I've set up my 'dream system' comprising Spotify Premium on a laptop, a meaty low distortion amp, excellent speakers and a USB DAC. I'm convinced the hardware is all working well. I'm just having difficulty with (my perception of) the audio quality of the source...
Spotify Premium allows the listener to select a 'high bitrate' option. Reputedly it is Ogg Vorbis q9 at 320 kbps as described on Spotify's web site:
What quality does Spotify stream in? - Spotify
"We use the Ogg Vorbis format for streaming. There are three quality ratings that we use:
* q3 (~96 kbps)
* q5 (~160 kbps)
* q9 (~320 kbps)
Note: not all tracks are currently available in high bitrate. "
Hmm. All well and good, but I notice the link talks about the quality of the streaming, not necessarily about the quality of the source. How are the tracks encoded when the record company makes them available to Spotify? (It may seem a silly question, but there isn't any chance that the music is being encoded twice is there?)
I'm not enjoying listening to classical music on Spotify at all. Everything is fine until a piece builds into a section with lots of strings, brass and/or woodwind playing. I'm then hearing very unpleasant distortion.
Now, I realise that that is precisely the effect that audio compression might be expected to have; the most 'complex' sounds suffering from a paucity of bits. It's just that I hoped that at 320 kbps it wouldn't be audible. I've been taking a distinct interest in the same sorts of music on FM Radio 3 in the car in the morning, and there I just don't hear the same unpleasantness - it sounds 'transparent'.
If I had to bet, it would be that the classical stuff is encoded at 160 kbps and that I shouldn't hold my breath for it to ever be 'upgraded'...
Spotify Premium allows the listener to select a 'high bitrate' option. Reputedly it is Ogg Vorbis q9 at 320 kbps as described on Spotify's web site:
What quality does Spotify stream in? - Spotify
"We use the Ogg Vorbis format for streaming. There are three quality ratings that we use:
* q3 (~96 kbps)
* q5 (~160 kbps)
* q9 (~320 kbps)
Note: not all tracks are currently available in high bitrate. "
Hmm. All well and good, but I notice the link talks about the quality of the streaming, not necessarily about the quality of the source. How are the tracks encoded when the record company makes them available to Spotify? (It may seem a silly question, but there isn't any chance that the music is being encoded twice is there?)
I'm not enjoying listening to classical music on Spotify at all. Everything is fine until a piece builds into a section with lots of strings, brass and/or woodwind playing. I'm then hearing very unpleasant distortion.
Now, I realise that that is precisely the effect that audio compression might be expected to have; the most 'complex' sounds suffering from a paucity of bits. It's just that I hoped that at 320 kbps it wouldn't be audible. I've been taking a distinct interest in the same sorts of music on FM Radio 3 in the car in the morning, and there I just don't hear the same unpleasantness - it sounds 'transparent'.
If I had to bet, it would be that the classical stuff is encoded at 160 kbps and that I shouldn't hold my breath for it to ever be 'upgraded'...
Last edited:
320 kbps OGG is good, I personally can in no way distinguish that from a good lossless FLAC.
Personally I don't use spotify but my mom does and I've noticed that far from all spotify track recordings are good, some are downright awful and then better bitrate won't help you at all.
The difference between 160 kbps and 320 kbps ogg is in my opinion actually quite small, the difference is not in distortion but that you can hear slightly more instruments at the same time and more depth.
My best recommendation would be to take a lossless track you have with lots of detail and instruments and use a free compressing tool and then compress and compare that track at say q9, q5 and q3 and then listen if and where you hear differences.
If you sit in ubuntu SoundConverter is a good tool to use for the convertion.
Personally I don't use spotify but my mom does and I've noticed that far from all spotify track recordings are good, some are downright awful and then better bitrate won't help you at all.
The difference between 160 kbps and 320 kbps ogg is in my opinion actually quite small, the difference is not in distortion but that you can hear slightly more instruments at the same time and more depth.
