Dolby LM100 and Discovery deliverables

[medmondson]

Hello-

I am Senior Audio Mixer and Audio Post Production Supervisor for Discovery Communications’ Production Center in Silver Spring, MD. This thread was recently brought to my attention and it is my hope to be able to shed some light on Discovery’s implementation of the LM100 in its own facilities, as well as its expectations and needs from audio vendors.

In addition, I will respond to a few key posts that are representative of issues that are relevant not only for Discovery deliverables, but for the television/cable broadcast industry as a whole. Quotations of posts are indicated by <>.
-----
Discovery installed Dolby LM100 units in all of our offline, online, QC, mix, dubbing and machine room facilities in Spring/Summer 2005 as our first tangible step towards resolving the escalating problem of uneven broadcast levels (level shifts between programs) as well as dynamic compression issues at affiliates. This rather expensive and technically challenging endeavor (over 80 LM100s were implemented) was undertaken after conducting serious research into the nature of viewer complaints.

It was and is our feeling that the best starting point for solving broadcast level discrepancies is for major networks to begin adhering to a standard means of measurement- one that accounts for perceived loudness as opposed to pure electrical levels.

While it certainly can be debated as to whether the LM100 is the ideal tool for providing a system that will bring us all closer to the holy grail of resolving all broadcast level issues, we felt that it does in fact provide a level of objectivity in assessing a mix that can not be accomplished by means of traditional VU, Peak, and RMS metering.



<< Personally, I hate someone telling how to mix my shows. I believe if I feel a program sound have loud DX, I should be able to mix it like that>>

First and foremost, we train our users to recognize that the values provided by the LM100 do not, under any circumstance, represent a “good mix/bad mix” reading. While it is certainly tempting to desire such a unit, especially in a high-volume network, currently no such tool exists.

More importantly, we do not feel that the NEED for such a tool exists. It is not Discovery’s desire to stunt or eliminate artistic and creative freedom from audio mixing.

At the same time, there IS a certain level of objectivity that CAN be applied to audio; specifically, the thresholds of audibility and inaudibility. One way this can be measured is through careful consideration of viewer surveys and complaints.

Consider the following:

* A “hot” mix (defined as a mix where dialog is difficult to comprehend due to over-loud music or effects/M&E) is likely to make a viewer change channels within 4 minutes when sustained at a level of 83 dB SPL, the average home listener’s level.

* Audio is the single most important technical factor governing viewer retention; bad video edits or color correction, etc. are not nearly as influential in keeping a viewer tuned to a program as are bad audio edits, compression or EQ.

* For every 1 viewer that takes the time to write and complain, nearly 1000 simply switch the channel.




It is a common belief among some Producers that a hotter mix creates a more positive viewer response. All research indicates that this is not true.



<< I find Discovery's new needs laughable! They still use so much dynamics processing during broadcast that it does not matter what you send them! Maybe they are reconsidering this practice, but I doubt it because they still have all of their existing materials to recycle.>>

Discovery does not apply dynamic compression or limiting to any of its programming at transmission for any of its networks. All programming is passed through at a dynamic ratio of 1:1, having first been assessed by our QC facilities. If we discover a show’s dynamic range is too wide, or its overall level is too low/high, an attempt is made to allow the vendor to correct this. If this is not possible, we send the program through our in-house audio department. However, none of these methods employ severe dynamic processing; more likely, the compression you are hearing is being applied by a local cable affiliate. Some affiliates pass Discovery’s audio nearly perfectly, while others, frankly, crush it.

It is important to note that we at DPG do not attempt to absolve ourselves of responsibility for this and simply blame someone else, actually the contrary is true- and this points to the entire reason for the purchase of the LM100s. Because of this tool, we feel we CAN do something to PREVENT or MINIMIZE the artifacts caused downstream by affiliate compression.



<>

Most affiliates apply broadband dynamic compression. By nature, broadband compression can’t and does not affect RELATIVE levels between music, effects, and dialog. Some affiliates, however, implement MULTIBAND dynamics, like the Aphex Dominator. Such devices compress sections of the frequency spectrum independently, hence a bass-heavy mix can actually end up drowning out the midrange, because low frequencies are being brought up to the same level as the mids. A mix that is “on the fence” or “hot”, i.e. the ratio of M&E to DX is close, can easily be tipped into the negative.

