Is the DR database really accurate for vinyl?

Discussion in 'Music Corner' started by Mij Retrac, Oct 3, 2013.

Thread Status:
Not open for further replies.
  1. ad180

    ad180 Forum Resident

    Location:
    USA
    @Ian Shepherd: Does the summing that happens with vinyl occur at different frequencies? I assume not all are summed equally.
     
  2. Ian Shepherd

    Ian Shepherd Forum Resident

    Louder doesn't mean more dynamic. The width processor I used maintains equal overall gain, so the level of the centre "M" changes relative to the side "S" channel, but it doesn't make the drums punchier, or "repair" their transients, or otherwise create any real dynamics, as we've said.

    The existing dynamics are re-distributed within the stereo image - different, but not more.
     
    Last edited: Oct 7, 2013
  3. Ian Shepherd

    Ian Shepherd Forum Resident

    I actually don't know, for sure - this to probably depends on the stylus. But even if the amount of crosstalk is even across all frequencies, because of the complex phase relationships some frequencies will cancel and others will build up, so even if the summing is even, the result won't be. Again, you'll end up with something that is just... different. How it will sound will be different for every piece of music, even without the variable of the stylus.
     
  4. back2vinyl

    back2vinyl Forum Resident

    Location:
    London, UK
    What I'm thinking is that, suppose the sounds that formed the dynamic peaks tended to be at the centre of the soundstage, those peaks would be emphasised and exaggerated relative to the rest of the recording and hence that's where you could get your extra dynamic range from. For memory, the heightened peaks were clearly visible on the waveform - they're not just a product of the TT meter's imagination!
     
  5. tmtomh

    tmtomh Forum Resident

    You are correct that if you narrow the stereo image of the CD/digital file, you may very well boost the DR reading. BUT - and here's where Ian's point comes in - then not only will the CD no longer be the "real," original CD - but it also will no longer be a comparable mastering to the vinyl. So to compare apples to apples, you'd have to narrow the stereo image of the vinyl rip, and you'd be back to square one: having different DR readings for CD and vinyl.

    Narrowing a CD rip's stereo image is indeed "artificial" in that it's extra processing beyond what the master is supposed to sound like. But narrowing the vinyl's stereo image before cutting the vinyl master is not "artificial" in the same way. The image is narrowed as part of an effort to keep the sound as originally intended as much as possible within the constraints of the vinyl format.
     
    Ian Shepherd likes this.
  6. Ian Shepherd

    Ian Shepherd Forum Resident

    The drums (kick and snare) will be the biggest contributor to what the TT meter measures - peak to RMS ratio. Anything that's in mono won't cancel in the way I described, only what's at the edges. So the relative size of those peak-to-RMS measurements won't be affected, only their relative level to what was at the edges of the image.

    So for an extreme example if all the drums were at the edges, in anti-phase (say) they would disappear when mono-ed, as I mentioned in a previous reply. This would affect the measured dynamic range, but in a very un-musical way. In this case, the drums, bass, vocals etc are in phase - it's mainly string pads and reverb at the edges. We're also only talking about a 10-15% reduction in width, by my guess.

    Overall I don't believe (or hear) a real 4dB change in dynamics can happen in this way. When I listen to the vinyl in comparison to the CD, I hear less impact, less depth, less clarity, less richness, less beauty.

    But I know you guys won't take my word for it, so with the kind permission of the band, here are the audio files - knock yourselves out !

    https://dl.dropboxusercontent.com/u/8441718/PA/artemis CD.wav

    https://dl.dropboxusercontent.com/u/8441718/PA/artemis vinyl.wav

    Both are 24-bit, even though the CD was originally 16-bit, obviously. This is to make the comparison even and maintain the highest possible quality in the adjustment in gain.

    The levels were matched using the integrated R128 meter in Wavelab.
     
    Last edited: Oct 7, 2013
    eyeCalypso and Mij Retrac like this.
  7. Thurenity

    Thurenity Listening to some tunes

    Thanks! This is really the meat of the matter -- two lossless files to compare.

    The CD version actually sounds better to me as well (I'm dead serious) - it's just a tad "fatter", but it's also a little brighter than the vinyl rip - more "alive" to my ears. This is one of those situations where, if it were me, I'd be pulling out my AT440Mla to redo the drop as it would likely match the CD and possibly bring out a bit more detail. Of course I'm likely colorizing the sound that way too, but if I am...then I am. :)

    I have a few carts that I use - primarily it's the AT440Mla for the added brightness and detail, but at times I will use my AT120e (if the source LP is too bright to begin with) or, for certain older LP's and/or mono LP's I might pull out my Nagaoka MP-110 - it's usually a quick judgement call on what I think sounds best.
     
