|
Post by johndyson on Nov 1, 2019 8:42:18 GMT 10
Gang -- the new version of the DHNRDS (actually, I haven't officially produced a release yet, will do so tomorrow) is still working very well :-). PLEASE ENJOY, and feel free to criticize (kindly :-)).
I'd suspect that most people reading this already know the context of the DHNRDS DA decoder, basically intended to resurrect older recordings, eliminating the DolbyA HW as a quality limitation. Note that the DolbyA HW isn't all that bad for mixing down, because the material being mixed tends to have relatively simpler dynamics and signal statistics. The DolbyA doesn't do too bad for mixing down, but could be improved on. Where the DolbyA really falls down is handling the final mix. All of the interactions in the individual signals added together are too much of a challenge for DolbyA HW. (Lets keep in mind, the DolbyA was designed in the middle 1960s, and Ray Dolby was pretty much a magician designing the DolbyA itself.) The DolbyA design has definite indications of genius in the circuitry -- it is MUCH MUCH more intricate than an initial read of the schematic might suggest!!!
* Most of my decoding is done on the brighter side... It can be toned down, but choosing the correct EQ is tricky as hell!!! Also, the recordings could stand a little better balance -- again I chose a brighter decoding result, but can be tamed.
I unleashed the DHNRDS DA from being 'Bug for Bug' compatible with the DolbyA, and basically changed it to be 'as good as possible'. The sound might be VERY different than you expect, esp for ABBA. LindaR should seem REALLY good also. Of course, my taste might not be quite right. My own initial opinion is 'astonishing' sound quality. Mega Caveat: IS very tricky to undo the EQ that was done to skip the DolbyA decoding step -- I can provide the parameters that I used, and you can try passing the material through a true DolbyA if you wish -- given the parameters that I provide. I doubt that you would like the sound of the DolbyA in comparsion, but each of us has our own opinions. :-). Normal professional use of the DHNRDS does NOT need for the EQ correction, but decoding feral DolbyA is a very tricky thing to do!!!Please be tolerant of me about the Carpenters recordings. You might notice more sibilance than normally desired. I think that part of the issue is that the true DolbyA cannot track the signal, so it was enhanced to compensate. Since the DHNRDS DA can track ANY audio signal, the boosted sibilance is passed through cleanly. If mastering the Carpenters material, it might be a good thing to run it through a sibilance processor or maybe change the pre-decoding EQ a little on the most egregious songs. Usually a little bit of manipulation in the 6kHz range can help, but the pre-decoding EQ is totally vanilla and is intended to be correct, not just to sound good!!! I have produced some snippets (full recordings on request) of some stuff that has probably not been heard as cleanly for at least 30yrs. So far, there are some 'Linda Ronstadt' recordings, and some 'ABBA'. Carpenters is coming in another hour or so, and also I plan to produce some Olivia Newton John snippets. The Linda Ronstadt examples are NOT 'manipulated' after decoding, and I don't plan any manipulation of ONJ or Carpenters either. (More will likely be coming, e.g. Bread -- which REALLY sounds good now.) The ABBA stuff has been slightly manipulated after decoding because their tonal balance is too biased away from the bass range, so I have added about 1.5dB below 360Hz on ABBA for easier casual listening. Of course, LindaR and ABBA are done, and I have the correct decoding parameters for Carpenters and running the decode just this second. I'd suggest checking the repository now (or whenever) and looking again every day or so for new stuff.
|
|
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 Nov 1, 2019 9:11:55 GMT 10
Hi John This is great. and with your permission I will repost it in Martin Colloms Hi Fi Critic forum where I am a moderator. This work from you deserves to be recognised by the Audiophile press too, as it may lead to requests to the Recording Companies
Kind Regards Alex
|
|
|
Post by johndyson on Nov 1, 2019 9:33:14 GMT 10
Hi John This is great. and with your permission I will repost it in Martin Colloms Hi Fi Critic forum where I am a moderator. This work from you deserves to be recognised by the Audiophile press too, as it may lead to requests to the Recording Companies Kind Regards Alex Okay -- good idea... Also, for examples, I have full copies of everything that is on the Dropbox site as snippets... (most are 88.2k/24bit.) Note also I have supplied most in mp3 form also, making it easy to do a quick listen before a download.
I am still doing decodes -- right now, ONJ... Next will be 'Bread'. I am wondering of 'The Cars' will be worth it, because they are so very processed anyway, and I found the material that I had previously used was already whacked pretty well with some processing. Material that is already very mangled just doesn't show all that much improvement.
Even though already fairly processed -- I am thinking about Petula Clark and Anne Murray. If I can get a really clean copy of Carly Simon, will try that also.