Here is a quote from our training documentation:

“Like all cable networks, Discovery distributes its signal via satellite to thousands of cable systems around the country. Each cable system takes the satellite audio signal and converts it into a cable signal that is distributed to their affiliates. As part of this conversion process, signal compression is often applied to the signal… While this process generally preserves the relative dynamics of the mix it allows the audio signal to fit into the smaller transmission space provided by the cable system… In order to ensure that audio is passed transparently to the home consumer through the cable system, the audio needs to be confined primarily to a “safe area” in the middle of the transmission range. In particular, the spoken language elements of the program need to be confined to this safe area to ensure that they are passed on to the consumer without adjustment. While DCI has always maintained its technical specifications with this aspect of cable broadcast in mind, measuring the dialogue level as part of a full program mix has always been a technical impossibility until recently, making this part of the evaluation very subjective and inconsistent.”



Our research shows that program material (NTSC) averaged to a dialog level between -26 and -28 places dialog in a safe “pocket”; one that is a good compromise between a) resisting affiliate compression, b) resisting modulation errors caused by the multiplexing of several AC3 signals together and c) preserving the original relative levels of the mix.

Broadcast mixing differs significantly from music mixing in this regard- while it is common for people to listen to “master quality” audio at home on CD, broadcast Mixers must build a certain degree of “safety” into their mix that anticipates the myriad dynamic and spectral processing that occurs in television.


<< Or even you can make the first half 6 db hot and the other half 6 db low and get the same result. Unless the algoritm has some "intelligence " to detect "eveness" (is that a word?) thru time>>

This is true. Of course, it is my hope that people use the LM100’s measurement as a tool to help them create mixes with consistent levels, rather than simply regarding the mix process as a pursuit of an LM100 rating.

We should, as Audio Mixers, hope that our ears will always be a better guide than any meter; and that our mixing decisions should always be trusted over those of a machine. As Mixers, we pride ourselves on the quality of the sound we create; making a mix that simply “tricks” a machine is neither challenging nor productive.

Another quote from our internal LM100 training document:

“Dialogue levels are only one element in determining the relative loudness and equality in a mix. The LM100, while it provides useful information, does not provide a magic score that determines whether or not a mix is good or even balanced. Dialogue levels need to be combined with the peak and VU measurements to form a useful picture of the audio levels for a program, and the relationship between all of these measurements needs to be combined with critical listening and judgment to determine the validity of a mix. Average and peak measurements must continue to fall within the levels set by current standards. The Dialogue Level score provided by the LM100 is an additional point of data that allows DCI to more objectively judge the volume level of spoken language in the program.”



<< Meters like the LM100 will in my opinion allow a mix to be done without look ahead limiters being put on the final output.>>

We regard the LM100 as part of the Mixers’ and Broadcasters’ toolbox, not the end-all solution to broadcast level discrepancies. We believe that this tool does a respectable job at providing data about a mix that no other meter can. We have also found that when we keep mixes within a certain safe zone, they translate better to the viewer.

<< However its a shame that the tool for measuring this level is tied to one manufacture>>

I do not disagree that some very good ethical questions arise from making a private company’s proprietary technology an industry standard. However, thus far, no other manufacturer has, to my knowledge, tackled the issue as effectively as Dolby. Most post houses use Sony-format tape machines, Digidesign/Avid editing platforms, Microsoft data formats for text, Adobe for graphics, etc. Most would probably agree that this standardization has greatly eased inter-facility operations and opened the door for creative freedom, not the other way around.

As Mixers, I think we’d all rather spend our time focusing on creating great mixes rather than fighting the unpredictability of the broadcast chain.

I, personally, would like to see Dolby make more of an attempt at making the LM100’s technology available to more vendors and smaller post houses, in the form of a Pro Tools plug-in. Thus far, I know of no such plans. Perhaps enough requests from users will change that.

While it can be debated as to whether Discovery’s choice of using the LM100 as a standard is the right one, initial results are very positive: viewers’ audio complaints have decreased significantly for us. But the LM100 standard should only be seen as a first step in controlling the extraordinary amount of level discrepancy that exists today in broadcast. A long road lies ahead.

Our motives are nothing more than the desire to allow maximum creative freedom while operating in a medium subject to hundreds of variables, while generating a minimum of viewer complaints, thereby (hopefully) achieving the goal we all want: for viewers to hear everything we intend them to hear.

Thank you for the opportunity to share this information. I hope you find it helpful.

