IPB

Welcome Guest ( Log In | Register )

2 Pages V  < 1 2  
Reply to this topicStart new topic
TT Dynamic Range meter "lies" about vinyl's dynamic rang, plus why you can't trust waveforms either
Engelsstaub
post Oct 8 2013, 01:37
Post #26





Group: Members
Posts: 545
Joined: 16-February 10
Member No.: 78200



QUOTE (DVDdoug @ Oct 7 2013, 17:12) *
QUOTE
Why do all (by all I'm speaking from my personal experience thus far as well) the eighties LPs measure almost exactly the same as their CD counterparts? Both are equally dynamic but for some reason always have pretty much the same DR "value" as one another.

...these records should, in theory, be "scoring" 3-4 dB higher on that meter as well. The records have the same issues of crosstalk, soundstage, etc. that we're taking Ian's word is changing the the "value" of his own recording.
I believe it's the lack of excessive compression (or lack of excessive compression). If you take an "pure" uncompressed recording of an orchestra and run it through an all-pass filter, the DR "measurement" or crest factor should be unaffected, because all of the frequency-components sum-up randomly either way.

But a compressed/limited file has been limited wherever the frequency-components "randomly" sum-up to exceed the threshold/limit. When you all-pass, the timing of the various frequency components is "randomized" again...


That seems like a reasonable explanation.

Thanks also for your willingness to discuss this. smile.gif I'm actually interested in this particular topic without some a priori expectations or agenda to gloat about. I just want to know the truth about it all and nothing more or less.

That's why I tend to trust Ian Shepherd's findings and (lack of) motives for looking into this. He comes across as one who likes both digital and vinyl for different reasons and doesn't seem to have some axe to grind.


--------------------
The Loudness War is over. Now it's a hopeless occupation.
Go to the top of the page
+Quote Post
krabapple
post Oct 8 2013, 03:34
Post #27





Group: Members
Posts: 2159
Joined: 18-December 03
Member No.: 10538



QUOTE (xnor @ Oct 7 2013, 16:09) *
Why does anyone use the TT DR Meter? Where is the algorithm specified? Why isn't it based on BS-1770?


there was a link in that thread at one point to a brief white paper describing the TT Meter's workings....not code though, if that;s what you mean.

Go to the top of the page
+Quote Post
extrabigmehdi
post Oct 8 2013, 20:01
Post #28





Group: Members
Posts: 401
Joined: 15-August 09
Member No.: 72330



You can download the music samples related to the video to do you own experiments in the following link.
They were "volume matched" using R128 in wavelab , so you can abx them directly if you wish:
http://forums.stevehoffman.tv/threads/is-t...14#post-9485602

My thought regarding the TT meter : ok , a big DR value isn't a warranty that the record has a good dynamic range.
However I think the reverse is true, a too low value (let's say below 6) indicate that the dynamics has been too much squashed.
So the TT meter is still doing it's job to some extent.
Go to the top of the page
+Quote Post
DVDdoug
post Oct 8 2013, 22:45
Post #29





Group: Members
Posts: 2441
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



QUOTE
My thought regarding the TT meter : ok , a big DR value isn't a warranty that the record has a good dynamic range.
However I think the reverse is true, a too low value (let's say below 6) indicate that the dynamics has been too much squashed.
So the TT meter is still doing it's job to some extent.

I'm pretty sure the TT meter is calculating something that correlates with perceived dynamic range to some extent...

But, we know there is some uncertainty in the calculation... Maybe we should allow for up to 4dB "error" or uncertainty. i.e., If two recordings measure within 4dB, there may be no real difference.

If someone wants to kill some time, I'll propose a casual experiment:
- Listen to a batch of recordings (at least 10) and give each recording a "dynamics" score. It would probably be best to use a 0-10 or 0-100 scale rather than trying to use dB. But if you have a lot of experience with the TT meter, it would be OK to predict the TT meter result.

Use recordings that you have not measured with the TT meter. And, it would probably be best if all of the sample recordings are either vinyl or digital.

