5.1: gimmick, or closer to reality?

[Martin Kantola]

Originally Posted by Dan Lavry

If the singer is willing to swallow the mic, they should have their ear 1 foot away from the speaker... The ear does not need 1msec delay.

Just a little comment; in a typical recording situation with a vocalist he or she will more or less "swallow the mic" while listening through headphones (placed over the ears) so the acoustic delay is indeed very small. Also, remember that a vocalist hears the voice through the head too. Have found that even 1msec of delay in the monitoring does bother me when playing electric bass as an example. So this is where latency does matter.

Martin

[Dan Lavry]

Originally Posted by Martin Kantola

Just a little comment; in a typical recording situation with a vocalist he or she will more or less "swallow the mic" while listening through headphones (placed over the ears) so the acoustic delay is indeed very small. Also, remember that a vocalist hears the voice through the head too. Have found that even 1msec of delay in the monitoring does bother me when playing electric bass as an example. So this is where latency does matter.

Martin

Instead of saying that a "vocalist needs to "swallow the mic" while listening through headphones", it would be good to state clearly how much delay is required. In my experience, a guitar player (sharper attack then a vocalist) can live with around 3-4 msec. I do not wish to argue about 1msec. My point is that at 44.1KHz, each sample is around 22usec, so a 10 sample delay is 220usec, and 20 samples are around 440usec. That is around 5 inches of acoustic delay. Surly you are not going to suggest that we hear delays when the vocalist mic is 5 inches away, listening to a headphone.

If one is that concerned, they can always use ANALOG circuitry (bypass the conversion), and that would yield just a few microseconds. If one needs to listen to themselves, they can do just that.

Some of the new converters offer just a few samples of delay, so they can advertise "low latency".

Again, my point is that one should quantify their statements. Once you state how much delay is required, and the NUMBER of msec is based on REAL requirements, we can talk about the various ways to achieve the requirement.

I would think that the use of headphones is to hear oneself, so that other loud sounds do not cover your singing.

Your statement seems to me to be very exaggerated. I go to concerts, and you have large orchestras perform where the distances between players as well as listeners are not fixed to be 3 inches, or even a few feet. I play music, and my piano strings are not inside my ears. I often play through amps, and I play with other musicians, including fast attack drums. The speakers are a few feet away, (thus a few milliseconds) and there is no latency issues. Say a spot monitors are 3 feet away, so going from say 1msec to 500usec delay (6 inch savings) is not all that significant. If the mic is 0.5 foot away and the speaker (monitor) is on the floor, at say 5 feet away (short singer), the converter delay is pretty insignificant.

When you say that even 1msec delay bothers you, you are suggesting that acoustic guitar players, acoustic drums, acoustic bass, piano and more players would place their head a few inches away from the sound source. Can it be that the 1 msec that is bothering you is IN ADDITION (in series) to other accumulated delays that you did not account for?

Regards
Dan Lavry

[Martin Kantola]

Originally Posted by Dan Lavry

When you say that even 1msec delay bothers you, you are suggesting that acoustic guitar players, acoustic drums, acoustic bass, piano and more players would place their head a few inches away from the sound source.

No, I'm not suggesting that! All I said was that I hear/feel the effect of a 1 msec delay in my studio headphones with my own voice or with an electric bass guitar. So I prefer analog circuitry as you suggested in that specific situation. My theory is that it's because in both examples I have a very direct mechanical connection to the sound source, and with headphones the sound (including other instruments) is practically speaking "inside my head".

But just as yourself, I'm perfectly comfortable playing a gig through a bass amp some 10msec or so away. Let's get back on topic...

Martin

[Audiop]

Off course a 1ms delay is audible when you also have the direct sound thru the air/scull or whatever. If we assume the direct sound from mouth to ear is 17cm (ignore the bone conduction for now) and we have a 1ms "electric" delay that means about 0.5ms/17cm effective time difference. This (if my tired brian is working right now) results in cancellation/combfiltering from 1kHz and up all the way thru the audible spectrum.

A 1ms delay between tracks is off course insignificant.


For the Topic, I think two channel will dominate for another 50 year or so but like there are vinyl/SACD/DVD-A people and such, surround will co exist and grow. I have upgrade my recording gear from 2 to 8 channels due to my curiousity on this. With few exeptions I'm not much for discrete sounds from the rear but ambience. In general I think I'd like a 100-180 degree "soundstage" with ambience speakers behind and above. Eight channels is a compromise but that'll have to do for my own experiments as a starter.


/Peter

[mohthom]

Dan and others who may be able to answer this one,

I read in a number of technical and journal papers that lower end multi-channel ADC and DACs will often not link in the way that we want them to, meaning that one channel will be played back fractionally (~10 samples) before the other in a stereo setup. Is this problem exacerbated in multi-channel setups (i.e. is this problem six times worse in a 5.1 setup)? Just a curiosity . . .

