Raising sound levels

[Liffy99]

Hope this is in the right forum.

as would be expected, my music plays back at very different volume levels depending on the album. I know I can apply 'replay gain' to intelligently equalise this to a large extent, but in my experience this generally reduces levels to a more uniform setting.

would I be better increasing the levels of many tracks ( taking care to avoid clipping) instead ? Is this the same as normalisation ? Would something like Audacity be the right software to use ?

Thanks

[bdiament]

Hi Liffy,

 

If you listen primarily to modern popular music, most recordings of this tend to be dynamically compressed, making the difference between the loudest sounds and the softest sounds, smaller than it is when the music is actually performed. The end result is the average level of these recordings is higher (sometimes much higher) than recordings where the differences between loud and soft are left natural.

 

The recordings where these differences (the dynamics) are left natural will tend to have a lower average volume. However, if those recordings are properly mastered, their loudest peaks will be as loud as can be without adding distortion. So if you raise the level on these, you will likely be driving them into distortion.

 

Much better to drop the level on those recordings that are dynamically compressed and hence, sound louder.

This will cause you to advance your playback volume control and that is a good thing. Your system will likely perform better with that control "opened up" a bit.

 

My best suggestion is to experiment with a few tracks and listen to the results.

 

Normalization is something else and is generally used (in my opinion MISused) to raise the level. Audacity can be used to raise or lower the level but this is best done by listening and adjusting the level accordingly (usually dropping louder tracks, not raising apparently lower tracks).

 

The thing to watch out for is to not be mislead into thinking those louder, dynamically compressed tracks are the "normal" ones and others are "too low". My experience has been the reality is that it is the "loud" tracks that are in fact "too loud". But try this out for yourself to find out how *you* hear it.

 

Hope this helps.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[Liffy99]

Thanks Barry

I appreciate your comments, esp. around the whole 'Loudness Wars' issue. But I'll explain in more detail what I'm thinking of - which is partly driven by a psychological reaction to using my stereo gear.

Like everyone else, music files play at different levels and while some have high average levels many others (like vinyl rips) tend to have lower average, and peak, levels. Rather than using Replay Gain across the board which usually seems to reduce volume levels to a lowest common denominator, I thought I'd investigate the possibility of raising some of the track levels (OK, a lot of them as I've over 11,000 in total). Of course I can simply advance the playback volume control but this is where it 'feels' strange. Let me explain.

 

In the good old days of my analogue system the volume control usually resided at about 11-12 o'clock (assuming the control went from say 8 to 4 on a clock face) for normal listening. This always gave the 'feeling' that the amp was cruising and there was plenty more to come of needed. This was probably true although I expect that clipping may start to creep in long before listeners would normally expect (eg around 2.30 on that clockface analogy). Anyway, it just felt right. Now I have all digital amplification (Lyngdorf) with a logarithmic volume scale. Full output is nominally at 88dba and my normal listening levels about 20db below this at high 60s. In other words, 1/100 of full output. So all is fine but it just 'feels' strange to be running the amp at a 'perceived' 3/4 (68 out of 88) of full whack. So this started me thinking of, essentially, providing a higher volume source and going some way to equalising the differences between music files where needed.

 

Anyway, I've used the Amplify effect of Audacity to try this with a couple of tracks successfully. Some have allowed up to 3db more headroom. However the big problem comes when looking to Amplify a whole album, where track volume levels need to be kept in relation to each other (esp. in the case of a classical symphony and its movements for example). The 'quiet' tracks allow plenty of scope of course whereas the 'louder' ones don't. For example one movement may allow an increase of just 0.5db whereas another as much as 5db. I need to maintain the relationship between the two movements so I can only really Amplify both by 0.5db. As far as I can tell, Audacity will only let me handle one stereo track at a time which means I have to first analyse each track to identify the maximum headroom available, and then Amplify and Export each track individually. I could be here for some time !

 

Any ideas / comments welcome but perhaps I'll see this as a gradual labour of love over the next couple of years !

[bdiament]

Hi Liffy,

 

Thanks Barry

I appreciate your comments, esp. around the whole 'Loudness Wars' issue. But I'll explain in more detail what I'm thinking of - which is partly driven by a psychological reaction to using my stereo gear.

