Is the DR database really accurate for vinyl? (pt2)

Continued from here:

Is the DR database really accurate for vinyl?

 

I can't see how to reply to a post from Part 1 of this thread but this is what Antares asked:

Antares, I listened to Thurenity's needledrop from Love of the Game and tested exactly the section of music, using the needledrop I did directly from my copy of the vinyl test pressing. I get the same result as Thurenity - DR11:

foobar2000 1.2.9 / Dynamic Range Meter 1.1.1
log date: 2013-11-02 21:52:56

--------------------------------------------------------------------------------
Statistics for: ?-loveof the game extract
Number of samples: 3329328
Duration: 1:15
--------------------------------------------------------------------------------

Left Right

Peak Value: -8.54 dB --- -8.46 dB
Avg RMS: -21.77 dB --- -21.88 dB
DR channel: 10.79 dB --- 11.06 dB
--------------------------------------------------------------------------------

Official DR Value: DR11

Samplerate: 44100 Hz
Channels: 2
Bits per sample: 16
Bitrate: 715 kbps
Codec: FLAC
================================================================================

So, yes, I agree that whatever the reason may be for the big difference in DR values between the vinyl version and the CD version, it's not the equipment used in the needledropping - because everyone is getting the same result.

There's something odd about the vinyl version that I'll mention in passing. On the track "So Long", there's noisy light crackle all the way through - it sounds just like an old 78rpm record. And the following track, Listen, has a different kind of distortion through the middle section, like a lot of digital clicks and crackles. Now here's the really odd thing - exactly the same faults appear on the digital download, so it's nothing to do with the vinyl. I've asked Keith what's going on but haven't had a reply yet.

 

Many thanks! So we can safely say the difference is unrelated to the needle drop setups in this case. This would sure make more sense if the vinyl had been cut from the unlimited master.

I suppose the crackle on that one track isn't an artificial vinyl effect?

 

No, I got a reply from Keith and he said it was "artifacts of some of the sampling that was done for each of those songs". I don't really understand it at all but it's probably not relevant here so I'll leave it at that. I'm not sure from what he said whether it's on the CD version but it's very noisy indeed on the LP and the digital download.

 

I've just realised - if you go to Artemis's home page you can hear all the tracks from the album in full. So if you go to the following link for the Sephyra album, and scroll down to the playlist, and click on So Long, you can hear exactly what I'm talking about right from the outset. Please do - it only takes a second and I would really like to know whether or not you think I'm exaggerating about the crackle.

http://music.artemis.fm/album/sephyra-4

If you go on to the next track, Listen, it's fine to start with (there are some sound effects but that's OK) but at about 1.30 a clicking and crackling noise sets in, becomes worse, and continues to the end of the track.

Again, I'm really curious to know, do you think it's OK and nothing to worry about or would it bother you?

BTW it appears that you can actually download the tracks, possibly at CD res, for next to nothing, if you want to do any tests.

Also, if anyone would like the vinyl test pressing for testing purposes, I would be happy to send it on for the cost of postage (from the UK). Just PM me.

 

Judging by the results of comparisons of 2 different vinyl rip excerpts (posted in part 1 of this very thread) to alleged limited digital master, I 'd have to say the TT DR values for vinyl appear to be accurate. We have to realize that all it takes for the TT DR value to jump 4dB is to have a single peak value (second highest) in the whole waveform to rise 58% over average RMS value of its 20% loudest blocks. This is exactly what we saw here, IMO...

 

On the penultimate track it's definitely an added vinyl effect that I can live with, on the last track I don't know what the intention was but it's not really pleasant to me either. I remember on the last Melody Gardot album there is a track with a seemingly random tick which bothered some people, but turned out to be intentional fingernail tapping or something like that. This is much more intrusive though.

 

No, really, it's not an intentional effect on either track. Keith said it was "artefacts of some of the sampling that was done for each of those songs." I don't want to quote any more from him without permission but he made it clear that the sounds weren't intentional on either track.

 

^ Okay, I would hope it can still be corrected then - the music is quite nice. Probably won't change the DR reading though.

 

I've been doing some experimenting. It seems to me that if we're going to compare DR values between a digital file and a recording of vinyl playback, it's important to even up the playing field a bit. As some have noted earlier in this thread, vinyl playback has a pile of variables that affect the resulting waveform. So it stands to reason that comparing it to a computer-generated estimation of the data in a digital file is not really useful. After all, that data will be used by a DAC to reconstruct an analog waveform. In the case of Artemis (nice album by the way, I was happy to buy it), we know Ian's master was used to produce both the CD and the vinyl. The data in the master was put through a DAC then sent to the cutting head (plus whatever else was done to compensate for vinyl playback). However, what happens if you look at the reconstructed waveform from the digital file by recording the analog output from a DAC? Here's what I got for Love of the Game:



Compare this to the Flac file I purchased from the Artemis web site, which I've reduced to match the RMS level of my rerecording:



As for the DR, the original file is DR8, my rerecording, DR11!

I did not process this other than to put the original flac file on a USB stick, plug it into my Onkyo receiver, assign its playback to the analog output of the receiver and record it into my computer the same way I do for needledrops. Granted, the Burr-Brown DACs in my receiver sound quite nice to me, but other than playing the file back through them, I did nothing else but set the level.

So, really the DR results from the vinyl are not far off from those of the reconstructed digital file.

Sure, different DACs might give different results, but the principle is still the same, the vinyl version has had its waveform reconstructed in a DAC, so to be valid for comparison, the digital file should undergo a similar process.

I've been spending the weekend doing this rerecording for a variety of albums I own that suffer from varying degrees of brickwalling but contain music I like. In all cases, the DR increased by 2-3 and the value I find more reliable for measuring actual perception, the Replaygain, ended up around -4 to -5 instead of -9 or -10.

