|
Post by jsdyson on Feb 5, 2024 20:46:40 GMT
Here are some 'decoded' examples with the last phase of mastering at least partially undone. These examples include some ABBA, Linda Ronstadt, Steely Dan, Carpenters, Herb Alpert (AFAIR). I forget all. ABBA probably has the biggest representation. The decoder will be available FOR FREE in source and binary form, about 2wks. For practical use, at least 4 core CPU needed, but much easier at 8-10 recent CPU cores. (I use an i10900X.) A more complete set of demos will be available in approx 1 wk once the nits from the current version are resolved. Sorry for the delay beyond my original estimate, but I have been having problems with my hearing, so the final A/B tests have been tricky at best. There were finally two candidate versions, and one problem that I have been worried about is an excessive dynamics stretch. The design of the system offers 'steps' instead of 'tweaks', but there are still combinations of 'steps' (original encoding probably designed by R Dolby). Over 6yrs ago, I had underestimated what it really takes to decode this last step, but finally the current structure WILL accurately decode recordings. As mentioned above, there is still a final set of choices, but except for the sound of slightly excessive 'stretching', I doubt that you'll hear any distortion 'tells' from the processing itself. IF YOU DO HEAR A PROBLEM, PLEASE BE KIND AND TRY TO DESCRIBE IT TO ME. This thing CAN and WILL be nearly perfect. The decoder is mind numbingly complex and uses over 2000 equalizers, 24 Hilbert transformers, a myriad of FIR filters, and a dispersive audio 'compressor/expander' (along with 6 DolbyA emulations.) Here are some early demos (sorry, again must be snippets!!!) PLEASE DOWNLOAD THE EXAMPLES FIRST -- the Dropbox player will not produce the best quality.
www.dropbox.com/scl/fo/zw5omel88g5x1ywh2ks8n/h?rlkey=egakjobeiyucr6xhvqf1cfj5b&dl=0John
|
|
|
Post by jsdyson on Feb 3, 2024 3:32:28 GMT
charly, THANKS FOR YOUR POTENTIAL INTEREST!!! WOW, it has been a long time since I have been on this board!!! Your timing is perfect because the 'FA decoder' is starting to work like I promised 5yrs ago or more. Many many premature announcements have been made, and crediblity is lacking. It is likely that this release in the next few weeks will be functioning correctly. Demos coming in days, not 'weeks'. As of 2017, the software was a vain attempt at the desired function. The software has been more completely developed over the years, to the point where few settings (if any) are required. The software is still free in source and binary form. Plans are for source to be distributed along with binaries. Patents and proprietary knowledge has been totally avoided, and any offers would have been rejected. The program development has take so very long because I started with no knowledge about the precise techniques used for distributing commerical audio recordings... The final design might be one of the most complex audio processing programs that exists. IT IS A BEAST.... READ FURTHER FOR A PARTIALLY DEFECTIVE (but probably worthwile) demo... Less defective demos are coming in approx 4 days or less (late sunday 04Feb2024 or maybe a day later...) The project was many many times more complex than I had originally guessed. It uses both normal dynamics processing (e.g. attack/release type gain processors), but also a rather eccentric dispersive processor which provides infinitely fast attack/release (really -- sounds impossible, right?) After about 5 yrs since my last post here, there have been many, many, many embarassments with my mistaken proclamations about success. This version is working and initial reviews are showing positive results. I have learned to be more proactive about accepting CONSTRUCTIVE and respectful criticism. Nowadays, I use lots of sources for demos and testing, while ABBA albums are still very important. The sources for the testing are totally pristine, often using sources that are not normally available (e.g. the earliest Polar version, which sounds very little like the normal Polar stuff.) After many years of work, being 100% persistent, the first fully functioning version is being finalized.I have been finishing a new, finally fully functioning version. For input, it requires a pure CD or digital download .wav file only (44.1k, 48k, 88.2k, 96k along with multiples.) Mp3 (or opus or anything like that) encoded/decoded recordings produce suboptimal results. The decoder will sometimes work pretty well on that material.[/div]
IMPORTANT: when playing the snippet, PLEASE download first. The dropbox online player is much lower quality than the recording itself.
ALSO -- VERY IMPORTANT, the demo IS defective. Since I am not totally happy with it (it was intended to be a quick example of functioning on defective material), I'll produce a more complete example in the next day. Sometimes, I get excited and prematurely demo something that might be misunderstood. The new release should be final enough to produce a casual more complete demo tomorrow....
On my home forum, I have promised a series of approx 5-10 demos for late this weekend, then the entire demo sequence later on in the week. I'll inform this forum when it is ready, but the demo above is a first offering.
I'll add several ABBA tracks to the short list for this weekend, The 'normal' set of demos is between 95 and a few 100 tracks, depending on how much time I decide to spend in making the test demos.
