Talk:A-weighting

I've restored this page to it's proper meaning, moving content from Weighting filter which has also been restored to it's proper (more general) meaning. I've attempted to clarify the fact that this page is about A,B,C and D weighting, by refering to these early on, and explaining them more. I don't think it would be appropriate to call the page A,B,C and D weighting, as no one is likely to search for such a title! I have fixed the redirect pages for B-weighting etc to point here now rather than to weighting filter.

The important point is: although this page is about a family of weightings, it most certainly is not about 'weighting filters' in general, as they are not confined to the field of audio. Please do not merge again! --Lindosland 14:47, 15 September 2007 (UTC)

That picture of the ear seems to be totally unnecessarry, and a bit stupid, too. Csab 16:01, 12 October 2007 (UTC)

I disagree with Csab. It provides an opportunity for the caption text. Besides, the ear and hearing is what this is all about. Pzavon 01:53, 13 October 2007 (UTC)

Time to change name of this article

This article has been substantially changed in structure and contents over the last few months. As it now has sections dealing with the B, C, and D weightings as well as with A, it seems to me that a name change is needed. I suggest "Accoustic Weighting Networks" but would be delighted with something clearer and less technical if it were also reasonably descriptive of the article's contents. Any ideas? Pzavon 02:17, 8 November 2007 (UTC)

I strongly oppose such a change. What people will search for is 'A-weighting'. That is what is widely used, but since it forms part of a set their inclusion here is appropriate. The other weighting's in the set do not warrant their own page, in fact the whole set need to be seen in the context of each other. They therefore re-direct here. The title 'acoustic weighting networks is far too broad, as it includes many other weightings used in broadcast, acoustics and telephony. Some of these, like 468-weighting and phsophometric weighting already have full pages and links from here. --Lindosland (talk) 17:20, 21 November 2007 (UTC)

Then let's find another way to deal with this. If C-Weighting does not deserve its own article, neither does A-Weighting. If C-weighting can redirect here, then so can A-Weighting. THIS article is clearly no longer only about A-Weighting and needs a better, title that is more correctly descriptive of its contents. How about "Noise Measurement Weighting Networks"? Pzavon (talk) 02:47, 22 November 2007 (UTC)

This is simply not true. Many articles contain sections that could become articles in their own right, but they are kept as a unit because they are closely related, and hence benefit from a common introduction and other common material. The various weightings known as A,B,C,D share an important commonality - they are a related set of curves defined in a single international standard document. "Noise measurement weighting curves" is not a good title as it would have to include a lot more curves than the ones here. There is already an article on Noise weighting, which links to this and other curves (Psophometric, ITU-R 468 etc). A-weighting is a phrase that is referred to very frequently, being the only one of the set of curves in the standard described in this article that has achieved ubiquitous use in many areas. It therefore deserves to be the title of this page. 'A-frequency weighting' is a clumsy phrase that I have never seen or heard used - I will therefore revert it. --Lindosland (talk) 16:51, 7 April 2008 (UTC)

"The various weightings known as A,B,C,D share an important commonality - they are a related set of curves defined in a single international standard document. " That is precisely why the article should have a title that covers all of them. If you don't think "Noise measurement weighting curves" is an appropraite title, please suggest something else, because I'll me changing the title pretty soon. Pzavon (talk) 02:23, 8 April 2008 (UTC)

