Home |
Search |
Today's Posts |
#81
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Lord Hasenpfeffer wrote: Erwin Timmerman wrote: Lord Hasenpfeffer wrote: Despite all of this processing, all 8 final WAVs sound "normal" to me. By the time I got to -40dB, I'd have expected the final MP3-WAV to sound pretty crappy but it doesn't - at least not by any degree which I would expect it to sound. What does this tell you? Exactly. It tells me that Normalize seems to treat WAVs with "kid gloves" compared to Audacity. I've done some more tests in search of the noise floor. Do you know if Audacity does dithering, and if so, whether you've turned it on or not? If "Normalize" can reduce 90 dB in 16 bit and still produce something that even resembles to a song, it must use dither. That by itself can make a huge difference. If you don't know that dither means: http://www.pcrecording.com/dither.htm All other reactions and conclusions are void if you didn't turn on dither in Audacity (or if it doesn't have the option to do so). With Audacity, I repeated (as suggested) a short segment of a WAV with 5 seconds between each segment. I then saved this original WAV to disk. Then I highlighted all but the 1st segment, reduced amplification by -5dB, hightlighted all but the 2st 2 segments, reduced by -5dB again, highlighted all but the 1st 3 segments, reduced by -5dB again, and so on. This is not a good way of doing it. Either you're adding dither noise with every step, or you're adding quantization noise with every step. That by itself will make the more often processed wavs sound worse. To do this right you should amplify each segment by itself with the amount needed. So don't select the whole bunch, select only the segment that you're going to do, and then amplify with -5dB, -10 dB, -15 dB, whatever is needed. Especially with very low level audio the quantization or dither noise will become more and more audible, so it is increasingly important to not add it cumulatively. Meanwhile using just the starter WAV segment that I'd repeatedly duplicated for the above test, I was able to drop the gain on it by as much as -90dB using "normalize -g -90dB starter.wav" at a shell prompt, Yeah, but giving the wavs a huge headstart by processing them only once. http://www.mykec.com/mykec/images/20..._Me_Tender.png Looks familiar. I hope you will treat your generated mp3's the same way, as far as the original copyright owners are concerned... Absolutely. My MP3s stay with me. I have shared a few files with close friends on occasion when the need has arisen to clarify or reinforce a point brought up in conversation but never have I traded MP3s or posted some online for all my friends to enjoy, etc.. OK, I wrote it just to make sure. You have to admit, putting names of all the mp3's you have on a web site, saying you have a business with storing huge amounts of mp3's on line, taking someone who talks about downloading mp3's off-group, adds up a bit of suspicion. Good to see it was unfounded. Personally, the last thing I would like to do is to help someone improve the quality of the mp3's he's going to put on kazaa. But you've stated clearly enough that this isn't the case. Thanks. Greetings, Erwin Timmerman |
#82
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Man, Jonas, I am *so sorry* there isn't a way to correct so obvious a
mistake after one has clicked the Send button! Just keep reading. Much brighter stuff still lies ahead! I really do wish I hadn't have written what I did this time. What you were actually trying to tell me didn't exactly click until about 10 seconds too late! I'm so sorry. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#83
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
How's that for a newbie? Pretty good. :-) Thanks. I *am* at least trying, y'see, to get all of this figured out before everybody completely vacates the thread! And you know what? It never occurred to me until late last night just how much I'd forgotten the first rule of thumb every successful programmer has to learn... Never enter into a den of engineers without being fully dressed for battle! (And the same goes in reverse for those on your side of the equation as well!) If you *really* *wanna* *know* the *one thing* that's been seriously flawed with both this thread - and the other one in the other newsgroup before it - from the absolute very beginning, that's it! Hehehe!!! :-D Maybe this is what Mr. Olhsson was trying to tell me all along! The way I understood the docs, the final RMS level of the loudest parts of the tracks combined will become -10dbFS. What by his definition constitutes a "loudest part"? Where exactly does he draw the line to even define what that is or what that means? It would be instructive if you analyse a file you've told Normalize to normalize (not in batch mode) to -10dBFS and see the average RMS of the whole file really is -10dBFS or if, as I suspect, it is the loudest parts of that file wich become -10dBFS. Hmmm... Lemme post to you here what I get on my terminal display from it. Let's just take out ol' Elvis Presley's "Burning Love" here for a bit. Hold on just a sec... rip rip rip rip rip... OK! The "-n" option causes Normalize to just read the peaks and levels of the file and that's it. This is a 2 minute, 57 second song, btw.... [mykec@sillygoose Elvis2]$ normalize -n track19.cdda.wav Computing levels... level peak gain -16.9302dBFS -1.0545dBFS 4.9302dB track19.cdda.wav There. That took all of 2 seconds to scan. Note how the level plus the gain = -12dBFS. OK, now let's Normalize the file to my personally preferred level: [mykec@sillygoose Elvis2]$ normalize -a -10dBFS track19.cdda.wav Computing levels... track19.cdda.wav 100% done, ETA 00:00:00 (batch 100% done, ETA 00:00:00) Applying adjustment of 6.93dB to track19.cdda.wav... track19.cdda.wav 100% done, ETA 00:00:00 (batch 100% done, ETA 00:00:00) OK, there. All of 8 seconds were required to rescan and normalize this particular WAV. Now, let's repeat step 1 to see how things look now... [mykec@sillygoose Elvis2]$ normalize -n track19.cdda.wav Computing levels... level peak gain -10.1832dBFS -0.0244dBFS -1.8168dB track19.cdda.wav And there you go. The level of the song is -10.1832dBFS. Would you like to see some screenshots of this as well? Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#84
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Wow, Erwin! What a fascinating idea! I never thought of it.
