Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: I think Spotify premium has pre-EQ'd all the songs (Read 8626 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

I think Spotify premium has pre-EQ'd all the songs

So recently I signed up for Spotify Premium because I was sick of buying CD's and just wanted to listen to anything I want. I thought it was pretty great for a while, until I accidentally did a Spotify vs CD comparison. The Spotify's songs are noticeably "clearer". It doesn't sound like a straight treble boost, but kind of like a noise sharpener. I've tried it with a few other CD's and the result is the same.

AFAIK, there is no EQ option on the desktop version of Spotify. Has anyone else noticed this on Spotify's songs?


Re: I think Spotify premium has pre-EQ'd all the songs

Reply #2
If you have Spotify, try "Lean On" by Major Lazer against the YouTube or CD version. The Spotify version has noticeably "clearer" and airier vocals than the CD or YouTube version, but the CD version has fuller bass and more low vocal register weight than Spotify. There is no way this is because it is MP3 as I cannot hear the difference between 192Kbps and CD, there is something fundamentally different.

For a very specific and noticeable example, jump to 2:11, the first 3 bass note is accompanied by distortion that sounds like sand shaking on Spotify. This is nearly inaudible on YouTube or the CD.

In fact, try it with any song. I hear the same thing on everything I've tried so far.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #3
Spotify uses Vorbis, not mp3.  Regardless, you're right in that lossy compression isn't responsible for the differences you hear.

Regarding your choice in language regarding ascribed differences, please refer to the rules you agreed to follow when registering, specifically #8.  In the spirit of rule #8, you should provide samples of each version you hear which are no longer than 30 seconds so others can have any hope of helping you beyond sharing anecdotes which do not serve the purpose of progressing a discussion.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #4
Spotify uses Vorbis, not mp3.  Regardless, you're right in that lossy compression isn't responsible for the differences you hear.

Regarding your choice in language regarding ascribed differences, please refer to the rules you agreed to follow when registering, specifically #8.  In the spirit of rule #8, you should provide samples of each version you hear which are no longer than 30 seconds so others can have any hope of helping you beyond sharing anecdotes which do not serve the purpose of progressing a discussion.

I would, but I have no way of getting the Spotify version of the song. Any advice on ways to record audio from the sound card to WAV?

Plus, wouldn't it be easier for users to pick any song they're familiar with on Spotify and compare it to their CD? It should be fairly obvious in a A/B comparison? This way, they can hear it with any number of songs they're familiar with.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #5
It should be fairly obvious in a A/B comparison?
Perhaps not as obvious as you think. There are a lot of things that can go wrong.

 If you are playing the tracks with two different pieces of software then one of them could be applying some modifier that the other isn't.

Most commonly they are not level matched (levels must be matched very very closely to make a proper comparison).

I won't even go into the problems of sighted comparisons.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #6
wouldn't it be easier for users to pick any song they're familiar with on Spotify and compare it to their CD?
You're essentially asking that Spotify massively increase their overhead by purposely keeping multiple pressings.  Besides that, there could be additional financial and legal complications arising from the ownership of distribution rights.  For example, what happens when Spotify wants to make a version available that was sourced from a master that was neither created nor possessed by the label(s) who currently own the rights and are already distributing content sourced from a different master?  I'm not even sure Spotify will even be able to obtain your personal pressing through practical means.  This idea seems highly unreasonable to me.

Anyway, I took the time to capture a sample.  I cannot attest to the precision in level matching, however It is very clear to me that the differences between this and the original do not corroborate your suspicion as you have expressed it.

First paragraph was edited for clarity.


