Recordings Indistinguishable From Real Thing

[RobbieC]

Well, any music we purchase is lossy - inevitably some information gets lost in the recording process between the sound of the instrument and/or artist and the recording.

 

I'm going to have to disagree. Human hearing cannot detect everything coming from an instrument and/or artist. Therefore, so long as a recording does capture that information that a human being can detect and does so in a faithful way, then for all intents and purposes given a good enough playback mechanism a reproduction can be made indistinguishable from the original.

 

A truly perfect recording (and subsequent transmission) is impossible due to random variations at a molecular level all of which cannot be accounted for much less controlled. However, since we are talking about human detection, so long as the following four requirements are met a recorded waveform can be considered "perfect" (i.e., one that is indistinguishable from the original from a human detection standpoint).

 

(1) Amplitude-all detectable amplitudes must be preserved (generally considered from below 0 dB to above 120 dB SPL)

 

(2) Frequency-all detectable frequency components must be preserved (generally considered from 20 Hz to 20 kHz)

 

(3) Phase-the relative phase between frequency components must be preserved

 

(4) Dynamic range-(similar to amplitude with an added subtlety) the lowest amplitude component of the waveform detectable must be preserved. Has to do with noise.

 

It is a common misconception that quantization alone loses information since waveforms are continuous and quantization discrete. (Not saying that you're saying that.) Fortunately, that is not true from both a mathematical point of view and a practical one. Also, random noise of the same type is indistinguishable to human hearing.

 

(In all fairness, you did say music that "we purchase" implying music that is generally available today and captured with generally available techniques so I guess I agree with that to some extent.)

[Julf]

A truly perfect recording (and subsequent transmission) is impossible due to random variations at a molecular level all of which cannot be accounted for much less controlled. However, since we are talking about human detection, so long as the following four requirements are met a recorded waveform can be considered "perfect" (i.e., one that is indistinguishable from the original from a human detection standpoint).

 

Robbie, I am in total agreement - that was actually one of the points I was trying to make. I am reluctant to call anything "perfect", but at some point we can draw a line and say "good enough" - or in your words, "indistinguishable from the original from a human detection standpoint". And that is why I am not sure it makes sense to talk about a "lossy" recording - all we care about is if we preserve the information we recorded in the further processing of the data.

 

One way of looking at it is that the human gearing is a rather lossy process that discards a lot of information.

[Jud]

It is a common misconception that quantization alone loses information since waveforms are continuous and quantization discrete. (Not saying that you're saying that.) Fortunately, that is not true from both a mathematical point of view and a practical one. Also, random noise of the same type is indistinguishable to human hearing.

 

(In all fairness, you did say music that "we purchase" implying music that is generally available today and captured with generally available techniques so I guess I agree with that to some extent.)

 

Robbie -

 

Be careful. Most folks here understand that Shannon-Nyquist pretty much works. There are people here who intimately understand DACs and phenomena that occur due to the fact that perfect Shannon-Nyquist conditions (e.g., infinite time) do not exist in the real world, such as pre- and post-ringing, phase shift, etc. So I hope, at least, that we are pretty much beyond saying you can't get an analog wave out of digital information. Plainly it's quite possible with no particular loss of information in the sense you were speaking of. There may be problems with loss of information (we've had threads lasting forever on sampling rates and ability to reproduce certain types of transients), but there indisputably are real world problems with adding information that isn't in the original signal (the aforementioned pre- and post-ringing and phase shift) that DAC and player software designers must work hard to resolve.

[The Computer Audiophile]

I moved this post to its own thread as its an interesting perspective and was off topic from the original thread.

[RobbieC]

Jud,

 

I am very familiar with the problems of which you speak. They are due to digital filter implementations and are not what mathematicians (myself to be included here) call phenomena. A true and applicable phenomenon in this context is the Gibbs Phenomenon but I don't want to digress. I'd like to simply state that digital filters can be made to have ringing and phase shifts that are below detection thresholds and are, therefore, capable of accomplishing the desired indistinguishability result we want. For instance, using a suitable IIR filter, one could apply that filter to a data set in its entirety in one direction and then reiterate that filter but in the opposite direction and completely cancel out phase shifts. There are also advantages that can be taken from symmetry to give yet another example and of course there is ongoing research in this field of mathematics.