As for the licensing is concerned... I think all a restaurant would need is either a BMI or an ASCAP license and a way to track what songs were played in the dining area or wherever. The publishing companies should be able to handle everything else from there. Are you an entrepreneur as well? Or was this just a lucky shot? I like the way you think. Myke Erwin Timmerman wrote: Lord Hasenpfeffer wrote: And in case anyone still thinks I might be doing these hits this way just as a hobby, that's only half correct. Once I've gone through all of my CDs, I'll finally be ready to think more in terms of "high school reunions" and "decade parties" - that is, if anyone'll have me to "do" theirs. A lot of small communities don't have big budgets for big DJ services at their small events and gatherings. You could sell computers with your stuff installed (including the MP3's) to restaurants, I bet. Of course that needs a lot of licensing going on, but it might be worth checking out. Good luck! Erwin Timmerman -- -================================- Windows...It's rebootylicious!!! -================================- |
#85
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Erwin Timmerman wrote:
Do you know if Audacity does dithering, and if so, whether you've turned it on or not? I have no clue. The version of Audacity I currently use is only v0.98 so the developer didn't even have it out of beta yet at the time I downloaded and installed it - which was about a year or so ago now. It seems to be sorely lacking in features for tests like these but for all the kind of WAV editing I ever need to do, it's more than met my needs so far. I'll look soon to see what version he's got it up to these days. If "Normalize" can reduce 90 dB in 16 bit and still produce something that even resembles to a song, it must use dither. That by itself can make a huge difference. If you don't know that dither means: http://www.pcrecording.com/dither.htm Thanks. I am familiar with the term "dithering" in only the graphical image sense. GIF images use dithering to try and simulate smooth-transitions (i.e. gradients) from one colour to another. But with a palette of only 256 colours available, the results always pretty pretty much suck. All other reactions and conclusions are void if you didn't turn on dither in Audacity (or if it doesn't have the option to do so). This is just another thing I've got to look into then. Geez, that stack gets taller by the minute it seems! This is not a good way of doing it. Either you're adding dither noise with every step, or you're adding quantization noise with every step. Very well. I opted to do it that way because it was difficult to tell where the 5 second breaks were located once the waveform is virtually invisible! I'll try again and be more deliberate and careful. I've got an idea for how to accomplish what you've said to do. OK, I wrote it just to make sure. You have to admit, putting names of all the mp3's you have on a web site, Oh, that's just a demonstration of one kind of thing I can do as a web-programmer. Most of my work I do for other people. It's therefore difficult to always come up with demonstration projects which are unique to only me. That was just one thing I knew that I could do so I did it. That's all. Plus the chronlogical arrangement of the lists is rather unique and probably quite useful to a lot of people who're aware of its existence on the web. That page doesn't really look like much but there truly is a lot behind-the-scenes which is required to make it happen, believe me! One thing it does is demonstrate my skill when called to develop an uncluttered, user-friendly, intuitive user interface. That's very important in my business. saying you have a business with storing huge amounts of mp3's on line, That's in reference to a project I'm slowly but surely working on building for with one of my clients. Any MP3s that go online for that will definitely be files that he personally owns or legally controls for the purpose of making it electronically available for download to the public. taking someone who talks about downloading mp3's off-group, When did I do that? I'm not familiar with that to which you're referring. Personally, the last thing I would like to do is to help someone improve the quality of the mp3's he's going to put on kazaa. That's understandable. But I don't use or even have kazaa - and I *won't* use or have it either. First off, I don't *need* it as I already own a rather extensive collection of real CDs for source. Secondly, there's a lot of effort involved in what I'm doing so the *last* thing I'm going to want to do is start giving my stuff away. Especially after I paid so much for the CDs in the first place and then worked so hard to rip and encode them over such a long period of time. It doesn't bother me that you've inquired about it either. We live in very suspect times with regard to all of this and I completely understand your concern. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#87
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Erwin Timmerman wrote:
Quite easy. Take the original wav. Trim to the part you want to use. Generate 1 (or 5, or whatever suits you) second of silence. Then select the whole bunch, copy, and loop paste the number of times you want to test. This will end you up with a wave file with visually good definable "chunks" of music. Display the complete wave. Select the second chunk, amplify by -5. Select the third chunk, amplify by -10 etc etc etc... As the left over (not yet processed) chunks are still at full volume it is easy to select them. Aside from the error in my attenuation method, this is how I did it. The problem I had was knowing where the breaks between parts were while I was trying to reamplify them. I suppose I should make 2 copies of the original and use one of them as a guide for knowing where the breaks between segments are actually located in the one that I'm actually working with. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#88
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