    Ian Shepherd likes this.
  8. Thurenity

    Thurenity Listening to some tunes

    OK, here's my test. Three files : 50 second samples from Fleet Foxes "Ragged Wood":

    https://dl.dropboxusercontent.com/u/9158449/ff01-cd.wav
    https://dl.dropboxusercontent.com/u/9158449/ff02-120e.wav
    https://dl.dropboxusercontent.com/u/9158449/ff03-440mla.wav

    24/44.1k WAV files all, similar to your samples. The first one is a CD rip and amplified down so it somewhat matches the other two, volume-wise. The second is from a vinyl drop I had done in May 2011 with my early setup (AT120e cart, AT-LP120 TT from possibly the USB out? I didn't document back then. Definitely the 120e cart as I didn't own a 440Mla until a month later). The third one is a drop I just did with my current setup, which is a SL-3200 TT, AT440Mla cart and my GT40 stage / ADC into a Linux workstation.

    I think the differences in sound are significant from the CD to the drops, but notice that even the two drops sound different - I prefer the last one as it has that more bright and detailed sound signature I'm used to hearing. The first sample is DR4 and the other two are both DR9 (my entire 2011 drop averaged out at DR10, and the CD averaged out at DR7).

    I use this as an example for a few reasons. 1) the CD and vinyl were released at the same time, and it's a modern album 2) I own both the CD and vinyl to do the test and 3) I recall this song, in particular, where I had that "aha!" moment where I thought there really was something about the dynamics of modern vinyl that made me stand up and take notice. And, at DR10, my old drop isn't exactly screaming "great dynamics!".

    Two years and many drops later, it's still possible that this could be more colorization than a different master (ie. the two drops themselves sounding different), or maybe it's a combination of both, I still don't know. But, using these three samples as a guide, I can tell you which sample I think sounds the worst and that's the CD version, so that's really the bottom line for me.
     
    contium likes this.
  9. Keith Crusher

    Keith Crusher Member

    Hey everyone! I'm the label owner, co-producer/engineer, blahblahblah that works with Artemis and I was the one that did the rip. I thought I'd sign up and drop in to give a few details about it, so that we can 'close the loop' on that subject.

    Here's what I used, with notes:

    Turntable - modified Audio Technica AT-LP120 (preamp removed).
    Cartridge - Shure M97xE (1.5gm tracking weight - measures using a scale with .01gm accuracy)
    Recorded through a Pro-Ject Phono Box via USB

    I believe I used the Technics protractor for alignment, but can't recall for sure - I tried several versions (Lofgren, Stevenson, etc) and this one sounded the best.

    While definitely not 'high end' by any stretch, I think it's a pretty solid sounding turntable - and I've had at least a dozen models in my 25+ years of turntable ownership, including nearly every Technics model, Stantons, Rega and a few others.

    Personally, I preferred the sound of the vinyl - at least when I was playing it directly from the turntable (I haven't actually listened to more than a few minutes of the rip on headphones after I ripped it to verify it was a good rip...) but hey - that's just me.

    Hope this helps in some way and thanks for the interesting discussion!
     
    ClausH, morinix, Starwanderer and 7 others like this.
  10. testikoff

    testikoff Seasoned n00b

    Here are some graphs/stats for the files Ian posted (both were very lightly modified: DC offset was zeroed and ReplayGain for both tracks was set to equal level of -3.3dB).

    The spectrograms for both tracks look a tad different:

    - digital:
    Love Of The Game [digital].jpg

    - vinyl:
    Love Of The Game [vinyl].jpg
    As you can see something is clearly missing in the higher frequency area (above 14kHz & also above 20kHz) of the vinyl rip compared to digital master.

    Now to the spectrum graphs (at equal perceived loudness):

    - both spectrum curves and their delta:
    Artemis - Comparison.jpg

    - spectrum delta only:
    Artemis - Delta.jpg
    Apparently, digital master has more low bass (under 40Hz), slightly less upper-low bass and mids (between 100Hz & 2kHz) & substantially more highs (above 4kHz) than vinyl rip.