John
|
|
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 Nov 1, 2019 10:09:51 GMT 10
Hi John This is great. and with your permission I will repost it in Martin Colloms Hi Fi Critic forum where I am a moderator. This work from you deserves to be recognised by the Audiophile press too, as it may lead to requests to the Recording Companies Kind Regards Alex Okay -- good idea... Also, for examples, I have full copies of everything that is on the Dropbox site as snippets... (most are 88.2k/24bit.) Note also I have supplied most in mp3 form also, making it easy to do a quick listen before a download.
I am still doing decodes -- right now, ONJ... Next will be 'Bread'. I am wondering of 'The Cars' will be worth it, because they are so very processed anyway, and I found the material that I had previously used was already whacked pretty well with some processing. Material that is already very mangled just doesn't show all that much improvement.
Even though already fairly processed -- I am thinking about Petula Clark and Anne Murray. If I can get a really clean copy of Carly Simon, will try that also.
John
Hi John I decided not to post it directly, but forwarded it to Martin Colloms via email, as what I forwarded will need editing, and perhaps Martin himself may wish to listen to your vastly improved versions too. I will also give Martin your contacts details if you are in agreement, and he asks for them. In the event that I don't hear back from Martin due to perhaps having an old email address, I will bring the subject up in HiFi Critic Forum
I can also help you searching for suitable material if need be.
Kind Regards Alex
Kind Regards Alex
|
|
|
Post by johndyson on Nov 1, 2019 11:56:04 GMT 10
I can understand your strategy. My note was a bit disorganized -- I should have been a little more structured in the creation. Also, it really wasn't written as a press release. The actual sales/marketing person on the DHNRDS team is 'Richard Hess', who is available at he DHNRDS.com website. (I'd suspect that a lot of people in the recording arena might know of him.) Richard is not offended by my interfaces into the audiophile or even pro forums, and would likely be appreciative of any additional mention made about the DHNRDS. (It had been mentioned at AES in recent years.)
ONE IMPORTANT COMMENT ABOUT THE DEMOS -- some of these demos are so clean that MP3 really falls down, and to get the full/correct impact, you REALLY DO need to use the .flac versions to review!!!
John
|
|
|
Post by johndyson on Nov 1, 2019 12:20:10 GMT 10
MEGA important -- the Carpenters examples were re-EQed before decoding. Still not perfect, but you can hear the subtle details. The sibilance is mostly gone!!!
John
|
|
|
Post by cj66 on Nov 1, 2019 16:39:49 GMT 10
Exciting stuff!
I'll have a chisel into the Carpenters material and give Bread a raise through later when I'm on the lappy.
|
|
|
Post by johndyson on Nov 1, 2019 22:33:12 GMT 10
Exciting stuff! I'll have a chisel into the Carpenters material and give Bread a raise through later when I'm on the lappy. I am doing a massive complete and full quality controlled decoding operation on full albums on Carpenters, ABBA, Bread and selected Carly Simon & a few others today. It is a big Unix/Linux shell script, and takes many hours on the high quality modes. The simpler modes (still sound good/covers up flaws in the recordings) run several times faster than realtime, but are less interesting except for certain pro usage needs.
One thing for sure, I did find a minor eccentricity in the decoder on the current examples. The highest decoding mode seems to apply a slight amount of either 'enhancement' or remove too much 'modulation distortion'. The results are NOT POOR at all, just something I need to better understand... The behavior appears to be 'desirable', but I do need to understand it. 'Question: why is the highest quality mode more clear than I'd expect over and above the second highest quality mode?' All decoding except ABBA was done in the highest quality mode, but the 2nd highest quality is really pretty darned good. It is very possible that the modulation distortion removal code can remove multiple generations of DolbyA HW distortion, and trying to figure out if I didn't overshoot the desire to remove only one layer?!?!?
You can actually hear the significant difference when using the non-anti-MD version of decoding vs the modes being demoed here. The more primitive DA modes (not used in these demos) have a 'veil' over them, very similar to the true DolbyA HW. It is a kind of 'smoothing' behavior that makes the vocals in a chorus meld together, while the high quality modes demoed here tend to keep the vocals separate in the chorus.
If there is something that you really enjoy, and you already have a copy of the recording, I'll happily make those recordings privately available. I do this simply because I want the music to be enjoyed in its full splendor as much as possible.
I have a hopeful desire that the distributors would better 'prepare' digital material for distribution. I give away free use of the decoder for moderate term evaluation purposes (several months), especially to very, very technically avanced hobbiests, and Richard Hess (check with him at DHNRDS.com) sells copies to professionals. I'd rather just give evaluation versions to consumers, unless they have established ability to use the decoder. It is very tricky for me to process 'feral' DolbyA myself, and I am relative expert at doing it. However, for pro purposes, the DHNRDS is childs-play to use, except it isn't directly integrated into the tape deck. The DHNRDS is INFINITELY more useful in processing material that has been archived onto digital form already, but there are groups/individuals who are successfully using it on multi-track recordings. The DHNRDS does have up-to 10 sample error in input/output file timing, but is totally accurate between tracks given the same settings. The DHNRDS also handles RF64/BEXT for true pro applications, can run full albums/etc.