"Lord Hasenpfeffer" wrote in message ... Wow, Erwin! What a fascinating idea! I never thought of it. As for the licensing is concerned... I think all a restaurant would need is either a BMI or an ASCAP license and a way to track what songs were played in the dining area or wherever. The publishing companies should be able to handle everything else from there. Actually there are zillions of companies already doing exactly that ! geoff |
#89
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Roger W. Norman wrote:
I was going to take you through a scenario that would put any ideas of LOUD being a method to make better mp3s, but I just figured that A) it's a holiday, and B) it's not worth the trouble, and C) everybody has been over this in about 15 different ways in the first place. Sounds perfectly reasonable to me. The results may well be different if extremes in volume are used, but the perception of an algorithm doing something different due to overall volume is plainly wrong. When you say "algorithm" here, are you referring to ATH filtering during lossy encoding? After conducting my own first test, I'm very *inclined*, though still not absolutely *convinced*, that Geoff's initial assertion that +4dB or +5dB isn't enough of a difference to make a difference, lossy or not. The only reason I ever considered higher amplitudes to be "helping" with regard to lossy compression is that the effects of ATH-based filtering are directly determined by the amplitudes of the frequencies being filtered or retained. To me, the relationships in play between a frequency's amplitude and its ability to survive passing through an ATH filter seem quite obvious and direct - but much larger differences in amplitude than those which I'm creating are required in order for the difference in fidelity to be appreciable. Geoff said this. My tests appear to confirm this. I would have been more willing to accept Geoff's message earlier had his delivery not been seasoned with so much emotion and "out of the blue" speculation. But my original point was that it's not up to you to determine how someone else's product, that you didn't have anything to do with during the production phase, would sound better if you just did THIS or THAT. If someone wants you to put up good, better, best mp3s, then it's not up to you to change the product they give you and THEN make an mp3 of it. It's up to you to go out and listen to all the possible products you could use in the job and then make that purchase to give your clients THEIR PRODUCT in the best mp3 form that you can. I said it before and I'm saying it again. You are not a part of the creative process, and to alter their music to your own tastes before you commit them to another process is destructive and is doing a disservice to your clients. And it's not going to sit well for future business because no one will be able to trust your judgement that THEIR music is THEIR music and not subject to your whims and fancies. And if it doesn't sit well with them, your reputation is going to be squat. To this end I simply consider how I would feel if someone decided to alter my own music in similar ways, therefore, you are quite correct in terms of reputation preservation and the like. With my particular project in mind, however, the assumption that I am not a part of the creative process will not always be accurate. In many cases, depending on the financial state and the technical savvy possessed by the client being served, I may indeed be *entirely* responsible for the decisions required to produce the original recording from which all subsequent MP3s will result. I'm not saying that your desire to do a good job for your clients isn't a good thing. I'm saying that you have no rights, much less the experience, to make judgement calls that affect your client's interests. Because my role would not always be limited to just "webmaster" and "MP3 encoder", the truth of this should actually fluctuate on a client-by-client basis. Nevertheless, I clearly understand what you're saying and why you're saying it, and your advice is sound. Perhaps the smarter thing to do so that you don't dig yourself into a grave is to simply require that all product an artist wants put up on your website be submitted as an mp3. The current plan already allows for this, however, we will also need to make "MP3 production" available as an additional service. A lot of people don't understand how MP3s are made and may also not have the resources available to create them on their own. And again, if you aren't working directly for clients that have the right to make the determination of having their works put out on the web in mp3 format, you're simply part of the problem and you won't find a solution here. I can't make it any simpler. An essential part of the plan is to be certain that the owners/maintainers of the website have complete legal authority to make available the files which are placed online for distribution to the public. If said authority is not obtained, the files in question simply will not be placed online. And again, if you want the specifics on the algorithm, then by all means, go to someone who does the coding and understands the principles. Of course. You're just spinning all our wheels here and it doesn't appear that you've learned a single thing. Sure I have. There's just been so much to take into consideration that I've not had time yet to absorb everything that's been said and suggested. Take that as abusive if you want, but it's about as plain as I can speak without coming down on you like some newbie that wants to tell me about how I'm wrong because I love my Crest and he's right because he love's his Behringer. The obvious answer to that conundrum is that both of us are right, but I have more experience invested in my purchase. Nothing hard nosed, nothing negative, just different perspectives and experience levels. Your input has not been perceived as being anything even remotely resembling abusive or hard-nosed commentary. Honestly, I'm confused as to why you think it may have been. It's all perfectly reasonable and wholly understandable on this end. Thank you! Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#90
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Erwin Timmerman wrote:
Well, the one doing the actual copying (you in this case, if you are preparing the system for the given restaurant) has to pay a license fee as well. Yes, I'd say that, um,... You, sir, are corrrrrrect! (a la Ed MacMahon) :-) I'm not an entrepeneur. I have plenty of ideas, but never take actions to make them into money. Been there! I have a nice day job. As they say, "Necessity is the mother of invention." For example: I don't consider my playing or my music releaseworthy. Yet I see people actually making money with stuff I wouldn't dare to release (quality-wise, not music-style-wise as that is just a matter of taste). They are doing something right that I am doing wrong. Hmm-hmm-hmm... Whaddaya think it is (or might be)? Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#91
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
What by his definition constitutes a "loudest part"? Where exactly does he draw the line to even define what that is or what that means? That's a good question. Here's two quotes from the documentation: I'm going to spend a little more time on this before posting any real reply to it. This is indeed kinda weird - and probably commands an email to the author of the program. Meanwhile, I'll grab that WAV and see what happens... Thanks for the link. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#92
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
Download this file: http://vvv.truls.org/temp/RMS-test.wav Run it through Normalize and tell us what numbers Normalize showed. These are the values shown by CoolEdit's statistics: --8-- window = 50ms: 0dB = FS Sine Wave Minimum RMS Power: -31.47dB Maximum RMS Power: -6dB Average RMS Power: -22.35dB Total RMS Power: -15.79 --8-- window = 50ms: 0dB = FS Square Wave Minimum RMS Power: -34.48dB Maximum RMS Power: -9.01dB Average RMS Power: -25.36dB Total RMS Power: -18.8 --8-- [mykec@sillygoose mykec]$ normalize -n RMS-test.wav Computing levels... level peak gain -9.0106dBFS -6.0002dBFS -2.9894dB RMS-test.wav Looks to me like we're talking *Maximum* RMS power here. Without really thinking too much about it, I'd venture to guess that that's why the limiting, if any, which I've seen in my "improvements" to my original WAVs is minimal at best compared to what would be occurring otherwise if Average RMS Power levels were being considered. And your assessment is? Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#93
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Geoff Wood wrote:
-- -================================- Windows...It's rebootylicious!!! -================================- |
#94
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
"Lord Hasenpfeffer" wrote in message ... Geoff Wood wrote: -- -================================- Windows...It's rebootylicious!!! -================================- Sorry, I don't understand this rash of posts such as that quoted in full above. Are they accidental posts, or some sort of childish snub ? geoff |
#95
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Lord Hasenpfeffer wrote: P.S. By the way, how's the ol' tolerance for gore holding up these days? ;-) Whaddaya t'ink, I'm some kinda pigeon or sumpm'? :-) Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein |
#96
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Looks to me like we're talking *Maximum* RMS power here.
Yes, it does. We should note that the wav file I created was a very simple file though, and very different from music. Without really thinking too much about it, I'd venture to guess that that's why the limiting, if any, which I've seen in my "improvements" to my original WAVs is minimal at best compared to what would be occurring otherwise if Average RMS Power levels were being considered. Exactly. That was why I thought you didn't really want your music normalized to an average -10dBFS RMS. And now we see that I probably was correct in this. The sound you like is closer to the reasult of normalizing to a maximum of -10dBFS RMS. Huge difference. Regards /Jonas |
#97
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
On Sat, 05 Jul 2003 13:04:29 GMT, Jonas Eckerman wrote:
As I can see, the relative loudness of all the tracks are not well preserved. I don't know what that picture shows, but according to the documentation of Normalize it's "batch" mode is meant specifically for preserving the relative loudness of all tracks. In this mode it treats all the tracks as one long audio stream. But that's not what he said he did. He said he normalised each track individually, when he apparently whipped the arse of MSFL by 'remastering' DSOTM, though he didn't mention that he didn't use batch mode when he originally made this claim. While normalising any track to 0dbFS is harmless in itself, doing what you've done to DSOTM (normalise and potentially limit in order to make the RMS level of each individual track -10dbFS), is truly idiotic. And, our little limited test indicates that Normalize does not do that. It seems that with Myke's settings it normalizes the tracks, seen as one single stream, to a *maximum* (or close to maximum) -10dBFS RMS. OK, I admit to not reading all the posts in the multiple threads about this issue, and to skim reading many of the others, but Normalize (sic), as used by Myke in recent times, but perhaps not at the moment, will limit or clip, depending on user settings, when normalising to an RMS value if the file contains such samples. Are you saying that Normalize (sic) will normalise to a lower RMS value than specified by the user if the file contains samples that would otherwise cause clipping or limiting? If so, fine, but that's not how Myke was using it when he whipped the arse of MSFL - I specifically recall him mentioning the limit option around that time. |
#98
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
That was why I thought you didn't really want your music normalized to an average -10dBFS RMS. And now we see that I probably was correct in this. I thought dBFS was a peak voltage indication reading and in contrast, RMS - root mean square - was a mathmatical average power reading. It would seem to me that dBFS and RMS don't belong in the same measurement. I never dreamed this thread would go this far... AT The sound you like is closer to the reasult of normalizing to a maximum of -10dBFS RMS. Huge difference. Regards /Jonas |
#99
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
But that's not what he said he did.