A true release should be available in a month or so. I lost the machine where I do Windows builds, so I still have to rebuild the build environment for the decoder. The 'Windows port' has already been done, just that I cannot build it anymore. (Machine failed.) It will take up to 1wk to re-create the build environment (more likely a better part of a day, but gotta leave slop in the schedule.)
The decoder is now trivial to use, most of the time requires only the input/output file name along with decoding quality settings, often use a fixed set of command options. Previously, many years ago (e.g. 2017), even when the decoder was working, it required a complex set of settings. I have learned A LOT about the ubuitious final processing used on consumer commercial recordings.
Will return with more examples in approx 4days!!!
(This message might be a bit scrambled because I have been working hard while finishing up the decoder.)
THANKS SO MUCH!!! I'll be back in a few days with more info...
|
|
|
Post by jsdyson on Jun 12, 2019 21:20:12 GMT
Wow -- trying to get full quality for Dreamworld is incredibly difficult -- and I havne't been as successful as I initially thought.
I have included a very well done 'SuperTrouper' for reference - I really tried to make Dreamwold sound as good as the SuperTrouper... Spent almost a full day on this, after realizing that my previous example was so embarassing...
Dream World is apparently encoded 3 times in sequence!!! The multiple/nested 'DolbyA encoding' is the reason why Dream World is such a dense mass of sound. I ran Dream World through the processor that undoes DolbyA encoding -- three times in sequence (carefully calibratining for each step.) I still didn't get the results that I like -- but it MIGHT sound better after decoding to some people.
The DolbyA system was never meant to used to nest the encoding, and the original hardware would sound even worse than the results that I got (the hardware had limits on the kind of clean-up and processing that it could do.)
The originals for these examples are from the ABBA Gold and more ABBA Gold disks from 1992/1993, which were leaked with DolbyA encoding (giving that intense high frequency sound.) I kind-of like 'Dream World' and would like to have a nice, clean copy of it some day. I am showing the SuperTrouper example to show what I was trying to make Dreamworld sound like.
So -- My very very aggressive attempt of getting a clean copy of 'Dream World' (I guess not as good as I had hoped):
For a reference -- the best possible recovery of 'Super Trouper' from the same series of CDs. (This one actually sounds 'okay' - to me, this is acceptable):
John
I sure hope that some-day that they would truly remaster the old ABBA recordings and distribute normally (not doing extreme dynamic range compression like is done so often nowadays.)
|
|
|
Post by jsdyson on Jun 11, 2019 15:06:26 GMT
Wow -- did I make a bit of a mistake :-). The song didn't sound fully decoded -- they seemed to have run multiple DolbyA encoding passes by mistake... The first example (very nicely) removed only one pass. I found that there is at least one more pass, and possibly two more that are needed!!! Doing the 2nd pass improved it further, but the third test pass was a little unstable sounding. It is a matter of tweaking calibrations. It will be EVEN NICER in a day or so. (It takes awhile to get the calibration correct without DolbyA tones.)
So, figure about 12Jun, I should be posting a more-clean yet version.
|
|
|
Post by jsdyson on Jun 11, 2019 10:43:42 GMT
I have been working on my mainline project for pretty much the last few years -- dropping in from time to time. Thank you for tolerating some of my previous attempted improvements. The normally available copy of 'Dream World' is undecoded DolbyA (the reason for the shrillness), and also is kind of a worst-case hit against DolbyA -- it is a kind of fragile noise reduction system, and certain kinds of material (like Dream World) can excite 'odd' behaviors. As a 'guilty pleasure', I have been trying to do the best possible to extract the most of Dream World from the normally available recording. If someone knows where/how to get a fully decoded version elsewhere, please let me know. This version was decoded from a Japanese copy of an ABBA Gold type collection (AFAIR.) Some ABBA releases aren't actually decoded, and this mostly causes a brighter sound with more high frequency density.
This is for research and interest-in-the-group purposes -- and if someone knows how to put it on 'YouTube' or whatever, go ahead and do it -- then the monetization would be done properly, and everyone can more properly benefit. These works are owned by ABBA and their assignees -- and I want for them to benefit. (Up until now -- it has been pretty much impossible to fully decode this material. I am participating on a project that extracts as much as possible from old DolbyA encoded recordings.)
(Don't expect my result to sound as 'bright' as the normally avaiable version -- but it is probably much closer to the pre-DolbyA version. It is still too smushed.)
|
|
|
Post by jsdyson on Apr 26, 2019 21:43:32 GMT
Kind of opposite topic -- mentioning 'Thank You For the Music' -- I actually like the 'Thank You For the Music, Doris Day edition', only because it is a cute almost kind of parody. About not liking some material on Waterloo -- it *does* seem like some albums are much more listenable than others. I find that my taste moves around from album to album though.
