably the few times they dropped nukes on their own troups and villages to get data.

Actually, blue/white shirt and jeans. Ceremonial magick is something I'm familar with but which I do not normally practice.

Well, look at the gridwire of a 1950's triode or pentode and then calculate how much EM Field you need to damage that. That newly fangled solid state stuff is easily blown out though. Hence a lot of crucial stuff in soviet era military gear was valve based, from at least the receiver input stages in almost any military radio (including radar in Jets) to the bleeding servo motor drivers in Mig Jets (the legendary 6S33).

In fact, the penchant of the Eastern Block Military for using Valve electronics as simple and primitive methode of EMP hardening their gear is responsible for valves still being made, otherwise the factories would have been closed in the 1970's!

Ciao T

 

You can measure below the noise floor if you do it right and employ suitable signal processing ;-) I bet the ear cant go down to -160 anyway. You can even improve the performance of radio receivers below the noise floor by using stochastic resonance. Have you heard of spread spectrum techniques?

Planes used valves for EMP protection for their avionics - thats fine. However radios amplify an input signal - unlike avionics - and a nuclear pulse can cause even a valve input stage to overload, though of course at closer range. That russian plane may keep flying but I bet its radios wont work. Plus disturbances in the propagation environment will last for hours anyway. RF noise is the worry.

 

Hi,

Well, this works on certain assumptions that IMHO are neccesarily supportable. It's sorta like assuming there are atoms and constructing a measurement to illustrate them. It gives you some result, but not neccesarily one that in the "absolute" sense is relevant

No, but to 40-50db with a noise overlay of 80db and depending upon conditions and familiarity quite possibly even more.

I am passinly familar with spread spectrum from SMPS design. I am also aware of a variety of DSP Tricks, averaging multiple measurements and so on, which may be employed to "look into the noisefloor". But any of these are only valid for a reliably repeatable stimulus. The ear & brain "get it" single pass.... ;-)

I think we can extend the discussion endlessly, but that is missing the point.

The way the human auditory system differs drastically from traditional mechanisms and hence it can do things mechanical systems cannot (yet) do. Mechanical systems on the other hand can outperform the human in many situations too, but tend to be non-learning and non-multi-purpose.

Yup. but in AM RF noise is AF noise.

Anyway, enough of this pleasant banter, we are now way off topic and don't cntribute much usefull.

Ciao T

 

Yes but the snr is measured relative to audio in one and rf carrier in the other. Quite different

Some of your figures are wildly optimistic. Anything that the ear can measure can be done (better) by science also. But we dont understand the process of "hearing" and "cognition" well enough to have perfect audio knowledge, but the fault does not lie in the ability to measure. It is a question of what we measure.

The ear doesnt always get it right and "prior knowledge" such as grammer is employed in the process which is not really to do with this argument. Measurement and recognition are not the same... and measurement is not the same as copying the ear mechanically.

OK I will shut up now

 

Hi,

Always have to posit the ultimate argument, do you now.

BTW, your points all miss my actual argument and in several cases are completely wrong. ;-)

OK, NOW I will shut up

Ciao T

 

You are guilty as me You missed all my points with knobs on!

Ok NOW I will stick my head in a pig!