- Check with the TT meter to see what kind of correlation you get.

Go to the top of the page
+Quote Post
greynol
post Oct 8 2013, 23:02
Post #30





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (DVDdoug @ Oct 8 2013, 14:45) *
up to 4dB "error" or uncertainty.

How many data points were used to derive this number?

Why is this metric even being given any credence in the first place?

I like your proposal, especially the part about listening before measuring. Suffice it to say, in the spirit of the experiment, listening before looking at waveforms should also be a requirement.

This post has been edited by greynol: Oct 8 2013, 23:39


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
pisymbol
post Oct 9 2013, 05:07
Post #31





Group: Members
Posts: 43
Joined: 22-March 09
Member No.: 68274



QUOTE (krabapple @ Oct 7 2013, 22:34) *
QUOTE (xnor @ Oct 7 2013, 16:09) *
Why does anyone use the TT DR Meter? Where is the algorithm specified? Why isn't it based on BS-1770?


there was a link in that thread at one point to a brief white paper describing the TT Meter's workings....not code though, if that;s what you mean.


I can upload it somewhere later but essentially (without formulas):

Measuring DR

Each channel of the audio signal is splitted into blknum blocks of 3 seconds length. The RMS
is calculated for each block as the square root of the double sum over all input samples
squared, divided by the block size in samples. The Peak is calculated as the maximum of the
absolute value from the block.

Each RMS and Peak value is entered into a histogram with 10,000 discrete values ranging
from -100 dB to 0 dB in steps of 0.01 dB.

The DR value for each channel j can then be derived as the difference between the second
largest Peak value and the RMS-sum over the upper 20% histogram values converted to dB.

The overall DR is finally calculated as the average of the channel DR values rounded to the
next integer value.

Using the RMS-sum in (4) results in the overall RMS of the upper 20% of the input material,
eliminating the contribution of small Peaks. This method also ensures that the resulting DR
value is virtually independent from the block size used (3s in this example) as long as this is
small compared to the overall input material length.

Limiting the DR-measurement to the upper 20% of the blocks with maximum RMS is a
compromise that allows to somewhat compare a wide variety of different material in a
quantitative way. Also in highly dynamic Material only the loudest parts, which usually best
reflect the processing of the material (compression etc.), contribute to the DR measurement.


This post has been edited by pisymbol: Oct 9 2013, 05:08
Go to the top of the page
+Quote Post
pisymbol
post Oct 9 2013, 05:28
Post #32





Group: Members
Posts: 43
Joined: 22-March 09
Member No.: 68274



QUOTE (extrabigmehdi @ Oct 8 2013, 15:01) *
My thought regarding the TT meter : ok , a big DR value isn't a warranty that the record has a good dynamic range.
However I think the reverse is true, a too low value (let's say below 6) indicate that the dynamics has been too much squashed.
So the TT meter is still doing it's job to some extent.


What I want to know is this...

If I have a CD at DR5/6 and a vinyl rip at DR11/12, can they really be the same master? And even more importantly, what exactly is throwing off the algorithm I posted above?
Go to the top of the page
+Quote Post
mjb2006
post Oct 9 2013, 09:20
Post #33





Group: Members
Posts: 706
Joined: 12-May 06
From: Colorado, USA
Member No.: 30694



QUOTE (pisymbol @ Oct 8 2013, 22:28) *
If I have a CD at DR5/6 and a vinyl rip at DR11/12, can they really be the same master?


Ian Shepherd's video is talking about a specific situation:

The same peak-limited master as was used for prepping the CD/digital releases was also given to the vinyl mastering house, where it was then subjected to an unknown amount of additional processing that's required for putting audio onto playable vinyl.

At the very least, presumably, the bass was made more mono, and bandpass filtering was applied to roll off frequencies outside of the 30 Hz to ~20 KHz range (maybe not even that high). Either of these things alone can increase the measured DR value, effectively acting as an expander, reintroducing a few dB of "peakiness" to the audio. So whether you consider the vinyl master to be the "same" depends on whether you're taking this into consideration.