|
|
|
Post by jsdyson on Apr 22, 2019 21:43:47 GMT
I rather like Super Trouper although, as with most of ABBA’s 1980s songs, it doesn’t match up to the songs from the peak years (1976 to 1978). With regard to its recording quality, I wonder if it was rushed (unusually for B&B)? It was the last song recorded for the album, apparently when B&B decided Put On Your White Sombrero wasn’t quite good enough. As for songs I never listen to, I usually play albums straight through from start to end. B&B’s standards were so high that there are only a couple of songs in ABBA’s entire output that I actually dislike. I am sure B&B would never have let through some of the dross on The Beatles’ White Album, for example. If you are sometime interested in the cleanest possible copy of SuperTrouper, I can produce one for you -- but i have mostly used SuperTrouper for 'if my DolbyA compatible decoder can cleanly decode SuperTrouper', it is doing pretty well ;-). SuperTrouper is VERY VERY difficult to properly convert from the DolbyA copy to something that should be on a CD. (When I claim -- 'cleanest possible', I really mean it from a mathematical/signal processing standpoint.)*
I cannot say that I really don't like SuperTrouper -- it is just my hearing is so distorted because of listening for audio problems -- sometimes all I hear are problems with audio, and don't really hear music anymore. I used to really like music
John
|
|
|
Post by jsdyson on Apr 22, 2019 16:29:18 GMT
There is no such song. I don't skip even I Saw It In The Mirror, which I consider their worst song ever, when I listen to the Ring Ring album (which I do seldom, to be honest). I tried to compile my least favourite ABBA tracks in the "Bottom 20 ABBA songs" thread (http://abbachat.boards.net/thread/1090/bottom-20-abba-songs) but even those 20 I listen to now and then. Well -- yea, we all end up playing songs that we don't listen to. However, there are songs that I don't participate in listening - even if they are playing/being played. I tend to agree with you -- I should add 'Mirror' to my 'doesn't sound characteristic' list.
|
|
|
Post by jsdyson on Apr 22, 2019 16:18:09 GMT
My favorite worst is SuperTrouper -- recording quality really bad, I don't know what they did to it. It is so bad that I don't even know if I like the song or not. I do use it as one of my worst-case tests -- one requirement is that my software can process SuperTrouper to have no weird 'splats' in it. I don't think that I could tolerate listening to the entire song (even though ABBA is one of my favorite guilty pleasures in general.) Others on my non-favorites are Move-On, Rock and Roll Band, and Watch Out. I think that those kind of bother me because they are uncharacteristic of typical ABBA songs.
|
|
|
Post by jsdyson on Apr 22, 2019 15:23:33 GMT
Was in the middle of a discussion about some kind of processing matter, and needed to produce an example. Just randomly grabbed 'Soldiers' from my collection of leaked DolbyA copies.
Good for my benefit, during the discussions of the sound quality on the processed copy of 'Soldiers', I found ANOTHER bug in the DolbyA decoder -- resolved nicely, but after all of that, i actually noticed the song instead of the processing. I found that (in my opinion) the song is really kind of pretty. Good ole me, I decided to put my non-existent mastering skills to the project -- what can I legitimately do to the song? Can I do something to make it sound more 2000's instead of 1970's? My original copy is DolbyA compressed, so I can undo that compression -- which gives a very dry sound to the recording. The result wasn't nearly as pretty as I had hoped. Played around a little (didn't take all that much work), and did the following:
Original -> DolbyA process -> 3 band RMS compression -> small amount of EQ -> time-limit/create mp3 at 320K
EQ choices from the decoded copy:
Bass +1.5dB @120hz/Q=1.0 Bass +1.5dB @360hz/Q=0.707 Treble +3.0dB @6.3kHz/Q=0.707 Treble -3.0dB @12khz/Q=0.707
The results (IMO) sound more current -- less of the AM radio optimized sound... Also, made a snippet of the Polar CD available to compare.
Cannot make the entire thing available, so I cut it so that it is less than 1/2 of the original (sorry.) Please, please enjoy. I'll keep this online for a week or so.
I think that IF (big if) I was allowed to remaster some albums -- could probably make them sound more 'current' -- I do have some proprietary software just going on the market. The new decoder is the only way to get the velvety clean sound -- old DolbyA is kind of grainy. Oh well -- I'd be willing to do it for free :-).
John
|
|
|
Post by jsdyson on Apr 13, 2019 13:14:01 GMT
Here -- I wanna show what I think that ABBA should sound like (maybe a little brighter -- I am not a mastering engineer), but this general sound. The 'complete studio recordings' are so dynamic range compressed that there is little of the original 'life' in the music. It is all crunched. I am demoing the same dynamics as the source ABBA material (pretty much the same), and re-did the EQ just a little bit for 2000's sensibilities. I think that a little more brightness than my demos would be desirable, but there ARE tradeoffs that an actual mastering engineer might want to make. I tried to provide the absolute cleanest sound, which is NOT the same as clear. Trying to be a better citizen than in the past, I limited the demos to about 45 seconds (maybe 20 or so short demos) -- but here is what I mean (this sounds more like what I think I remember back in the 1970s when the music appeared on the scene.):
PS -- just updated with a slightly brighter sound (minor changes can make major differences.)