Well... That's one of the main problems with this thread. Myke hasn't been clear on either the terminology or how his Normalize app works. So a lot of the time he has said he did one thing while in actuality he did something else. OK, I admit to not reading all the posts in the multiple threads about Wich is quite understandable. This thread has been extremely convoluting and confusing at times. At least we have actualluy gotten through a lot of it, and some things have become clearer in recent posts. For example what Normilize actually does. It hasn't become clear, but a lot clearer than it was at the start of the thread. :-) will limit or clip, depending on user settings, when normalising to an RMS value if the file contains such samples. Are you saying that Normalize (sic) will normalise to a lower RMS value than specified by the user if the file contains samples that would otherwise cause clipping or limiting? No. I'm saying that Normalize will not normalize based on the average RMS of the file. It normalizes based on the loudest parts of the file, wich of course is a completely different thing. Exactly what it counts as the loudest parts of the file is unclear. In a very limited test we just did, the RMS level Normilize based it's operation on was the same as CoolEdit showed as the maximum RMS of the file. That file wasn't representational of music though (it was just a sine way, 1 second at -6dB and 9 seconds at -31.46dB). The Normilize app does limit in order to avoid digital clipping if necessary. But obviously normalizing to a maximum -10dBFS RMS (or somthing similar) is not nearly as brutal as normalizing to an average -10dBFS would be. If so, fine, but that's not how Myke was using it when he whipped the arse of MSFL - I specifically recall him mentioning the limit option around that time. Myke did say that he normalized to an average 10dBFS RMS with limiting applied, but he wasn't. He normilized to something similar to a maximum of -10dBFS with limiting applied. He simply didn't know how the Normalize application really worked. He did try to tell us that when he normalized a complete CD he did it in a way that kept the relative loudness of the tracks, but as he thought that kind of operation was implicit in the use of term "batch" he completely failed to explain it to us in the beginning of the thread. As the thread then became so large, the explanations and findings necessary to understanding what Myke actually did became hidden in variuous branches of it. One revelation was finding the documentation for Normalize. It doesn't only explain that Myke wasn't doing what he said he was doing, it also partly explained why Mike's terminology was different from everyone else's. An important quote from the docs: --8-- Please note that I'm not a recording engineer or an electrical engineer, so my signal processing theory may be off. I'd be glad to hear from any signal processing wizards if I've made faulty assumptions regarding signal power, perceived volume, or any of that fun signal theory stuff. --8-- The rest is at: http://www1.cs.columbia.edu/~cvaill/...ze/README.html Regards /Jonas Eckerman |
#100
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Martin Tillman wrote:
I "batch Normalized" the entire set of 10 separate WAVs from that CD to -10dBFS and then loaded then all into Audacity end-to-end so I could then take the pretty picture. Ah, so that's how you managed to bugger up the dynamics. Shame you didn't respond to me when I pointed this out buggeration days and days ago. The dynamics are not "buggered up". Correct me if my memory has failed me, but there aren't many gaps between tracks on DSOTM. How do the jumps in level at the joins sound, given that each track has almost certainly had its gain increased by a different value from the one that precedes and succeeds it? You need to read the thread before you post this kinda crap. While normalising any track to 0dbFS is harmless in itself, doing what you've done to DSOTM (normalise and potentially limit in order to make the RMS level of each individual track -10dbFS), is truly idiotic. You need to read the thread before you post this kinda crap. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#101
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
We should note that the wav file I created was a very simple file though, and very different from music. But the principles are the same. Without really thinking too much about it, I'd venture to guess that that's why the limiting, if any, which I've seen in my "improvements" to my original WAVs is minimal at best compared to what would be occurring otherwise if Average RMS Power levels were being considered. Exactly. That was why I thought you didn't really want your music normalized to an average -10dBFS RMS. And now we see that I probably was correct in this. Well, until these discussions began, I didn't know or even understand the term "RMS" so all I knew based upon what I've been *hearing* is that -10dBFS is usually a perfectly fine "level" for my tastes. I believe it was Geoff who suggested that I start using the term "RMS level" for clarification, so to please him I did. I didn't know until you submitted your readings for that WAV yesterday that there was also Maximum and Minimum RMS levels in addition to Average RMS! So, Jargon strikes yet *again* and causes even more confusion and frustration. The sound you like is closer to the reasult of normalizing to a maximum of -10dBFS RMS. Huge difference. Yeah, I guess *so*. So, despite the inherent evil in my habitual tamperings with the sound from a professionally technical point of view, does this "huge difference" make what I've been saying all along seem just a little more sensible and harmless (at least for my purpose in making "louder/better" MP3s) than it ever did before? Because based upon all the listening to my files that I have done, I've yet to create an aurally offensive, pop/rock MP3 by going with my beloved -10dBFS setting - and I have maintained this all along, despite all the rantings and condemnation. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#102
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Roger W. Norman wrote:
OR, obviously, if the band doesn't have a problem with you taking a CD of their product and doing whatever you want with it to make mp3s, then you're fine. Most if not all of the band's we're going to be working with are unsigned and looking for a better way to obtain any exposure at all. The extent of our production services would to be to assist in creating affordable, amateur demos which would are better than *nothing* if nothing is all they currently have by which to present themselves; a "first steps" kinda thing at best. Once they have enough resources available to afford hiring professionals, they can do so. However, as you can see, I'm still coming up with somewhat different scenarios because out of all these posts, you really haven't specifically said what your goal was, whom you were working with (titles, not people's names) and such. I certainly understand the difficulty in your position in that regard. It's a little hard for me to since I'm still working with the client and using a lot of what's being discovered via this discussion as information upon which to base more specific future decisions. A lot is still "up in the air" with this project as for exactly what we're going to be doing and how we're going to do it. Right now we have the basic idea down and are working to formulate a strategy for implementation. If we all look like we're talking off the topic, perhaps it's because we haven't really had this type of specific information to work with in formulating answers. This is true - and I have been taking this into consideration all along. Sometimes the replies are off-topic yet still useful for other additional topics which relate to my other projects. Other times, they're "dead on". The only times I risk losing my patience is when people post unwarranted, rude comments stating that I'm an idiot, etc. Those people are just looking to prove themselves superior and are not here to help. The sooner they leave the discussion, the better. I mean, over this thread I've recommended for you to talk with the developers of the mp3 specs, the coders have been suggested by others, talking with people whom are in the know about how to mix for mp3, and even downright calling you on some perhaps misinformed assumptions. Unfortunately, the thread has kept me so occupied, however, that I simply haven't had time enough in each day to do all of that - but I will. I definitely have my homework cut out for me! And, of course, getting called on our own misinformed assumptions didn't help any. As long as truth prevails in the end, it doesn't matter to me. This thread has so far been a great example of miscommunications throughout, or maybe that's just my perception. No, you're quite right - and my experience with the other group made this quite clear to me before I spawned this discussion with my "hypothesis". I didn't want David to even begin looking at my screenshots before I knew that everyone involved was "clear on the terms" but he Googled them out anyway so there wasn't much I could do about it. But ifyou're working WITHIN the creative process, then that was either my greatest misconstruence, or your worst presentation of information! g My sin of omission, perhaps? g So still, what's it gonna be? Theory of perceptual encoding, or pratical application? g Hehe... Being not an audio professional, I generally weigh in on the side of practical application because that's from the money's ultimately going to come. However, understanding the theory is always helpful as well when it comes to making decisions about things where practical application says "either way is fine". In such cases, I always prefer to go with what is more technically correct than "I don't know. Guess I'll do this instead of that" - even if the difference is not readily apparent. There's always a certain level of satisfaction in knowing that you've "done the right thing" even if no one else can tell. Thanks for all your input. It has not been wasted on me. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#103
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
As I can see, the relative loudness of all the tracks are not well preserved. I don't know what that picture shows, but according to the documentation of Normalize it's "batch" mode is meant specifically for preserving the relative loudness of all tracks. In this mode it treats all the tracks as one long audio stream. You are absolutely correct about that, Jonas. And, our little limited test indicates that Normalize does not do that. It seems that with Myke's settings it normalizes the tracks, seen as one single stream, to a *maximum* (or close to maximum) -10dBFS RMS. Thank you. As other's have pointed out, this whole thread has been an excanple of miscommunication. If we had had access to the docs for Normalize from the start, the thread would probably have been very different. I have noticed in both threads - though definitely more in the other, previous one - that a lot of the miscommunication comes from people posting immediately to older messages from when terms were still misunderstood and problems were still unresolved. If they'd just keep reading a bit before they spout off on an already resolved issue, they'd see just how pointless some of their rantings and ravings have been. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#104
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
On Sat, 05 Jul 2003 10:36:00 -0500, Lord Hasenpfeffer wrote:
Martin Tillman wrote: I "batch Normalized" the entire set of 10 separate WAVs from that CD to -10dBFS and then loaded then all into Audacity end-to-end so I could then take the pretty picture. Ah, so that's how you managed to bugger up the dynamics. Shame you didn't respond to me when I pointed this out buggeration days and days ago. The dynamics are not "buggered up". Explain your levels at 11 min and 24.5 min with respect to the original. To those reading this who haven't seen the screenshot, in the original MFSL CD, the level at 11' is significantly below that at 24.5'. In the arse whipping Hasenpfeffer remaster, the relative levels are reversed. As there is no scale on the screenshot it is impossible to know real values, but I'd guess that the level at 11' is 4dB below that at 24.5 on the original, and 2dB up on the Hasenpfeffer remaster - a whopping 6dB difference. So, you don't feel like calling that buggering up the dynamics? Think carefully before answering, you *are* in the company of professional audio people, including me. Correct me if my memory has failed me, but there aren't many gaps between tracks on DSOTM. How do the jumps in level at the joins sound, given that each track has almost certainly had its gain increased by a different value from the one that precedes and succeeds it? You need to read the thread before you post this kinda crap. Did you not say that you normalised each wav individually, then joined them up again? ------------------------------------------------------------------------ Message-ID: Well, just so ya know... That screenshot of "Dark Side..." with the clock, the cash, and the paper? Well, *that* is what happens when you "batch Normalize" an entire set of WAVs from a single album -10dBFS. I didn't run Normalize across the whole album as a single WAV. I "batch Normalized" the entire set of 10 separate WAVs from that CD to -10dBFS and then loaded then all into Audacity end-to-end so I could then take the pretty picture. ------------------------------------------------------------------------ Hmm... You didn't... Serves me right for skimming. So, explain the level discrepancy. Deny it, and I'll have to post a damn screenshot, together with ugly graphics. |
#105
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