 

That said, the actual implementation may be very computational taxing and perhaps impractical for implementation into standard mass produced DAC chips meant for realtime processing of data. However, I must make the point that the problems you mention are not impossibility problems but more problems of making compromises.

[The Computer Audiophile]

Hi Robbie - Just so I have your perspective correct are you saying recordings should be indistinguishable from the real thing because of the limits of human hearing and technology that captures recordings?

[Jud]

I moved this post to its own thread....

 

It is good to be the King! ;-)

[Jud]

Jud,

 

I am very familiar with the problems of which you speak. They are due to digital filter implementations and are not what mathematicians (myself to be included here) call phenomena. A true and applicable phenomenon in this context is the Gibbs Phenomenon but I don't want to digress.

 

OK, would certainly like to be properly understood - so what would you call these problematic occurrences (agreed not as a result of limitations to theory but rather of implementation)?

 

I'd like to simply state that digital filters can be made to have ringing and phase shifts that are below detection thresholds and are, therefore, capable of accomplishing the desired indistinguishability result we want. For instance, using a suitable IIR filter, one could apply that filter to a data set in its entirety in one direction and then reiterate that filter but in the opposite direction and completely cancel out phase shifts. There are also advantages that can be taken from symmetry to give yet another example and of course there is ongoing research in this field of mathematics.

 

Sounds interesting. But I'm not completely clear on whether you are saying making the problems of ringing, phase shift, etc., inaudible would depend on the ability to, e.g., reiterate filtering in the opposite direction?

 

That said, the actual implementation may be very computational taxing and perhaps impractical for implementation into standard mass produced DAC chips meant for realtime processing of data. However, I must make the point that the problems you mention are not impossibility problems but more problems of making compromises.

 

Clearly none of this is impossible mathematically. I'm wondering, though, whether anyone in the audio engineering field agrees that with current technology we can make the results of filtering inaudible in terms of reconstructing the original analog waveform that was presented to the A/D conversion process. There seem to be quite a number of different solutions in DAC firmware and player software, and even quite expensive solutions aren't identical. (I do recall Barry Diament saying he can get results out of his equipment that are at least in the same ballpark as, if not indistinguishable from, the live mic feed. But other folks seem to prefer different equipment than Barry has.)

[RobbieC]

Barry has said in the past that at the 192 kHz sampling rate for the first time in his opinion the reproduction is indistinguishable from the live mic feed. Sampling rate affects the amplitude and frequency of the ringing for lack of a better way to say it and Barry's is a good example of what I meant about ringing being made below detection thresholds and therefore obtaining that indistinguishability result we want.

 

About what I said about IIR filters: It is a common misconception that all filters introduce phase shifts; however, this is not true. As an example, I mentioned the use of a suitable IIR filter. If we were to run an entire signal through a suitable IIR filter, take that result and then run it through that same filter backwards we would cancel out phase shifts completely. The requirement for suitability, by the way, is symmetry at a minimum. It should be obvious that the attenuation a filter affords is independent of the direction in which data is processed (the formula remains the same and is unchanged; therefore, its action remains the same). The limitation is that there would be a significant number of samples to process both in the forward direction and then in the reverse before a single final resultant sample is even produced. This means either a very, very computationally strong processor or the need for an off-line process. This is what I meant when I hinted about mass produced DAC chips being made with compromise in mind.

[RobbieC]

Hi Robbie - Just so I have your perspective correct are you saying recordings should be indistinguishable from the real thing because of the limits of human hearing and technology that captures recordings?

 

In a sense, yes, I am saying that it is possible should the requirements I enumerated be met. Remember that those requirements must be met at transmission as well as at capture.