 

Great post, Stefan - and thanks for putting in the time and effort on this. Very interesting.

 

I wouldn't have expected such a significant difference. One possible exmplanation might be that the analogue path between the output of the DAC chip in the receiver and the input of the ADC chip in the laptop is making changes to (maybe) frequency response or (more likely) relative phase. This could be down to something as basic as the values of coupling cpacitors used.
Or, maybe the digital data contains "intersample peaks" and the DAC is reproducing them. I think you could test this by resampling the digital file to a higher rate and seeing what it looks like.

 

Interesting - I'm surprised that the DAC made such a big difference to the numbers, too. I suspect it's down to frequency-dependent phase issues, as Don suggested and we saw with applying non-linear digital processing in the original thread.(There are no inter-sample peaks in the digital file, btw)It doesn't really change my overall conclusion, though - comparing DR values made from vinyl with digital ones isn't useful - and therefore, the TT Meter isn't useful for vinyl, in my opinion.Thanks for taking the time to do the tests and post the results, though !

 

So it's got nothing to do with vinyl after all. If you play the digital file, you get the same DR as the record...

 

Could well be phase related. I still think converter design may play a part too. Re-recording from the DAC output is a trick I read surround sound engineer Ian Wilkes advocate in a surround-related forum several years ago. Mind you, he was suggesting it be done using high end DACs, such as n RME unit. I've tried rerecording before using the DAC in my Yamaha CD player and didn't get results quite as dramatic as this one, but neither did I get complete flat tops either. I also used a tube buffer for awhile in the re-recording chain and that produced some pleasant sounding results but it tended to roll off everything under 50Hz rather sharply so that played a part (as was discussed in this thread).

I haven't tried upsampling with the current setup, but the Yamaha player mentioned above can do resampling to both 88.2kHz and 176kHz. I tried both with rerecording and didn't see any noticeable difference.

Oh I agree with your premise. I just felt if folks are going to compare DR values, they should both be put on as fair a playing field as possible. I've felt for awhile now that graphic representations of waveforms done with computer software are not entirely accurate, especially considering the variances in how software displays them, not to mention how people misread them. I don't know how many times I've seen people misuse the description "clipped" on here simply because they don't understand what constitutes real clipped digital audio and because they don't realize that zoomed out limited audio will look "clipped." I know some folks here would look at your Artemis waveforms and think you should be condemned to eternal damnation!

As for Dr, I know you find the TT Meter useful for digital, but I feel DR is a flawed approach. I prefer Replaygain levels for judging overall loudness, and most of all my ears.

 

There are variables, but I believe they're a bit closer than some think. Put it this way, the equivalent of a waveform graphic for vinyl would be a computer-generated graphic representing the groove modulations in the vinyl. Both it and the waveform graphic display the variances of amplitude in the source but show nothing of the waveform produced. Digitally recording the output from a cartridge/phono stage combo or a DAC and looking at the resulting waveform is I believe a fairer comparison.

 

I might add that I believe the amount of "reconstruction" possible by a DAC depends a lot on the degree and type of limiting applied. I re-recorded about a dozen CDs over the weekend. Some resulting waveforms showed the same or even better results than Artemis but others were much flatter with shorter peaks across the flat. Those with heavier compression/limiting seemed to show less improvement, although in all cases, they sounded better.

 

Interesting post @Stefan and something I've wondered about, mainly due to SACD (where I might use the analog hole to record the few that I have, since I don't have an old unmoodded PS3).

But here's a thought - perhaps the DR meter is also inaccurate for CD's and digital files as well? If we're essentially trying to capture the analog output from the DAC to our ears, wouldn't recording the CD / digital file from the analog outputs be the way to do that? Tricky question.

 

My analog rip of Brothers In Arms Mofi sacd layer shows a different DR than the one posted in the database

 

Okay just completed an analog rip of Killing Jokes Absolute Dissent and re-ran the DR meter and sure enough it gave a different (better,not much) analysis

 

I wanted to clear up the confusion about the 'artifacts' and noise in a couple of the tracks that was discussed earlier in this thread - they're intentional in that we didn't do anything to remove them. During the process of recording this record we used various samples that Artemis captured in her travels as elements in the songs. Some of them had scratchy ticks, pops, etc. or other noises that we decided to keep for various reasons. Some samples were used to create some of the synth instrumentation (mapping samples across the keyboard and so on) while others were processed in various ways.

In hindsight we probably should have done pop/click removal for 'So Long' for the vinyl release. C'est la vie. Live and learn. Everybody makes a mistake now and then. Hopefully it's forgiveable...!

 

I've been following this thread for a while and thought I'd share a quick bit of analysis I conducted on the DR database. Below is a graph of all the vinyl values I could easily identify (1783 of them) plotted against time for 1970 onwards.

What may be of interest is the difference between vinyl, lossless and lossy shown here:
http://www.tristancollins.me/computing/dynamic-range-analysis/

I'm no expert in this, just an interested party, so would be keen to get your thoughts / input on this.

Tristan

 

Interesting graph. As I looked at it this morning, I thought it could also be a plausible graphic representation of how the role of popular music has changed from a primary focus of entertainment to background, that is, folks used to sit down and actively listen to an album in its own right with their entire focus on the listening experience (which many of us who frequent this forum still do), whereas in the post-2000 world, for many folks, music is something that plays in the background while they're focused on something else. Thus compression and lack of big dynamic swings is desirable so as not to distract focus.

Just a thought.

 

Yep - potentially. That could be seen in the different trends for 2000 onwards between lossy and lossless values. Of course, there are many, many caveats to all this! If there is a positive to all this, it is that lossless doesn't seem to be getting worse and vinyls have somewhat plateaued...

Tristan