Lies about vinyl vs digital

[John Dyson]

1 hour ago, fas42 said:

Hmmm, better ... but the "feral" still gives the clearest rendition - in particular, the acoustics of the percussion, voice, and piano very nicely separate, are 'cleaner'.

Remember, you will NEVER get the compressed, enhanced highs of the feralA with properly decoded material.  That is the whole problem with getting used to the defective sound -- it creates a kind of expectation of that defect.  It isnt' that the defective feralA always sounds bad, but it is NO WHERE near the intent of the artist or recording engineer, except in very limited cases.  Normally, it wouldn't be feralA at that point, but used as part of a mix for vocal enhancement - which is one of the side-effects of feralA anyway.

 

In a way, it can be hard to get used to good sound.  I have my own problem, because my hearing learns VERY quickly, and if I listen to defective results for too long, then the defective sound starts sounding good.  That is the major reason for sometimes choosing the wrong corrective EQ.

 

This is also a primary reason for resistance to properly decoded material -- people get used to defective sound.

 

I know that SOME audiophiles think that they have absolutely correct hearing -- and maybe some do, but most people who work with audio must be very careful to avoid polluting their hearing with defective sound - e.g. feralA.  Sometimes, getting away from the processing for a day or at least several hours brings hearing back to baseline.  Perhaps if being used to feralA for decades, it might take longer to unlearn the bad habit...

 

Lucky for me, I seldom listen to feralA, compressed FM radio or similar material like that.  This helps to keep the hearing from being polluted.

 

John

 

[fas42]

Just to check how polluted my hearing might be, I downloaded a 2L demo, http://www.lindberg.no/hires/test/2L-145_01_stereo_01.cd.flac - hopefully, that deeply feared feralA hasn't seeped into the building there ...

 

Hmmm, piano sounds like a piano, especially in the right hand; nice sparkle in the percussion, good transients - oh dear, I fear it has been contaminated, 😁.

[John Dyson]

49 minutes ago, fas42 said:

Just to check how polluted my hearing might be, I downloaded a 2L demo, http://www.lindberg.no/hires/test/2L-145_01_stereo_01.cd.flac - hopefully, that deeply feared feralA hasn't seeped into the building there ...

 

Hmmm, piano sounds like a piano, especially in the right hand; nice sparkle in the percussion, good transients - oh dear, I fear it has been contaminated, 😁.

Most people's hearing has been contaminated -- good to understand it, because generally chosing compressed feralA over decoded material shows that fact.  Sometimes the DolbyA compression helps the material if you like harder, less natural vocals -- sometimes that is good, but seldom so.

 

Frankly, listen to clean material (e.g. the master tapes in my posession), the decoded material sounds most like it.  Some of the stuff includes currently popular Jazz legends (the material has tones and all.)  I have both true DolbyA decoded copies direct from the tape and also DHNRDS decoded versions -- the DHNRDS is much more natural and clean/clear.  The decoded feral matches the true, accurate master tape material much moreso than the feral does. (I seldom mention access to actual master material -- but it sits in a different place, and I don't normally allow convenient access even to myself for proprietary reasons.)

 

The fact that the DHNRDS does the 'right thing' has been tested completely, by somewhat well known people other than me.

 

PS:  I have some master tape quality ABBA results nowadays.  It has been tricky -- they almost 'encrypt' the actual, somewhat more pretty sound than what is normally heard on CD.  That is the next, upcoming POP project atter the Tijuana Brass stuff.  I have been a little confused by the Brasil'66 right now, getting some help on that in private discussions.

 

John

 

[marce]

10 hours ago, sandyk said:

 

 Tell that to the quite a few members who are reporting hearing differences between different types of RAM especially the more expensive " industrial"  types.  The information that it represents may be the same, but that does NOT necessarily mean that the output ends up being a " text book" perfect waveform with no noise riding along with it.

Do you mean the analogue output...

You are hearing a 0,1dB difference or measuring it, is this at the speakers or voltage output of the DAC, and is this the difference between two identical files...