And that's only talking about what audio goes onto the wax. When you rip it, you're introducing many more variables and slight changes to the audio via your particular playback and recording system, which can nudge the rip's peak-to-RMS ratio in one direction or the other. All together, a ~6 dB difference in the measured DR value probably isn't out of the question.

This post has been edited by mjb2006: Oct 9 2013, 09:21
Go to the top of the page
+Quote Post
pisymbol
post Oct 9 2013, 13:08
Post #34





Group: Members
Posts: 43
Joined: 22-March 09
Member No.: 68274



QUOTE (mjb2006 @ Oct 9 2013, 04:20) *
QUOTE (pisymbol @ Oct 8 2013, 22:28) *
If I have a CD at DR5/6 and a vinyl rip at DR11/12, can they really be the same master?


Ian Shepherd's video is talking about a specific situation:

The same peak-limited master as was used for prepping the CD/digital releases was also given to the vinyl mastering house, where it was then subjected to an unknown amount of additional processing that's required for putting audio onto playable vinyl.

At the very least, presumably, the bass was made more mono, and bandpass filtering was applied to roll off frequencies outside of the 30 Hz to ~20 KHz range (maybe not even that high). Either of these things alone can increase the measured DR value, effectively acting as an expander, reintroducing a few dB of "peakiness" to the audio. So whether you consider the vinyl master to be the "same" depends on whether you're taking this into consideration.

And that's only talking about what audio goes onto the wax. When you rip it, you're introducing many more variables and slight changes to the audio via your particular playback and recording system, which can nudge the rip's peak-to-RMS ratio in one direction or the other. All together, a ~6 dB difference in the measured DR value probably isn't out of the question.


It sure seems a lot to me given the algorithm. Most needle drops I've seen are around the same TT score (+-1dB).

I'm trying to understand WHAT is causing spurious peak variations to offset the DR?
Go to the top of the page
+Quote Post
2Bdecided
post Oct 9 2013, 13:24
Post #35


ReplayGain developer


Group: Developer
Posts: 4945
Joined: 5-November 01
From: Yorkshire, UK
Member No.: 409



QUOTE (pisymbol @ Oct 9 2013, 13:08) *
I'm trying to understand WHAT is causing spurious peak variations to offset the DR?
An all-pass filter makes the peak value of a clipressed digital master jump up by about 3dB, and doesn't change the RMS at all - hence the measured DR value "improves" by about 3dB.

A phono cartridge and complimentary RIAA filtering are (at best!) an all pass filter (plus some attenuation of the frequency extremes, plus complete blocking of DC, plus hopeless response at very high frequencies, plus distortion, plus noise etc etc ...).


The biggest problem with the DR meter isn't that it gets vinyl so "wrong" - it's that it measures a certain kind of numerical (and, in the waveform: visible) property without any regard for how audible and/or inappropriate to the content it is. Even that isn't really a "fault" with the meter - it's a fault with the way people sometimes unthinkingly use the results.

Put an all pass filter in the algorithm, and you'll solve the vinyl vs CD problem. However, this will then downplay the "badness" of clipression and clipping (despite the fact is can be audibly objectionable), while still penalising "normal" compression (which reduces the dynamic range without introducing clipping) as much as now.

I don't like to criticise because I can't think of a better way of doing it. In what it is (a numerical measurement of peak to average ratio), it's working fine. The problem is, it's not a reliable measure of unwanted distortion (which is what most people object to), it's only a half-decent measure of clipping if there's no post-processing, and it can have no concept of what a particular recording "should" sound like in terms of dynamic range.

Cheers,
David.

This post has been edited by 2Bdecided: Oct 9 2013, 13:41
Go to the top of the page
+Quote Post
pisymbol
post Oct 9 2013, 13:38
Post #36





Group: Members
Posts: 43
Joined: 22-March 09
Member No.: 68274



QUOTE (2Bdecided @ Oct 9 2013, 08:24) *
QUOTE (pisymbol @ Oct 9 2013, 13:08) *
I'm trying to understand WHAT is causing spurious peak variations to offset the DR?
An all-pass filter makes the peak value of a clipressed digital master jump up by about 3dB, and doesn't change the RMS at all - hence the measured DR value "improves" by about 3dB.