Re: I think Spotify premium has pre-EQ'd all the songs

Reply #7
I don't hear a difference there.  I didn't ABX it, just shut my eyes.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #8
The residual from subtracting the two only sounds like error from lossy encoding to me.  I see absolutely no evidence of any change in audible timbre below 16kHz (not that I am claiming to hear any change above 16kHz), dynamics or anything else not related to lossy encoding.  I'm not hip to all the meaningless audiophile speak, but I'm pretty sure audible differences in "air" can still exist a few kHz below 16k at the very least.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #9
I didn't ABX it, just shut my eyes.
Even if perfectly level-matched, I wouldn't at all be surprised to hear that ABXing these samples was quite easy; though, again, it would only be due to the listener's ability to detect lossy artifacts.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #10
wouldn't it be easier for users to pick any song they're familiar with on Spotify and compare it to their CD?
You're essentially asking that Spotify massively increase their overhead by purposely keeping multiple pressings.  Besides that, there could be additional financial and legal complications arising from the ownership of distribution rights.  For example, what happens when you choose to make a version available that was sourced by a master that wasn't created by the label(s) who own the rights which have a different master?  I'm not even sure Spotify will even be able to acquire your personal pressing.  This idea seems highly unreasonable to me.

Anyway, I took the time to capture a sample.  I cannot attest to the precision in level matching, however It is very clear to me that the differences between this and the original do not corroborate your suspicion as you have expressed it.



How did you record the Spotify? I would like to make the same thing and upload it.

Your upload sounds identical to me.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #11
I thought the same thing on the free Spotify but never really gave it further attention. Free Spotify uses vorbis 160k and premium is said to be 320k .  Vorbis v1.0 until certain aotuv versions had a HF boost bug and wasn't hard to abx even @ Q6 ~ 192k. To make it worse, the clowns at Xiph never fixed or gave attention to the damn thing (or any other quality matter). They did eventually merge a version of aotuv (vorbis1.1) and I hope that fixed the issue or reduced it??. Though unlikely , The old pre-1.1 encoders may still be in use.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #12
I see absolutely no evidence of any change
I guest DJ'd at a commercial station in the early '00s, ran my own station for years, and used to participate in forums for hobbyist and small webcasters, and it surprises me that Spotify apparently isn't doing anything special. Hobbyists vary widely in practices, and maybe things have changed in the satellite radio and smartphone era, but it has been pretty common in commercial online radio since the early days to sweeten the outbound sound, just like they do in terrestrial radio. Sometimes it's just for volume normalization (via dynamic range compression), sometimes it's EQ and stereo expansion to make your station sound as "alive" (and crappy earbud-friendly?) as the competition, the competition being other stations and the sound of modern mastering on CDs/downloads. Also, if there are knobs to twiddle, the temptation is too much to resist. Everyone likes to think they can improve on whatever they're starting with.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #13
I thought the same thing on the free Spotify but never really gave it further attention.
Update: As it turns out, because I made my capture from Spotify's web player, it was of the ~160kbit variety.  Inspection should show that it was not subjected the bug reported in the post linked from the quote above.

Not that it has much relevance in this context, I was initially I was under the impression that my capture was from Spotify's premium content because I was using a premium account.

This was my original reply to the above quote:
Quote
This stirred my curiosity.
The earlier clip was from Spotify premium* (per the subject line), but I went ahead and captured the same thing using Spotify free.

I won't go so far as to say that none of their other free tracks are now or ever were encoded using a different codec from that used to create this one.

There was also an attached file, but it was removed because it was redundant.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #14
OK I finally found a way to record what I hear on Spotify for the song Lean On. I did a ABX blind test and I got it 10/10. I've attached my logs. Can someone listen to both and tell me what they hear?

Thanks

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #15
I ran them through Masvis. Here's the output: (it's the normalized average spectrum that shows the EQ)

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #16
For the most part, the Spotify graph looks like what I'd expect to see if the CD were run through lossy coding and the volume dropped,

But just doing some A/B listening with ReplayGain applied, I easily hear what you are talking about. The "we need someone to lean on" vocal, especially, is relatively muted on the CD, and "clearer" via Spotify.

Does someone have an ancient version of oggenc to try, to see if it's the bug mentioned by shadowking?

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #17
I ABXed CD.flac and Spotify.flac, listening to the beat that comes in at 0:09. In Spotify.flac I heard less bass there.