[sandyk]

48 minutes ago, marce said:

Do you mean the analogue output...

You are hearing a 0,1dB difference or measuring it, is this at the speakers or voltage output of the DAC, and is this the difference between two identical files...

I added the attached after the usual reactions.

 "For the benefit of the Richard Dale's and Mansr's of the forum, I was referring here to the levels of the differences that John and several other members are now routinely hearing between different changes that John has made with his Decoder. " 

I was not referring to bit identical files, although I was pointing out just how small differences may be heard, as evidenced by the extremely small measured changes that John often makes with his decoder settings.

 As I also said, ask John about this, as he agrees that we are all (John's small P.M. group from both sides) indeed hearing differences of a small fraction of a dB. 

Given that Checksums were never designed to need this kind of precision, it shouldn't be a surprise if they are unable to show minute audible differences like these that we are routinely noticing with John's minor corrections.   

[semente]

On 1/10/2020 at 2:01 AM, John Dyson said:

Suzanne Vega 1987

 

Definitely a feral A+! 1985 also feralA. Not sure about the others.

 

On 1/10/2020 at 2:01 AM, John Dyson said:

Simon & Garfunkel (parsely, bridge, sounds

 

The remasters sound fine to me. Are you referring to these (2001) or the original releases?

[esldude]

1 hour ago, sandyk said:

I added the attached after the usual reactions.

 "For the benefit of the Richard Dale's and Mansr's of the forum, I was referring here to the levels of the differences that John and several other members are now routinely hearing between different changes that John has made with his Decoder. " 

I was not referring to bit identical files, although I was pointing out just how small differences may be heard, as evidenced by the extremely small measured changes that John often makes with his decoder settings.

 As I also said, ask John about this, as he agrees that we are all (John's small P.M. group from both sides) indeed hearing differences of a small fraction of a dB. 

Given that Checksums were never designed to need this kind of precision, it shouldn't be a surprise if they are unable to show minute audible differences like these that we are routinely noticing with John's minor corrections.   

I don't understand.  If you process a signal which produces a different output even into the tenths of a db, it won't be a bit perfect match.  So what are you going on about in regards to checksums being designed for this kind of precision?  Checksums weren't designed for audio at all, and work at a few orders of magnitude higher precision.

[sandyk]

6 minutes ago, semente said:

 

Definitely a feral A+! 1985 also feralA. Not sure about the others.

 

 

The remasters sound fine to me. Are you referring to these (2001) or the original releases?

 

How about Suzanne Vega -Solitude Standing  1987 ?  (Luka etc.)

[sandyk]

6 minutes ago, esldude said:

I don't understand.  If you process a signal which produces a different output even into the tenths of a db, it won't be a bit perfect match.  So what are you going on about in regards to checksums being designed for this kind of precision?  Checksums weren't designed for audio at all, and work at a few orders of magnitude higher precision.

 

Dennis

I was talking about 2 different things, and TBH, I doubt that your precious Checksums are anywhere near as precise as you claim, otherwise they would show the differences that I hear, AND have been verified by numerous members.

 

Alex

[semente]

6 minutes ago, sandyk said:

 

How about Suzanne Vega -Solitude Standing  1987 ?  (Luka etc.)

 

That's the one I was referring to as being a feralA+

 

It sounds like someone turned on the floodlights...

[sandyk]

13 minutes ago, semente said:

 

That's the one I was referring to as being a feralA+

 

It sounds like someone turned on the floodlights...

Agreed.

 I don't remember it sounding so bad on FM radio , as it does on CD though.

[mansr]

40 minutes ago, sandyk said:

I doubt that your precious Checksums are anywhere near as precise as you claim, otherwise they would show the differences that I hear

Why do you keep banging on about checksums? If you don't trust a checksum to detect a file difference, you can simply compare them one bit/byte/whatever at a time.

[sandyk]

15 minutes ago, mansr said:

Why do you keep banging on about checksums? If you don't trust a checksum to detect a file difference, you can simply compare them one bit/byte/whatever at a time.