John
|
|
|
Post by johndyson on Nov 1, 2019 23:26:37 GMT 10
Super-duper, mega-challenging material to decode: ABBA: DreamWorld, and the copy that I have has been encoded by DolbyA 3 times sequentially. Yes, not 3 encode/decode cycles, but 3 encode cycles sequentially. At that level, where decoding the material requires 3 sequential decoding operations with the same settings -- it is a very tedious operation. Also, unless each step of the decoding effort is precisely tweaked to have exactly the same input/output levels, there will be a drift in calibration, therefore causing more distortion/odd effects. Only one layer of decoding error doesn't cause a lot of trouble, but multiples in sequence make the first calibration in sequence VERY critical. My test was to run three DHNRDS DA decodes of the 3-level Dreamworld material without ANY variation in calibration between each step, so any drift in gain will cause errors to accumulate. Also, the DHNRDS DA was not carefully tuned for accurate input/output gain, so there is (AFIAR) about 0.1-0.2dB of output gain error on the DHNRDS (not usually very consequential, a true DolbyA is MUCH MUCH worse.) Here, I am pointing to a copy of Dreamworld, fully decoded with the same parameters on each step (calibration is -12.60dB, and the input EQ is applied to the first layer only, because it is the corrective EQ for 'listenability'.) Also, the highest quality decoding mode was used, mostly to make sure that it does not cause MORE errors than expected. This copy is NOT 'beautiful' sounding, and in fact a bit ugly, however there are few severe expansion artifacts -- again, three sequential decoding steps on material that is probably impossible to otherwise decode cleanly at all ANYWHERE ELSE... www.dropbox.com/s/1c43vo42hlg6am3/Dreamworld-V0.9.7Z0.flac?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 Nov 2, 2019 8:23:53 GMT 10
Super-duper, mega-challenging material to decode: ABBA: DreamWorld, and the copy that I have has been encoded by DolbyA 3 times sequentially. Yes, not 3 encode/decode cycles, but 3 encode cycles sequentially. At that level, where decoding the material requires 3 sequential decoding operations with the same settings -- it is a very tedious operation. Also, unless each step of the decoding effort is precisely tweaked to have exactly the same input/output levels, there will be a drift in calibration, therefore causing more distortion/odd effects. Only one layer of decoding error doesn't cause a lot of trouble, but multiples in sequence make the first calibration in sequence VERY critical. My test was to run three DHNRDS DA decodes of the 3-level Dreamworld material without ANY variation in calibration between each step, so any drift in gain will cause errors to accumulate. Also, the DHNRDS DA was not carefully tuned for accurate input/output gain, so there is (AFIAR) about 0.1-0.2dB of output gain error on the DHNRDS (not usually very consequential, a true DolbyA is MUCH MUCH worse.) Here, I am pointing to a copy of Dreamworld, fully decoded with the same parameters on each step (calibration is -12.60dB, and the input EQ is applied to the first layer only, because it is the corrective EQ for 'listenability'.) Also, the highest quality decoding mode was used, mostly to make sure that it does not cause MORE errors than expected. This copy is NOT 'beautiful' sounding, and in fact a bit ugly, however there are few severe expansion artifacts -- again, three sequential decoding steps on material that is probably impossible to otherwise decode cleanly at all ANYWHERE ELSE... www.dropbox.com/s/1c43vo42hlg6am3/Dreamworld-V0.9.7Z0.flac?dl=0Hi John That is WAY better ! Unfortunately I was unable to locate the original CD that you have for a direct comparison.
