sandyk (RIP Alex, 1939 - 2021)
Global Moderator
Posts: 226
About Me: Retired ex Principal Telecommunications Technical Officer with 43 years at Telstra (Australia)
I am also a Moderator in Hi Fi Critic Forum
Electronics hobbyist for >65 years with DIY projects including Loudspeakers, Stereo FM tuner, S/W Regen Receiver, Superhet AM ,
Synchrodyne PLL AM tuner (Phase Lock Loop),Stereo Tape Deck, Amplifiers including I.C. types, Class A, Class AB 100W/Ch. (ETI5000) 240W/Ch. Mosfet (AEM6000) ,several DACs , numerous PSUs including VERY low noise (<4uV) types etc.for myself and friends
Audio Industry Affiliation: NIL
|
Post by sandyk (RIP Alex, 1939 - 2021) on Dec 14, 2019 8:24:02 GMT 10
A recent Update from John
8 hours ago
I am still 'disappeared', but came up with a few breakthroughs -- one was that the DA was being a little too aggressive in trying to mitigate MD, so created a bit of an indelicate sound... I am modifying the behavior to be more agile, yet still sound good. One problem was Linda R's (Youre no good), where the chorus was 'chunky', almost distorted sounding. I don't have the correct balance yet -- but it is definitely a tweaking game to mitigate MD, but also to make sure that the DA can track the signal very closely. I ran my test of a 3 times encoded copy of Dreamworld, and one version of test code where I tweaked up the anti-MD algorithms -- the result was ludicrous, but almost tolerable. That tells me that with the more weak filtering was even too strong (the 'Youre no good' problem.)
The attack/release coefficients for the nonlinear attack/release code for BOTH attack/release layers is tedious, and I think that I have it tuned to work on difficult cases... The attack/release portion of the program is one of those things that should not be modified for 'effect', and must be dead-on correct. (The attack/release is not a time constant, but rather an exponential relationship that mimicks the conductivity of a diode.)
There is a problem of tracking the needed attack/release to make the music CORRECT, but also making sure that during those fast gain skews, that distortion isn't created. The problem with the old DolbyA is that both the attack/release had a delay factor in it when decoding (fixed), and also the attack/release speed was poorly controlled (still a work in progress.) If you let the attack/release go nuts, then you get those foggy effects... Strange - but intuitively like the frequency/time resolution tradeoff.
I'll be back in a few dtys. This requires full concentration and deliberation, not just tweaking right now. This is way too complex to come up with quick answers. In the worst case, I have all approx 500 previous versions of the DA!!!!
John
|
|
|
Post by johndyson on Dec 17, 2019 6:44:05 GMT 10
Super Duper Happy!!! I think that there is a really good decode of 'Crime of the Century' (better than anything else.) There have been several problems -- the last ones were a matter of difference in opinion and stressing the DolbyA compatible DHNRDS DA decoder beyond its design. First, there is all kinds of special filtering that is intended to very carefully shape the fast gain control transitions so that distortion products are minimized. Also, the 'Crime' album, when decoded in the way that is expected, does provide a signal that isn't a standard DolbyA encoded signal. Because of tradition on that album, it appears that it was 'specially decoded' by the DolbyA those many years ago. It is incredibly difficult to decode at that level of intensity while avoiding the creation of distortion. I conceptually designed the DHNRDS DA to almost be modulation distortion proof, but as we know - it is a very complex and tedious design. The 'Crime' album pushes the decoder even beyond what it was designed to decode. So, I tend to avoid wanting to push the decoder as hard as what the audiophile community prefers. All I wanted was about 0.75 dB cut at about 9kHz (about 1.5dB cut at 20kHz), and the decoder was pretty happy. However, that amount of cut was deemed 'not good'. Anyway, I found a class of bugs in the DA decoder which, if corrected, DOES eliminate the distortion created by intense DolbyA material. SO, as of a few hours ago, the DA decoder can properly decode the extremely intense 'Crime' album without distortion. How do I determine whether or not the DA produces too much distortion? On the 'Crime' album, sometimes the vocals seem to 'crackle' a little bit, like with a slightly harsh edge. To me, that sounds like distortion, but isn't always true distortion -- my Beyerdynamic DT770 Pro/80 headphones and my hearing both don't like too much HF either. I perceive the vocals with too much HF as 'distorting' also. How do I determine if it isn't distortion? Easy -- I do a treble cut of perhaps 3dB at 9kHz with a mild Q of 0.5 or 0.707. If that remedies the harsh edge, then my perception of distortion was simple headphone/hearing overload. With the above criteria now, I haven't found a case where the harsh vocal that