Probelm with K-weighting in BS1770

[adru]

I've been having a tough time understanding the K-wieghting filter in the BS-1770.

The shelving filter is used to simulate the diffraction effects of the head, and the RLB filter to account for the fact that our ears are less sensitive to sounds of lower frequencies.

But why do we boost the higher frequencies in the K-weighted filter?

According to the equal loudness curves, we notice that at lower freq, for the same loudness, the intensity must be greater than that of freq around 1-6khz.
This is because our ears are less sensitive to lower freq, and this has been modelled rightly in the RLB filter.

But, the equal loudness curves also show that the higher freq also have to have high intensity in order to sound as loud as freq in range of 1-6khz(which can be of lesser intensity).
If this is the case, then the K-weighted filters should not boost higher freq, but rather diminish it.

Have i gone wrong in reasoning this out somewhere?

[tom_lowe]

I don't know the ins and outs, all I do know is that both the ITU and EBU P.LOUD group put a lot of research and development into this, so doubt they will have got it wrong.

Maybe someone with more information would like to chime in.

[danijel]

I think you'd get more informed responses at the DSP section of the KVR forum.

[tpad]

...the ITU and EBU P.LOUD group put a lot of research and development into this, so doubt they will have got it wrong....

I wouldn't be so sure about that. So far, none of the architects of LKFS have been able to explain why RLB, which is essentially flat through most of the bass range, gives more accurate results than other loudness models that rolloff the lowend consistent with equal loudness curves. Furthermore, the HF shelving curve that is supposed to compensate for "head diffraction" is a joke! It's almost like someone threw that in at the last minute for whatever reason. When is the last time you saw someone using a weighting curve to compensate for "head diffraction"? Never?

Bob Orban has a white paper on his web site that shows how much better the CBS Loudness algorithm is than LKFS. Ben Bauer and Emil Torrick developed the CBS algorithm throughout the 70's and it has stood the test of time. When the ITU did their loudness studies, they never included the CBS Loudness model! They never had a very good explanation for that omission, either!

[cananball]

I can't answer your question but will add that it is really irrelevant as long as everybody uses the same standard.

Sent from my DROIDX using Gearslutz.com App

[adru]

Originally Posted by danijel

I think you'd get more informed responses at the DSP section of the KVR forum.

Danijel... I'm a bit of the noob. What exactly is the KVR section?

[tom_lowe]

Originally Posted by tpad

Bob Orban has a white paper on his web site that shows how much better the CBS Loudness algorithm is than LKFS. Ben Bauer and Emil Torrick developed the CBS algorithm throughout the 70's and it has stood the test of time. When the ITU did their loudness studies, they never included the CBS Loudness model! They never had a very good explanation for that omission, either!

Could it be the ITU didn't want to license any algorithms and wanted to come up with their own? I've not read Bob Orban's white paper, but just because 1 man's research says it's better doesn't necessarily make it better.

*UPDATE*
I've just had a quick read of the aforementioned paper, something that immediately stuck out was that Orban's own products use the CBS algorithm. Now, I'm sure he conducted very thorough research, however, he does state that ITU.1770 would be more effective if it used gating. Gating has been implemented in both ITU 1770 and EBU R.128, so perhaps the findings of his white paper should be re-examined.

[tpad]

I'm not aware that there is any licensing involved these days, since CBS Labs went bust quite awhile ago (not when Orban initially decided to use their technique). Skovenburg and Nielsen briefly commented in their AES paper that the CBS model wasn't included in the ITU tests because there was no software implementation. Pure bull, because Orban had created a software version of the (analog) CBS algorithm for his own Optimod product. If you read the Jones/Torrick/Bauer IEEE papers on the development of the CBS algorithm, you'll see what the real meaning of -thorough- research is.

[insomaniac]

Originally Posted by cananball

I can't answer your question but will add that it is really irrelevant as long as everybody uses the same standard.

Sent from my DROIDX using Gearslutz.com App

This is kind of my thought. As long as we're all on the same level, is it really a problem?