|
|
|
Post by johndyson on Nov 2, 2019 10:58:05 GMT 10
Super-duper, mega-challenging material to decode: ABBA: DreamWorld, and the copy that I have has been encoded by DolbyA 3 times sequentially. Yes, not 3 encode/decode cycles, but 3 encode cycles sequentially. At that level, where decoding the material requires 3 sequential decoding operations with the same settings -- it is a very tedious operation. Also, unless each step of the decoding effort is precisely tweaked to have exactly the same input/output levels, there will be a drift in calibration, therefore causing more distortion/odd effects. Only one layer of decoding error doesn't cause a lot of trouble, but multiples in sequence make the first calibration in sequence VERY critical. My test was to run three DHNRDS DA decodes of the 3-level Dreamworld material without ANY variation in calibration between each step, so any drift in gain will cause errors to accumulate. Also, the DHNRDS DA was not carefully tuned for accurate input/output gain, so there is (AFIAR) about 0.1-0.2dB of output gain error on the DHNRDS (not usually very consequential, a true DolbyA is MUCH MUCH worse.) Here, I am pointing to a copy of Dreamworld, fully decoded with the same parameters on each step (calibration is -12.60dB, and the input EQ is applied to the first layer only, because it is the corrective EQ for 'listenability'.) Also, the highest quality decoding mode was used, mostly to make sure that it does not cause MORE errors than expected. This copy is NOT 'beautiful' sounding, and in fact a bit ugly, however there are few severe expansion artifacts -- again, three sequential decoding steps on material that is probably impossible to otherwise decode cleanly at all ANYWHERE ELSE... www.dropbox.com/s/1c43vo42hlg6am3/Dreamworld-V0.9.7Z0.flac?dl=0Hi John That is WAY better ! Unfortunately I was unable to locate the original CD that you have for a direct comparison.
Thanks for the heads-up about Dreamworld being on the 2005 boxed set -- it is part of the HORRID 'The Complete Studio Recordings' with their 10-12dB PEAK-RMS and Crest factor of 3.5-4 (really really really compressed.) I listened to it, and it is blaringly loud -- WHY DO THEY DAMAGE RECORDINGS SO TERRIBLY?
Actually, I didn't try to tweak-in Dreamworld because it is so tedious to do so, and I was just demoing doing 3 decode operations sequetially -- which is a worst case test. It is very likely with a little patience, I can tweak it in to sound almost normal. When the DHNRDS DA runs in full quality mode, it cannot run three decoders simultaneously, but if I run in the lower quality modes, I can run three at a time at 48k sample rate. If I tweak the levels in/out of the each of the three decoders, I can probably make it sound ALMOST the same as a normal ABBA recording.
My guess is that they destroyed the recording by mistake somehow by doing 3 encoding operations without decoding. Maybe they destroyed it on purpose -- I am fairly sure that a DolbyA cannot decode it at all. The 'The Complete Studio Recordings' dont' appear to be decoded at all, and are simply further compressed with a radio station type processor.
Maybe some day, I'll spend an hour or so -- tweaking it in and getting rid of some of the roughness!!!
John
|
|
|
Post by cj66 on Nov 3, 2019 17:03:17 GMT 10
I've had a rummage through some of these latest decoded tracks and have to say it's a massive change. The Abba songs are sounding quite full and far more natural. The Carpenters are also much better, vocal tone is incredible now Carly sounds far more intimate than before. My previous comments of a closed-in feel don't count here, no detection of that at all. I think the change of tak on decoding/eq has been very worthwhile, these tracks are showing new personalities with a great sense of natural tone.
The main tracks I used were; Abba, Super Trouper Carly, You're So Vain Carpenters, Mr. Postman & Top Of The World
|
|
|
Post by johndyson on Nov 4, 2019 6:19:19 GMT 10
I've had a rummage through some of these latest decoded tracks and have to say it's a massive change. The Abba songs are sounding quite full and far more natural. The Carpenters are also much better, vocal tone is incredible now Carly sounds far more intimate than before. My previous comments of a closed-in feel don't count here, no detection of that at all. I think the change of tak on decoding/eq has been very worthwhile, these tracks are showing new personalities with a great sense of natural tone.
The main tracks I used were; Abba, Super Trouper Carly, You're So Vain Carpenters, Mr. Postman & Top Of The World
Before worrying about the quality with the ABBA results -- they are the worst, most difficult material that I have access to. The upcoming results are more beautiful than I even expected. I will be gone for about 1wk after a day or so, trying to get this stuff done!!!
Got some more REALLY MEGA news about the Decoding... The next decoding attempts WRT the carpenters should be super itneresting (I found that mu compensating EQ was all screwed up -- my Carpenters copies are almost pure DolbyA with some bass boost!!!) Too bad that the decoder takes many hours when trying to produce the best possible copies.
For ABBA -- the latest FULL demo decodes (Normally, wont' be able to show the entire files in public, but I'll do this for a day or so): (The full albums are coming tonight -- but these should be 'VERY INTERESTING' to a person who likes ABBA -- these are NOT snippets, but will disappear in about 1-2days.)
|
|
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 Nov 4, 2019 11:11:02 GMT 10
They sound SUPERB, but something went wrong with the last track Waterloo(English Version)
|
|
|
Post by johndyson on Nov 4, 2019 15:16:55 GMT 10
Here is some Carpeters that sound great, even though they are MP3. (I'll make flac available very soon.) Carpenters doesn't push against mp3 limits, so still sounds okay, but of course, .flac will probably be better.
The sound has the traditional 'lush' Carpenters sound. Finally I found the key!!!
John
|
|