appears when decoding overly intense material as being true distortion!!! It now is just lots of HF!!! The ambience has also been a problem. There are two kinds of problems -- the layer one and layer two attack/release algorithsm, and also the attack/release shaping algorithms. Both must be excruciatingly correct -- like within 0.5% or less. The attack/release calculations are indeed very correct and very accurate -- more like 0.2% or less. The filtering had a bias bug, where a certain coefficient was tuned for the 66% point instead of 50% point. There is a historical reason for the old 66% number instead of 50%, but now it isn't needed, and is in fact destructive. Once that bias point was changed to the correct and optimum value, the ambience for pianos didn't have odd shifts or -- in this last case, almost like a compression type effect. It is now gone, the ambience is now natural. Alex and I had some disagreements on this -- but it is very tricky, and I have a rule of not tweaking the signal much after decoding. I feel that the DolbyA output is theoretically 'pristine', and any other modifications need to be done with great care. I almost never tweak on a per cut/song basis in any event -- that kind of mastering is for experts. I am not a mastering expert. So -- the results I have produced have ZERO filtering after decoding. The fact that there are zero splats above 20kHz in the signal doens't come from filtering after decoding. The files I provide are conservatively manual dithered, 24bit copies. Since the BW of the input is 20kHz, there are zero noticeable splats outside of 20kHz on the output. Any time that you do gain control, then modulation occurs, That modulation with fast attack/release stuff like the DolbyA behavior will create splats. However, the DHNRDS DA does not do splats. I don't even know how to coerce the code to create splats right now, it would require careful breakage!!! Below is probably the best possible decoded copy of 'Crime' (well, you know what I mean -- it can always be better.) I do not hear any ambience bounces, and I am incredibly sensitive to that kind of stuff. Tell me what you think -- this has required many iterations (this is the 20th main version that I have produced, with many more attempts), and much feedback -- mostly from Alex. The files now have ID information in them, and they are the 'somewhat-official' version 20. Previous versions are NOT comparable. I'll keep these available for a few days.. Alex will probably have them after that, or you can tell me that you'ed like to try them. www.dropbox.com/sh/ewsi70ful3gt577/AADLvpgJxW5fXVHgFcbiDneya?dl=0
|
|
sandyk (RIP Alex, 1939 - 2021)
Global Moderator
Posts: 226
About Me: Retired ex Principal Telecommunications Technical Officer with 43 years at Telstra (Australia)
I am also a Moderator in Hi Fi Critic Forum
Electronics hobbyist for >65 years with DIY projects including Loudspeakers, Stereo FM tuner, S/W Regen Receiver, Superhet AM ,
Synchrodyne PLL AM tuner (Phase Lock Loop),Stereo Tape Deck, Amplifiers including I.C. types, Class A, Class AB 100W/Ch. (ETI5000) 240W/Ch. Mosfet (AEM6000) ,several DACs , numerous PSUs including VERY low noise (<4uV) types etc.for myself and friends
Audio Industry Affiliation: NIL
|
Post by sandyk (RIP Alex, 1939 - 2021) on Dec 19, 2019 8:36:35 GMT 10
If anybody would like to hear some samples of the most recent corrected versions of Abba-Ring Ring from John, which further increase the exuberance of these recordings without added distortion, please send me a PM
Alex
|
|
|
Post by cj66 on Dec 19, 2019 13:54:16 GMT 10
I must admit I struggle with Abba as more than background or party music.
Have there been advances on the Carly Simon or Carpenters material? More than happy listening closely to them. 😉
|
|
sandyk (RIP Alex, 1939 - 2021)
Global Moderator
Posts: 226
About Me: Retired ex Principal Telecommunications Technical Officer with 43 years at Telstra (Australia)
I am also a Moderator in Hi Fi Critic Forum
Electronics hobbyist for >65 years with DIY projects including Loudspeakers, Stereo FM tuner, S/W Regen Receiver, Superhet AM ,
Synchrodyne PLL AM tuner (Phase Lock Loop),Stereo Tape Deck, Amplifiers including I.C. types, Class A, Class AB 100W/Ch. (ETI5000) 240W/Ch. Mosfet (AEM6000) ,several DACs , numerous PSUs including VERY low noise (<4uV) types etc.for myself and friends
Audio Industry Affiliation: NIL
|
Post by sandyk (RIP Alex, 1939 - 2021) on Dec 19, 2019 14:40:36 GMT 10
I must admit I struggle with Abba as more than background or party music. Have there been advances on the Carly Simon or Carpenters material? More than happy listening closely to them. 😉 Hi Chris I believe so. I understand that John was also doing Phil Collins, so perhaps if we ask nicely we may get to hear that too ? A recent PM from John
|
|
|
Post by johndyson on Dec 20, 2019 3:29:20 GMT 10
Okay, I'll shelve ABBA right now -- got a technical problem with matching the sibilance vs. getting the correct EQ on the output.