[tpad]

Not a problem if you want to be like everyone else and use a flawed algorithm. The reason that the so-called "loudness" problem exists is because the same mentality prevailed when Dolby sweet talked everyone into playing Dialnorm and they responded by taking all of the automatic level controls off. There never was a "loudness" problem in the first place; just an inconsistent level control problem due to poor operating technique. There was no need to re-invent the wheel and spend all of the time "discovering" how to measure loudness when a more accurate model (the CBS model) was already in place. Another example of what you might call flawed science.

[nucelar]

hi tpad if you're making bold claims, please provide examples of audio that is significantly louder than another while measuring the same LUFS, before calling it flawed.
The 1770 was designed to 1) be simple algorithm 2) provide good correlation between subjective and measured loudness. The filters and weightings were found empirically, personally I don't care if they call it head diffraction or ear lobe scattering, as long as the correlation is good.
I don't underestimate the work of CBS, they have contributed to the advancement of this science just as many others:
http://www.tcelectronic.com/media/sk...loudness_m.pdf
cheers

[danijel]

Originally Posted by adru

Danijel... I'm a bit of the noob. What exactly is the KVR section?

KVR: Forum - DSP and Plug-in Development

[tpad]

nuclear.

There is no such thing as "measured loudness". Loudness "is a psychological term used to describe the magnitude of an auditory sensation [Harvey Fletcher and W.A. Munson]". It's the sensation of what you hear, not what is carried in a pair of audio program wires or in an audio data file.

LKFS was being discussed at an AES convention years back and I asked one of the architects of LKFS what it was they were listening to for evaluation. He said they were using samples of off air broadcast material. When I further asked if the material was significantly compressed (as with a multiband broadcast processor), he said yes. A conventional volume indicator (aka VU meter) is perfect for this. You don't need any fancy meters or algos to balance this stuff; just peak it at zero VU and it all comes out nice and uniform sounding. Amazing how well the old technology still works!

[tom_lowe]

Originally Posted by tpad

nuclear.

There is no such thing as "measured loudness". Loudness "is a psychological term used to describe the magnitude of an auditory sensation [Harvey Fletcher and W.A. Munson]". It's the sensation of what you hear, not what is carried in a pair of audio program wires or in an audio data file.

LKFS was being discussed at an AES convention years back and I asked one of the architects of LKFS what it was they were listening to for evaluation. He said they were using samples of off air broadcast material. When I further asked if the material was significantly compressed (as with a multiband broadcast processor), he said yes. A conventional volume indicator (aka VU meter) is perfect for this. You don't need any fancy meters or algos to balance this stuff; just peak it at zero VU and it all comes out nice and uniform sounding. Amazing how well the old technology still works!

Yes, but we all know a VU is Virtually Useless :-P besides, this is starting to sound like you have a general problem with technology that has been introduced since the 1970s, specifically metering technology. The whole reason ITU.1770 was developed was because broadcasters have always had issues with matching the level of programmes and adverts. If this was achievable using VU, why was it not done?

[tpad]

It's beginning to sound like YOU have a problem accepting the fact that technology introduced as far back as the 70's may have been more thoroughly and intelligently researched than what is currently in vogue. You're extrapolating to assume that just because I'm not enamored with BS-1770 (ONE instance of "modern" technology) that "I have a general problem with technology that has been introduced since the 1970's". Do Gearslutz a favor and don't try and psychoanalyze me or other members over the internet, because you don't know WTF you're talking about!

If broadcasters have been having problems trying to balance program material using the CBS algorithm, I would like to see where that has been discussed in -legitimate- broadcast literature. I don't think you'll find any, because I don't think anyone other than Orban actually gave it a real try.

Actually, broadcasters were NOT having problems matching levels when they were using intelligently applied processing. The whole problem started when they took the processing off and started aligning unprocessed program segments by using the single number measurement of program level (loudness), aka Dialnorm (LKFS). You can't claim that VU meters didn't work because the broadcasters gave up on them likewise, and started using PPMs and then eventually transitioned to using the LM100.

I don't agree that the VU meter is totally useless. They never were accepted in the UK and through most of Europe either, and so that is your apparent bias against them.