I can't understand what you think is gained by changing the title. An encyclopedia needs articles about things people are likely to want to look up. They are very likely to want to look up 'A-weighting' or 'A-weighted'. They are pretty unlikely to look up 'noise measurement weighting curves', and if they do then they are probably interested in all such curves, in which case this is not the best page for them - they want "noise weighting curves" which will show them the choice available, acting effectively as a disambiguation page. A more correct title for this page would be 'IEC61672:2003 noise weighting curves', but who's going to look that up? In any case, that title would not be correct, since that is only one of many standards that define the same curve, since these curves have been defined in many international standards in different countries (BS ISO ANSI etc). In fact the original standard that defined the curves is now obsolete. So, yes, you could go for 'Noise weightings A,B,C, D,E and Z as defined in IEC61672:2003 and other international standards'. Not very attractive though surely? The one thing that characterises all these standards, and which started them all off, and is quoted very widely indeed is 'A-weighting'. I hope you will understand this and leave the title alone. You could criticise many other articles as not strictly describing what they are about. Take Lamarckism which has a redirect from Neo-Lamarckism, or Industrial revolution which some have criticised as refering to a specificy western revolution. The latter is accepted because most people looking up the term will be using it as shorthand for 'the nineteenth century industrial revolution which began in Britain, and spread to the Western World.'--Lindosland (talk) 16:21, 9 April 2008 (UTC)

There will still be an article on A-weighting. It will be this one. There will still be an article titled A-weighting, which people will find in searches. It will just redirect here. I strongly support this move proposal. — Omegatron (talk) 01:43, 10 April 2008 (UTC)

No there won't be an article titled A-weighting, there will be a redirect of that title to this article under a name which I am suggesting will be highly innapropriate for the reasons I have given. The title 'Noise measurement weighting networks' is very bad for three reasons. Firstly, what do networks have to do with it? Yes, weighting can be done with networks, but it can also be done in DSP as pure calculation, and these days it is commonly done that way. Secondly, there are and have been many many noise weightings, and to try to explain everything about all of them, their history, their application and their specification on one page is to add confusion and limit future development. Thirdly, not all 'noise weighting' is about measurement. A-weighting is, and ITU-R 468 noise weighting is, and psophometric weighting is, but then there is noise weighting for noise shaping in digital audio. And not all noise weighting is about audio, as noise weighting can be used in video, and EMC measurement, and infrasound measurement, and radio astronomy - it's a big topic, and we have a page already on 'noise weighting' that acts as a disambiguation for the subject. Renaming this page, which is well established, would be a big change, and upset many wikilinks that would have to be changed. It should not be done on a consensus of two, especially as I have seen no reasoned answers to my objections. I hope to create articles about Zwicker, and more advanced weighting methods that have recently been described to cope with all levels and spectra by simulating human hearing. We will then have maybe six articles or more on different methods, all worthy of their own space. I suppose the only alternative title I would accept as properly describing this page is 'A, B, C, D and Z-weighting' but it's too cumbersome and problematic (do you include the commas or not? The dash?) It's not worth the hassle of all the changes to redirects and links. --Lindosland (talk) 10:09, 11 April 2008 (UTC)

Oh, and let's not forget the fundamental fact that this page is not even about NOISE weighting anyway! The A-weighting curve was invented for sound level measurement, and defined in a standard relating to sound level measurement. It is based on pure tone audiometry, not noise, and though it is now commonly used for noise measurement it is well known that it is not really suitable for that purpose. The other curves are not featured in all the standards, so the point about A-weighting is that it is the common feature in many standards and many applications, and the starting point historically. --Lindosland (talk) 10:19, 11 April 2008 (UTC)

Use of the gain functions

It remains completely unclear how the gain functions in the article are used. They are displayed as functions G(s). But what is s and what is its unit? The frequency (Hz)? Or any function of it (e.g. the wavelength or wave number)? Please clarify this because the reader cannot be expected to have expert knowledge.--SiriusB (talk) 22:56, 2 February 2008 (UTC)

I have added the weighting functions of the frequency for A, B and C from [1] (these are what most readers are probably looking for rather than for complex transfer functions), where one can guess how frequency weighting is related to those transfer functions. However, I do not have translated D into a weighting function yet since the formula is much more complicated and has to be tested before posting. Maybe I find time do to this, but perhaps someone else may add it from another source.--SiriusB (talk) 10:30, 27 February 2008 (UTC)

Update: I have added a formula from [2] that lists an algorithm for all four filters. It also contains a graph that is well reproduced by the formula (tested!). However, the resulting graph (as well as that shown in the paper) differs from that in the article for frequencies above about 10 kHz. Since I have found the same graph in several other sources, we should think about replacing the graph in the article by an updated one. Since I didn't find any tool so far that creates valid SVG images from a graph (maybe Gnuplot's SVGs are valid, but, however, are not properly displayed in any browser I have tested) may I ask the original author of the graphs to redraw the SVG with the new formula?