Let me check Phil Collins and see if it is good enough yet. There is a cut that really tested the attack/release, and still might be a little odd (you might hear it -- might not be easy to fix -- subtle to most people.) The Phil Collins cut actually spurred on the latest improvmemnt. The cut was '04 In the air tonight' from the 'very best' album.
I have to get 'Phil' done very soon for my doctor, so will seee what I can do today. I usually get about 1/2 way through and then find something that I don't like. As I remember, the last time that I stopped it -- just ran out of time, and wanted to do something else :-).
About the 'Carpenters' -- wow, I don't know if I have made comments about decoding their material, but they did WEIRD stuff that took me 2yrs to figure out. THere was a method to their madness, but it was madness indeed!!! Actually, it appears that their 'madness' was about getting around imbalances in the DolbyA units and avoiding excess distortion... If it wasn't madness, it was certainly genius -- someone who understood what DolbyA did make their choices.
John
|
|
|
Post by johndyson on Dec 20, 2019 7:53:39 GMT 10
Okay, I'll shelve ABBA right now -- got a technical problem with matching the sibilance vs. getting the correct EQ on the output. Let me check Phil Collins and see if it is good enough yet. There is a cut that really tested the attack/release, and still might be a little odd (you might hear it -- might not be easy to fix -- subtle to most people.) The Phil Collins cut actually spurred on the latest improvmemnt. The cut was '04 In the air tonight' from the 'very best' album. I have to get 'Phil' done very soon for my doctor, so will seee what I can do today. I usually get about 1/2 way through and then find something that I don't like. As I remember, the last time that I stopped it -- just ran out of time, and wanted to do something else :-). About the 'Carpenters' -- wow, I don't know if I have made comments about decoding their material, but they did WEIRD stuff that took me 2yrs to figure out. THere was a method to their madness, but it was madness indeed!!! Actually, it appears that their 'madness' was about getting around imbalances in the DolbyA units and avoiding excess distortion... If it wasn't madness, it was certainly genius -- someone who understood what DolbyA did make their choices. John ----
Oh well, I give up on the 'very best'. After several decoding attempts, the 4th cut cannot be zeroed-in. With subsequent careful review of the material, and noting the date. I am wondering if the material in my posession is SR instead of A. When decoding the material, it comes back as 'still compressed' in an especially strange way. Even though I have all of the schematics of SR (shush.... no one knows that), I am still not 100% sure, but given what I do know about SR -- the material JUST MIGHT be SR. Also, DolbyA decoding supposed to 'sort of' work with SR. Those are the results I am getting -- 'sort of'. Not good enough.
So, I pulled out an album image of 'No Jacket Required'. You guys have a choice: would you like an album image or a bunch of music files? I can do either... I might truncate the files a little if you dont' get the full album image... However, I don't usually miss much if any..
Let me know -- I am lining up the album right now. Just might decode it as the album, then split it afterwards.
John
|
|
|
Post by ROWUK on Dec 20, 2019 8:25:41 GMT 10
In the Air tonight is a cut from 1981. Dolby SR came in 1986. Maybe the remaster used SR to decode dolby A? I do not have the album or CD to compare. I have fast internet so file size is no issue.
|
|
|
Post by johndyson on Dec 20, 2019 8:55:55 GMT 10
In the Air tonight is a cut from 1981. Dolby SR came in 1986. Maybe the remaster used SR to decode dolby A? I do not have the album or CD to compare. I have fast internet so file size is no issue. My guess is that SR might have been used for 'temporary' copies during mastering. The material simply doesn't line-up. The DA can do 'impossible' things. It COULD just be a bug also.
If the result had only one fault then I could guess it to be a bug. But I noticed two things -- couldn't line up the MF band (80-3k) and there was left-over compression that almost sounded like botched DolbyA decode.