 

Wait...perhaps I misunderstand your question: If you mean that all CD quality recordings, for example, should theoretically be indistinguishable from their respective original signal, that is not what I'm saying. What I am saying is that it is possible to use current technology and achieve an indistinguishable reproduction from the standpoint of human detection. (And I'd like to point out that a signal need not be sampled at a bit depth of 24 bits. I believe 16 bits could suffice in many instances though perhaps not all if we are to be 100% inclusive of all possible human ears listening to all possible signals).

[RobbieC]

Perhaps there are a few things that are important to consider in terms of modern technology.

 

First, level accuracy of the ADC stage is important (as well as its precision) so as to avoid non-linearity problems and so that results are ultimately acceptably consistent. However, should errors arise here that in itself does not necessarily mean an "imperfect" recording in the sense we have been talking about. If the resulting quantization error results in random noise below the required minimum as specified by the required dynamic range of the original signal then we are golden. However, if non-linearity happens to be present then that results in non-linearity "correlated" distortion. So long as the amplitude of that distortion is also below the minimum noise threshold then we are golden here too. (That's not entirely true since it is possible for humans to discern organized signals buried beneath a noise floor...but that has a limit too to perhaps a few dB below.)

 

Second, timing accuracy and precision is also important. Errors here result in the well-known issue of jitter. And of course jitter in and of itself is not a disqualification for a "perfect" recording. Jitter manifests itself audibly as sidebands (I've dabbled in the maths of this in the past briefly when I investigated asynchronous sample rate converters, which are by the way mathematical formulae just like digital filters are contrary to popular belief). So long as those sidebands are of low enough amplitude then again we are golden.

 

The actual conversion of digital to analog also introduces some areas of concern none of which cannot be overcome...but to be discussed later.

[bdiament]

Hi Rob,

 

If I may interject a perspective, before we get to the converters, there are the microphones (and how many of them and where they are placed... and why). The best A-D in the world can only be as good as what it is fed.

 

I'd add the mic cables too. Even being the cable aficionado I am, I was pretty shocked when I first substituted a pair of Nordost Valykrjas for my old "top of the line" Mogami Neglex. In short, I keep the latter as emergency backups, lest they be used to bundle newspapers for recycling.

 

But #1 for me, far and away is the microphones and how they're used. This, for me, determines the ultimate quality *potential* (i.e., ceiling) of the recording.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[The Computer Audiophile]

Hi Robbie - I think it's necessary to state that what's heard via mic feed is different from the actual live event.

 

I like your enthusiasm for this topic and the fact you have a deep technical understanding of the theories. However I disagree that a live performance can be captured and played back transparently. I.e. live v. recorded are indistinguishable.

 

Lets use the London Symphony Orchestra as our test subject. How would you go about recording and playing the material from LSO so it's indistinguishable from the actual event?

 

I'm seriously interested in the topic. Please don't take my question and disagreement as an attack at all.

[bdiament]

Hi Chris,

 

This is the area that most fascinates me about recording and is why I started Soundkeeper. Having worked in studios early in my career and seen how it is typically done, having cut tracks, overdubbed, mixed and then mastered, I came to believe it all happens with the mics, long before any recorder of any format is engaged.

 

Is it possible? This is what I seek to pursue.

First, we have to look at how humans hear and the types of cues our brains use. (I wrote a bit about this in my "Recording in Stereo" articles.)

We have to look at which types of microphones and what sort of microphone implementations (quantity, placement, etc.) best emulate this.

 

I think we can do a whole lot better than 99% of the records sold to us. Many minimally mic'd recordings point the way. I find it interesting though how no two engineers (I'm aware of) do things the same way - even among the purists.

 

To answer the LSO question, I think it isn't simply the recording. The playback chain (and environment) will determine how much of what is captured in the recording gets heard. I don't believe we're at a point where playback can completely fool an experienced listener into thinking they are in the presence of the performance. However, for the past few years, we *have* been at a point where the best gear used to capture the mic feed can present it in a way that I've not yet been able to distinguish from that mic feed. To me, that is a huge step. With the best gear, many of the requisite cues *can* be heard in playback to, if not necessarily convince, at least provide a great deal of the *feeling* I'm standing at the position of the microphone array at the recording session.