Like everyone else, music files play at different levels and while some have high average levels many others (like vinyl rips) tend to have lower average, and peak, levels. Rather than using Replay Gain across the board which usually seems to reduce volume levels to a lowest common denominator, I thought I'd investigate the possibility of raising some of the track levels (OK, a lot of them as I've over 11,000 in total). Of course I can simply advance the playback volume control but this is where it 'feels' strange. Let me explain.

 

In the good old days of my analogue system the volume control usually resided at about 11-12 o'clock (assuming the control went from say 8 to 4 on a clock face) for normal listening. This always gave the 'feeling' that the amp was cruising and there was plenty more to come of needed. This was probably true although I expect that clipping may start to creep in long before listeners would normally expect (eg around 2.30 on that clockface analogy). Anyway, it just felt right. Now I have all digital amplification (Lyngdorf) with a logarithmic volume scale. Full output is nominally at 88dba and my normal listening levels about 20db below this at high 60s. In other words, 1/100 of full output. So all is fine but it just 'feels' strange to be running the amp at a 'perceived' 3/4 (68 out of 88) of full whack. So this started me thinking of, essentially, providing a higher volume source and going some way to equalising the differences between music files where needed.

 

Anyway, I've used the Amplify effect of Audacity to try this with a couple of tracks successfully. Some have allowed up to 3db more headroom. However the big problem comes when looking to Amplify a whole album, where track volume levels need to be kept in relation to each other (esp. in the case of a classical symphony and its movements for example). The 'quiet' tracks allow plenty of scope of course whereas the 'louder' ones don't. For example one movement may allow an increase of just 0.5db whereas another as much as 5db. I need to maintain the relationship between the two movements so I can only really Amplify both by 0.5db. As far as I can tell, Audacity will only let me handle one stereo track at a time which means I have to first analyse each track to identify the maximum headroom available, and then Amplify and Export each track individually. I could be here for some time !

 

Any ideas / comments welcome but perhaps I'll see this as a gradual labour of love over the next couple of years !

 

It is a call you will make for yourself of course but I would suggest the following considerations:

 

Change your perception of the volume control and see it without obfuscation. For example, at no level is the volume control itself responsible for clipping. Clipping results from inadequate amplifier power (or an overload in the recording).

 

Think of the faucet in your kitchen sink. It does not increase the flow of water, it *decreases* it. Remove the faucet and does the water stop? Of course not -- the water will flow at full force. Similarly, the volume control is used to *decrease* the playback volume. The lower the playback volume, the more of the control that is in the signal path.

 

The only exception is something like a ladder volume control using discrete resistors. If the control is a variable resistor type, as most are, you use "more" of it as volume is turned down. With digital volume control, you are losing word length as the volume is turned down. (Some will use digital processing to mitigate the effects.)

 

If you could start with a recording played at a low enough level to bypass the volume control entirely, the audible result is a revelation. (I experimented with this several years ago.) Getting rid of the volume control gets rid of one of the major weak links in the electronic chain. The sound just opens up in every area I know how to describe -- it is like removing a thick layer of dirt from a window.

 

So my best recommendation is to try and get past any misapprehension you have about the playback volume setting. If it achieves sufficient level without clipping, there is (in my opinion) no problem at all.

 

The alternative is to learn mastering skills and re-master all of your own recordings - in the process risking any sonic damage that may be incurred along the way. (Of course, if you want to invest the time and energy, you can always work with copies, so the originals remain intact as a safety.)

 

Also, I tried to keep things simple in my first post but it should be realized that when you perform a process on a digital file, *any* process, even raising the level 1/1000th dB (!), you are increasing the word length of that file. If the "container" can not accommodate the added length, you lose low order bits (low level information). Translated into English, if you process a 16-bit file, you will end up with something less than 16-bits. So, prior to any process, the word length of the file must be increased. For a 16-bit target, 24-bits is a reasonable minimum for processing.

 

Much consumer software (and some pro software) will process internally at 32-bits. That sounds like a good thing on the surface. (The best pro software will process internally at 64-bits and even 80-bits.) They key is that much software, especially consumer applications (but some pro software too) will save its temporary files at the word length of the source file. If you start with a 16-bit file, you will invariably low information (low level detail, spatial cues, etc.), even if all you are doing is the tiniest level change.