I keep banging on about them because so many think they are infallible . They are NOT infallible any more than you believe you are in all matters electronic. 

 Let's see you find any differences in that area if Mani does a rerun and repeats his performance again.

[John Dyson]

1 hour ago, semente said:

 

Definitely a feral A+! 1985 also feralA. Not sure about the others.

 

 

The remasters sound fine to me. Are you referring to these (2001) or the original releases?

I forget the context, but I have both -- and both are feral A.

Whet I'll do a little later on is to do decodes of both versions that I have of a few songs, and give snippets here.

 

The S&G stuff sounds good though -- but it does need much more TLC than what I did in the past when decoding.

Also, there could be other releases.  Not all releases of the same timeframe might might/not be feral -- never really know without testing.  I owe it to you to do some examples though.  Give me a few hours to get situated.  Also, I'll try to be careful!!! 🙂  It is so easy to make mistakes, it really is.  This undoing the EQ (the corrective EQ) is sometimes tedious and tricky to do right.  Sometimes, it is really easy.  I seem to remember that S&G can be a bit tricky.

 

EDIT:  sorry... I haven't seen the CDs in a long time, but both were remasters of slightly different genres.   I don't have an original release -- I thought I did.   If you have an original -- if you send me about 55 second snippet, I'll check it out!!!  SORRY!!!

 

John

 

[semente]

17 minutes ago, John Dyson said:

I forget the context, but I have both -- and both are feral A.

 

John

 

 

You find the Simon & Garfunkel remastered releases 2001 feralA?

 

I was quoting this message:

 

On 1/10/2020 at 2:01 AM, John Dyson said:

Unfortunately, I have no database of 'bad' CDs.  I have found it to be hit or miss on CDs before 1994, usually a hit (bad.)  Not always, but very often.

I can state that my ABBA albums are all 'feralA', ABBA Gold (1992) and more ABBA Gold(1993) are a different kind of FeralA, Phil Collins No Jacket, Phil Collins very best, Brasil 66, Herb Alpert Very best, Burt Bacharach collection 2 CD set, Carpenters albums (all), Bread, Nat King Cole Story (Analog Productions), Carly Simon Greatest hits, every Linda Ronstadt that I have, ONJ 48 early singles EMI, Carpenters Singles from HDtracks, Paul McCartney Band on the Run unlimited HDtracks, The Cars Complete greatest hits, Suzanne Vega 1987, 1985, 1992 albums, Simon & Garfunkel (parsely, bridge, sounds), a sample from Sheffield Labs (I have the music in me) is also feralA.

 

Frankly, almost everything that I have is FeralA -- except a remastered Supertramp and a few esoteric like that.

 

My taste probably hits the feralA problem perfectly.  I cannot believe that everything is feralA.  However, when I try to decode material that is NOT feralA, it sounds BAD.

 

John

 

[John Dyson]

11 minutes ago, semente said:

 

You find the Simon & Garfunkel remastered releases 2001 feralA?

 

I was quoting this message:

 

I am sorry -- I have the HDtracks and another 2007 remaster.  I made a mistake, I apologize!!!  I haven't seen my CDs in a long time (a decade or so.)  Everything is online nowadays.  I am happy to do a full or partial decode attempt of anything you want, if you are curious.  Also, much of the time, I can tell by listening.  About 55seconds of interesting material is uaually enough, sometimes less.  I have some S&G examples here, and can use them as a base.  Maybe Frank LLoyd Wright or some similar kind of material would be useful.

Again, I really regret making misstatements, all we have is our veracity and truthfulness when we don't know each other eye-to-eye!!!

 

Here is a quick, but careful decode of 'The Boxer', HDtracks version (192k, but decoded at 66.15k, converted to mp3 -- just to save time, sound should be close.)  The decoding quality level is the normal professional maximum quality.  (I have secret super anti-MD modes also, but not needed here.)  I am not 100% sure as to the EQ, but it seems fairly close.

 

I noticed on S&G, the most important thing is more 'hiss' and 'room rumble' rather than high level sound, but sometimes it helps, dulling the feral a bit though...