My best recommendation would be to take a lossless track you have with lots of detail and instruments and use a free compressing tool and then compress and compare that track at say q9, q5 and q3 and then listen if and where you hear differences.
If you sit in ubuntu SoundConverter is a good tool to use for the convertion.
Maybe I am doing Spotify a disservice. I am beginning to wonder if what I am hearing is the result of standard orchestration combined with hyper-critical listening and expecting the worst from my audio system!
When a composer wants to make the orchestra 'louder' it seems he usually adds instruments as a general background wall of sound, as well as getting the already-playing instruments to play 'louder'. It occurs to me that if I am listening intently to the violins and they begin to play a bit louder, but in the background the composer adds a wall of brass and woodwind playing the same notes, it may be that I am being tricked into believing that the violins have changed their 'tone' as they have got louder. The sound of a clarinet, for example, contains high levels of odd harmonics (it can be imitated quite well with a synthesiser using a square wave) so if I blend it with a violin, it may begin to sound as though the violin is beginning to clip.
Not the sort of thing one would notice at a 'live' concert, but when listening critically to an audio system...
In other forms of music, the volume often changes by sliding a knob on a mixing desk, rather than adding more instruments, so the effect is different. Maybe I am just too used to this.
When a composer wants to make the orchestra 'louder' it seems he usually adds instruments as a general background wall of sound, as well as getting the already-playing instruments to play 'louder'. It occurs to me that if I am listening intently to the violins and they begin to play a bit louder, but in the background the composer adds a wall of brass and woodwind playing the same notes, it may be that I am being tricked into believing that the violins have changed their 'tone' as they have got louder. The sound of a clarinet, for example, contains high levels of odd harmonics (it can be imitated quite well with a synthesiser using a square wave) so if I blend it with a violin, it may begin to sound as though the violin is beginning to clip.
Not the sort of thing one would notice at a 'live' concert, but when listening critically to an audio system...
In other forms of music, the volume often changes by sliding a knob on a mixing desk, rather than adding more instruments, so the effect is different. Maybe I am just too used to this.
Not quite. Remember that 1440 can be compressed down to about 720K without loss. So 320K is not such a big step from there. Not ideal, but not awful.
If you're referring to a lossless codec, the amount that it can be reduced is wholly dependent on how much unique information exists on the original. I have Apple Lossless files that compress hardly at all, while some compresses very highly. You can't say that 720K is a standard value "without loss".
Most compact 16/44 ALC file: Ethel Waters: Sentimental Jazz: 3:07 238 kbps
Least compressed 16/44 ALC file: The Beatles: I Wanna Hold Your Hand: UK Single: 2.27 1164 kbps
If I create a smart folder in iTunes and select sample rate of 44.1 and Bitrate above 1000 kbps I have 1400+ songs out of a library of 11700 songs, ie 1 in every 8 songs, that cannot be losslessly compressed below 1000 kbps.
If I change the criteria to 44.1 and above 721 kbps, 9918 songs, or 84% of my library, cannot be compressed to 720 kbps without loss.
There is little variance amongst lossless codecs regarding how much compression is available .... after all they have to keep the same amount of unique information, so a FLAC or other similar codec won't be significantly different.
Interestingly, classical music in my library does compress well; 850~650 kbps is common. However these are average bitrates per second over the entire track, and I would expect complex passages to be significantly higher than relatively quiet passages, which themselves are far more common in Classical music than other genres. If you are subject to a strict bitrate limit, as a streaming service is likely to be, I would be far from surprised to learn complex passages sound congested and lacking in dynamics and detail.
Last edited:
I use Pandora|One (192kbps) service and I have to go in on occasion when the music is sounding flat, and adjust the setting upward to high rates. It seems that when bandwidth is limited, it automatically sets itself to a lower level. Even at the higher settings, I still like my Apple Lossless files better. Pandora sounds good, but the Lossless files sound great.
Paul
Paul
A bit of an OT, to test the hypothesis that things at -20dB compress better, I tried 3 files converted to FLAC in Audacity, one untouched version and one amplified -20dB. All 3 -20dB-amplified files had filesizes of ~70±2% compared to the untouched FLAC encodes.