If I disabled decoding the MF band, then the results were okay, but still some upper-midle frequency compression/poor-decoding above the 3k band. It sounded like really badly tuned DolbyA decoding. If I could have gotten by with disabling the MF band, I would have slipped it in, but there were still the other misdecoding problems -- sounded like a mild case of the messed up details from ABBA when mistuned. (It is easy to disable and offset the calibration/gain bands on the DHNRDS decoder, used to emulate pulling a card on the old A301 units.) Richard requested that every possible professional feature be included in the DHNRDS DA (but I stopped when he also asked for SR :-)).
The troublesome source is below: (as is, without damaging by attempt to decode, it seems to sound like the usual DolbyA compression with the feral-EQ.)
John
|
|
|
Post by johndyson on Dec 20, 2019 23:22:28 GMT 10
Just let Alex have a quick listen to both Phil Collins Very Best and No Jackets -- just to make sure no botches. My explanation about 'Very Best' was wrong. It wasn't 'SR', but instead they used the super dynamic range calibration of -20.47dB instead of the more common -12.67dB. Some audiophile quality stuff has the very low calibration level so that there is wider high level dynamic range. The 'I've got the music in me' CD from Sheffield Labs also needed that level of calibration for proper decoding. Very uncommon to use that level, so I didn't even consider it whlie trying to decode the 'Very Best' CD. That extreme calibration error certainly allowed the small compression errors appear as the latent compression, and also made cut 4's very difficult to track dynamics impossible to get reasonable results. Cut 4 is still not quite perfect (sub 0.1dB transient error), but better tracking is almost impossible to get, that is just a very problematical recording. +
Sadly, the more impressive dynamics that were attained by the mis-decoding were somewhat less impressive when decoding correctly, but the 'Very Best' album should be interesting to listen to nonetheless. I am not hiding the results, but simply worry about wasting more peoples time if there is a botch. Instead of hiding them away, I'll tell you where they are, but unless very curious, don't download until we get a 2nd opinion or at least some time has passed (maybe at least a few hours.)
If there aren't any red-flags, then the material is probably okay. Don't expect the quality of 'Crime'!!!! :-). No Jacket is available at: www.dropbox.com/sh/rf2a8jy5xgts179/AACLnFS7AiJiJKPK_1JNP1hJa?dl=0Very Best is available at: www.dropbox.com/sh/twy83ghvrzd2vn0/AAAd2Dja_xTcMwqURNPT3z5ea?dl=0 John
|
|
|
Post by johndyson on Dec 21, 2019 2:39:05 GMT 10
Just let Alex have a quick listen to both Phil Collins Very Best and No Jackets -- just to make sure no botches. My explanation about 'Very Best' was wrong. It wasn't 'SR', but instead they used the super dynamic range calibration of -20.47dB instead of the more common -12.67dB. Some audiophile quality stuff has the very low calibration level so that there is wider high level dynamic range. The 'I've got the music in me' CD from Sheffield Labs also needed that level of calibration for proper decoding. Very uncommon to use that level, so I didn't even consider it whlie trying to decode the 'Very Best' CD. That extreme calibration error certainly allowed the small compression errors appear as the latent compression, and also made cut 4's very difficult to track dynamics impossible to get reasonable results. Cut 4 is still not quite perfect (sub 0.1dB transient error), but better tracking is almost impossible to get, that is just a very problematical recording. +
Sadly, the more impressive dynamics that were attained by the mis-decoding were somewhat less impressive when decoding correctly, but the 'Very Best' album should be interesting to listen to nonetheless. I am not hiding the results, but simply worry about wasting more peoples time if there is a botch. Instead of hiding them away, I'll tell you where they are, but unless very curious, don't download until we get a 2nd opinion or at least some time has passed (maybe at least a few hours.)
If there aren't any red-flags, then the material is probably okay. Don't expect the quality of 'Crime'!!!! :-). No Jacket is available at: www.dropbox.com/sh/rf2a8jy5xgts179/AACLnFS7AiJiJKPK_1JNP1hJa?dl=0Very Best is available at: www.dropbox.com/sh/twy83ghvrzd2vn0/AAAd2Dja_xTcMwqURNPT3z5ea?dl=0 John Quick heads - up. Even though the decodes are okay -- I am working on something better -- I THINK...