MohThoM

[d_fu]

Originally Posted by Dan Lavry

At 20KHz, the shape of the wave may not be relevant from non audible harmonics standpoint but the shape is still very relevant, because you need to know the amplitude and the phase of the fundamental.

Dan, I know - I was just referring to harmonics. People look at post-DA 5k "square" waves with an oscilloscope, complaining that at 44.1k sampling rate, the wave doesn't look square at all...

Originally Posted by mohthom

I read in a number of technical and journal papers that lower end multi-channel ADC and DACs will often not link in the way that we want them to, meaning that one channel will be played back fractionally (~10 samples) before the other in a stereo setup. Is this problem exacerbated in multi-channel setups (i.e. is this problem six times worse in a 5.1 setup)?

No. Usually converters come in stereo, i.e. one chip for two channels. I don't know about cheap converters, but in the olden days of digital audio (eighties), e.g. some CD players used to have converters with only one actual DAC for both channels, causing a slight offset between channels. I don't know whether this kind of thing still exists. But even if, in a multichannel setup, tis would only affect pairs of channels and not cause increasing delay from 1 thru 6 or whatever.

Daniel

[Dan Lavry]

Originally Posted by d_fu

Dan, I know - I was just referring to harmonics. People look at post-DA 5k "square" waves with an oscilloscope, complaining that at 44.1k sampling rate, the wave doesn't look square at all...

Daniel

My point is that they are NOT looking with a scope at the DA output. If they used a scope, they will see a clean 5KHz sine wave.

I do not think they are looking at the point where the DA conversion leads to the output filter, because the wave there is not too far from a clean 5KHz. Probing that point will show some steps, but most likely very small steps, and the higher the upsampling, the smaller the steps. But looking at the final DA output (after the analog filter will show a clean wave.

So what are they looking at? Probably some computer simulation of a DA without upsampling, and without an analog filter. But I already explained it in a previous post...

Regards
Dan Lavry

[andy_simpson]

Originally Posted by Dan Lavry

Hi Andy,

One common mistake that throws people off is the "famous picture" of a single 20KHz sine wave, with 2 only sample points. When one looks at that, they just can not see how it is possible to know what kind of a wave will be "going through" those 2 points.

The fact is - 2 points is not good enough to yield the needed information. The reconstruction of a signal is based on many samples ("points"). The reconstruction utilizes very many sample values that happened BEFORE one looks at the pair of samples of a single cycle. Say you chose to take into consideration the 1000 previous samples PLUS the last 2 samples. Now, with 1002 samples, you can reconstruct the wave very well, with better then 0.00001% error.

It is correct to say that the samples further in time contribute less to the information (weighed less), and nearby samples contribute more to the information (more weight) that yields the outcome. But taking too few points makes the error bigger, and 2 points is ridiculously hopeless. (Of course I used 1000 points just as example, the number may be larger or smaller).

One may ask - what about the first 2 samples of a music track? Where are the previous 1000 points? The answer is simple: The earlier 1000 samples are assumed to be of zero value. That assumption is very appropriate, because the wave form before the beginning was indeed zero (no sound).

You said:
"Given a sufficiently high noise-floor in the test, I would expect 12bit to do as well as 24bit".

I did not read the whole thread, but more bits is NOT only for better noise floor. It is also about fewer distortions. This is not easy to grasp, and it does not show up in THD+N (total harmonic distortions plus noise) measurements.

However, if you separate the error energy into two parts, one part is random noise (un correlated to the music) and the other part is harmonic distortions (correlated to the music), you find that fewer bits alter the sound by introducing harmonic energy.

Dynamic range limitations in analog tend to be "just more noise". But dynamic range limitations in digital (not enough bits) also impact the harmonic content, thus the sound.

Regards
Dan Lavry

Hi Dan, having clarified my position with you via PM I just thought to return here and clarify similarly.

Where I refered to noise-floor I should perhaps have said acoustic noise-floor of the listening environment.

What I meant is that given a sufficiently high acoustic noise-floor in the listening environment, the difference between 12bit and 24bit would be effectively masked.

In other words, the harmonic distortion & 'random noise' of 12bit would fall below the acoustic noise-floor and not be audible.

In my conversation with Dan I described a 'test' where I took a 16bit recording and compared it with an 8bit truncated version of the same file.

By introduction of sufficiently high acoustic noise-floor I was able to mask the differences.

While the topic of masking is not quite so simple, in principle this is what I meant and this can put in perspective the significance of bit-depth.

Andy

[mohthom]

Maybe I should just put white noise generators in all of my listening rooms rather than record in 24bit then . . .



I don't record for people to listen with high acoustic noisefloors - I record with the assumption that people are going to listen in a well sounding room on fair equipment. Obviously it would be negligent to assume that my recordings were only going to be listened to this way, but if you prepare for the playback that shows every flaw. Or maybe I should be getting people to drive past my listening room window to mask the poor fidelity . . .