[tom_lowe]

Originally Posted by tpad

It's beginning to sound like YOU have a problem accepting the fact that technology introduced as far back as the 70's may have been more thoroughly and intelligently researched than what is currently in vogue. You're extrapolating to assume that just because I'm not enamored with BS-1770 (ONE instance of "modern" technology) that "I have a general problem with technology that has been introduced since the 1970's". Do Gearslutz a favor and don't try and psychoanalyze me or other members over the internet, because you don't know WTF you're talking about!

If broadcasters have been having problems trying to balance program material using the CBS algorithm, I would like to see where that has been discussed in -legitimate- broadcast literature. I don't think you'll find any, because I don't think anyone other than Orban actually gave it a real try.

Actually, broadcasters were NOT having problems matching levels when they were using intelligently applied processing. The whole problem started when they took the processing off and started aligning unprocessed program segments by using the single number measurement of program level (loudness), aka Dialnorm (LKFS). You can't claim that VU meters didn't work because the broadcasters gave up on them likewise, and started using PPMs and then eventually transitioned to using the LM100.

I don't agree that the VU meter is totally useless. They never were accepted in the UK and through most of Europe either, and so that is your apparent bias against them.

The point I was making was that in the recent past, varying loudness between programmes became an issue. Clearly the CBS system never really took off. Also, Dialnorm is not the same as LKFS/LUFS. Dolby introduced that way before the ITU started their work. The VU statement was intended to be taken in a light hearted manner, based on how BBC engineers used to describe it, hence the :-P at the end.

Also, how do you KNOW CBS did more research than TC Electronic, ITU and P.LOUD did in to this?

[Sonsey@mac.com]

Originally Posted by tpad

It's beginning to sound like YOU have a problem accepting the fact that technology introduced as far back as the 70's may have been more thoroughly and intelligently researched than what is currently in vogue.

I can't speak to Tom's like or dislike for modern versus vintage. I can however state that a great deal of the research for ITU 1770 was done up here at the CRC (just down the road as a matter of fact), and several of my colleagues were involved in EXTENSIVE listening tests as part of said research - my partner at the studio at the time being one of them. The Communications Research Center of Canada has pages of documentation on just how much research was done as merely PART of the standard. So while we can debate the relative merits of CBS vs, ITU 1770 (and perhaps that's not a bad idea) to claim it was not "thoroughly and intelligently" researched is not correct.

Audio Perception | Communications Research Centre Canada

[tpad]

Howard and Tom

I KNOW what CBS Labs did because they published the information in a number of IEEE and SMPTE papers (which I have). I know CRC did a lot of research on the ITU program but it always comes across as hearsay. Maybe you should post or publish the material so we can all see it. Then maybe I'll have to eat my words!

The intelligence part refers to the lack of expertise on program balancing and level matching that was commonplace knowledge back in olden days. I've got a I.R.E paper written by the legendary Howard Chinn from way back in 1947. Funny how they didn't have any problem balancing program material back then. They knew exactly what they were doing. Why is it that no one seems to know how to do it in modern times?

[Sonsey@mac.com]

Hi Tpad,

CRC Publication Database ~ Search Results | Communications Research Centre Canada

All can be requested from CRC.

I hear red wine goes well with white papers...

[tom_lowe]

Originally Posted by Sonsey@mac.com

Hi Tpad,

CRC Publication Database ~ Search Results | Communications Research Centre Canada

All can be requested from CRC.

I hear red wine goes well with white papers...

Let me clarify, I don't dislike 70s technology, but I don't think the claim that the ITU and EBU didn't do thorough research was accurate. The only documentation about CBS vs ITU.1770 wasn't accurate either since it was written before the -8 (now -10) gate was implemented.

Anyway, this is all elementary since the CBS method wasn't widely adopted and ITU.1770 in it's various guises is gaining a lot of momentum on both sides of the pond and elsewhere too.

[tom_lowe]

Originally Posted by tpad

Why is it that no one seems to know how to do it in modern times?

Automated playout and cost cutting. This was also made worse by the loudness wars in commercials and even some programmes and broadcasters metering only the peaks, not overall loudness, certainly here in the UK at least.

[UnderTow]

Originally Posted by tpad

Bob Orban has a white paper on his web site that shows how much better the CBS Loudness algorithm is than LKFS.

Is this the same Mr Orban that published a white paper on the problems of loud music masters and how they suffer at the hands of radio processing without ever acknowledging that the brutal radio processing itself might be part of the problem?