OK, I admit to not reading all the posts in the multiple threads about Wich is quite understandable. This thread has been extremely convoluting and confusing at times. Yes it has. I've had a hell of a time just keeping up with it myself - and all the rude interjections from people such as Martin who've *not* worked as hard as I have to keep up with it have been extremely annoying to say the least. I'm saying that Normalize will not normalize based on the average RMS of the file. It normalizes based on the loudest parts of the file, wich of course is a completely different thing. Just *how* different it is, though, is what's got me bugged now. So far you've said it's a "huge difference" - and as polite and honest as you've been with me throughout this ordeal, I'm certainly willing to believe you - but I'd still like something a little more specific to go on before we call this whole thing a wrap. Myke did say that he normalized to an average 10dBFS RMS with limiting applied, but he wasn't. He normilized to something similar to a maximum of -10dBFS with limiting applied. He simply didn't know how the Normalize application really worked. But I've got a much tighter grip on its reality now than I've ever had before, thanks to you! He did try to tell us that when he normalized a complete CD he did it in a way that kept the relative loudness of the tracks, but as he thought that kind of operation was implicit in the use of term "batch" he completely failed to explain it to us in the beginning of the thread. Gee it feels nice to be so completely understood for once! As the thread then became so large, the explanations and findings necessary to understanding what Myke actually did became hidden in variuous branches of it. I'll say they did! One revelation was finding the documentation for Normalize. It doesn't only explain that Myke wasn't doing what he said he was doing, it also partly explained why Mike's terminology was different from everyone else's. Bingo. Please note that I'm not a recording engineer or an electrical engineer, so my signal processing theory may be off. I'd be glad to hear from any signal processing wizards if I've made faulty assumptions regarding signal power, perceived volume, or any of that fun signal theory stuff. Again, it all comes down to programmers vs. engineers. These two really do need to learn how better to get along with one another - especially as we all depend on each others specialized talents and abilities just to make it through our daily lives. I'm sorry I don't dream of logarithms in my sleep, but that's just the way I'm made and I can't say I've ever felt a need to change in that regard. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#106
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Lord Hasenpfeffer wrote in
: No, not at all. And in fact, I've done it already. Please see my next reply to your message. Yes, I replied before I saw that. Do you have these files to listen to anywhere online? Regards, Mark -- http://www.marktaw.com/ http://www.prosoundreview.com/ User reviews of pro audio gear |
#107
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Mark T. Wieczorek wrote:
Do you have these files to listen to anywhere online? The WAVs? Or The MP3s? Or both? Unfortunately, I've deleted them already. But I can make them again if you'll tell me what you'd like to examine. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#108
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
On Sat, 05 Jul 2003 19:58:49 -0500, Lord Hasenpfeffer wrote:
Martin Tillman wrote: Explain your levels at 11 min and 24.5 min with respect to the original. snip So, explain the level discrepancy. Deny it, and I'll have to post a damn screenshot, together with ugly graphics. When you have that much data being packed into a single frame, it is not practical to believe *everything* you see because the resolution of the data being represented exceeds the resolution of the image. Hmm... If that screenshot is distorted so much due to the scaling that the points I'm talking about AREN'T so hugely different, then, well, words fail me. If you'd like, I will re-rip and re-batch-normalize the WAVs from that CD and provide close-ups of those points in particular. Yes, I'd like, (or, even better, email me mp3s of the two versions those two areas so I can see for myself in CoolEditPro). But haven't you still got your 'remaster'? Even the mp3 would do - it's not going to look so different from the wav. I'd hate to think that a recreation of your 'remaster' would be any different from the original remaster - can you be absolutely certain of your original settings in Normalize? |
#109
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
"Artie Turner" wrote in message m... Jonas Eckerman wrote: That was why I thought you didn't really want your music normalized to an average -10dBFS RMS. And now we see that I probably was correct in this. I thought dBFS was a peak voltage indication reading and in contrast, RMS - root mean square - was a mathmatical average power reading. It would seem to me that dBFS and RMS don't belong in the same measurement. I never dreamed this thread would go this far... Given that the RMS of a single cycle sinewave is 0.7071 of the peak voltage, there is a simplistic relationship there ... geoff |
#110
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Geoff Wood wrote:
"Artie Turner" wrote in message m... Jonas Eckerman wrote: That was why I thought you didn't really want your music normalized to an average -10dBFS RMS. And now we see that I probably was correct in this. I thought dBFS was a peak voltage indication reading and in contrast, RMS - root mean square - was a mathmatical average power reading. It would seem to me that dBFS and RMS don't belong in the same measurement. I never dreamed this thread would go this far... Given that the RMS of a single cycle sinewave is 0.7071 of the peak voltage, there is a simplistic relationship there ... Then it's either redundant, misleading or both to have both RMS and dBFS in the same reading, especially if you're talking about the subjective loudness of program material and not sine waves. geoff |
#111
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Martin Tillman wrote:
Hmm... If that screenshot is distorted so much due to the scaling that the points I'm talking about AREN'T so hugely different, then, well, words fail me. Many times with Audacity I've seen peaks which appear higher than they really are because pixel resolution is only so good. Zooming in on the waveform solves this. If you'd like, I will re-rip and re-batch-normalize the WAVs from that CD and provide close-ups of those points in particular. Yes, I'd like, (or, even better, email me mp3s of the two versions those two areas so I can see for myself in CoolEditPro). We'll start with the screenshots. But haven't you still got your 'remaster'? Not the WAVs, no. If I want to hear CD-quality audio, I play the CD. I have stated many times that my effort here with Normalize is to create "better sounding" MP3s (and MiniDiscs on occasion where MP3Gain is useless). Even the mp3 would do - it's not going to look so different from the wav. Yeah, but what ****wit would send you an MP3 in a discussion like this? I'd hate to think that a recreation of your 'remaster' would be any different from the original remaster - can you be absolutely certain of your original settings in Normalize? More specifically? Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#112