 

 

John

 

HD-SandG-06.Boxer.mp3

[mansr]

1 minute ago, sandyk said:

I keep banging on about them because so many think they are infallible . They are NOT infallible any more infallible than you believe you are in all matters electronic. 

Let's talk about checksums for a bit then. The simplest checksum is a single parity bit indicating whether the number of ones in the message is even or odd. This is guaranteed to detect a single-bit error, but it fails if there are two errors. As such, it is only suitable for very short messages. A common application is per-byte parity in RS-232 communication. To reliably detect multi-bit errors, a more complex method must be used, a popular choice for moderate message sizes being some CRC32 variant. With messages of up to a few kilobytes transmitted over a reasonably good link, the vast majority of errors will be caught. Ethernet uses a per-frame CRC32 for this reason. For large files, cryptographic hashes such as SHA-256 are often used.

 

Regardless of how it is calculated, a checksum shorter than the message can never detect all errors. This follows trivially from the fact that there are more possible messages than there are possible values for the checksum (the pigeon-hole principle). The goal when choosing a checksum is to make likely errors unlikely to result in collisions.

 

To determine whether two files in local storage contain the same data, the most efficient and most reliable method is to compare them a byte (or a few bytes) at a time. The only way this can produce a false match is if a random errors within the computer (e.g. in reading the file from disk) happen to change one file such that it perfectly matches the other. The probability of this happening in an otherwise working system is vanishingly small. The probability of it happening every time the comparison program is used while not affecting the audio player is as close to zero as one cares to imagine.

 

To sum up, if two files compare equal, they are equal. If two files produce the same SHA256 hash, the chance of them actually being different is 1 in 340282366920938463463374607431768211456 (that's 2^128). The chance of two audio files with the same SHA256 hash being different yet sounding almost the same is unimaginably small.

 

If you regularly find hash collisions in your data, you should probably check for the presence of an infinite improbability drive in your vicinity.

[semente]

28 minutes ago, John Dyson said:

I am sorry -- I have the HDtracks and another 2007 remaster.  I made a mistake, I apologize!!!  I haven't seen my CDs in a long time (a decade or so.)  Everything is online nowadays.  I am happy to do a full or partial decode attempt of anything you want, if you are curious.  Also, much of the time, I can tell by listening.  About 55seconds of interesting material is uaually enough, sometimes less.  I have some S&G examples here, and can use them as a base.  Maybe Frank LLoyd Wright or some similar kind of material would be useful.

Again, I really regret making misstatements, all we have is our veracity and truthfulness when we don't know each other eye-to-eye!!!

 

Here is a quick, but careful decode of 'The Boxer', HDtracks version (192k, but decoded at 66.15k, converted to mp3 -- just to save time, sound should be close.)  The decoding quality level is the normal professional maximum quality.  (I have secret super anti-MD modes also, but not needed here.)  I am not 100% sure as to the EQ, but it seems fairly close.

 

I noticed on S&G, the most important thing is more 'hiss' and 'room rumble' rather than high level sound, but sometimes it helps, dulling the feral a bit though...

 

 

John

 

HD-SandG-06.Boxer.mp3 2.1 MB · 1 download

 

The tonality seems nicely balance, but the sibilance (mp3?) is a bit overwhelming.

[John Dyson]

2 hours ago, sandyk said:

 

How about Suzanne Vega -Solitude Standing  1987 ?  (Luka etc.)

Here is a decoded Luka snippet.  I didn't do the entire album with the new style of formulas, so there might be more quality issues than if I consider the entire album.  (Note that when I do decodes, I am not looking for 'good sound' as much as avoiding defects.)  Sometimes, I don't even remember the material when concentrating for problems -- weird, huh?