John
|
|
|
Post by johndyson on Dec 21, 2019 19:59:05 GMT 10
I found another one of the ancient recordings requiring M+S decoding -- Sergio Mendes & Brasil'%%... It is MUCH nicer sounding when decoded correctly. Doing lots of Christmas presents (My sister is getting a nice memory stick full.)
|
|
|
Post by johndyson on Dec 22, 2019 3:01:49 GMT 10
I am delayed a bit on Phil Collins, Carly Simon and Linda R -- they are in the script that is currently running. Doing Christmas presents for family and a doctor-friend of mine. However, got some 'cute' material. I found that all of my Brasil'66 stuff and Herb Alpert were M+S encoded and not L+R, which means that any previous decoding attempt was subpar. The recordings are not really audiophile, but these fun examples should be higher quality than anything other than vinyl. I actually spent time on the Alpert and Sergio Mendes stuff, because the music puts me into a bit nicer mood. Here, I offer some decodes of 'Spanish Flea', 'Taste of Honey', and 'Casino Royale'. Don't be dissuaded by the 44.1k/16bit, my space usage is bumping up against the limit, but the quality loss relative to the actual recording quality is truly NIL. Alex will attest that the DHNRDS DA decoder produces digital files that don't degrade much when downsampled, and it is because the files are so very clean. There isn't excess junk in the files that heeds to be downsampled. I don't claim perfection, but these are truly FULL AUDIO BANDWIDTH recordings, even though they are packed into a 44.1k/16file. Don't feel like I am foisting junk on you :-). These are FUN, not serious audiophile, but are very good quality (take a gander at the spectrum. BTW. my normal decoding process I can easily remove 6dB at 20kHz if needed -- it won't be wrong to do so, but by my limited hearing, I didn't have to chop these files up like usual. These actually MIGHT have TOO MUCH HF as I cannot hear about 14kHz. www.dropbox.com/s/kcimapc1x6tl2i0/02.A%20Taste%20Of%20Honey.flac?dl=0www.dropbox.com/s/ke1j3li4quek1kc/04.Spanish%20Flea.flac?dl=0www.dropbox.com/s/64reay0n7t9bzgg/07.Casino%20Royale.flac?dl=0John
|
|
sandyk (RIP Alex, 1939 - 2021)
Global Moderator
Posts: 226
About Me: Retired ex Principal Telecommunications Technical Officer with 43 years at Telstra (Australia)
I am also a Moderator in Hi Fi Critic Forum
Electronics hobbyist for >65 years with DIY projects including Loudspeakers, Stereo FM tuner, S/W Regen Receiver, Superhet AM ,
Synchrodyne PLL AM tuner (Phase Lock Loop),Stereo Tape Deck, Amplifiers including I.C. types, Class A, Class AB 100W/Ch. (ETI5000) 240W/Ch. Mosfet (AEM6000) ,several DACs , numerous PSUs including VERY low noise (<4uV) types etc.for myself and friends
Audio Industry Affiliation: NIL
|
Post by sandyk (RIP Alex, 1939 - 2021) on Dec 22, 2019 9:52:22 GMT 10
I found another one of the ancient recordings requiring M+S decoding -- Sergio Mendes & Brasil'%%... It is MUCH nicer sounding when decoded correctly. Doing lots of Christmas presents (My sister is getting a nice memory stick full.) I hope the Pooches enjoy them too. Years ago, I had a cat that loved The Blue Danube waltz
|
|
|
Post by ROWUK on Dec 23, 2019 8:14:31 GMT 10
I don't claim perfection, but these are truly FULL AUDIO BANDWIDTH recordings, even though they are packed into a 44.1k/16file. Don't feel like I am foisting junk on you :-). These are FUN, not serious audiophile, but are very good quality (take a gander at the spectrum. BTW. my normal decoding process I can easily remove 6dB at 20kHz if needed -- it won't be wrong to do so, but by my limited hearing, I didn't have to chop these files up like usual. These actually MIGHT have TOO MUCH HF as I cannot hear about 14kHz. John In spite of the bad hype, 44.1/16 can be VERY convincing. In many cases, higher resolution does not offer more music, but sometimes only a bit softer top end (which has NOTHING to do with bandwidth). John, are we sure that the vinyl masters were really that much different - after all, back then music was sold over the radio, on 8-Track and cassettes - all subject to non audiophile pimping to make them "spectacular" and therefore saleable! I would just like to take this opportunity to say thank you one more time. This is way over the top as a service to this community.
|
|