|
|
|
Post by jsdyson on Apr 8, 2019 23:40:51 GMT
Here is what I mean when I say that the TCSR is too 'compressed'. The peak signal levels of the two examples below are identical within 0.01dB (very close to the same peak levels), but the peak-RMS level on the 'goodmaster' version is just under 18dB, and for the TCSR version it is about 12.3dB. Anything below 14dB peak-RMS is very compressed (levels all the same -- really intense.) 18dB for peak-RMS isn't great, but is pretty good for pop music. I grabbed a copy of a leaked DolbyA copy (commercially sold, but left DolbyA encoding on it), and decoded it -- basically did a simple remaster. I tried to make it sound close to the TCSR, except I didn't compress it. So, except for the vastly superior DolbyA decoding and the lack of compression -- the versions are rooted from the same source. The DolbyA decoding is NOT important here, but the sound of compression -- WHERE THE PEAK SIGNAL LEVELS ARE THE SAME. The difference in loudness is *astounding* -- and all life is taken away from the compressed version. *The 'goodmaster' version has the original dynamic range, but is decoded VERY differently from the original, so WILL sound more like the TCSR frequency response balance. If you play these, PLEASE play the TCSR version first, because otherwise the TCSR version will be too loud if you set the volume for the 'goodmaster' version. I limited the lengths of the demo recordings to avoid taking away sales of the recordings (whether or not someone likes TCSR or the original dynamic range.) My copy has LOTS of electronic ticks and pops (not vinyl pops, but electronic glitches.) SO, please be tolerant of those -- I don't ahve actual masters, but just copies that are close to masters.
www.dropbox.com/s/ksb4f5aoczxcpt4/Bobby-tcsr.mp3?dl=0
One more -- I didn't add a TCSR version of disillusion -- but to show how simple clean/clear ABBA can sound (ignore electronic glitches):
John
|
|
|
Post by jsdyson on Apr 8, 2019 22:21:36 GMT
In my opinion, even though the content is great -- the sound quality sucks and is severely overly compressed (4.5-5 crest factor, 13dB+- peak-rms), that would be The Complete Studio Recordings. I have The Complete Studio Recordings, and it sounds fine. I have it also -- when the crest factor is only 4.6 and the peak to RMS is 14.5dB approx -- it is too compressed. Did you notice that it is VERY loud? It is all mashed up -- dynamics are really restricted. GOOD ABBA copies don't sound that same vanilla loudness all of the time. Normal well made ABBA recordings have crest factors in the 6+ range -- sometimes in the 9-11 range. The peak to RMS on the real recordings is seldom less than 17dB. Regrettably, I have archived TCSR off onto a secondary back-up media (something is not right with that kid -- I mean, compilation.) Some of the selections on TCSR are difficult to find -- but mixing the super loud TCSR versions with the more traditional "real' sounding versions just isn't right.
The actual ABBA recordings sound very different from Album to Album -- On un molested material, Arrival has a bit of a grainy quality, and Ring Ring (the way that the original vinyl sounded) was too muffled, a properly mastered SOS has wonderful high end -- the sibilance is actually clean sounding. It is all in the mastering -- and TCSR basically sounds like they used broadcast style loudness wars processing.
The TCSR squeezed ABBA to death -- ALL LOUD ALL THE TIME!!!
John
|
|
|
Post by jsdyson on Apr 6, 2019 14:44:43 GMT
PS -- one reason why it is obvious that there is more 'edginess' or seems like more 'highs' in the TCSR copy is that they use multi-band compression, which keeps the 'tone control' boosted when it can. It is like a graphic equalizer keeping the bands relatively equal. If you listen to the original polar version (or vinyl versions) it sounds more like the examples that I showed, except the versions that I showed extract more detail from the original recording. The TCSR brute forces loudness (probably used a broadcasting processor squashing the sound to be loud all of the time.)
|
|
|
Post by jsdyson on Apr 6, 2019 14:29:19 GMT
In my opinion, even though the content is great -- the sound quality sucks and is severely overly compressed (4.5-5 crest factor, 13dB+- peak-rms), that would be The Complete Studio Recordings. It is so very sad, because it would be a FANTABULOUS collection if done correctly. For an idea as to how I think ABBA should sound from the Voulez Vous Album -- two snippets: www.dropbox.com/s/si1hw14mi3em5ej/GoodAsNew1.mp3?dl=0www.dropbox.com/s/jy0euuts1f8971t/GoodAsNew2.mp3?dl=0(Decoded from commercially (legally) leaked DolbyA material, decoded by a new commercial decoder.) John
|
|
|
Post by jsdyson on Dec 18, 2018 7:12:16 GMT
Wonderful news -- came up with a new version of the DolbyA compatible decoder, where there used to be a dynamic frequency response hump between 3k-9k just like DolbyA HW. I removed that hump because the result is a sound like distortion (esp on material like ABBA.) The 'hump' is actually due to an eccentric attack time, and manifest very similar by the real DolbyA. That is actually a 'bug' in the design, but my design doesnt need that bug. Here is the result -- a WONDERFULLY clear sound -- even .mp3 sounds better. I think that I got the first-ever good decoding of Dreamworld (not perfect, but MUCH better.) Dreamworld is hideous to DolbyA decode, and that is probably one reason why it was relegated to like a secondary status. With this new software, hard sound stuff like SuperTrouper and MamaMia (and even Waterloo) is better. Sample listening repo: spaces.hightail.com/space/yDG3L339Rn
|
|
|
Post by jsdyson on Dec 16, 2018 23:04:37 GMT
Hey -- I apologize -- I just fixed the EQ on the recordings. I made a mistake!!!