Comment -- I do believe that Vega's recording should be a little more hot/brighter than something like S&G, so I tried to make sure that I didn't dull the material too much.  I did a straight -1, -2, -1, -2, -1, -2, -1, -2 EQ -- which seems to be a relatively common EQ value (the freqs are (4250, 4500, 5750, 6000, 8750, 9000, 11750, 12000), all Q=0.50.  Almost all (I mean ALL) decodes need EQ at these frequencies or 0dB sometimes for 4250/4500 and once in a LONG LONG while, 0dB at 8750 on up.  Note that the 250 Hz offsets are totally critical -- amazing, eh?

 

 

Luka.mp3

[John Dyson]

3 minutes ago, semente said:

 

The tonality seems nicely balance, but the sibilance (mp3?) is a bit overwhelming.

I understand the sibilance.  That is usually contained in the 5k-7kHz range, and that is one of my foibles when decoding.  There are people who criticize the sibilance when I decode, and it is a flaw in the way I hear or my perception.  I tend to feel reassured by a bit extra sibilance, but it is probably incorrect, and it appears that many people are irritated by it (probably rightfully so.)  I might have a dip in the 5kHz-7kHz range, but do know that I can hear reasonably up to 14kHz, so I do make mistakes for a couple of reasons.

Here is another decode attempt, instead of the original (-1,-2, -1, -2, -0.5, -1.0, -0.5, -1.0) decode, this one is (-1.5, -3.0, -1.5, -3.0, -0.5, -1.0, -0.5, -1.0) EQ.   (I mentioned the frequencies of the EQ in a previous posting.)   There is also some MF EQ, but those seem usually to be a relatively consistent set of values.  (There are usually 3 choices of those, and I just happened to probably correctly pick the 2750, 3000Hz pair.)

 

John

 

HD-SandG-06.BoxerA.mp3

[semente]

12 minutes ago, John Dyson said:

Here is a decoded Luka snippet.  I didn't do the entire album with the new style of formulas, so there might be more quality issues than if I consider the entire album.  (Note that when I do decodes, I am not looking for 'good sound' as much as avoiding defects.)  Sometimes, I don't even remember the material when concentrating for problems -- weird, huh?

Comment -- I do believe that Vega's recording should be a little more hot/brighter than something like S&G, so I tried to make sure that I didn't dull the material too much.  I did a straight -1, -2, -1, -2, -1, -2, -1, -2 EQ -- which seems to be a relatively common EQ value (the freqs are (4250, 4500, 5750, 6000, 8750, 9000, 11750, 12000), all Q=0.50.  Almost all (I mean ALL) decodes need EQ at these frequencies or 0dB sometimes for 4250/4500 and once in a LONG LONG while, 0dB at 8750 on up.  Note that the 250 Hz offsets are totally critical -- amazing, eh?

 

 

Luka.mp3 2.1 MB · 0 downloads

 

 

 

 

[semente]

15 minutes ago, John Dyson said:

I understand the sibilance.  That is usually contained in the 5k-7kHz range, and that is one of my foibles when decoding.  There are people who criticize the sibilance when I decode, and it is a flaw in the way I hear or my perception.  I tend to feel reassured by a bit extra sibilance, but it is probably incorrect, and it appears that many people are irritated by it (probably rightfully so.)  I might have a dip in the 5kHz-7kHz range, but do know that I can hear reasonably up to 14kHz, so I do make mistakes for a couple of reasons.

Here is another decode attempt, instead of the original (-1,-2, -1, -2, -0.5, -1.0, -0.5, -1.0) decode, this one is (-1.5, -3.0, -1.5, -3.0, -0.5, -1.0, -0.5, -1.0) EQ.   (I mentioned the frequencies of the EQ in a previous posting.)   There is also some MF EQ, but those seem usually to be a relatively consistent set of values.  (There are usually 3 choices of those, and I just happened to probably correctly pick the 2750, 3000Hz pair.)

 

John

 

HD-SandG-06.BoxerA.mp3 2.1 MB · 1 download

 

I prefer the previous version, this one sounds too "dull" in the treble.

 

Perhaps the exaggerated sibilance is an artifact from the mp3 conversion..

[John Dyson]

45 minutes ago, semente said:

 

I prefer the previous version, this one sounds too "dull" in the treble.

