Perhaps the terminology is different in the realm of audio technology, but in mathematics, electrical engineering and computer science, "polarity" is not an accepted alternative to "sign", where the latter indicates whether a number is positive or negative. It is used in electrical engineering to denote the sign of a physical quantity, such as voltage, but not the sign of an abstract number.

--------------------
Robert Eidschun

Quote Robert Eidschun:

---

Perhaps the terminology is different in the realm of audio technology, but in mathematics, electrical engineering and computer science, "polarity" is not an accepted alternative to "sign", where the latter indicates whether a number is positive or negative. It is used in electrical engineering to denote the sign of a physical quantity, such as voltage, but not the sign of an abstract number.

---

His use of the term 'polarity' is a little cavalier, but the essence of his point is correct in so far as the MSB indicates whether the digitised audio signal is in the positive half or negative half of the total range (ie, above or below the nominal zero mid-line) -- or in other words, which way up it is at any time, which could loosely be related to describing a signal's polarity.

Hugh

--------------------
Technical Editor, Sound On Sound

Point taken re. "polarity". However, I'm sure Watkinson himself actually uses it interchangeably with "sign" in The Art of Digital Audio. Still, best to avoid any more confusion; my head is starting to hurt with all this talk of digital things.

Or I suppose it could just be the lack of sleep. . .

Edited by rhd_toyl (06/03/09 01:15 AM)

I think one should consider what is clearly his fundamental point, which is that somehow, my statement may be incorrect, that statement being that in a 16-bit system, the largest sample value, denoted in standard binary, is a string of 16 ones. That number may be represented in a variety of ways (e.g., 2's complement), but it can still be considered to have the value that I indicated, using the representation that I used; and certainly, it is much easier to use that representation to illustrate my original point concerning dynamic range, rather than some "encoded" version.

Furthermore, some sampling systems actually store sample values in standard binary. In such systems, it is only during mathematical operations, in the digital domain, on those sample values that alternative representations of those sample values are used.

"Encoded" versions (e.g., 2's complement) then are useful to facilitate representation and calculation using digital circuits. However, they are not useful for the purposes of conceptualization, where one can always use standard binary -- or decimal, for that matter -- and it is indeed more convenient to do so. I think that rhd_toyl should be aware of this.

--------------------
Robert Eidschun

I'm not 100% sure about this, and i might just be missing something, but if you're using unsigned binary numbers, then wouldn't the "0000 0000 0000 0001" you mentioned earlier represent a large negative-going voltage? Whereas in two's complement I'm pretty sure that that number actually would represent a positive-going peak of the lowest possible amplitude (assuming no dither) - like you said, "minimum scale". If that number does represent "minimum scale", I can't really see how a long string of 1s could be full scale, since there'd be no numbers left to represent negative-going voltages.

Reading back over what I've just written has made me feel very confused - time for bed I think!

The total number of different values of an n-bit word is 2^n, which is 65,536 different values in all, which you might enumerate consecutively from 0 to 65,535. (This you might do conceptually, i.e. in your head, but that doesn't mean that you have to represent them that way in the register of a microprocessor...)

Conceptually, you might map your signal voltage to those numbers in a variety of ways. So, for example, you might assign 0 to your "largest negative going peak", which might be, say, -1 volts, and 65,535 to your "largest positive going peak", which might be, say, +1 volts.

Or you might enumerate the different values as going from -32,767 to +32768. It doesn't matter for the purposes of asking questions such as, "what is the dynamic range of the system?". For a question such as that, the only thing that matters is the number of different values that your data word can take on. Even more fundamentally, you're talking about the number of different states of 16 transistors, e.g., "they can all be off; all on; the first on and the rest off"; etc;

By the way, the actual representation, inside of a machine, of those numbers can be any of a variety of sorts, e.g., 2's complement, but is often such that manipulation and calculation is efficient, using the procedures available. Inside of a digital computer, those procedures can only be, fundamentally, OR, AND, XOR and so on, given that a transistor assumes only one of two different states, etc.

--------------------
Robert Eidschun

Quote Hugh:

---

Dolby Digital uses data reduction (often erroneously called data compression) as a means of making the file sizes more manageable for storage or transmission. But this has nothing to do with dynamic range compression.

---

I see. By the way, as you may know, in information technology, it is standard to refer to the kind of data reduction that you refer to as "data compression". So, for example, mp3 is referred to as a "compressed format", created by "data compression". But it is really, as you indicate, a way of encoding in order to accomplish data reduction.

Quote Hugh:

---

The metadata carried with Dolby Digital soundtracks has encoded information which states which of the several different dynamic range compressions curves that are available to impose if the user seleects DRC mode.

---

Am I correct in assuming that this works the other way 'round too in that on the set-up menu of any DVD player, DRC is made available as an option only if Dolby is enabled, or that even if DRC is made available as an option and is enabled when Dolby is not enabled, DRC will only be implemented during playback if Dolby is in fact enabled? I want to be sure that I can simply tell a viewer, a projectionist, or whomever to "turn off Dolby" in order to turn off DRC as well as any other processing that Dolby might do.

Quote Hugh:

---

Can I suggest you do yourself a huge favour and either spend the time and effort reading through the full set of Dolby's papers on Dolby Digital and its metadata, or just ignore it completely.

---

The problem is that I am, in fact, forced to deal with it, as I'm forced to use Apple's Compressor, and unfortunately, reading up on Dolby at Dolby's Web site is not going to help much. I will explain...

Perhaps you're not aware of this, but for audiovisual works outside of major feature films and TV programs, Apple's Final Cut Studio has become the de facto standard for post production (although Pro Tools is used to generate soundtracks for all sorts of works). Part of Final Cut Studio is Compressor, which encodes video and audio for DVD. Given the circles that I move in, I must use Compressor, as the other folks with whom I'm working simply won't accept my using anything else. This is why I asked if you would please look at Compressor's instructions for choosing Compressor's various settings for Dolby processing, to see if they make sense, to see which settings should be chosen, etc. It's not quite clear in Compressor in light of what I now know about Dolby, thanks to you. (Thank you.)