Code: [Select]
foo_abx 2.0.1 report
foobar2000 v1.3.9
2016-07-23 16:26:43

File A: CD.flac
SHA1: 3ee1beef1f26e475fa9ff62ffd2e16540dc5fab4
Gain adjustment: -6.65 dB
File B: Spotify.flac
SHA1: 6729631d60b5fc164915592e386debe9681271c2
Gain adjustment: -4.56 dB

Output:
DS : Primary Sound Driver, 16-bit
Crossfading: NO

16:26:43 : Test started.
16:28:05 : 01/01
16:28:20 : 02/02
16:28:31 : 03/03
16:28:39 : 04/04
16:28:47 : 05/05
16:29:15 : 06/06
16:29:25 : 07/07
16:29:47 : 08/08
16:30:19 : 09/09
16:30:34 : 10/10
16:31:01 : 11/11
16:31:07 : 12/12
16:31:29 : 13/13
16:31:40 : 14/14
16:31:49 : 15/15
16:31:57 : 16/16
16:31:57 : Test finished.

 ----------
Total: 16/16
Probability that you were guessing: 0.0%

 -- signature --
92a3d4c9489bd427bd56a5d6187dd2dbe20126d0

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #18
Be careful with RG adjustments when it comes to level matching of different sources, especially if (IIRC) R-128 doesn't model human HF response as well as the original algorithm and from what I can tell, the difference here is in the treble.

After looking into it I found a couple different masters of the original, non-remixed version of this track each with varying amounts of treble, though I didn't find one that I could confidently say was used for Spotify.  However, I'm sticking to my guns: differences in mastering rather than encoding bugs or secret sauce.

That Spotify premium would use an old codec with known issues at low bitrates (if they ever did) for their premium content on a 2015 release is beyond me.  I warned about anecdotes earlier, and think this very much applies here

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #19
For the most part, the Spotify graph looks like what I'd expect to see if the CD were run through lossy coding and the volume dropped,

But just doing some A/B listening with ReplayGain applied, I easily hear what you are talking about. The "we need someone to lean on" vocal, especially, is relatively muted on the CD, and "clearer" via Spotify.

Does someone have an ancient version of oggenc to try, to see if it's the bug mentioned by shadowking?

Theres the GT3b1 version on rarewares. It is modified for Q5 ~ 10 for better pre echo handling and the bitrate higher. I believe the HF issue was there at the time.

http://www.rarewares.org/files/ogg/oggenc2.3-GT3b1.zip

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #20
OK I finally found a way to record what I hear on Spotify for the song Lean On. I did a ABX blind test and I got it 10/10. I've attached my logs. Can someone listen to both and tell me what they hear?

I hear and see two files that are probably the same artistic work but technically very different - not level or time matched. Of course you got 10 out of 10 different!  

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #21
I noticed this as well and then I took the time to correct for the discrepancy.  If you had done the same you might have found the same thing that I and others did.  Perhaps you could put forth a bit of additional effort* before criticizing next time.

(*) at which point you may have noticed a difference in timbre as you attempted to adjust for any perceived difference in level.

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #22
OK I finally found a way to record what I hear on Spotify for the song Lean

Measurements show, that once the timing and level differences are adjusted out:

* the CD has about 2.7 dB more bass at 110 Hz and below.

* there are some differences in dynamics processing as their peak-to-average sums are different.  The CD version is a bit more dynamic.


Re: I think Spotify premium has pre-EQ'd all the songs

Reply #23
Cool. That you measured a difference in bass definitely draws some attention to the crude method I used to level match.  I probably should have just used replaygain. :/

Re: I think Spotify premium has pre-EQ'd all the songs

Reply #24
Theres the GT3b1 version on rarewares.
Spotify.flac still sounds "brighter" than any .ogg I can produce with this version of oggenc2; I hear no treble boost with -b160, -q4.5, -b320, or -q8.3 (those q values producing approximately 160 and 320 kbps, respectively). I would still like to hear the boost that the buggy versions made, but I stand by my initial guess that Spotify is just sweetening the sound on their end, especially now that it has been determined there's a measurable bass difference as well.

(I thought I noticed the bass difference, but not consistently enough to mention it. Everyone thinks they hear bass differences where there are none, so I don't remark on it even to myself unless there's no question about it.)