Mark Edmondson
Supervisor, Audio Post Production
Discovery Communications Production Center
Silver Spring, MD

[1150Post]

Mark,

Happy New Year! Many thanks for your time and info. Hope all is well!

Your posting is wonderful, and goes a long way towards advancing the credibility of standardizing broadcast levels. Information is certainly the key for this concept to gain widespread acceptance. But it can be difficult when a network (any network, not just DCI) puts vendors in a push-pull scenario.

Production budgets are decreasing. Mixers are required to do more, with less, faster. HD/Surround has not produced the increases in time/budget that many of us hoped for. To that end, many of us are indeed quite proud of the "tricks" we've challenged ourselves to devise and implement to balance the budgets, creative expectations, technical requirements, and personal creative satisfaction requirements present in mixing the various programs that run through our rooms. Some of us refer to these tricks as "innovation to meet market demands." But it still reflects the same desire to mix good audio you mentioned.

What I feel this community is looking for is an authoritative bit of information that lays out what and how the LM-100 is measuring. Mark, you know me well enough to know that, while I can be hard headed, I'm no schmuck! And I've been fortunate to pass 6 shows through the new standard with no repercussions. But I've got another potential series coming up that will involve a lot of run and gun, camera mic quality production dialogue, with little to no narrator. I'm sure it will be squashed and noisy. So right out of the gate, as I perceive this measurement standard, I've got problems. But that's based on discussion here and elsewhere, and reading a rather vague manual on the device.

A better understanding of how the box knows if someone is talking, and how it deviates from a standard Leg(A) measurement would be great in reinforcing the knowledge, experience, and confidence of seasoned mixers who want to make this work.

We have discussed the LM-100 implementation idea, both here in the DC office, and also in the larger Arlington facility. I know I would prefer to wait until it can happen in the box, as a plug in. A better understanding of how the system works would buy us all the time we need to see if such a plug will happen, and possibly allow us to adapt existing tools in our facilities to create a more positive real world simulation until we know for certain if the technology will be licensed to to other vendors and/or made available in different forms other than the LM-100. Perhaps DCI could/should put pressure on Dolby to make this easier for vendors to implement by providing additional information, in writing, or additional methods of implementing the measurement algorithm beyond the LM-100. DCI is primarily the force that took this meter from "nice-to-have" status, to "must-have" status. And you've no doubt been the catalyst for additional sales since then. No one wants to plunk down $3K for a box that is difficult to implement into multi-room environment, unless you buy multiple boxes, only to turn around and find out there will be a plugin for 1/4 as much that could live on each system. The ability to derive similar readings, even if they are not exact, through existing hardware by utilizing a greater understanding of the process would certainly help smooth the waters as well as to buy a little more time to see how this all shakes out equipment wise..

I believe that, for the majority of us at least, we all want to make this work. It's the smoke and mirror, don't look at the man behind the curtain aspect of the unit itself that is frustrating so many of us. Dolby certainly has the lock on most aspects of multichannel audio, storage and transmission and delivery, both on DVD and in the newly developing digital TV transmission scheme, and they should be doing quite well financially for it. And after all of that money is counted, frankly, a $3K single purpose meter is a touch insulting.

Thanks again, Mark, for stepping up and taking the lead on this. I fear for your messenger status, though, and hope that you don't get mauled for delivering this info. I know I've tried not to, and know fully that, as usual, we both just want the same things. Feel free to call me at 1150 if you want to discuss this post offline, or anything else, further.

Take Care!

[Richard Fairbanks]

Mark,

Thank you so very much for your exceptionally clear and well written thoughts. Discovery's efforts, your efforts, are laudable. You'll find no argument from anyone in this forum about that. I personally thank your company for trying to get the ball rolling and improve the uneven volume problems.

If you can, would you mind telling us whether you are measuring 5.1 streams or stereo streams, or both, in your QC installations? As it is possible that the measured values will be different between them, I am not the only one who is unclear about which we should monitor during mixes for your network.

[bigbadhenchman]

Allthough I think it's greta bthat you ahve coem here to explain the reasoning why, I again will have to pint back to 2 problems.

1. It's the enormous jump in level when going to commercials that pisses people off more than any difference between programs.

2. Most mixers I know strive to produce mixes that meet the old standard, at a level where there are pretty comparable levels between programs across the board.
In my very personal opinion, the problem has gotten worse because production companies are having the audio portion of the show done by very inexperienced people because they're cheap, or in Avid suites by the picture editor. I know for a fact that shows have and are being done at a local Audio school by the students.
When this problem started poking it's head up, Discovery should have let it be known that this would not be accepted. Now, facilities have to spend a considerable amount of money to have another level of QC for shows where the budgets, at least here in town, have dropped to half of what they used to be or less.