Quote Hugh:

---

In film mixes, the dialogue is generally lower (to allow more headroom for the explosions), typically around -24 to -12dBFS... The replay level is adjusted by the dialnorm metadata to set the nominal dialogue to -31dBFS. That is entirely separate from the industry conventions for the source recording.

---

I realize that. But if I'm including a normalized audio track on the DVD of my project, and am not indicating any Dolby processing when setting up for encoding to MPEG2 (for creating a DVD), then should I prepare my audio track such that my dialogue is around -31 dBFS, or rather, in the range of -24 to -12 dBFS?

If I understand you correctly, in a mix for film, dialogue will be placed between -24 to -12dBFS -- say, at an average of -18 dBFS; and so during the creation of the DVD of that film, "-13" will be indicated as the "dialnorm offset", thus putting the dialogue down at -31 dBFS during DVD playback. But that would suggest that in mixing the audio for my project, I should place the dialogue at -31 dBFS, so that it plays back from DVD at that level too, right?

Thanks.

--------------------
Robert Eidschun

Quote Hugh:

---

Two's Complement is a binary counting system in which the upper half of the binary number range has been redefined to represent negative quantities.

---

I'm not sure why you refer to the "upper half" of the range, since, as you indicate yourself, it's only the left-most bit that represents the sign of the number.

By the way, when referring to what should be considered a number (as opposed to, say, status bits) that is represented in two's complement form, the most significant bit is the one just to the right of the left-most bit, since by definition, a significant bit is one that conveys information about the magnitude of the number in question. By contrast, the left-most bit indicates only the number's sign.

Quote Hugh:

---

The other significant advantage that stems from this redefined centre zero is that it makes the mixing of audio signals very simple, as concurrent samples can simply be added using pure binary calculations

---

This, or more generally, efficient calculation, is by far the the most important advantage, from the standpoint of design of a digital system, and was really the only reason why two's complement arithmetic was developed, allowing, for example, the multiplication of n and m through the addition of just two numbers, rather than having to compute m + m + m + ...

--------------------
Robert Eidschun

Quote Hugh:

---

Can I suggest you do yourself a huge favour and either spend the time and effort reading through the full set of Dolby's papers on Dolby Digital and its metadata...

---

I read through most of the white papers concerning encoding for stereo playback from DVD, and downloaded, installed and ran the encoder for Macintosh, reading all of the instruction pages and making sure that I understood all of the parameters to be set (for stereo playback from DVD). All seems clear, as it really isn't that complicated.

But what a viewer or projectionist might encounter is what still concerns me, as those instruction pages state the following:

"Full-featured decoders allow the consumer to select whether to use DRC at all, and which mode to use. The consumer sees options such as Off, Light Compression, and Heavy Compression instead of None, Line mode, and RF mode. Advanced decoders may also allow custom scaling of the DRC metadata."

I can just see a viewer or projectionist choosing the wrong combination of decoder settings...

By the way, I notice that Dolby uses the term, "decoding", in many contexts where there is no encoding of the audio data at all, but rather, only the creation of accompanying metadata that directs the playback device to process the audio on the fly, during playback. But I suppose that "decoding" refers to what is necessary in order to undo the "compression" that was applied (as a function of the data rate, etc.) during the "encoding" process.

--------------------
Robert Eidschun

Quote Robert Eidschun:

---

...his fundamental point, which is that somehow, my statement may be incorrect

---

That's not what he was saying. He said that he thought most digital audio systems used two's complement coding rather than straight binary that you were using in a rather laboured way to debate dynamic range -- and he is correct. They do. That's not to say that your examples of maximum and minimium quantising levels was wrong within the context of straight binary counting.

Quote:

---

and certainly, it is much easier to use that representation to illustrate my original point concerning dynamic range, rather than some "encoded" version.

---

Actually, it would have been easier to debate dynamic range using dBFS values, in my opinion -- a far more practical and meaningful context.

Quote:

---

"Encoded" versions (e.g., 2's complement) then are useful to facilitate representation and calculation using digital circuits. However, they are not useful for the purposes of conceptualization

---

Tell it to Dr John Watkinson...

Hugh

--------------------
Technical Editor, Sound On Sound

Quote rhd_toyl:

---

I'm not 100% sure about this, and i might just be missing something, but if you're using unsigned binary numbers, then wouldn't the "0000 0000 0000 0001" you mentioned earlier represent a large negative-going voltage?

---

Yes.... But can I suggest we drop this line of unhelpful and potentially very confusing argument.

Hugh

--------------------
Technical Editor, Sound On Sound

Quote Robert Eidschun:

---

By the way, as you may know, in information technology, it is standard to refer to the kind of data reduction that you refer to as "data compression".

---

...and as I said: ...data reduction (often erroneously called data compression)...

The term 'compression' implies that 'decompression' is possible. In a lossy codec such as mp3 or AC3, data have been lost and cannot be restored. These are data reduction codecs. There are forms of loss-less audio codec which are truly data compression codecs, such as MLP and Dolby True HD, but that's not what we were talking about.

Quote:

---

So, for example, mp3 is referred to as a "compressed format", created by "data compression".

---

Only by people who know no better

Quote:

---

...DRC is made available as an option only if Dolby is enabled

---

The Dolby DRC functionality is only applicable to a Dolby Digital soundtrack. Some home theatre systems, TVs and DVD players may have their own independent form of dynamic range reduction. It is important to recognise and differtiate the two, where both are present. The equipment handbook should make it clear what facilities are available.

Quote:

---

I want to be sure that I can simply tell a viewer, a projectionist, or whomever to "turn off Dolby" in order to turn off DRC as well as any other processing that Dolby might do.

---

If you aren't encoding a Dolby Digital soundtrack, then Dolby's DRC functionality is irrelevant and plays no part in the audio signal path. However, as I said, some TVs, players and other replay equipment may well have some alternative built-in proprietary dynamic range reduction system, which would normally be configured by the user during install if required.

Quote:

---

This is why I asked if you would please look at Compressor's instructions for choosing Compressor's various settings for Dolby processing, to see if they make sense, to see which settings should be chosen.

---

If you are not using a Dolby Digital soundtrack -- you stated your intention to use linear PCM -- then the Dolby Digital settings in Compressor are irrelevant.