 

The quest continues.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[MikeJazz]

But #1 for me, far and away is the microphones and how they're used. This, for me, determines the ultimate quality *potential* (i.e., ceiling) of the recording.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

 

Barry, I have the impression that each Microphone brand has it's own voicing, is this correct?

 

What do you think is the most neutral mic? Do you pursue a neutral mic, or do you prefer to work with a certain voice.

 

Interesting subject...

[mitchco]

Is Binaural+ the current state of the art for reproducing a live event as faithfully as being there? Manually switch to 720p HD, it would not embed for some reason.

 

[video=youtube_share;gzZX5V8gCAs]

 

 

Dr. Chesky's Sensational, Fantastic, and Simply Amazing Binaural Sound Show

[Audio_ELF]

Other recording techniques...

 

4channel ISOMike from Kimber...

 

Eloise

[Julf]

Hi Robbie - I think it's necessary to state that what's heard via mic feed is different from the actual live event.

 

Chris,

 

I am pretty sure (but I could be wrong) that Robbie approaches this from the slightly narrower view of accurately reproducing the signal from the mic feed, and is not addressing the very interesting, but very separate, problem space of actually capturing all aspects of the performance using traditional transducers (microphones).

[bdiament]

Hi MikeJazz,

 

Barry, I have the impression that each Microphone brand has it's own voicing, is this correct?

 

What do you think is the most neutral mic? Do you pursue a neutral mic, or do you prefer to work with a certain voice.

 

Interesting subject...

 

The "Recording in Stereo" articles I cited in a previous post talk about this a little.

 

Not only do the different brands have a family sound but each pickup pattern has a profound effect, as does the size of the diaphragm itself.

 

I never subscribed to the idea of certain mics for certain instruments, as a number of my colleagues do. Perhaps because my own recording techniques have taken me in a direction where I simply don't mic instruments anymore. I find the close positioning works against my goal of helping the listener suspend disbelief. I prefer to mic the *event* instead.

 

In order to help with that suspension of disbelief, i.e., to make the recording as convincing as possible (that the listener is in the presence of the performance, in the room where the event took place), I want my mics, along with the rest of the gear, to "get out of the way" as much as possible.

 

Of all the mics I've tried over the years, it is the small diaphragm omnis (omnidirectional pickup pattern) that have the least "character" of their own. Most commonly used mics tend to roll off the bass, peak in the lower treble and roll off the high top. The better small diaphragm omnis have frequency responses that look more like that of a good power amp - flat and wide. Perhaps even more importantly, their *time* response is much faster than the type of mic used in typical studio recordings; they tend to settle much more quickly.

 

The Danish DPA microphones (formerly B&K) were favorites for years. Then I auditioned the Earthworks QTC-40 (aka QTC-1) and was won over immediately by how much better a job it did of "disappearing" than the B&Ks (revealing just a tad of time smearing in the latter). I still love the B&Ks too; they're wonderful. But I haven't heard anything that sounds as much like what occurs in their presence as my matched pair of QTC-40s.

 

I mentioned earlier that every engineer I know, even those who take a purist approach, does it differently from every other. Each has his own favorites.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[sandyk]

Hi Barry

Going by the Spectrum Analysis in Sound Forge 9 that I just looked at for "Maria" from "Americas-Paul Beaudry and Pathways" , your mics are very wideband too ! There even seems to be some signal content into the 50kHZ area.

Kind Regards

Alex

[bdiament]

Hi Alex,

 

Hi Barry

Going by the Spectrum Analysis in Sound Forge 9 that I just looked at for "Maria" from "Americas-Paul Beaudry and Pathways" , your mics are very wideband too ! There even seems to be some signal content into the 50kHZ area.

Kind Regards

Alex

 

I believe some tunes on "Equinox" will show content up around 67 kHz (from the trumpet and percussion).

The mics are spec'd flat from about 4 Hz to 40 kHz ±1dB. At ±3dB, they do something like 3 Hz to 50 kHz.