 

Therefore, I suggest starting any process by copying the source file and creating a longer word length version. For example, do a Save As on a 16-bit source file and save it as a 24-bit (or longer) file. This will make the "container" large enough to accommodate any processing. Perform any processing on this longer word length copy.

 

But wait, as they say, there's more. These longer word length copies which you've processed will take up appreciably more storage space, being at least 50% larger than the source files. If you want to use them as 16-bit files again, after all processing is complete, the very last step will be to reduce their word length back to 16-bits. To do this without throwing away all the information in the low order bits, you need a good dither algorithm, perhaps one with noise shaping. In my experience, most dither/noise shaping algorithms tend to alter instrumental timbres and cloud the soundstage. Only the best seem to create results that sound much like the longer word length sources.

 

So doing the volume adjustments without incurring losses in quality involves a bit more than simply changing the level.

In fact, the most transparent way to do this involves adding to the word length of the source file, converting the sample rate of the source file upward (especially if it is only 44.1k), processing, converting the sample rate back to 44.1 (if desired) and dithering/noise shaping while reducing the word length (if desired). Now most sample rate conversion algorithms in my experience will brighten and harden timbres. Only the very best avoid this.

 

Of course, you could also just leave the processed files at the longer word length (and perhaps higher sample rate) -- if you don't mind the larger files.

 

Just some things to consider. As always, my best suggestion is to *not* simply take my word for any of this but instead, to try things out for yourself. Do some experiments to find out how things sound to *you*.

Have fun!

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[Liffy99]

Hi Liffy,

 

 

 

It is a call you will make for yourself of course but I would suggest the following considerations:

 

Change your perception of the volume control and see it without obfuscation. For example, at no level is the volume control itself responsible for clipping. Clipping results from inadequate amplifier power (or an overload in the recording).

 

Think of the faucet in your kitchen sink. It does not increase the flow of water, it *decreases* it. Remove the faucet and does the water stop? Of course not -- the water will flow at full force. Similarly, the volume control is used to *decrease* the playback volume. The lower the playback volume, the more of the control that is in the signal path.

 

The only exception is something like a ladder volume control using discrete resistors. If the control is a variable resistor type, as most are, you use "more" of it as volume is turned down. With digital volume control, you are losing word length as the volume is turned down. (Some will use digital processing to mitigate the effects.)

 

If you could start with a recording played at a low enough level to bypass the volume control entirely, the audible result is a revelation. (I experimented with this several years ago.) Getting rid of the volume control gets rid of one of the major weak links in the electronic chain. The sound just opens up in every area I know how to describe -- it is like removing a thick layer of dirt from a window.

 

So my best recommendation is to try and get past any misapprehension you have about the playback volume setting. If it achieves sufficient level without clipping, there is (in my opinion) no problem at all.

 

The alternative is to learn mastering skills and re-master all of your own recordings - in the process risking any sonic damage that may be incurred along the way. (Of course, if you want to invest the time and energy, you can always work with copies, so the originals remain intact as a safety.)

 

Also, I tried to keep things simple in my first post but it should be realized that when you perform a process on a digital file, *any* process, even raising the level 1/1000th dB (!), you are increasing the word length of that file. If the "container" can not accommodate the added length, you lose low order bits (low level information). Translated into English, if you process a 16-bit file, you will end up with something less than 16-bits. So, prior to any process, the word length of the file must be increased. For a 16-bit target, 24-bits is a reasonable minimum for processing.

 

Much consumer software (and some pro software) will process internally at 32-bits. That sounds like a good thing on the surface. (The best pro software will process internally at 64-bits and even 80-bits.) They key is that much software, especially consumer applications (but some pro software too) will save its temporary files at the word length of the source file. If you start with a 16-bit file, you will invariably low information (low level detail, spatial cues, etc.), even if all you are doing is the tiniest level change.

 

Therefore, I suggest starting any process by copying the source file and creating a longer word length version. For example, do a Save As on a 16-bit source file and save it as a 24-bit (or longer) file. This will make the "container" large enough to accommodate any processing. Perform any processing on this longer word length copy.

 

But wait, as they say, there's more. These longer word length copies which you've processed will take up appreciably more storage space, being at least 50% larger than the source files. If you want to use them as 16-bit files again, after all processing is complete, the very last step will be to reduce their word length back to 16-bits. To do this without throwing away all the information in the low order bits, you need a good dither algorithm, perhaps one with noise shaping. In my experience, most dither/noise shaping algorithms tend to alter instrumental timbres and cloud the soundstage. Only the best seem to create results that sound much like the longer word length sources.