Quote:

---

But if I'm including a normalized audio track on the DVD of my project, and am not indicating any Dolby processing when setting up for encoding to MPEG2 (for creating a DVD), then should I prepare my audio track such that my dialogue is around -31 dBFS, or rather, in the range of -24 to -12 dBFS?

---

The Dolby conventions are irrelevant if you are not using a Dolby Digital soundtrack, and in any case, don't affect the nature of the source recorded material.

Quote:

---

...and so during the creation of the DVD of that film, "-13" will be indicated as the "dialnorm offset", thus putting the dialogue down at -31 dBFS during DVD playback.

---

Only if a Dolby Digital soundtrack is being created, and that dialnorm value provides the optimal loudness correction.

hugh

--------------------
Technical Editor, Sound On Sound

Quote Hugh:

---

Actually, it would have been easier to debate dynamic range using dBFS values, in my opinion -- a far more practical and meaningful context.

---

My original point was only that I had adjusted my audio such that the maximum value of the waveform was represented by the upper limit of the system. By the way, in 16-bit AIFF data, isn't the upper limit in turn represented by 16 ones, and isn't AIFF data just a series of 16-bit words, each a sample value?

Quote:

---

"Encoded" versions (e.g., 2's complement) then are useful to facilitate representation and calculation using digital circuits. However, they are not useful for the purposes of conceptualization

---

Quote:

---

Tell it to Dr John Watkinson...

---

If he understood what I was saying, then I'm sure he'd agree with Dr. Robert Eidschun

--------------------
Robert Eidschun

Quote Hugh:

---

But if I'm including a normalized audio track on the DVD of my project, and am not indicating any Dolby processing when setting up for encoding to MPEG2 (for creating a DVD), then should I prepare my audio track such that my dialogue is around -31 dBFS, or rather, in the range of -24 to -12 dBFS?

---

Quote Hugh:

---

The Dolby conventions are irrelevant if you are not using a Dolby Digital soundtrack

---

So you're saying that having speech at -31 dBFS is appropriate if the recording is being Dolby processed during playback, but that if it's not, then it's more appropriate for speech to sound louder or quieter?(!).

The fact that dialnorm is designed to place dialogue at -31 dBFS during playback implies that the folks at Dolby have determined that -31 dBFS is indeed an appropriate level for dialogue during playback. In fact, they explicitly state that in one of their white papers, and explain that that is why they designed dialnorm to work the way it works. If I could find the citation again, I'd reproduce it here.

So, all I would like to know is if you think that -31 dBFS is an appropriate level for dialogue in my non-Dolby soundtrack as part of an audiovisual work for DVD. If not, which level would you suggest?

Thanks.

--------------------
Robert Eidschun

Quote Hugh:

---

...and as I said: ...data reduction (often erroneously called data compression)... The term 'compression' implies that 'decompression' is possible. In a lossy codec such as mp3 or AC3, data have been lost and cannot be restored. These are data reduction codecs.

---

Hugh, if you really want to debate symantics, then your use of the term "data reduction" is also erroneous The question is whether or not there is a loss of (or reduction in, if you prefer) information, and a reduction in data may or may not result in such a loss.

--------------------
Robert Eidschun

Quote Robert Eidschun:

---

So you're saying that having speech at -31 dBFS is appropriate if the recording is being Dolby processed during playback, but that if it's not, then it's more appropriate for speech to sound louder or quieter?(!).

---

No, that's not what I'm saying. I thought you said you'd read the Dolby papers on this.

In order to achieve reasonably consistent dialogue loudness levels between different programmes and materials, the Dolby Digitial decoder can apply a level offset controlled by the Dialnorm metadata value. In essence, the purpose of this level offset is to match all incoming material such that on the output, the average dialogue loudness equates to -31dBFS.

Dialogue is used as the nominal refernece material becaue of its commonality across most programme genres. Heavily compressed programmes will tend to have high dialnorkm values and so be turned down quite a lot, while quieter programmes won't be turned down as much.

Quote:

---

The fact that dialnorm is designed to place dialogue at -31 dBFS during playback implies that the folks at Dolby have determined that -31 dBFS is indeed an appropriate level for dialogue during playback.

---

They determined that -31dBFS is a reasonable point to aim at, covering the widest range of programme types with an attenuation-only system to match loudness levels.

Quote:

---

So, all I would like to know is if you think that -31 dBFS is an appropriate level for dialogue in my non-Dolby soundtrack as part of an audiovisual work for DVD. If not, which level would you suggest?

---

I have already given you the benefit of my opinion on that several times.

Hugh

--------------------
Technical Editor, Sound On Sound

Quote Robert Eidschun:

---

Hugh, if you really want to debate symantics

---

No I don't, but you clearly do enjoy it.

Quote:

---

then your use of the term "data reduction" is also erroneous

---

No, it is not.

Quote:

---

The question is whether or not there is a loss of (or reduction in, if you prefer) information, and a reduction in data may or may not result in such a loss.

---

Data reduction implies a reduction in data -- and that is exactly what systems like mp3 and AC3 are designed to do -- it is their entire raison d'etre.

Whether or not there is a reduction in informtion content is covered very nicely by the terms 'lossy' and 'loss-less'

The fact that a reduction in data may not result in a loss of inforamtion (becaues a loss-less codec has been used) doesn't detract from the fact that the amount of data has been reduced -- a data-reduction process has occured.

But at the end of the day, it really doesn't matter that much what terms we each choose to use provided there is sufficient context to extract the indemded meaning. And I think we probably both know what we are talking about within our own fields of expertise.

hugh

--------------------
Technical Editor, Sound On Sound

I remind you that you are the one whose raised the topic by criticizing the term, "data compression". By the way, I actually agree with your conclusion, but for different reasons...

In this most recent posting of yours, you're using the term "data reduction" to mean a reduction in the "amount of ones and zeros", if you will -- which is how most folks use the term, including me.

But in your previous posting, you used the term "data reduction" to mean a loss of information. I believe that most people who read that posting and take the time to think about it will interpret it in the same manner as I do, as your language is quite unequivocal.

In considering your two most recent postings, you have clearly drifted from explaining the merits of your positions, and so I suggest that we stop now, for the sake of remaining cooperative.

Sincerely, I really do appreciate all of your help.