 

More important than this, though I'm sure this is a relevant factor, is their time response and above all, how they *sound* when one listens to the output and compares it with what is occurring in the air. I became a fan very quickly.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[Jud]

Hi MikeJazz,

 

Barry, I have the impression that each Microphone brand has it's own voicing, is this correct?

 

What do you think is the most neutral mic? Do you pursue a neutral mic, or do you prefer to work with a certain voice.

 

Interesting subject...

 

The "Recording in Stereo" articles I cited in a previous post talk about this a little.

 

Not only do the different brands have a family sound but each pickup pattern has a profound effect, as does the size of the diaphragm itself.

 

I never subscribed to the idea of certain mics for certain instruments, as a number of my colleagues do. Perhaps because my own recording techniques have taken me in a direction where I simply don't mic instruments anymore. I find the close positioning works against my goal of helping the listener suspend disbelief. I prefer to mic the *event* instead.

 

In order to help with that suspension of disbelief, i.e., to make the recording as convincing as possible (that the listener is in the presence of the performance, in the room where the event took place), I want my mics, along with the rest of the gear, to "get out of the way" as much as possible.

 

Of all the mics I've tried over the years, it is the small diaphragm omnis (omnidirectional pickup pattern) that have the least "character" of their own. Most commonly used mics tend to roll off the bass, peak in the lower treble and roll off the high top. The better small diaphragm omnis have frequency responses that look more like that of a good power amp - flat and wide. Perhaps even more importantly, their *time* response is much faster than the type of mic used in typical studio recordings; they tend to settle much more quickly.

 

The Danish DPA microphones (formerly B&K) were favorites for years. Then I auditioned the Earthworks QTC-40 (aka QTC-1) and was won over immediately by how much better a job it did of "disappearing" than the B&Ks (revealing just a tad of time smearing in the latter). I still love the B&Ks too; they're wonderful. But I haven't heard anything that sounds as much like what occurs in their presence as my matched pair of QTC-40s.

 

I mentioned earlier that every engineer I know, even those who take a purist approach, does it differently from every other. Each has his own favorites.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

 

Channel Classics lists mics used among the various pieces of information in the thorough provenance they have for each of their DSD downloads. Would love to see the rest of industry follow this example insofar as it's possible.

[The Computer Audiophile]

Channel Classics lists mics used among the various pieces of information in the thorough provenance they have for each of their DSD downloads. Would love to see the rest of industry follow this example insofar as it's possible.

 

Hi Jud - I didn't know that. Very cool. However, listing equipment is similar to listing the sample rate. It's a piece of the puzzle that may be unimportant without good engineering. Maybe we should push for engineer résumés :~)

[MikeJazz]

Hi Jud - I didn't know that. Very cool. However, listing equipment is similar to listing the sample rate. It's a piece of the puzzle that may be unimportant without good engineering. Maybe we should push for engineer résumés :~)

 

Almost there, You also get the name of the engineer.

And then you can search. For example, Jared Sacks recorded Stravinsky for Channel Classics.

I just found this very cool piece of text from Jared, about our topic.

 

“80% of a good recording is the ambience of the hall. If you don’t have a good hall, you can screw around as much as you like and you can make it good but it will never be great. It’s also about distance – someone who’s only interested in sound will put a microphone above the hammers of a piano or the strings of a violin, but if you want to really hear the music, you need distance. You need to capture the overtones, not just the direct sound.”

In http://www.onlythemusic.com/People/Jared_Sacks/.

 

This talk about distance is in line with Barry great description of his way of work.

 

And is something that I am now more aware off, especially after moving to my QUAD's... ambiance and a sense of distance is now very easy to perceive, when recordings have that information.

[Paul R]

Hi Jud - I didn't know that. Very cool. However, listing equipment is similar to listing the sample rate. It's a piece of the puzzle that may be unimportant without good engineering. Maybe we should push for engineer résumés :~)

 

I think of it in the same way I think of great Chefs. A really great Chef (audio engineer) can take mediocre less than perfect source materials and creat a very pleasent dining experience. They can take top quality materials, and make dinner an experience one will never forget.

 

Conversely, a mediocre chef may take the best materials and creat a mediocre result.

 

-Paul