This I have a problem with. If Disovery wants to ensure that a certain percent of the Budget is spent on Audio as part of their contract with the producer, then at that point it's justifiable to force studio's to by the gear. Right now the deliverables are getting out of control, while budgets are shrinking.

[1150Post]

Mark,

One more thing. I understand everything you've written about standards and training, etc. at DCTC. If you could explain the following comment I received on one of the episodes that I passed through QC and how it constructively fits into the goals of DCI, it would be appreciated.

From the QC report.

In the comments section of the report, it is stated that the audio is unacceptable and should be remixed because...

"Remix audio due to high LM-100 reading at -24, low vu's at -26 dbfs and peaks at -12 dbfs."

I realize the show passed despite these comments, and that these are just the comments of a QC operator. But the LM-100 puts my average dialogue 2 db's hot, the operator is compaining that the average vu on the show is too low, and I've got to explain all of this to a client!

Thank god they've already aired!

Help a brother out!



edit....

Oh yeah, they were primarily voice and music, with archival television promos cut in as well.

Thanks!

[medmondson]

Hi there-

Brad, I hope the New Year finds you and your family well!

I don't claim to be the voice of the industry on this, and I am not a Dolby representative. I hope my first post did not come across as such. I was only hoping to share Discovery's applications of the unit and help those of you who are our vendors to meet the new spec. It is my hope that we can all learn from each other.

<>

4/5 of our audio complaints (2,200 complaints logged over 16 months) are for buried dialog and overly dynamic mixes.

I hope it was clear in my first post that our spec is a first step- for us. I'm personally happy that we took the iniative, but Discovery can not fix broadcast audio alone. We require all of our internal promos to also have consistent LM100 scores, but we can not control inserts. Here in DC, the local Fox News inserts play back at full code on Starpower HD. Ouch. We went with the LM100 because it provides information about a mix that no other meter does; information that does seem to actually be useful and have direct results.

Someone, at some point, has to start striving for audio level standards that actually work.

Take Dialnorm in 5.1 broadcasts: it's incredibly easy to enter a false metadata value, say, -31, when your material is actually much closer to -21. This will result in a nearly FULL CODE output at the set-top. There is no policing agency that can stop a Mixer or Producer from entering "malicious" metadata like this; we have to, at some point, all agree that what is best for ALL of us is when viewers don't have to touch the volume knob, rather than focusing on trying to make our mixes play hotter than the last guy's.

How long is it before Producers get hip to the fact that they can tweak their material to play hotter just by encoding inaccurate metadata? It didn't take long for brickwall limiters to make a Mixer's primary raison d'etre to crank up mixes to +2. Metadata is just the next freight train on the horizon unless we all start agreeing that there should be a standard for perceived loudness and matching our programming to that standard.

All research, and I do mean ALL research, indicates that viewers are more likely to pay attention to consistent commercial/show levels than to ones that fluctuate madly. Producers are constantly asking Mixers to crank things up, when really this does nothing in their favor.

It's like some kind of herd mentality- all commercial/promo Producers "know" that this is what they should ask their Mixer to do, without ever taking the time to look at the research and see how viewers actually react to commercials that blast (they reach for the mute button!).

<>

Brad, I couldn't agree more. The "pay no attention to the man behind the curtain" stance of Dolby is quite frustrating. When Richard here discovered a >4 dB inconsistency in the loudness measurements made by the DP570 vs. the LM100, Dolby's response was "yes, we knew about that already"... which left us wondering if we were ever going to have been told about it. They very casually suggested that we re-encode our entire library to the 'new', correct metadata. I think there is a disconnect between users and designers.

We Audio people are the type that make a living fixing problems- technical problems- and we by nature want to know how things work.

I unfortunately have no technical details whatsoever on how the LM100 works, other than what I have been told: it is based on a learning algorithm that was designed and modified over several years to 'learn' and identify patterns that quantify human speech of many languages. It uses a combination of temporal, spectral, formant and frequency readings.