--------------------
Robert Eidschun

Quote Robert Eidschun:

---

I remind you that you are the one whose raised the topic by criticizing the term, "data compression".

---

Because it appeared you had confused the terms 'data compression' with 'dynamic range compression' when you suggested that using Dolby Digital with DRC enabled would result in the sound track being 'compressed twice.'

I think we have resolved that confusion now, thankfully.

Quote:

---

in your previous posting, you used the term "data reduction" to mean a loss of information.

---

The only occasion I previously used the term 'data reduction' (as far as I can see) is this one:

Quote:

---

Dolby Digital uses data reduction (often erroneously called data compression) as a means of making the file sizes more manageable for storage or transmission. But this has nothing to do with dynamic range compression.

---

Which seems entirely accurate and consistent to me. I can't see it here, but I may have used the term 'data reduction' in relation to lossy codecs if the context was such that it was implicit, but I'm normally fairly fastidious about such things.

Quote:

---

Sincerely, I really do appreciate all of your help.

---

No problem. It can be frustrating to try to explain complex topics remotely and with a time lag, and I apologise if some of that frustration that has shown through in the posts. That wasn't my intention.

regards

hugh

--------------------
Technical Editor, Sound On Sound

I would just like to say thanks guys for an e conversation that went so far over my head I was looking at my brain.......but I read it to my 10 year old and she went straight to sleep!

--------------------
My head hurts!

Quote Hugh Robjohns:

---

Quote rhd_toyl:

---

I'm not 100% sure about this, and i might just be missing something, but if you're using unsigned binary numbers, then wouldn't the "0000 0000 0000 0001" you mentioned earlier represent a large negative-going voltage?

---

Yes.... But can I suggest we drop this line of unhelpful and potentially very confusing argument.

Hugh

---

Absolutely - another topic for another thread (maybe one day!)

Hi Hugh,

Initially, I wasn't sure if DRC was part of Dolby. Now I know that turning it on invokes the particular compression curve that is indicated in the metadata.

However, I have a DVD player that plays back Dolby audio but does not allow one to turn DRC on or off, and I've read that such players indeed apply the compression curve indicated in the metadata.

So, in the beginning, before I had ever seen "DRC" on a DVD player's set-up menu, I thought that perhaps the compression curve in the Dolby metadata was always applied to Dolby audio by all DVD players, and that any option on the DVD player's menu to turn something called "DRC" on or off was perhaps an option to apply compression "further downstream" to any audio on the DVD, Dolby or otherwise. That then would result in "double compression", if the audio playing back had been Dolby encoded and if the metadata's compression curve was being applied to it (before the additional "DRC" applied "further downstream"). But, as I say, I now know that DRC is not compression in addition to that resulting from the application of the compression curve in the metadata.

However, as you pointed out, some TVs allow you to turn on or off the compression of any audio, regardless of its type or origin, and so if applied to Dolby audio that has already been compressed by DRC, then the result will be "double compression", it would seem.

By the way, I've made some additional inquiries concerning film festivals, and it seems that at first, you submit a DVD for consideration. The judges then watch that and determine if your entry will "participate" in the festival. If indeed it is chosen for participation, then it will be screened, i.e. projected or shown somehow in a large room or theater for all to watch, in which case you'll be asked for your audiovisual work on HDCAM or some other high-end format.

Well, for the DVD that you submit, there seem to be no standards. I guess this is because the judges are probably going to watch on plain 'ol TVs, perhaps at home, whenever they have time, etc. and so you should simply aim for a dynamic range that seems appropriate for a TV in the living room of a house. In this case, I see no reason not to normalize the audio to full scale, as the judges can simply adjust the volume accordingly and things should be fine, provided that your dynamic range is narrow enough to allow a volume setting that is neither too loud nor too soft. Thus my figure of 45 dB, maximum, determined experimentally.

But concerning screening, there are horror stories out there of the volume being turned down too much, or the speakers sounding almost blown, or mono playback only, or so much noise among the audience as everyone keeps talking or incredibly loud hiss from noisy amplifiers that you'd better have your audio normalized and highly compressed such that everything is right near the top so that it can be heard -- especially the traditionally quiet sounds, such as the rustling of clothing. And yet on the other hand, there are also stories of how the audio reproduction at a certain festival was really good, but that unfortunately, the film maker had his audio highly compressed and so it sounded like a "squashed television program", rather than a "fine film worthy of projection in a proper cinema".

So, in short, it seems that the quality of the audio reproduction at film festivals is all over the map and that there is no way to anticipate what it will be. In fact, I contacted one such festival recently and was told, "there are no standards here... just do what you think is best and things will be fine...", which thus left me totally in the dark, at least concerning that particular film festival, its playback equipment and its practices.

Regards,

Robert

--------------------
Robert Eidschun

I think in such circumstances the best you can do is to make the most professional product you can, conforming to the relevant standards. If the replay system is poor, that's the organiser's problem, but at least they can't blame the source material.

Hugh

--------------------
Technical Editor, Sound On Sound

Quote rhd_toyl:

---

I can't really see how a long string of 1s could be full scale, since there'd be no numbers left to represent negative-going voltages.

Reading back over what I've just written has made me feel very confused - time for bed I think!

---

You're correct - in two's complement 16 bit, sixteen consecutive 1's (1111 1111 1111 1111) would actually be -1, with the largest magnitude negative number represented by 1000 0000 0000 0000 (-32768) and the largest positive is 0111 1111 1111 1111 (+32767).

For bits 0 to 14, they have their standard binary values (1, 2, 4, 8 etc) and bit 15 has the value -32768.

To negate a two's complement number, simply flip all the bits and add 1 (although this doesn't work for the largest negative, as it simply stays at -32768!)

This is convenient as it makes binary arithmetic on two's complement numbers largely the same as for unsigned numbers.

--------------------
* Soundcloud *
* Bandcamp *

OK. Thanks, Hugh.

--------------------
Robert Eidschun

Once again, the results of the computations done in digital signal processing are independent of the form of the representation of the amplitudes of the waveforms, and independent of the form of the representation of the numbers used in the computations, provided that that form uniquely enumerates each of the possible states that a "data word" can assume. For a 16-bit word, one can take the smallest value to be 16 zeros and the largest to be 16 ones -- or you can use two's complement representation, or a variety of others, which only makes the computation unnecessarily complex for humans.