    The audio stats for tracks are:

    - digital:
    Code:
    Artemis Robinson - Sephyra
    
    Peaks
    =====
    61.5
    
    Peak Table
    ==========
    1| 61.5
    
    Full Statistics
    ===============
    #|Chan|Peak %|  Peak dB|AvgRms dB|TotRms dB|MinRms dB|MaxRms dB| MinSample| MaxSample| DCOffset
    --+----+------+---------+---------+---------+---------+---------+----------+----------+---------
    1|  1|  61.5|  -4.2239| -13.4110| -13.0089| -23.7889|  -6.4797|-0.6149021| 0.6146790|  -0.0000
      |  2|  61.5|  -4.2239| -13.9711| -13.5499| -23.5778|  -6.6958|-0.6148996| 0.6146810|  -0.0000
    
    0 dBFS: Sine
    RMS Window Size: 50 milliseconds
    
    - vinyl:
    Code:
    Artemis Robinson - Sephyra
    
    Peaks
    =====
    98.2
    
    Peak Table
    ==========
    1| 98.2
    
    Full Statistics
    ===============
    #|Chan|Peak %|  Peak dB|AvgRms dB|TotRms dB|MinRms dB|MaxRms dB| MinSample| MaxSample| DCOffset
    --+----+------+---------+---------+---------+---------+---------+----------+----------+---------
    1|  1|  97.8|  -0.1969| -13.2034| -12.8478| -23.0754|  -6.7526|-0.9775865| 0.9287746|  0.0000
      |  2|  98.2|  -0.1574| -14.3013| -13.8612| -24.2132|  -6.9788|-0.9820442| 0.8025596|  -0.0000
    
    0 dBFS: Sine
    RMS Window Size: 50 milliseconds
    
    Finally to the DR stats:

    - digital:
    Code:
    
    --------------------------------------------------------------------------------
    Statistics for: 02-Love Of The Game [digital]
    Number of samples: 3521244
    Duration: 1:20
    --------------------------------------------------------------------------------
    
      Left  Right
    
    Peak Value:  -4.22 dB  ---  -4.22 dB 
    Avg RMS:  -13.00 dB  ---  -13.53 dB 
    DR channel:  7.46 dB  ---  8.06 dB 
    --------------------------------------------------------------------------------
    
    Official DR Value: DR8
    
    Samplerate:  44100 Hz
    Channels:  2
    Bits per sample:  24
    Bitrate:  1501 kbps
    Codec:  FLAC
    
    - vinyl:
    Code:
    --------------------------------------------------------------------------------
    Statistics for: 02-Love Of The Game [vinyl]
    Number of samples: 3521244
    Duration: 1:20
    --------------------------------------------------------------------------------
    
      Left  Right
    
    Peak Value:  -0.20 dB  ---  -0.16 dB 
    Avg RMS:  -12.83 dB  ---  -13.84 dB 
    DR channel:  10.92 dB  ---  11.23 dB 
    --------------------------------------------------------------------------------
    
    Official DR Value: DR11
    
    Samplerate:  44100 Hz
    Channels:  2
    Bits per sample:  24
    Bitrate:  1492 kbps
    Codec:  FLAC
    
    Track DR values differ by 3dB, not 4, actually...
     
    Ham Sandwich likes this.
  11. Thurenity

    Thurenity Listening to some tunes

    There could have been some loss in the actual recording process, hard to say? A lot of factors when it comes to making a needledrop (good call on the modded LP120 @ Keith Crusher - I never modded mine but a few forum members here have and have reported significant sound improvement).

    I still think, between the two recordings presented, the CD is the superior one - it has a bit more "life" to it, can't explain it any better than that - could be choice of cart, or perhaps the USB input into the PC doing the recording, no way to know.
     
    Ian Shepherd likes this.
  12. Ian Shepherd

    Ian Shepherd Forum Resident

    That's probably because my comparison was for the whole song, whereas yours is for this section only, ie. only 25% (approx.) of the whole thing. But those particular values aren't telling us anything musically useful anyway, as we've already established !

    Nice to see that the frequency plots agree with my ears very closely :)
     
    Mij Retrac likes this.
  13. Ulli

    Ulli Forum Resident

    Location:
    Germany
    Wow, this thread is moving so fast, I can't keep up. Several posts here that I'd love to reply too, but you've already moved on...

    First of all, thanks a lot to Ian and Keith for providing the files and the ripping information, respectively!