John
|
|
|
Post by jsdyson on Dec 16, 2018 17:39:00 GMT
I got some feedback that the sound was a bit dead -- so instead of showing some experiments, and also simply decoding. I just did a not-quite full remaster (I didn't do frequency response limits withn the audible range -- nothing like that.) The results are bright like people normally expect -- maybe too bright, and the vocals are as clear as can be done. I did find out that in some cases, there were multiple passes of DolbyA (or other extreme compression) for effect, and I was going to demo that -- but most people here are mostly just interested in the listening experience. Youll notice that the material is very clear -- and even during 'honey honey', when one of the guys says 'high' -- the voice actually seems to move 'high' when on phones. So -- I put up some really clear, as-expected examples. They are effectively 'remastered' this time, rather than experiments. Have fun. (One warning -- the material is very bright -- you might want to be ready for a slight tone control -- maybe about 1.5dB of treble cut... The material is a 'kind of' full remaster, but I didn't do a final contour of frequency response (I'd actually do about 1.5-2.5 dB of a 20kHz cut with a Q of 0.500 -- which is a very mild/smooth cut.) I did NOT 'tweak' the results beyond my planned results. Here is my repo that I listen to: spaces.hightail.com/space/yDG3L339Rn
|
|
|
Post by jsdyson on Dec 11, 2018 13:54:59 GMT
The previous results weere pretty raw -- just the raw material, before a lot processing. Now, I have produced/finalized examples of the actual material that I listen to. They aren't full quality due to space limitations, but better than .mp3 files. One thing that you might notice are the midrange tones from the vocals arent suppressed. Again, these are what I listen to. I might change a few of the copies as I want something different. I also found much better source material -- they played games with it, and I had to compensate a little. Much cleaner results. repo: spaces.hightail.com/space/yDG3L339Rn
|
|
|
Post by jsdyson on Dec 10, 2018 2:51:00 GMT
My repository contains some redecoded examples -- with the correct distortion sideband removed. You should notice a significantly more mellow sound, but still ABBA. The 'S' sound is cleaner, and generally the voices come through more clearly. This examples will probably be up for a few days - am intending only to show examples of what ABBA can sound like. My previous examples WERE improvements, but these go a few steps further. This is exactly the same processing, but with corrected parameters. Absolutely no compression/expansion/ambience recovery. My repo only has a few samples that I have space only for a few days. Please enjoy and compare with the distributed copies. In some cases, mine might be much naturally clearer, but sometimes the distributed copies did get it right. I wish that they treated the ABBA material better, but nowadays 'LOUDER IS BETTER'. I don't think so, but seems like everyone else does. I do wish that there was a *legal* & *ethical* way to distribute the 100+ ABBA recordings in my posession, but cannot because it just isnt practical. My DolbyA compatible decoder is also too complex for the average listener, and is not meant to work in a HiFi system. So, letting people use the software for themselves is also not practical. If it would be beneficial or useful, I would offer free consumer versions of the decoder(without some of the advanced quality) -- however, again, it is too complicated to use. It IS being used by some recording pros -- interest is developing, but even then -- I will NOT be able to make money, and it is even not convienient for them to use. (I have started working on a GUI, but it would be foolish to believe that the DolbyA compatible decoder will be adopted by very many pros either.) I did this decoder effort for the love of the music and the listener. Hopefully, the recording distributors will start taking better care of their product!!! I'll have to clean out this repo by Tuesday, but I really hope that this somehow opens eyes to improve what we buy. (I am working on some other projects and need the space.) I have spent mucho money over the years for my collection, and ABBA has been hundreds of dollars. I want to help the ABBA people and the listeners both. Maybe I am just being silly -- I just don't know!!! Repo: spaces.hightail.com/space/yDG3L339Rn
|
|
|
Post by jsdyson on Dec 9, 2018 17:39:12 GMT
Man -- I am embarassed. Getting technical, my DolbyA compatible ABBA examples were decoded with the wrong IMD sideband suppression direction. Basically, it means that the results were distortion reduced, but not anything like what is possible. I'll try to produce some examples in the next day or so. It takes approx 2-3Hrs to process my entire repository (of 106) ABBA album recordings. (Actually, I have more of them, but this is my album repository.) I also have some 'extras' sitting around with some of the more obscure songs (e.g. The 'Doris Day' version of one of the songs.) However, processing 106 songs takes appox 3Hrs -- so it will be at least that long before I can produce any demos. (I also have to select the examples, then mp3 encode those, etc.) Takes some time for sure. I'll probably be able to edit this post with the announcement in about 5-6Hrs. (BTW -- it took me 20+yrs to collect my repository!!!)