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
David Morgan (MAMS) wrote:
It is still *very* unclear as to what kind of Linux 'command line' options there are in this software called "Normalize". Ask and ye shall receive. (I can post the entire man page which came with my version of Normalize if you'd care to see it.) Unfortunately with these screenshots, we haven't been seeing the same song(s) used as a continuous reference. Nobody's made any particular suggestions yet in that regard. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#113
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
"Lord Hasenpfeffer" wrote in message ... David Morgan (MAMS) wrote: It is still *very* unclear as to what kind of Linux 'command line' options there are in this software called "Normalize". Ask and ye shall receive. (I can post the entire man page which came with my version of Normalize if you'd care to see it.) Sure. Unfortunately with these screenshots, we haven't been seeing the same song(s) used as a continuous reference. Nobody's made any particular suggestions yet in that regard. Destroy something on purpose. RMS normalize an older song to say -6dBFS. Find the highest average RMS value area of the normalized song and start some screen shots from about 30 seconds in duration, expanding the .wav in each subsequent shot until you reach only a second or so in duration. Provide several incremental screen shots of this alonf with a before and after of the entire song. ....Just an idea to start with so we can try to see what the software is doing and then ask for other, supportive screenshots later. -- David Morgan (MAMS) http://www.m-a-m-s.com http://www.artisan-recordingstudio.com |
#114
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
Below are the values CoolEdit shows for the file http://mmm.truls.org/temp/AAahAAA-She(3).wav --8-- 0dB = FS Sine Wave Using RMS Window of 50 ms Left Right Minimum RMS Power: -56.77 dB -56.24 dB Maximum RMS Power: -3.92 dB -3.68 dB Average RMS Power: -11.81 dB -11.95 dB Total RMS Power: -10.64 dB -10.81 dB --8-- 0dB = FS Square Wave Using RMS Window of 50 ms Left Right Minimum RMS Power: -59.78 dB -59.25 dB Maximum RMS Power: -6.93 dB -6.69 dB Average RMS Power: -14.82 dB -14.96 dB Total RMS Power: -13.65 dB -13.82 dB --8-- [mykec@sillygoose mykec]$ normalize -n AAahAAA-She(3).wav Computing levels... level peak gain -8.1064dBFS -0.2637dBFS -3.8936dB AAahAAA-She(3).wav Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#115
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
On Sun, 06 Jul 2003 16:51:49 -0500, Lord Hasenpfeffer wrote:
Martin Tillman wrote: Hmm... If that screenshot is distorted so much due to the scaling that the points I'm talking about AREN'T so hugely different, then, well, words fail me. Many times with Audacity I've seen peaks which appear higher than they really are because pixel resolution is only so good. Zooming in on the waveform solves this. It's the opposite in CEP, but we're hardly talking individual samples here. Even the mp3 would do - it's not going to look so different from the wav. Yeah, but what ****wit would send you an MP3 in a discussion like this? The peaks won't be significantly different from the wav. I'd hate to think that a recreation of your 'remaster' would be any different from the original remaster - can you be absolutely certain of your original settings in Normalize? More specifically? How can I be more specific than 'can you be absolutely certain of your original settings in Normalize?' ? |
#116
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Left Right
0dB = FS Sine Wave Maximum RMS Power: -3.92 dB -3.68 dB Average RMS Power: -11.81 dB -11.95 dB [...] 0dB = FS Square Wave Maximum RMS Power: -6.93 dB -6.69 dB Average RMS Power: -14.82 dB -14.96 dB level peak gain -8.1064dBFS -0.2637dBFS -3.8936dB AAahAAA-She(3).wav Well... Now we really don't know what Normalize bases it's calculations on, wich is what I suspected a test on a real music file would result in. :-) Does the above numbers mean that Normalize will lower the amplitude of this specific file with 3.9dB? Regards /Jonas |
#117
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
On Mon, 07 Jul 2003 16:43:15 -0500, Lord Hasenpfeffer wrote:
How can I be more specific than 'can you be absolutely certain of your original settings in Normalize?' ? Are you meaning in terms of recreating my steps? If so, absolutely certain. There's only one setting that I specified: -10dBFS as a target "level". That's OK then. |
#118
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Jonas Eckerman wrote:
P.S. Have you sent your song there to Dr. Demento? Dr. Demento? Is that some guy I'd recognize if I'd been to the US? I thought surely Dr. Demento was an internationally recognized phenomenon by now... Weekly radio show... Lots of oddball various artists compilations... He has a penchant for playing a lot of old novelty 78s and a lot of new recordings submitted by his fans and listeners; stuff like that to which virtually no other commercially-minded program director / air personality would ever consider giving airtime. He played a key role in helping Weird Al Yankovic get his start back in 1979. Why I'd even betcha some of the regulars in this newsgroup know him personally! http://www.drdemento.com/ Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
#119
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Lord Hasenpfeffer wrote:
Why I'd even betcha some of the regulars in this newsgroup know him personally! http://www.drdemento.com/ Cool site! I had forgotten that the Dr. had played "Lee Harvey Was a Friend of Mine" by my old buddy Homer Henderson back in '89. Artie Myke |
#120
|
|||
|
|||
Louder _ISN'T_ Better (With Lossy)
Martin Tillman wrote:
We'll start with the screenshots. (Of DSOTM) And are we going to see them anytime soon? I'm sorry, my hard drive's full of a whole bunch of other WAVs right at the moment and I'm back to work now. Things take longer than they do when I'm on vacation. Myke -- -================================- Windows...It's rebootylicious!!! -================================- |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Humming Sounds From Speakers just keeps getting louder? | General | |||
Lossy Compression | General | |||
Lossy Compression | Audio Opinions | |||
Subwoofer direction | Car Audio | |||
Advantage of tape over MD? | General |