A phono cartridge and complimentary RIAA filtering are (at best!) an all pass filter (plus some attenuation of the frequency extremes, plus complete blocking of DC, plus hopeless response at very high frequencies, plus distortion, plus noise etc etc ...).

Cheers,
David.


But 6dB difference David? Is that typical?

My assumption has been that a 6+ dB difference is MOST LIKELY caused by an altogether different master, not RIAA correction or a particular cart's FR...etc.
Go to the top of the page
+Quote Post
markanini
post Oct 9 2013, 15:04
Post #37





Group: Members
Posts: 534
Joined: 22-December 03
From: Malmö, Sweden
Member No.: 10615



QUOTE (greynol @ Oct 9 2013, 00:02) *
Why is this metric even being given any credence in the first place?


Probably cause it's the first tool of it's kind in a time when loudness wars and dynamic range became a hot button issues to audiophiles.

I did some comparisons between TT DR results and BS1770 loudness range (using Hofa 4U Meter) on a few titles that I had in both CDDA and needle drop versions. First I noticed the BS1770 numbers didn't tell me that analogue recordings from the '70s had tons of dynamic range when I could subjectively tell there's moderate compression on both(probably from the master tape). Second on titles that sounded similar the numbers were identical or the vinyl reported one value smaller dynamic range, closer to what's expected. Just a small test but if it's enough to draw any conclusions it seems the BS1770 is more accurate.

With this in mind and since people will keep using these tools could there be any merit to implementing a BS 1770 loudness range version of foo_dynamic_range?

This post has been edited by markanini: Oct 9 2013, 15:06
Go to the top of the page
+Quote Post
ianshepherd
post Oct 9 2013, 22:02
Post #38





Group: Members
Posts: 2
Joined: 19-February 13
From: Cambridge, UK
Member No.: 106755



QUOTE (markanini @ Oct 9 2013, 14:04) *
With this in mind and since people will keep using these tools could there be any merit to implementing a BS 1770 loudness range version of foo_dynamic_range?

The equivalent value in R128 terms is the "peak to loudness ratio", or PLR.

We're experimenting with displaying it in my Perception plugin - it gives values that are similar to, but different from , the DR values.
Go to the top of the page
+Quote Post
greynol
post Oct 10 2013, 17:48
Post #39





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (DVDdoug @ Oct 8 2013, 14:45) *
up to 4dB "error" or uncertainty.

For giggles I took the CD version of the file found here and ran it through Audition's subsonic rumble filter with 6dB of attenuation to prevent clipping.

The DR value jumped from 5 to 10.

EDIT: I bumped the treble a little bit in an attempt to more closely match the vinyl version and the DR value increased to 11.

This post has been edited by greynol: Oct 10 2013, 18:22


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
skamp
post Oct 10 2013, 18:32
Post #40





Group: Developer
Posts: 1343
Joined: 4-May 04
From: France
Member No.: 13875



QUOTE (greynol @ Oct 10 2013, 18:48) *
ran it through Audition's subsonic rumble filter


What does it do?


--------------------
caudec.net
Go to the top of the page
+Quote Post
greynol
post Oct 10 2013, 18:40
Post #41





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



It's simply a steep high-pass filter intended to remove subsonic rumble. I couldn't find an all-pass filter in Audition so I chose one that wasn't going to alter the audible frequency response.

Hopefully google can answer any additional tangential questions.

This post has been edited by greynol: Oct 10 2013, 18:53


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
greynol
post Oct 11 2013, 01:23
Post #42





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (extrabigmehdi @ Oct 8 2013, 12:01) *
However I think the reverse is true, a too low value (let's say below 6) indicate that the dynamics has been too much squashed.