|
|
|
Post by jsdyson on Dec 8, 2018 10:26:55 GMT
I had to free up the space for the decoding examples. Just wanted to show how 'nice' ABBA could sound if properly handled. In some cases, by DolbyA decoding some of the old CDs that I had i my collection -- the sound of the old vinyl (or better) could be retrieved. In a way, the lazy record labels gave us a 'diamond in the rough', it is just a matter of further processing some of the old leaked-DolbyA material.
The sad news is that more recent CDs tend to be hyper compressed -- for example, it isn't uncommon on the old CDs to get a peak-RMS dynamic range of approx 20dB (pretty tolerable for pop music), but the 'The Complete Studio Recordings' tend to have a peak-RMS of 13dB (terribly compressed), or the crest factor differences of perhaps 7-10 for the old CDs vs. the hyper compressed recent stuff of 4-5 at best.
Admittedly, the hyper compressed stuff can sound 'okay', but loses some of the music.
I am not claiming that the decoding operations that I did were perfect -- I am not a recording engineer, but an EE/DSP/Software person, so I don't have the skills or equipment to pull a full release together, but we (as listeners) are being subjected to inferior quality material in the name of 'louder is better'.
For example, some of the recent releases appear to be 6-10dB louder, even though electrically the signal peaks are the same!!! That is a HUGE amount of compression. I have a simple 3 band RMS compressor that tightens up the sound a little, and might add 1-3dB of loudness, and even then -- the music is damaged by doing that simple/small amount of processing.
What can we (as listeners) do? I don't know, but (for example), my DolbyA compatible decoder is the first software decoder that comes really close to a real DolbyA unit, and previously the distributors would have to convert from digital (dolbyA encoded) to pass through a DolbyA unit in real time, then convert back to digital again. That is an expensive operation, when they could just get by with copying the digital master (perhaps sometimes with a little bit of tone control...) ABBA is compressed anyway, right? So, why not add a little bit of DolbyA compression on top? (DolbyA compression tends to sound ugly, and isn't really meant to be listened to directly.)
Oh well, I have just been trying to help -- and intend to keep work on decoding projects, including DBX type I and type II, DolbyB, and DolbyC. We are considering C4 and SR, but those are more complex devices and even though DolbySR was used for a while -- it is exceedingly complex. DolbyA was complex enough!!! Additionally, DolbySR does so much of the wrong type of compression, I doubt that much SR material has leaked to the consumer.
John
|
|
|
Post by jsdyson on Dec 6, 2018 20:46:41 GMT
Just cheked out my storage locker, and found some VERY INTERESTING imported (from Japan) ABBA CDs, and took a look at the contents. Until now, I only had reasonable DolbyA copies of ABBA Gold and More Gold, but not much else. I wasnt' always satisfied by the Gold/More Gold material -- seemed like there might be an extra DolbyA encoded/decode generation beyond what really sounds like the best possible. (the specialized/super-advanced DolbyA decoder did a lot of work to maintain whatever quality was left.) So, when I ripped the CDs, and checked the material -- lo and behold, all of the imported CDs had all of the hallmarks of DolbyA encoding (excess high frequencies, more hiss on a spectogram than the material should have, the ghosting in the hiss (on the spectogram) looked like it was somewhat muted, and most importantly -- the CD material sounded correct and MUCH MORE natural once decoded. I suspect that I had originally put the CDs aside because of the sound being too tinny (back in the '90's some time.)Now, I have (I think) all of the common ABBA albums in DolbyA encoded form along with the Gold/Gold2!!! Not only that - like I mentioned above, the quality of the old/imported CDs are a 'notch' better than my Gold/More Gold CDs. A few of the pieces (2 or 3 out of perhaps 50-80) appear to have an eccentric DolbyA setting -- but that is fairly common in recordings of late 60's through the 70's. One of the recordings (Kisses of Fire) needed the 80-3k band setting to be 6dB lower (which means higher up on the gain curve.) This change mitigated the gating resulting from standard decoder options. The other demo examples are pristine and beautiful without tweaking. (If material isn't DolbyA encoded, very often the sound is ugly and muted, with gating often obvious at lower levels.) The MamaMia example seems to be noticeably different from the vinyl -- even without a quick A/B comparison, where the digital/decoded copy sounds more full and natural. In fact, I noticed that the vinyl in general has a few dB less bass than my copies when decoded. It isn't an amazing amount of additional low-end, but it is there. My processing for these examples is essentially a DolbyA decode operation (with minor adjustemens when helpful -- seldom needed), then an ambience extraction (which was necessary to get the same spatial image as the vinyl copy.) The repository now has both mp3 examples and flac examples. Also, the examples don't just come from the 'more common' ABBA songs, but really interesting alternative selections. Have fun!!! (I had disappeared for a few weeks when completing/testing the DolbyA compatible decoder project, now I need to 'go away' and focus on the very desirable GUI for users that expect that kind of thing nowadays.) REPO: spaces.hightail.com/space/QecbMZKnWh