In addition, someone should validate the transfer function for D. I have tried to convert it into a weighting function (as far as I can see one has to calculate the absolute value of the complex coefficients where s=2i*pi*f is the imaginary part), but didn't get the correct function (but a function that looks similar to the graph). Maybe the D transfer function is wrong.--SiriusB (talk) 13:06, 28 February 2008 (UTC)

The Graphs are Correct

I drew the graphs in this article (and several others) very carefully, and I believe they are correct. When I decided to start manufacturing audio test equipment in 1978 (I founded Lindos Electronics) I was for a while misled by graphs that appeared in various publications that differed. This was particularly true of the CCIR-468 curve which people seemed to have their own variations on. So I obtained the standards and got to the truth. Please don't make the mistake of drawing graphs based on a poll of what is out there - much of what is out there is simply wrong because the standards are very expensive (and guarded from a copyright point of view) and so few people have them.

Almost all audio standards use weighting curves that are defined in terms of a table of values with tolerances. They do not concern themselves with the means of realisation. At first this surprised me, but it makes some sense. You build the filter how you like, and check it against the table. I would be interested to know where those equations came from in the article. As far as I am aware they are not from the standards, and so should be quoted only as 'a possible realisation' that meets the requirements laid down.

I've just pulled out BS6402 (Sound Exposure Meters) and it lists a neat table of values for A-weighting from 10Hz to 20000 Hz. This table occurs in many other standards. Looking at the higher frequencies it reads:

1250Hz +0.6 +/-1.5 1600 +1.0 +/-2.0 2000 +1.2 +/-2.0 2500 +1.3 +/-2.5 3150 +1.2 +/-2.5 4000 +1.0 +/-3.0 5000 +0.5 +/-3.5 6300 -0.1 +/-4.5 8000 -1.1 +/-5.0 10000 -2.5 no tolerance given 12500 -4.3 16000 -6.6 20000 -9.3

A note explains that since testing at the higher frequencies is not required no tolerance is given. So others may have their own way of realising these filters but it is the values in this table that are difinitive. I have realised weighting filters in many ways. A-weighting is easy and can be done with a purely R-C network without any second order sections. 468 weighting, with its 0.2dB tolerance at 6k3Hz is hard when repeatable manufacture is required, and I have used sections with zeros as well as poles to realise this simply with minimum sensitivity to component variation, without going to the 6-pole filter that is given as an example in that standard. If anyone can show that my graphs are wrong with reference to any STANDARD document, please quote the details.

I would add that as far as I am concerned the standards are innadequate, because they do not adequately reflect the rapid rolloff in our hearing around 20kHz and do not even specify any sort of falling response above 20kHz. It is therefore possible to make an A-weighting filter that complies with the standard but gives quite different readings from other filters that also comply when noise at 20k and above is present, even at quite low levels. The ITU-R 468 noise weighting filter is much better in this respect, with a high-order rolloff. I have made this point in many places and it is generally agreed that the A-D weightings should always be used with a steep low-pass filter around 20k for consistent results. I think the curves are overdue for replacement, especially now that we have revised Equal loudness contours standardised a few years ago. --Lindosland (talk) 10:37, 24 June 2008 (UTC)

Could you please repost the picture in a file format supported by Internet Explorer? .svg is not a natively supported image format and the pictures are all screwed up. This would be fine if I was the only one using IE, but such is not the case. The whole world is on IE and it would be nice if we could see the pictures. Thanks.