For example, if x is y times greater than z, then that is true regardless of the representation of x and z, whether it be in base 10, base 2 (binary), or base 2 two's complement.

However, logic gates can more efficiently deal with numbers represented in two's complement form, and in one's complement form, for that matter.

So, full scale inside of a computer might not be represented by 16 ones, but humans can choose that representation anyway for the sake of performing calculations more conveniently. If the result that a human thus obtains is converted to two's complement, then it will be the same as that obtained by the computer that was using two's complement representation. Likewise, if the result that the computer obtains is converted to the representation that the human was using (e.g., base 10), then it will be the same as that obtained by the human.

I was merely confirming the other posters theory about the numbers, unrelated to your previous postings.

But to be truly pedantic, logic gates don't deal with numbers at all, just ons and offs!

--------------------
* Soundcloud *
* Bandcamp *

Yes, and it's the "on and off operations" that are far fewer in number (in order to carry out, for example, subtraction and multiplication) if the "on's and off's" of a memory cell or a register correspond to the two's complement representation of a number, rather than some other representation.

--------------------
Robert Eidschun

Indeed - basic arithmetic on 2's complement and unsigned integers are identical.

Right, I'm taking off my anorak now

--------------------
* Soundcloud *
* Bandcamp *

Hi Hugh,

I've had extensive discussions about the preparation of audio for entries submitted to two prominent film festivals, speaking with the technical director of one of the festivals and the programming manager of the other festival. What do you think?

Thanks.

Robert

**********

According to the technical director:

Most of the entries, whether on DVD for programming consideration, or on HDCAM for projection, have most of the audio between -10 dBFS and -40 dBFS, although peaks might be at full scale, and ambience or “room tone” might be at -60 dBFS. Normal dialog should follow the Dolby specs, even for PCM audio, i.e. it should be at an average of -31 dBFS, or -27 dBFS, which are two levels often adhered to.

The DVD may be watched on anything from a laptop computer to some super system with great audio reproduction. [This would suggest that a narrow dynamic range should be used for suitable playback on mediocre systems, such as a computer, although, he said that the audio used on the DVD and the HDCAM tape usually originate from the same mix, with the same relative levels among the various constituents, dialog, Foley, music and ambience.]

The HDCAM tape must be prepared so that it is Pro Logic compatible. By default, Dolby audio, prepared according to Dolby’s “e” specifications, will be compatible, and indeed, any Dolby audio on the HDCAM tape must be of the “e” type. In addition, a PCM, two-track mix is also required on the HDCAM tape, but it too must be Pro Logic compatible. (Dolby audio is not required for the tape, but PCM audio is.) Basically, the Pro Logic system will attempt to place all dialog on an additional track, to be reproduced on speakers that are centered in the theater; however, when this done, problems may arise due to phase relationships in the original two-track mix. [He did not explain how the dialog of a two-channel mix is identified by the Pro Logic system. Just because it’s centered, or because, in addition, it’s at a certain level? But regardless, wouldn’t that mean that any sound (e.g., Foley) that was centered and at the same level as the dialog would also be processed as the dialog was and thus be placed on the “center track”?]

During projection, the volume setting on the Pro Logic system will first be set to “7”, then adjusted upwards or downwards, if required, as the HDCAM tape plays. The volume scale on such a system is from 0 to 10.

I suggested that someone using Pro Tools might create a mix that has three submixes, or “stems”, as they are called in the audiovisual world, and he said that that would be appropriate: a mono stem consisting of dialog only, where that dialog was all recorded in mono; a mono stem consisting of Foley and ambience, where those elements were recorded in mono only; and a stereo stem consisting of music, thus making for four tracks in total. It is these stems that one would take to a post production facility to create audio for HDCAM, and from those stems, Dolby audio as well as Pro Logic compatible two-track PCM audio could easily be created (at the post production facility), where the latter would be without the phase problems mentioned earlier. [I suppose that relative levels between stems could also be adjusted then, compression or expansion applied to stems, etc.]

He also said that any latency in the digital projection system (which results in the image lagging the audio by as much as two or three frames) will be compensated for by the projectionist, and that any need to advance the audio with respect to the image, due to the slower speed of sound and thus its delayed arrival at the audience, will be done, once again, by the projectionist.

He added that the sound at the back of the theater may be 3 dB or so quieter than that at the front, and that theaters have an “x-curve”, which refers to its frequency response, and which results in there being a roll-off at the high end and at the low-end.

The festival has a spec. sheet concerning the audiovisual requirements for the HDCAM tape.

**********

According to the programming manager:

He said that the DVD for programming consideration will be viewed by the judges in a variety of environments, using a variety of playback equipment. Some judges view DVDs at home, perhaps on an “ordinary” TV, or on a home-theater system. If the submission makes it far enough “up the ladder”, then several judges might view it together in a small theater, having perhaps a 12-foot diagonal screen. The sound quality in such a theater is as good as that of a commercial cinema.

The sound quality in the theaters used for projection during the festival is also as good as that of a commercial cinema. People are well behaved and so “crowd noise” is minimal. The noise from the amplifiers is low. If the volume has to be changed during projection, it is usually adjusted downwards. Those theaters seat anywhere from 200 to 1,400 people. The 200-seat theater has a projection screen that is perhaps 30 feet across, measured diagonally. Short programs are typically shown in the smaller theaters.

He suggests PCM audio only for the DVD, since Dolby requires one to fiddle with the playback system, and he implied that the judges will not do that.

The festival has a spec. sheet concerning the audiovisual requirements for the HDCAM tape.

--------------------
Robert Eidschun

16 ones is NOT positive full scale for a 16 bit PCM audio signal in any standard representation I have ever seen in use.

16 ones is in fact -1 (2s comp.) and the range is -32768 to 32767

While some telephony stuff uses 8 bit unsigned signals, every linear audio system except that one exception I have ever seen uses 2s complement representation. CD audio, DAT, DVD PCM, SPDIF, AES3 they all use 2s compliment.