|
|
|
Post by jsdyson on Dec 4, 2018 17:22:27 GMT
I was listening to the mp3s that I translated from the 96k/24bit flacs that the processor produces, and was disappointed, even with the reprocessed & improved .mp3s. A lot of the detail is smudged over -- especially when one is used to the hyper compressed versions normally available on recent distributions. In general, the results coming from my processor are very close to the original vinyl albums (sometimes even less compressed than some of the album recordings.) There are a few songs that are even moderatly compressed on the vinyl copies. Of course, there is nothing wrong with artful compression. The results that I have presented aren't meant to do anything but demonstrate what ABBA sounds like when it doesn't have nearly as much processing as what is normally on a recent CD. I promise you, that my processing is purely a decoding operation ONLY, and also I do bandwith limit even the .flac versions at high sample rates. For material like ABBA, I do a sharp linear phase filter at 20k. But what I describe here is the limit to the processing from leaked DolbyA source material. (Maybe about 1/3 of the ABBA releases appear to be DolbyA encoded on the consumer CDs/digital downloads/etc.) So, first -- I have reprocessed the examples where it would make a significant difference. The .mp3s are encoded with the best available mp3 encoder, and also at a bitrate of 320k. Even then, the .mp3s sound a bit crunchy to me, also with missing detail. Secondly, since the .mp3s are really not as good as I had expected, I have produced some 48k/16bit flac examples -- directly translated from my higher quality material. I find this hard to believe, but it seems that even the .flac results have lost a bit of detail, and I am an advocate that 96k/24bit isn't needed, but something psychologically or whatever is telling me that the 48k/16bit copies really do not sound as good. However, 48k/16bit is as good as I can show as demos. I have made improved mp3 examples, and flac examples -- can be played directly online, no need to download or anything like that. I only had enough space for a few flac examples, but they do sound okay (better than the .mp3s.) One more thing -- I did a survey of the normal consumer copies - and my memory was very faulty -- the ABBA stuff really does loose a lot of impact and detail with all of the final processing. I do hope that you enjoy the examples, and they'll probably disappear by Friday 7Dec -- maybe a day after, as I need to reclaim the space and don't mean to permanently distribute anything, but trying to help with a listening experience: Improved mp3 repo: spaces.hightail.com/space/QecbMZKnWhSmall (clear sounding) flac repo: spaces.hightail.com/space/IGsGMA6iiU
|
|
|
Post by jsdyson on Dec 3, 2018 13:12:59 GMT
As some might know, I am working on 'attempted to be very pure' audio processing software, and one of my test materials is ABBA. (I do use a vast collection of POP and other material from the early 1960s through the early 1990s.) ABBA is usually one of my 'ACID test' testcases. ABBA used lots of processing, wall of sound, and creative techniques. This heavy processing as used by ABBA tends to increase the challenge while recovering as much original studio sound as possible. These are examples that just happen to be on my repository, and I really would like to share a listening experience. *THESE ARE NEW RESULTS -- INCREDIBLY IMPROVED* UPDATE: a few examples (at least, 'the day before you came") was processed incorrectly. It has been updated as of Dec 3, 17:45EST/USAThis processing is meant to remove/clean-up processing that happens AFTER the recording studio, and avoid using destructive techniques (that is, undoing my processing is possible.) The processing that has been done after the original material is NOT destructive OR constructive, so what you are hearing was on the original disk. The difference is that the recordings have been DECODED -- not ANY 'creative additions.'The results that I get (ever improving) come from 'following the rules' at every step. The only 'rule' that my processing busts is 'common wisdom.' I have cooperation from very smart/intelligent professional restoration experts (who give talks at AES/etc.) I don't use his name simply because I don't want to mix him up in my 'hobby' activities. Testing against ABBA material is more hobby than part of the professional effort. What you might listen to on my repository (you can listen directly), I prefer that you simply listen/enjoy. There is MUCH MUCH less compressioin in the recovered material -- so if you do an A/B between say, an original recording vs. the material on the repository -- the material on the repository will seem more alive, and not sound like it has been pushed through a 'tin can'. I apologize for the demo selections as being .mp3 encoded, because I simply have limited space on my repository (I also use it for professional file transfer), and .flac files are simply very big. The mp3's are encoded directly from 96k/24bit-floatingpoint copies, no steps in between, and no 