Admittedly OT, and we should respect the OP and keep it on topic, so I won't be adding to this sub-topic after this post.
Im a little confused as to what you were trying to show (and what your results were) but my point was simply that Classical music is not often pegged at 0dB FS like pop music is, so it should compress better during the less than complex passages, but not during the complex ones.
" ... What is the lowest bitrate (or highest compression) achievable with FLAC?
With FLAC you do not specify a bitrate like with some lossy codecs. It's more like specifying a quality with Vorbis or MPC, except with FLAC the quality is always "lossless" and the resulting bitrate is roughly proportional to the amount of information in the original signal. You cannot control the bitrate much and the result can be from around 100% of the input rate (if you are encoding noise), down to almost 0 (encoding silence). ..."
-Soundforge FLAC specification
Lossless codecs will use a constant bitrate within a frame, but vary the bitrate from frame to frame based on the information complexity in the original. So, the bitrate reported for a track will be an average over time.
Also, see http://members.home.nl/w.speek/tables.htm to get an overview of how various lossless codecs compare. Short answer ... they all will generate a similar level of compression in comparison to each other based on the complexity of the original. Typically the median +/- 10%; eg 45~55% of the original if the median is 50%, or 54~66% if the median is 60%, etc
This would be in contrast to a lossy codec like that used in Spotify where the music is "squeezed into" (for lack of a better term) the fixed bitrate [ignoring for the moment that Variable Bitrate (VBR) lossy codecs do exist] regardless of the complexity of the original.
Now, I'm going to respect the OP and try not to derail the thread, so that's all I will say on the subject. Cheers.
Last edited:
One point expressed in this thread is that "complex" music requires a higher bitrate. In my experience, this is not generally correct when using a perceptual encoder such as mp3 or Ogg. I wrote a tool to display the bitrate when playing mp3 files and played a variety of VBR tracks. At a given quality setting, loud / complex music didn't require significantly higher bitrates than "quiet / simple" music. On the other hand, relatively "soft" and inaudible transients - light percussion, the beginning of piano notes - invariably spiked the bitrate to 320 KBPS. This result is somewhat counter-intuitive, but when considering the way perceptual coding works - it discards content that is "masked" or hidden by other content - it makes sense.
So I agree with OllBoll:
- Find a track that sounds bad in Spotify.
- Obtain a CD or lossless copy of it and confirm that it sounds OK lossless.
- Compress it to the 3 different rates with Ogg and see if any of them match what you hear from Spotify.
I wouldn't expect Spotify to increase the bitrate of a low-bitrate original. There exists the possibility that the original might have been supplied to Spotify in MP3 format and they then transcribed it to a similar quality bitrate Ogg format. For example, a 128Kbps MP3 mithght have been transcribed to a 96 kbps Ogg format.
So I agree with OllBoll:
- Find a track that sounds bad in Spotify.
- Obtain a CD or lossless copy of it and confirm that it sounds OK lossless.
- Compress it to the 3 different rates with Ogg and see if any of them match what you hear from Spotify.
I wouldn't expect Spotify to increase the bitrate of a low-bitrate original. There exists the possibility that the original might have been supplied to Spotify in MP3 format and they then transcribed it to a similar quality bitrate Ogg format. For example, a 128Kbps MP3 mithght have been transcribed to a 96 kbps Ogg format.
Most of what I hear on Spotify sounds pretty good.
I have the normal bitrate Spotify stream. Yesterday a friend and I compared a CD directly to the Spotify version. My remarks are in this thread:
http://www.diyaudio.com/forums/pc-based/192590-spotify-usa-2.html#post2688896
I have the normal bitrate Spotify stream. Yesterday a friend and I compared a CD directly to the Spotify version. My remarks are in this thread:
http://www.diyaudio.com/forums/pc-based/192590-spotify-usa-2.html#post2688896
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.