As a user there is however little point in worrying about this, just work in dB relative to some reference (0dbFS, +4dBu, whatever).

On the projection front, I don't know if you are shipping a 35mm print, but if so make good and damn sure the SVA track is present and correct (and at the right level), there are many faults that can lead to the dolby decoder temporarily cutting over to the analogue optical track, and if the level is wrong it tends to be bloody obvious.
If dubious about the projection rig then try to get this printed in silver rather then as a cyan dye track as there may still be some folks who have not replaced the exciter lamps with the lasers required to read the newer SVA tracks.

Also make sure that your track sounds reasonable via a phase matrix decoder as these often get accidentally left in circuit under the time pressure of a festival (Having the projector failover to SVA and having the phase matrix decoder decide that all the audio belongs in the surround chain is embarrassing).

Film festivals (even competent ones) make for horribly busy projection rooms and sometimes routine maintenance (Like say cleaning optics on the dolby head) gets missed. Just be glad when the film comes up on screen in the right ratio!

If you are printing to film, do not forget that frame rates differ between film and video! Seen it, it was embarrassing.

Regards, Dan (Who used to be a projectionist).

--------------------
Audiophiles use phono leads because they are unbalanced people!

Just saw your latest post, pro logic is 4 channels derived from a stereo pair by a combination of matrixing and filtering, the surround chain has a very limited frequency response (~400 - 7K IIRC) and relative phase and amplitude is used to steer the audio to the appropriate speakers.

Mono stems is the way to play, real (as opposed to panpot stereo will just confuse the decoder, as will any use of 'stereo enhancement' tools.

The normal layout is LCR + Sub behind the screen (which will be perforated) with a chain of small surround speakers around the auditorium.

Note that cinema audio will normally have pretty good bass response and this may well be better then your own monitoring chain, especially if your room treatment leaves a bit to be desired, you have been warned!

Regards, Dan.

--------------------
Audiophiles use phono leads because they are unbalanced people!

Thanks for your comments, Dan.

Playback for projection will be from HDCAM tape, not film prints.

As for the stems used to create the audio for the HDCAM tape, what about the incorporation of commercially recorded audio? I'm using it in my project, and it's all taken off of CDs, which means that it's all stereo. Should I simply combine the left and right channels to make it mono? If so, then in that case, I'd have the following:

- One mono stem consisting of dialog

- One mono stem consisting of Foley and ambience

- One mono stem consisting of music

(But it seems a shame to have to convert the music to mono...)

Thanks.

--------------------
Robert Eidschun

You may find that the commercial audio is already coded for Pro Logic, in which case it may actually have some surround content.

Try playing back the commercial audio via a pro logic decoder (very common as a mode on an AV system of the 5.1 variety) and see what happens.

The really key thing is to have that centre channel dialogue as a mono stem (possibly with additional stems for dialogue you want to place elsewhere on the screen), you normally want that routed to the centre speaker.

Regards, Dan.

--------------------
Audiophiles use phono leads because they are unbalanced people!

Hi Dan,

Thanks for your very practical advice.

I'm not so confident that the commercial audio will already be encoded for Pro Logic playback, as all of it is classical music, and in my experience, the folks in that domain typically create straight forward, "ordinary" PCM stereo mixes -- although they often sound great. For one thing, those folks dislike compression...

As for dialog, the technical director at one of the festivals said that if it was dialog that had been recorded in mono and was present on both the left and right tracks of a two-channel PCM mix, then it would sound strange in a theater if the theater had only ordinary stereo playback, which would result in hearing exactly the same dialog recording out of the left-channel speakers and the right-channel speakers. But I don't understand why that would sound strange, since ordinary stereo playback is all that theaters used to have -- say, up through the 1970s, and I remember watching films in theaters back then that sounded great, and surely some of the sounds had been recorded in mono and were present in identical form in both the left and right channels, no?

What about the levels that the technical director had cited? He said that he has sat in a projection booth and watched, observing, typically, that the levels are usually in the range of -10 dBFS to -40 dBFS, with occasional full-scale peaks, and ambience or room tone as low as 60 dBFS, which you can hear (and see on the meter) when there is no action. This would imply a total dynamic range of 60 dB, but I have it on good authority that the dynamic range for projection purposes is often more -- say, 80 or 90 dB. (Maybe the film maker was striving for a compromise, choosing a dynamic range that was suitable for both playback from DVD on a TV, and projection in a theater.) What have you observed as a projectionist, regarding levels?

Thanks.

--------------------
Robert Eidschun

Quote Robert Eidschun:

---

Most of the entries, whether on DVD for programming consideration, or on HDCAM for projection, have most of the audio between -10 dBFS and -40 dBFS, although peaks might be at full scale, and ambience or “room tone” might be at -60 dBFS.

---

Sounds entirely plausible.

Quote:

---

Normal dialog should follow the Dolby specs, even for PCM audio, i.e. it should be at an average of -31 dBFS, or -27 dBFS, which are two levels often adhered to.

---

I'm a little surprised he is suggesting dialogue should be that low, but it's workable.

Quote:

---

The HDCAM tape must be prepared so that it is Pro Logic compatible.

---

Oh dear... that's going to cause you some additional problems then.

Quote:

---

By default, Dolby audio, prepared according to Dolby’s “e” specifications, will be compatible, and indeed, any Dolby audio on the HDCAM tape must be of the “e” type.

---

Dolby Pro-Logic is an entirely different beast from Dolby E. There seems to be confusion here. Pro-Logic is an analogue matrixed surround system, producing four (not discrete) channels from a matrixed pair. Dolby E is a data reduced (but much better than Dolby Digital) digital discrete track system that also involves inherent processing delays that need to be addressed carefully if lip-sync is to be maintained. Dolby E also carries all the same metadata info as Dolby Digital, but it is not usually used to control the replay specifications.

Quote:

---

Basically, the Pro Logic system will attempt to place all dialog on an additional track, to be reproduced on speakers that are centered in the theater; however, when this done, problems may arise due to phase relationships in the original two-track mix.

---

He's not kidding. If you attempt to process a conventional stereo mix through Pro-Logic you'll find that the frontal image width is narrowed considerably, and any spacious or ambient atmos effects will be thrown to the rear -- often in quite a dynamic and disconcerting way.