'boosting' or 'extra love' modifications of the materal. It is purely dry coming from the minimum processing. The copies that I have produced are as 'natural' and 'pristine' as I think can be produced from material available through consumer sources. Some well known samples are included in my temporary repo -- every ABBA lover probably has over-processed copies of EXACTLY the same material -- but these are as 'virgin' as possible. I'd really like to see/hear the record producers provide the 'real stuff' rather than the hyper compressed/processed travesties as relatively recently made material. Remember: these are raw/natural sounding results -- if you are an ABBA lover, this material already exists on your CD or download... Please enjoy the potentially EYE OPENING experience.Again -- sorry that they have to be mp3!!! I'll probably also have to recover the space in a few days (I'll try to keep it for a week from today.) The non-ABBA flac examples were meant for my project collaborator... Also, on the non-ABBA repo, there is actually a fully processed (finalized) MamaMia, so it sounds brigher, prettier, happier than the 'forensic' copies on the ABBA repository. I finalized the MamaMia on the 'collaboration examples' repository without adding compression or anything outside of what was on the recording. The big difference on the 'to-collaborator' repo is that I unfolded the ambience and applied a little sweetening.REPOSITORY: spaces.hightail.com/space/QecbMZKnWhRAW non-ABBA decodes (happily flac) of some other examples: spaces.hightail.com/space/zqrnA5nqnV
|
|
|
Post by jsdyson on Nov 5, 2018 4:43:20 GMT
I spent some time to fully remaster a few ABBA songs. NOTHING in the material didn't come from the original source. It is possible to believe that some of the material was artificially produced --- it might have been artifically produced, but NOT BY ME. This is original material will a full remaster, and a beautiful sound. I haven't been 100% careful, but in general the results might be surprising. It almost doesn't sound like ABBA -- or, perhaps the best you have ever heard for most of the selections. The vocals are good -- clear 'S' sounds, the wall of sound is fully utilized, and acutally very listenable. If someone is interested, I can explain how it was done. I promise no more than 3dB of EQ in any place other than perhaps a rolloff or lowpass that might have a small peak. Again, even though it might sound like I added something -- absolutely nothing was done to add or remove something that wasn't already there (except for maybe a little EQ.) Frankly, this is amazing for ABBA (and I don't have an ego that lets me make such claims... It goes to show how good they really were): John REPO: spaces.hightail.com/space/QecbMZKnWh
|
|
|
Post by jsdyson on Sept 27, 2018 17:51:39 GMT
FYI -- I have added some more obscure examples -- What about Livingstone, What about the music-doris day version, Kisses of fire and a few others. The processing is all legit, and will probably be on the repository for a few more days. Enjoy -- this is truly very special because of the 100% proper DolbyA-style decoding and some mitigation of the severe wall of sound (which is a bit less applicable nowadays.) Also redid the other examples -- I diddn't 100% like the results, so did some thigns differently, but still the same DolbyA decode. spaces.hightail.com/space/xghqJodgrj
|
|
|
Post by jsdyson on Sept 27, 2018 6:05:31 GMT
At the same location as before -- also repeated below, I have reposted the examples -- they are much less 3d processed -- pretty mellow, and as clear/clean/full as I have ever heard ABBA from a CD. The full bandwidth and lowest distortion possible in this rendition. Not all of my collection is DolbyA encoded, so it is hard to get really good sound for everything. I also did minimal EQ, so you can do what you want with EQ. (Nothing hyped.) spaces.hightail.com/space/xghqJodgrj
|
|
|
Post by jsdyson on Sept 26, 2018 5:25:54 GMT
I have a 'gift' for ABBA lovers -- almost remix of some of the old songs. Much smoother that normal -- clearer in some cases, more defined. Perhaps too much ambience, but I'll probably update it again in the next few days. NOTHING is in the recording that wasn't already in it -- no synthesis of ANYTHING. Used the latest version of the DolbyA compatible decoder -- it is heading to AES and into the hands of some big recording engineers!!! Also, what I have posted does give an alternative sound to the old ABBA -- and again, NOTHING ARTIFICIAL IN THE EXAMPLES: spaces.hightail.com/space/xghqJodgrj
|
|
|
Post by jsdyson on Sept 12, 2018 22:34:19 GMT
Honestly -- the only thing that I care about is if people who like ABBA can enjoy some of results of things that I am playing with. ABBA is such a special group, with a very 'interesting' set of characteristics (not necessarily speaking of the women, but they are included.) Their music is so very unique, and enjoyable. Some of their music is more poignant than what might initially seem, and some seem to be mildly editorial (What about Livingstone?), but always kind and mostly happy sounding. Not only that, their 'wall of sound' style is incredibly difficult for audio processing -- and much of the material is available in DolbyA encoded form (probably by mistake.) All of the above are reasons why I like ABBA.
|
|