    I'd like to add the following observation: If you concentrate at the section between roughly 30 and 50 seconds, you'll notice that the CD and vinyl rip waveforms look very similar there - no exaggerated peaks for the latter like in the preceding and following sections of the track. And indeed, if you measure the DR values for this short section only, you'll obtain DR10 for CD and DR9 for vinyl (I trimmed the two sections so as to start and end at the same positions to within 1 sample, which isn't the case for the entire files). This comparatively low-level section was not affected by the (soft, non-clipping) limiting that leveled the peaks of the surrounding sections of the CD master, so this seems to support the idea that it's the compression and limiting that produces the "fake" peaks when cutting the vinyl. Perhaps this also explains why on my many classic rock vinyl rips I obtain DR values that are very close to their digital counterparts, simply because these recordings weren't compressed as much.

    Oh, by the way, I really like the music and will purchase the album.
     
    Last edited: Oct 8, 2013
    Matt I, Keith Crusher and Mij Retrac like this.
  14. Ian Shepherd

    Ian Shepherd Forum Resident

    Maybe - or perhaps it's because this is where the sub-bass kicks in and any phase shift caused by RIAA filtering will have a bigger effect - or maybe where the crosstalk has a bigger effect because of the increased centre content.

    Or all three :)
     
    Last edited: Oct 8, 2013
    Mij Retrac likes this.
  15. back2vinyl

    back2vinyl Forum Resident

    Location:
    London, UK
    I still think Ian's theory about the narrowing of the stereo image is the most likely explanation so far. Having bought the LP (which is on its way to me), I have a CD resolution digital download of the whole album. This the waveform of the track in question, Love Of The Game, in Adobe Audition:

    Love Of The Game CD.PNG

    As you can see, the track is limited all right but there's no actual clipping - at all times there's a tiny amount of headroom, about 0.25dB, between the musical peaks and the 0dB mark.

    If you then use Adobe Audition to narrow the stereo image, this is what you get:

    Love Of The Game narrowed.PNG

    OK, I know it doesn't look that different but you have to look at the peaks - they are now hitting the limit and trying to go over it. Sure enough, if you check out the amplitude statistics, you find the track is now clipping:

    Amplitude stats narrowed.PNG

    The point is, narrowing the stereo image doesn't affect the DR number when you do it to a digital track like this one because pretty much the whole of the dynamic range is already used up and any further increase in dynamic range simply sends it into clipping. BUT if you transfer this track to the analogue domain and narrow the stereo image, there are no such contraints - that increase in dynamic range can now reach its full extent and will be very much in evidence if you then needledrop the track and apply the TT meter.

    Just going back to an earlier point, it seems to me that narrowing the stereo image would naturally increase the dynamic range, because the loudness of the centre is artificially increased relative to the loudness of the sides. So regardless of the overall level of volume, there will now be a bigger difference between the loudness of the dynamic peaks in or near the centre and the average level for the whole track, IMHO.
     
    Keith Crusher likes this.
  16. Ian Shepherd

    Ian Shepherd Forum Resident

    I don't think it's valid to refer to this as "dynamic range", it's just an increased in the measured "DR" value.

    As we've covered extensively earlier in the thread, DR is not really measuring dynamic range, this is a misnomer.

    And even though the DR value (peak-to-RMS) has increased, there is no audible change in musical dynamics, as you can now all verify for yourselves.

    So, while I agree that the measured DR value has increased as a side-effect of the narrowing (just as with the vinyl rip), this doesn't mean the result is any more dynamic, in the musical sense we're all interested in.
     
  17. Ulli

    Ulli Forum Resident

    Location:
    Germany
    It would be great if you could do that analysis again but reduce the gain by, say, 6dB first before narrowing the soundstage to avoid the clipping. I would be interested to know by how much the DR value changes then. I doubt it will be anywhere near 4dB.
     
  18. Stefan

    Stefan Senior Member

    Location:
    Montreal, Canada
    Now try doing the same thing but as soon as you open the file, convert it to 32-bit floating point (or else set Audition to convert by default. Otherwise, anything processing do in Audition adds dither noise with each operation).

    Then after you narrow the file, use the normalize function in Audition to normalize down to, say -0.25dB. I think you'll find the results a bit more surprising. That's because Audition doesn't show it, but when in 32-bit floating point mode, it allows peaks above 0dBFS. If you normalize down, they appear. That will likely give you a more accurate visual view of the new peaks affecting the DR results.
     