Maybe not exactly an unbiased source...

Alistair

[tpad]

Yes, it's the same Bob Orban. He wrote the paper on loud music masters after noticing how bad they sounded with additional processing added on top. Brutal radio processing is totally the fault of the broadcaster, not the processor vendor. You can easily set an Optimod, Omnia, etc to process very mildy, to the extent that most listeners would not be aware of the processing at all. It's entirely up to the station management as to how the processing is being applied.

[UnderTow]

Originally Posted by tpad

Yes, it's the same Bob Orban. He wrote the paper on loud music masters after noticing how bad they sounded with additional processing added on top. Brutal radio processing is totally the fault of the broadcaster, not the processor vendor. You can easily set an Optimod, Omnia, etc to process very mildy, to the extent that most listeners would not be aware of the processing at all. It's entirely up to the station management as to how the processing is being applied.

Of course but that should at least be mentioned in a white paper about the effects of radio processing on loud masters wouldn't you think? It seems Mr Orban has a blind spot for his own products or the technology in his own products... That was my point.

Alistair

[tpad]

....It seems Mr Orban has a blind spot for his own products or the technology in his own products...

The problem is not unique to just his product(s). Same thing happens with the Omnia broadcast processor line (Orban's primary competition). Omnia probably has a tech note on their website saying the same basic thing. There have a number of forums and discussions on the hazards of loud mastering at past AES conventions. You need to do more research on this.

[UnderTow]

Originally Posted by tpad

....It seems Mr Orban has a blind spot for his own products or the technology in his own products...

The problem is not unique to just his product(s). Same thing happens with the Omnia broadcast processor line (Orban's primary competition). Omnia probably has a tech note on their website saying the same basic thing. There have a number of forums and discussions on the hazards of loud mastering at past AES conventions. You need to do more research on this.

Of course it isn't just his products. Where did I say it was? That still doesn't change the fact that these processors very often completely destroy music.

Sure user settings determine how much damage is done but why even include multi-band clippers if you have any respect for sound quality? No, it is about making and selling products that aim to make a signal hotter than anything else at any sonic cost and then trying to shift the blame elsewhere.

Anyway, you missed the point entirely: The point was to question the objectivity of Mr Orban's papers.

Alistair

[tom_lowe]

tpad, do you know Mr Orban personally or hold shares in his company?

[tpad]

nope. I have absolutely no connection with Orban or any of the other equipment vendors discussed here. I do own a couple of his limiters (also own a couple of TC Electronics limiters too). I have talked to Bob briefly several times at the NY AES conventions, but that is the extent of my involvement with him. I used to deal extensively with radio broadcasters years ago, and were always talking processing, so his (Orban's) designs and philosophies are sort of hard-wired in my brain as a result. Wish we could get him to participate on Gearslutz, he makes for some real interesting discussion and reading!

[tpad]

Undertow. I think YOU missed the point. FM broadcasters use processing - period. Whether it is Orban's or some other brand, they all do similar things to the signal. An undesired side effect of ANY of the popular brands is that hypercompressed rock music sounds pretty dreadful after it has been run through them. Rather than nitpick about Orban's alleged bias, you should be thanking him for identifying the problem cause. He originally published that observation on rec.pro.audio in the 90's, so it is rather dated material. Similar degradation occurs when you try and encode hypercompressed material with AAC, MP3, etc. So the finger points back to the hypercompressed material, not the audio processing. Incidentally, FM Optimods don't use mutiband clipping. That's why I suggested you do a little more research on the art of broadcast processing before trying to discuss here.

[tom_lowe]

Originally Posted by tpad

nope. I have absolutely no connection with Orban or any of the other equipment vendors discussed here. I do own a couple of his limiters (also own a couple of TC Electronics limiters too). I have talked to Bob briefly several times at the NY AES conventions, but that is the extent of my involvement with him. I used to deal extensively with radio broadcasters years ago, and were always talking processing, so his (Orban's) designs and philosophies are sort of hard-wired in my brain as a result. Wish we could get him to participate on Gearslutz, he makes for some real interesting discussion and reading!

Thanks for the clarification. Out of curiosity, has anyone ever implemented the CBS algorithm in software?