 

So doing the volume adjustments without incurring losses in quality involves a bit more than simply changing the level.

In fact, the most transparent way to do this involves adding to the word length of the source file, converting the sample rate of the source file upward (especially if it is only 44.1k), processing, converting the sample rate back to 44.1 (if desired) and dithering/noise shaping while reducing the word length (if desired). Now most sample rate conversion algorithms in my experience will brighten and harden timbres. Only the very best avoid this.

 

Of course, you could also just leave the processed files at the longer word length (and perhaps higher sample rate) -- if you don't mind the larger files.

 

Just some things to consider. As always, my best suggestion is to *not* simply take my word for any of this but instead, to try things out for yourself. Do some experiments to find out how things sound to *you*.

Have fun!

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

 

Wow - that's an education session I obviously needed ! Thanks for explaining the pitfalls so clearly and so well - take an immediate pay rise ! So, I may explore a bit of 're-mastering' just to see how involved I need to get but in the meantime your tap analogy seems entirely sensible (the Lyngdorf does something weird and wonderful with its approach to volume control and preserves low level detail like few amps I've ever heard). On a note of curiosity - will the use of Replay Gain of one flavour or another have the effect of losing word length ? Or does that work in a safe and different way ? Might just have to live with the levels of many rock albums having 7-8db shaved off their average levels at the expense of boosting some jazz and classical by probably just 2 or 3 db.

[bdiament]

Hi Liffy,

 

Wow - that's an education session I obviously needed ! Thanks for explaining the pitfalls so clearly and so well - take an immediate pay rise ! So, I may explore a bit of 're-mastering' just to see how involved I need to get but in the meantime your tap analogy seems entirely sensible (the Lyngdorf does something weird and wonderful with its approach to volume control and preserves low level detail like few amps I've ever heard). On a note of curiosity - will the use of Replay Gain of one flavour or another have the effect of losing word length ? Or does that work in a safe and different way ? Might just have to live with the levels of many rock albums having 7-8db shaved off their average levels at the expense of boosting some jazz and classical by probably just 2 or 3 db.

 

As I don't use Replay Gain, I can't speak to whether it engenders any loss of low level information or whether it takes steps to preserve same.

 

It is always worth experimenting to find out what you'll hear.

 

My own personal preference is to avoid any automatic adjustment of audio and would instead, take the "long" path I described if such were my intent. That said, I tend to listen to my music library as it is, with no further processing. Unfortunately, when moving between good recordings and casualties of the Loudness Wars, this means liberal use of the playback volume control.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[Julf]

will the use of Replay Gain of one flavour or another have the effect of losing word length?

 

Yes, no and maybe.

 

ReplayGain is not one single way of adjusting gain. ReplayGain is a combination of things - it is a standard for analyzing the perceived loudness of a track or album, and calculating how much the gain needs to be adjusted to achieve a consistent loudness level (either across the tracks in one album, or across your whole collection), and then a standard way to embed that information in the track metadata, so that player programs can adjust the gain accordingly.

 

ReplayGain is not a way to adjust the gain, it is a way to calculate how much it should be adjusted, and a way to communicate that value to your player program. It is up to your player program to adjust the volume - either by performing digital signal processing, or by adjusting the analog output gain / volume setting.

[MikeJazz]

the Lyngdorf does something weird and wonderful with its approach to volume control and preserves low level detail like few amps I've ever heard).

 

I would not worry about Lyngdorf volume control.

If you dont't want to normalize volume levels, then you'd better prefer to give Lyngdorf a clean bit perfect signal...in my opinion, of course...so better not to change the files...

 

For classical music and some jazz, with more dynamic range, I can happily raise volume up to 80's...The amp never sounded compressed, strained, "stressed" or distorting. Also the quad's do not simply distort to any noticeable level...amazing.

 

Yes, I continue amazed by the low level detail...

 

They say that from 100 until 65 the full resolution is preserved!

Do not know about that, but for very compressed albums sometimes it's already too loud at 65db.

Specially at "nigh mode" I have to decrease that until about 55 for compressed and 60 to 65 for less compressed music.

 

Having said that, I find JRMC replaygain very useful to assess the files..comparing wich are more or less compressed...