If you're using commercial classical music as a source, you'll find that a lot of it will thrown much of the orchestra and reverb to the rears. It might give a nice envelopment effect, or it might just be very annoying -- and if there happen to be closed mic soloiststhen when they start up the sound will probably swim about all over the place! Dolby Pro-Logic uses a process called 'active steering' and it is easily fooled into sending sounds to all sorts of unexpected places around the room!

As a result, it is a mandatory Dolby requirement that tracks intended for Dolby Pro-Logic reproduction MUST be mixed whilst monitoring the entire encode-decode chain in order to spot and rebalance any problem sections.

Quote:

---

He did not explain how the dialog of a two-channel mix is identified by the Pro Logic system.

---

Once again, you'll find all that information in Dolby's relevant technical papers

Quote:

---

But regardless, wouldn’t that mean that any sound (e.g., Foley) that was centered and at the same level as the dialog would also be processed as the dialog was and thus be placed on the “center track”?

---

Yes. Anything which is of the same level and phase in both source channels will be directed to the centre channel on reproduction. Anything which is the same level but opposite polarity (or very non-coherent) on the two channels will be directed to the rear surround channels on reproduction.

Quote:

---

During projection, the volume setting on the Pro Logic system will first be set to “7”, then adjusted upwards or downwards, if required, as the HDCAM tape plays.

---

7 eh? very professional...

Quote:

---

He also said that any latency in the digital projection system (which results in the image lagging the audio by as much as two or three frames) will be compensated for by the projectionist

---

I'd expect nothing less.

But there will also be a time difference of at least one frame and possibly two (depending on how the tape is produced) between the picture / Dolby pro-logic audio and the Dolby E audio. I trust that will be addressed too.

Quote:

---

and that any need to advance the audio with respect to the image, due to the slower speed of sound and thus its delayed arrival at the audience, will be done, once again, by the projectionist.

---

ha ha ha... How many thousands of metres away from the screen will the audience be then?

Quote:

---

He added that the sound at the back of the theater may be 3 dB or so quieter than that at the front

---

More confusion. Sound intensity diminishes with distance following an inverse square law. The level of frontal sounds at the back of the theatre will be a lot more than 3dB lower than as heard at the front.

What he is actually talking about is the replay level alignment of the rear channel speakers, which he appears to be suggesting follows the normal Film practice of being 3dB down, as opposed to the TV standard which is for equal levels all round.

Quote:

---

and that theaters have an “x-curve”, which refers to its frequency response, and which results in there being a roll-off at the high end and at the low-end.

---

A historical process to minimimse analogue SVA track noise on replay. Consequently a compensating inverse X-curve should be applied in the monitoring chain during film mixing so that the replay to the audience comes out right. This also compensates for the inherent reduction in HF when listening in a large room far away fromthe speakers. Again, TV mixes are made flat, as reproduction at the home is in a small room and much closer to the speakers, so atmospheric losses are irrelevant.

The world of audio dubbing and mixing for film and TV is a very complex one, as you are probably beginning to realise, and there is a huge amount of confusion amongst those on the periphery and even some misinformation or misunderstanding from some of those involved directly.

Dubbing mixers are not just skillful at creating an artistic balance, but also very highly trained in the complex technical requirements of all the different release formats.

hugh

--------------------
Technical Editor, Sound On Sound

Quote Robert Eidschun:

---

...all of it is classical music, and in my experience, the folks in that domain typically create straight forward, "ordinary" PCM stereo mixes

---

But they also tend to prefer spaced mic arrays which produce a lot of non-coherent sound, or coincident arrays that inherently produce high Side signals. In both cases, the reverb, ambience and some of the frontal image will be misdirected by the Pro-Logic decoder to the rears, often with varying image widths and spaciousness as the material dynamics change. As Dan says, listen to the material through a Dolby pro-Logic decoder to check if anything untoward happens.

Quote:

---

As for dialog, the technical director at one of the festivals said that if it was dialog that had been recorded in mono and was present on both the left and right tracks of a two-channel PCM mix, then it would sound strange in a theater if the theater had only ordinary stereo playback, which would result in hearing exactly the same dialog recording out of the left-channel speakers and the right-channel speakers.

---

The reason the film industry adopted the centre channel for dialogue in the 1940s was precisely to avoiud this issue. In a large theatre, pure stereo replay will result in people seated near the edges hearing dialogue from the closest edge of the screen due to the 'Haas effect.' It can be very disconcerting if there is a talking head in a big closeup on the screen with the voice coming from 20 feet to the left!

Quote:

---

But I don't understand why that would sound strange, since ordinary stereo playback is all that theaters used to have -- say, up through the 1970s

---

No. Most modest cinemas were mono before 1976 and the better ones were discrete 4 or 6 channel -- both with front centre speakers for dialogue.

Hugh

--------------------
Technical Editor, Sound On Sound

Thanks for your comments, Hugh. I've responded below, in-line.

Robert

*********

Quote Robert:

---

Normal dialog should follow the Dolby specs, even for PCM audio, i.e. it should be at an average of -31 dBFS, or -27 dBFS, which are two levels often adhered to.

---

Quote Hugh:

---

I'm a little surprised he is suggesting dialogue should be that low, but it's workable.

---

I've read that having the dialog that low for projection purposes is not all that uncommon in North America, where the film maker wants to leave lots of headroom for explosions and other loud sounds. Indeed, whenever I go to the cinema here in the States, I am struck by the difference in volume between normal dialog and loud sounds: the former is loud but not too loud, while the latter almost blows my eardrums out. In fact, I sometimes leave the seating area and go find an attendant in the corridor, then ask him to ask the projectionist to turn down the volume.

Quote Hugh:

---

Dolby E also carries all the same metadata info as Dolby Digital, but it is not usually used to control the replay specifications.

---

He also mentioned this.

Quote Robert:

---

Basically, the Pro Logic system will attempt to place all dialog on an additional track, to be reproduced on speakers that are centered in the theater; however, when this done, problems may arise due to phase relationships in the original two-track mix.

---

Quote Hugh:

---

He's not kidding. If you attempt to process a conventional stereo mix through Pro-Logic you'll find that the frontal image width is narrowed considerably, and any spacious or ambient atmos effects will be thrown to the rear -- often in quite a dynamic and disconcerting way.