I wouldn't be so sure about this either:
http://www.hydrogenaudio.org/forums/index....howtopic=102963


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
Engelsstaub
post Oct 11 2013, 04:46
Post #43





Group: Members
Posts: 545
Joined: 16-February 10
Member No.: 78200



QUOTE (greynol @ Oct 10 2013, 11:48) *
QUOTE (DVDdoug @ Oct 8 2013, 14:45) *
up to 4dB "error" or uncertainty.

For giggles I took the CD version of the file found here and ran it through Audition's subsonic rumble filter with 6dB of attenuation to prevent clipping.

The DR value jumped from 5 to 10.

EDIT: I bumped the treble a little bit in an attempt to more closely match the vinyl version and the DR value increased to 11.


That seems fair and appropriate considering the source. That Ortofon OM5e was a terrible cartridge with an obvious treble-bias at the expense of midrange. I've since happily replaced it, along with the Debut III, with a Denon DL-110 on a slightly modded Technics SL 1200.

...not that I'll likely be contributing any new samples to that thread as I'm going to just admit that I was most likely hearing things on many of those examples. They really do sound different to me but probably not usually for the reasons I thought they did.


--------------------
The Loudness War is over. Now it's a hopeless occupation.
Go to the top of the page
+Quote Post
greynol
post Oct 11 2013, 05:41
Post #44





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



It should be no great revelation to anyone. This stuff has been discussed in our community for years now.
http://www.hydrogenaudio.org/forums/index....showtopic=80679

This post has been edited by greynol: Oct 11 2013, 05:45


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
user1
post Apr 3 2014, 07:18
Post #45





Group: Members
Posts: 33
Joined: 18-January 14
Member No.: 114096



QUOTE (greynol @ Oct 10 2013, 17:48) *
QUOTE (DVDdoug @ Oct 8 2013, 14:45) *
up to 4dB "error" or uncertainty.

For giggles I took the CD version of the file found here and ran it through Audition's subsonic rumble filter with 6dB of attenuation to prevent clipping.

The DR value jumped from 5 to 10.

EDIT: I bumped the treble a little bit in an attempt to more closely match the vinyl version and the DR value increased to 11.
A subsonic (rumble) filter may be built in, as an option (like RIAA filter tweaks), into the cutting-lathe console. It would be useful if a cutting engineer can comment on this.



Also note what one vinyl mastering/engineering service states in their Q&A:

QUOTE
Q: These days, many independent productions are completed at home in project studios. What can mastering offer to enhance the final sound of these projects?

A: Often home productions are recorded and mixed under less than perfect acoustical conditions. I can make educated decisions and offer options to the producer that will translate well to the real world. My experience allows me to make major or subtle changes, while working in conjunction with the artist, producer or label, which really can improve the final sound of any project.

Refs:
http://ekadek.com/?page_id=14
http://www.gearslutz.com/board/4684881-post21.html
Go to the top of the page
+Quote Post
user1
post Apr 3 2014, 10:44
Post #46





Group: Members
Posts: 33
Joined: 18-January 14
Member No.: 114096



Another thing I forgot to mention was this peculiar note from a 2009 technical article in Stereophile:
QUOTE
....Neumann's name became attached to the correction because its lathes—on which a great many of the world's vinyl records have been cut down the decades—have been reported to incorporate an additional record EQ time constant of 3.18µs (50.05kHz). As we will see, this isn't correct....

Via little oversights like this .... negligent, ignorant or otherwise unaware mastering/cutting engineers may be creating records with "unwanted"/distorted content.
Go to the top of the page
+Quote Post
extrabigmehdi
post Apr 3 2014, 12:03
Post #47





Group: Members
Posts: 401
Joined: 15-August 09
Member No.: 72330



I'm wondering if multiplying "%replaygain_track_gain%" and "%replaygain_track_peak%" datas ,
doesn't provide a more reliable information about the dynamic range, than with the TT Dynamic Range meter .
Go to the top of the page
+Quote Post

2 Pages V  < 1 2
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 20th April 2014 - 03:43