  19. Ian Shepherd

    Ian Shepherd Forum Resident

    I just re-did my test and the DR increases to 11 - actually 11.59 on the right channel, given the extra headroom. Of course we don't know how much the actual crosstalk of the vinyl is, or what the frequency-dependance is, or how the RIAA phase changes will affect it, but I think +4 is perfectly plausible.

    In fact, I next did a test where I widened by 15% and then narrowed again immediately by 15% (in 32-bit floating point). This kind of change in a digital system is entirely transparent, no cancelation occurs, so the final result should (and does) sound indistinguishable from the original.

    Even so, this file also measures DR11 (+0.59) in the offline TT meter, over the whole song, with the added headroom to prevent clipping.

    I think that's very clear demonstration that these extra DR points are "phantom" readings - an inaudible "audio mirage".
     
    Last edited: Oct 8, 2013
    Ham Sandwich likes this.
  20. Ian Shepherd

    Ian Shepherd Forum Resident

    Now that, right there, is what we should be worrying about ! That's why I founded Dynamic Range Day, and that's the problem.

    That disgusting DR4 mess is a disgrace. The vinyl doesn't sound stellar either, but at least it's not pumping and distorted. It makes me so angry ! It's absurd to make audio sound like that on a format with the potential for over 100 dB of dynamic range...
     
    Ulli, Engelsstaub, jimhb and 5 others like this.
  21. Ham Sandwich

    Ham Sandwich Senior Member

    Location:
    Sherwood, OR, USA
    Wow. That is unexpected. Widening by 15% and then shrinking by 15% can increase the DR number? Who would have thunk it?

    Now that the secret is out, the Fleet Foxes next album could be a DR 11 but still manage to keep their same squashed sound. :eek:
     
    Mij Retrac and Ian Shepherd like this.
  22. Thurenity

    Thurenity Listening to some tunes

    I've written this a few times here on the forum, but if modern CD's weren't so horribly compressed, in general, I would have never started buying vinyl again. For me personally, it wasn't about the tactile experience or the large album art (although those things are nice too), it was primarily a way to squeak out more dynamic range, possibly, versus a comparable CD.

    It's why this topic is important to me as I have to literally look at each needledrop on a case-by-case basis and say "was it worth buying this LP, or did I just waste my time and money?". I wish I didn't have to even worry about this question. :(
     
    Leif, Engelsstaub, jimhb and 5 others like this.
  23. back2vinyl

    back2vinyl Forum Resident

    Location:
    London, UK
    Ulli and Stefan, those were great suggestions and I would have followed them up but I think Ian has saved me the trouble! It does now seem very clear that narrowing the stereo image has the effect of increasing the peak-to-RMS dynamic range - and that this phenomenon is actually nothing to do with vinyl per se but happens just as much in the digital domain too if there's enough headroom for the increased peak-to-RMS dynamic range to show itself.

    What I still can't still get my head around is Ian's view that the increased peak-to-RMS dynamic range is just a "phantom" DR reading and that it's inaudible. My feeling is that there's nothing phantom about an increased peak-to-RMS reading that clearly manifests itself as extended peaks in the waveform and that can clearly be seen in the amplitude statistics of Adobe Audition. I don't know why it wouldn't be audible and look forward to receiving the LP so I can hear for myself but in the meantime I'll give it some thought and see what other people have to say.
     
    Leif and SergioRZ like this.
  24. Ian Shepherd

    Ian Shepherd Forum Resident

    Turn the question around though - why should they be audible ? It's not as if someone overdubbed a very quiet - um... cowbell - which would genuinely add some transient information.

    All of this comes back to the fact that actually, peak information is pretty meaningless in terms of audibility. It tells you when clipping will happen, it allows us to calculate the DR and... not much else. All these readings have to be carefully interpreted - to use my favourite example, a legato solo flute might typically measure a DR of only 3 or 4 dB. Does that mean it sounds like Death Magnetic ? No...
     
    Engelsstaub and Mij Retrac like this.
  25. Ian Shepherd

    Ian Shepherd Forum Resident

    For those who don't believe me about the "make a change then un-make it" increasing the peak level without affecting the sound and don't want to try it themselves, here's the evidence.

    One of these is the vinyl, one is the "faked" digital version (made from the limited CD file) - if you look carefully you can spot which is which, but you get the point...

    The "faked" DR11 version sounds identical to the DR8 CD.

    [​IMG]
     
    Mij Retrac likes this.
Thread Status:
Not open for further replies.

Share This Page

molar-endocrine