[andy_simpson]

Originally Posted by mohthom

Maybe I should just put white noise generators in all of my listening rooms rather than record in 24bit then . . .

Actually, there have been a few audiophiles doing exactly that quite recently (deliberately involving an acoustic noise-floor), though the purpose seemed vague to me. They claimed it 'improved' their perception of dynamics (if I recall correctly).

I don't record for people to listen with high acoustic noisefloors - I record with the assumption that people are going to listen in a well sounding room on fair equipment. Obviously it would be negligent to assume that my recordings were only going to be listened to this way, but if you prepare for the playback that shows every flaw. Or maybe I should be getting people to drive past my listening room window to mask the poor fidelity . . .

I'm not saying that there is no point in recording at 24bit, only that the improvements are to an extent predictable and 'maskable'.

This is interesting in light of the various tests that have 'shown no perceptible difference between 16bit & 24bit'.

In these tests I would expect that the distortion/noise products of 16bit were either acoustically masked or below the threshold of hearing for the test.

Andy

[mohthom]

I'm afraid I'll not be buying a subscription to that brand of Audiophile Weekly. I'm afraid I can't believe that an Audiophile (trans: lover of sound) would knowingly introduce noise not present on the playback medium to improve appreciation. If it were required it would be burned into the disc. I don't burn this data into the discs I produce because every bit (to the sixteenth and beyond) is important.

Again, maybe it's because I've met with rather favourable acoustics (£2m purpose built recording/concert hall, £2.5m purpose built audio research lab) but the consensus among those using these facilities is that depths above 16bit CAN be heard.

You say that the improvements between 16 and 24 bit are 'predictable.' What else would they be? An unpredictable result would be any piece of music played back at 24bit turned into Wagner's Ride of the Valkyrie. A predictable result is that there is an increased level of dynamic resolution. Of course these improvements can be masked also - with earplugs, by playing too loud, by playing other music at the same time, by piercing the eardrum with a sharp object . . .

MohThoM

[Mulmany]

Sample rate effects the perceived audio content on a recorded medium.

We should always be recording at high sample rates whether the release is stereo or surround. We are capturing complex waves that are interacting in an acoustic environment, every source (instrument, speaker, reflections, etc...) is altering the pure anechoic tones of the source. Its pure physics. Why would we want to limit these interactions to the recording space only? For me I want to mix hearing these mutations, because this is what I hear while listening in the recording space. It is what makes a cello sound the way it does, for example.

While mixing a close mic'ed performance I want as linear a recording as possible so that these interaction will be preserved in my mix. I know that transducers are far from perfect, but any degree of linear transfer is better then just cutting them out all together.

Its pure acoustics.

For those people that say surround is less immersive and inadequate, Have you truly recorded or mixed in surround?

I install Home Theater systems for a steady pay check, and with a few thousand installs in the past 4 years I can say that surround is very alive. These are not bose/theater in a box systems, but full range Klipsch systems. Consumers are much more informed now then a few years ago. They are beginning to see value in the enjoyment of dedicated theater/music systems in their homes.

[The Byre]

Digital theory I have come to the conclusion that a little knowledge is a dangerous thing!

The fact that Dan comes to this and other places again and again and again to explain the basics and still there are a few out there that refuse to hit the books and "read, learn and inwardly digest" the simplest of principles, never ceases to amaze me.

5.1 Again, a little knowledge . . .

There are a few engineers, producers and musicians who see little or no point in 5.1. Bruce Swedien is perhaps the most notable one. There are those who are new to 5.1 (we've had 5.1 now for about 30 years, so if you are not working in this medium, what on Earth have you been doing for the past 30 years!)

In an ideal world, we would have one speaker for every position of every instrument or other sound source. In stereo we reproduce many positions with the spreading of the image from Left to Right.

With surround, we can take that into depth - however, there are some producers who seem to think that it is necessary to have something going on from all directions at all times.

They forget that the listener is already in an environment that gives him sounds from all sides in the form of his car, living room or wherever he or she is. I have heard some very high profile projects ruined in this way.

A little something (as Winnie-the-Pooh puts it) now and again is more exciting than a constant stream of swirling mush that is supposed to sound like ambience. Such a stream can clash with the existing ambience of the room and make the listener feel physically sick in extreme cases.

[mohthom]

Have any of you guys ever been to see a multi-channel electroacoustic music diffusion concert (32ch+)? Maybe that would help clarify how un-immersive surround can be when done poorly!

MohThoM

[aracu]

An exciting use of surround with music is in filmed opera productions. The singers can sound like they are on a virtual stage closer to the audience than in a stereo recording, with the orchestra surrounding the audience in a wider sound image
set furthur back in the hall.