 

Perhaps the exaggerated sibilance is an artifact from the mp3 conversion..

I didn't think that mp3 would make the sibilance worse, but it just might.  I also noticed a confusion in the edges of the highs (the mixing of the vocals.)  Almost like a kind of IMD type distortion on the mp3..  MP3 tends to have more temporal issues than others.  Also, I created the mp3 with sox, when I normally use lame in 'extreme' mode, with V0 and q0, therefore getting close to the most variable rate that it can do.  Instead of discussing mp3, I am posting a flac version of the original  (Which seemed to me to be the best, but as mentioned elsewhere, I have a high tolerance for sibilance.)   There is one other possible problem -- I might be using the wrong MF EQ.  I am posting a flac version of the original and a flac version of the original, but with slightly different MF eq.

 

I believe that it is *possible* that the problem isn't that there is too much sibilance, but instead too little lower frequency support for the highs.  That kind of problem is a common EQ mistake done by those 'boosting' their own material -- it creates all kinds of problems...

 

So, the first .flac is the full version of the first version, the 2nd flac is the same as the 1st, but uses an MF EQ 250Hz lower and side calibration 3dB higher -- it is also possible that the side calibration was too low (making it too hot.)  Both changes should incrementally change/improve the relative sensation of sibilance.  This stuff is incredibly tricky, esp on material like S&G...  Some material is easier, but I am starting to be a little challenged by the S&G stuff -- more challenging than I thought.  My standards are also much higher than they were at the last time I tried S&G.

 

I did listen to the source material -- and the sibilance is there, but on the decode version, the other associated sounds are relatively suppressed.  That *does* increase audible confusion if the sibilance is too strong relative to the supporting lower frequencies.  There are also some other EQ options, but I try to avoid using EQ outside of the norm.  That is, it might be helpful to use slightly stronger EQ at 6kHz, but to keep the same at 4.5kHz.  On the second earlier example, I hit both 4.5kHz and 6kHz at the same amount.

 

Also, there is a third, where both the 3dB increase in side channel calibration, plus a general decrease of -0.25dB in calibration.

 

There are two common values of base calibration:  -12.44 (I was using -12.42 because it was slightly cleaner), and -12.67.  Sometimes that slight 0.25dB difference really improves things.  It is interesting that the calibration levels are usually one of the two values, except for certain premium recordings, where they use calibrations that support much larger headroom (like -20.47dB or so.)

 

Again, the new examples are 1) flac example of exact same as original decode, 2) flac example of original decode, but the EQ freq is 250Hz lower, and calibration is 3dB higher (which makes the highs in the side channel less hot.)  3) EQ same as orig, but side 3dB higher, and general 0.25dB decrease.

 

* Note if I had a master tape of the material, all of this EQ stuff goes away, and all it takes is sometimes a bit of a tweak on calibration.

 

John

 

HD-SandG-06.Boxer.flac HD-SandG-06.Boxer-256-Splus3dB.flac HD-SandG-06.Boxer-Splus3dB-cal-0.25dB.flac

[kumakuma]

4 hours ago, mansr said:

If you don't trust a checksum to detect a file difference, you can simply compare them one bit/byte/whatever at a time.

 

That's what I did with the pairs of files that Alex sent me on a number of occasions.

 

They were bit-identical and I could hear no difference between them.

[John Dyson]

37 minutes ago, kumakuma said:

 

That's what I did with the pairs of files that Alex sent me on a number of occasions.

 

They were bit-identical and I could hear no difference between them.

You are lucky that your ears aren't fooled.  My own hearing is not reliable enough -- and stuff sounds different even if it is totally the same.   I have a very short audio memory, and try to keep my comparisons down to the 5-10 second length.  Any more length -- maybe at 30seconds, then I become unreliable to the point of getting totally random results.   Also, my hearing accomodates defects irritatingly quickly - so a defect repeated too often eventually becomes ignored.   I really wish that I had more stable/reliable hearing.  I don't know if it is age or not, but my memory of nuances is very poor.  This makes my project very frustrating when testing...

 

John