If you're using commercial classical music as a source, you'll find that a lot of it will thrown much of the orchestra and reverb to the rears. It might give a nice envelopment effect, or it might just be very annoying -- and if there happen to be closed mic soloiststhen when they start up the sound will probably swim about all over the place! Dolby Pro-Logic uses a process called 'active steering' and it is easily fooled into sending sounds to all sorts of unexpected places around the room!

As a result, it is a mandatory Dolby requirement that tracks intended for Dolby Pro-Logic reproduction MUST be mixed whilst monitoring the entire encode-decode chain in order to spot and rebalance any problem sections.

---

It sounds like I should plan on creating, or having created Dolby E audio for any projection version of my project that I might need, and simply forget about having two-track PCM audio run through Dolby Pro-Logic.

Quote Robert:

---

During projection, the volume setting on the Pro Logic system will first be set to “7”, then adjusted upwards or downwards, if required, as the HDCAM tape plays.

---

Quote Hugh:

---

7 eh? very professional...

---

I didn't understand his comment either, because he also mentioned having a tone at the beginning, where the level of that tone is the same as that of normal dialog. Now, as I understand it, such a tone is used to set the volume of the playback system, and so why would the projectionist simply assume that a volume setting of "7" was appropriate?

Quote Robert:

---

He added that the sound at the back of the theater may be 3 dB or so quieter than that at the front

---

Quote Hugh:

---

More confusion. Sound intensity diminishes with distance following an inverse square law. The level of frontal sounds at the back of the theatre will be a lot more than 3dB lower than as heard at the front.

What he is actually talking about is the replay level alignment of the rear channel speakers, which he appears to be suggesting follows the normal Film practice of being 3dB down, as opposed to the TV standard which is for equal levels all round.

---

I figured that this was due to the playback level at the rear.

--------------------
Robert Eidschun

Quote Robert:

---

...all of it is classical music, and in my experience, the folks in that domain typically create straight forward, "ordinary" PCM stereo mixes

---

Quote Hugh:

---

But they also tend to prefer spaced mic arrays which produce a lot of non-coherent sound, or coincident arrays that inherently produce high Side signals. In both cases, the reverb, ambience and some of the frontal image will be misdirected by the Pro-Logic decoder to the rears, often with varying image widths and spaciousness as the material dynamics change. As Dan says, listen to the material through a Dolby pro-Logic decoder to check if anything untoward happens.

---

As mentioned in my previous reply, it seems to me that I should just have Dolby E audio on any version of my project intended for projection. But if I really wanted to stick with a two-channel PCM mix, then could I get 'round the problem that you mention above by mixing all of the music (which is stereo) to mono and then placing the result on each of the two tracks? By the way, I've tried mixing the music to mono and listening to it, and it sounds fine, with no funny cancelations, etc.

Quote Robert:

---

As for dialog, the technical director at one of the festivals said that if it was dialog that had been recorded in mono and was present on both the left and right tracks of a two-channel PCM mix, then it would sound strange in a theater if the theater had only ordinary stereo playback, which would result in hearing exactly the same dialog recording out of the left-channel speakers and the right-channel speakers.

---

Quote Hugh:

---

Most modest cinemas were mono before 1976 and the better ones were discrete 4 or 6 channel -- both with front centre speakers for dialogue.

---

Does this mean that if there was only one channel of audio, then that lone channel was routed only to speakers that were up front and centered, so that all of the audio came out of them and no audio came from any place else?

Thanks.

--------------------
Robert Eidschun

Quote Robert Eidschun:

---

As mentioned in my previous reply, it seems to me that I should just have Dolby E audio on any version of my project intended for projection.

---

At least that way you would have absolute control of what comes out of what speaker. But you will need to clarify the picture/dolby E sync requirements beforehand.

Quote:

---

But if I really wanted to stick with a two-channel PCM mix, then could I get 'round the problem that you mention above by mixing all of the music (which is stereo) to mono and then placing the result on each of the two tracks?

---

Sure... if you want pure phantom centre mono music. Seems a waste to me!

Quote:

---

Does this mean that if there was only one channel of audio, then that lone channel was routed only to speakers that were up front and centered, so that all of the audio came out of them and no audio came from any place else?

---

Yes... for a 35mm mono SVA soundtrack film -- which was what most of them were in the provincial cinemas.

Before the introduction of Dolby Stereo (now more often referred to as Pro-Logic) which happened worldwide in 1975/76, 35mm release films generally had either a mono optical soundtrack, or a four-channel magnetic stripe (left, centre, right and effects).

70mm film generally had a six channel magnetic stripe (plus mono optical) giving left, inner left, centre, inner right, right and effects.

Magnetic stripes on films were very expensive to add and wore out quickly, which is why they were only used in the premier theatres.

Dolby revolutionised the industry by first introducing Dolby A (later SR) noise reduction to the optical soundtrack. This provided such an improvement in signal-noise that they could then afford to halve the track width whilst maintaining an acceptable signal-noise ratio. That allowed the printing of two audio tracks on the film -- which is easy, cheap and reliable.

But the film industry wanted at least four channels. The centre front was essential, stereo across the front was important, and the enveloping effects channel was nice.

So Dolby modified some of the ideas that were kicking around at the time to do with phase-amplitude matrix quadrophonic systems, and that lead to the Pro-Logic technology and the introduction of Dolby Stereo, providing the industry with the familiar LCRS format. (The first Star Wars film being one of the earliest really well known adopters).

However, because of the de-matrixing process, it is inherently a compromised system that does weird things under certain circumstances, which is why the mix has to be monitored through the complete encode-decode chain to ensure that any funnies are caught and corrected before the final print.

Dolby Pro-Logic does what will appear to be unpredictable things to conventional stereo material not intended for pro-Logic reproduction.

Later digital incarnations of the Pro-Logic system (such as Dolby's PLII, but there are others) manage to handle conventional stsreo material a little better, but it can never be as reliable as a discrete channel system... which is why Dolby Digital was introduced to the film industry, and then Dolby E to the TV production industry.

hugh

--------------------
Technical Editor, Sound On Sound