Being skeptical, I spent some time one day trying to figure out just where its threshold lies- exactly where it ceases to recognize speech- and it is, admittedly, very good. I tried different means of alteration/distortion; EQ, panning, amplitude, delays, etc, to try to mess it up. I found that it is almost impervious to spectral distortion and panning- you can EQ a voice all the way down to a single harmonic and the LM100 will still recognize it. A single 100ms tap of echo won't fool it, but I found that 2 taps is where it starts to fall apart.

This is the about all I know. But... I do have to say that I don't frequently disagree with its measurements. I've been using it for a while now, and if I balance a string of voices out using its readings, I'm almost always in agreement with it.

We Post Mixers really need to keep pushing Dolby for technical details and a more affordable alternative. While they resist it right now, perhaps with enough outpouring from the industry they will realize that giving us these things will benefit them.

Part of me understands that developers need to keep their technologies guarded in order to stay profitable, but this particular issue straddles a weird political line that continues to cause discontent. It's hard to follow technical instructions when you aren't given the details you want.

I also understand the frustration caused by doing something one way for a decade or more, and all of a sudden being told that one's methods must be altered to fit a new spec. But really, our spec is not THAT different. Most Mixers tell me that they discover their mixes jive with the new spec as long as they don't try to constantly push them to 0VU. If you think about it, the levels needed to make a good LM100 score are not terribly different than the levels we ran in the days before the Finalizer.

And really, I think that's what this is about, I see this spec almost as a backlash against the brickwall limiter- a tool which has gotten so out of hand that it has caused more problems than it ever solved. If your brickwall limiters are only attenuating -2 to -4, chances are your mix will fall directly in line with our spec.

<>

We QC stereo and 5.1 programs with the LM100. The units are set read Center, with Dialog Intelligence on, using long-term measurement. Our mix suites operate with the units in short-term while mixing, and long-term while monitoring a layback.

<<"Remix audio due to high LM-100 reading at -24, low vu's at -26 dbfs and peaks at -12 dbfs." >>

For the last several months, Josh and I have begun a much more vigorous and intensive audio QC training program. Audio vocabulary is one objective, as is understanding the technology. I'd say if this happens again please drop me a line.

-Mark

[1150Post]

Mark,

I read today that the 4 Provinces has a new owner and a new name. Perhaps we should check it out in Feb. and grab a pint or two?

Thanks again for the reply on my second note, and I apologize if my tone seemed a little off. I actually was going to "reword" it today in hopes that you hadn't read it yet. Oh well.

I am very pleased to hear of your QC training program. I am sure it will be factual and thorough. Obviously, by the mere fact that the shows weren't kicked back for remix, someone intervened in the process and evaluated the mix by more than just numbers, which is great!

After a longish conversation with one of the designers of the LM-100 yesterday, I now have the info I was looking for, and an avenue to have further questions addressed. I wish I could post it or share it, but I am not comfortable sharing information Dolby hasn't made overtly public on its own.

Also, thank you so much for the clear implementation information. As Richard had asked, it was unclear what was being measured in a 5.1 mix. Knowing it's a C reading is a tremendous help.

As an aside to your comments on the loudness war, it was nice during the past series of shows to be able to say, with the authority of DCI behind me, that, "No, we can't make it any louder if you wish this show to air." Thanks for the support!

Let me know about the pints, and say hi to all of my friends in and around those halls and edit bays.

[Brandonx1]

Hi All,
<<"Remix audio due to high LM-100 reading at -24, low vu's at -26 dbfs and peaks at -12 dbfs." >>
How do you raise the VU and peaks while lowering the LM-100 reading? Crank the music or something? Doesn't make any since to me. Thanks for explaining this.
Peace,
Brandon

[JRogers]

Mark,

Thanks for the information. How long have you been at DCI? It has been really frustrating for me at times with the deliverables always being different. I also have been frustrated with trying to explain people in QC in the past what their own deliverables state. About three years ago I had some mixes rejected because QC said the shows had dropouts. It was a mere fraction of a second without audio (fade to black). They didn't understand that it was a normal occurance sometimes. Now, this was three years ago. Anyway, I could go on and on about that stuff.

Thanks,

Jim Rogers
Sweet Atom Pie Sound Design

[Chief Technician]

Mark,

Thank you for taking the time to contribute to these threads. More pieces of the puzzle are beginning to fall into place for us, and we are all grateful for that.

Here is a quote from our training documentation:

Is it possible for us to get a copy of the training documentation, or is that confidential data for DCI employees only? While your excerpts have enlightened us, I think it may be more useful to read it cover-to-cover (at least for me).

Thanks for reading!