Preserving quality with amplify effect

I have just recorded and saved a load of vinyl in 16-bit / 44100hz with sample rate converter set to high quality sync interpolation and dithering set to none as I did not originally intend to apply any effects to the recordings just save each track as an exported selection loss free ready to burn to cd, however the tracks are all different volumes as you would expect from the various vinyl sources, some of the tracks are pretty quiet and could do with a boost, I set the pre-amp / line in levels right for the first track so the peaks don’t go close to 1.0 on the waveform scale (about 0.7) but lots of the tracks I recorded after are much quieter and only reach about 0.5 on the scale.

My question is now I have already recorded and saved the tracks in 16-bit / 44100hz, is it possible to re-open the files in audacity > amplify > save with the above settings with zero loss of quality? Or once the tracks have been recorded too quiet in the first place is it impossible to boost them without losing quality (even if it is not noticeable to the human ear). I don’t want to do this if there is some other conversion process despite the above dithering settings and saving in the same bit rate and frequency, I understand with amplify the waveform is being changed and it is not a simple editing operation like delete / cut / paste, but I would think this is a fairly straightforward effect to apply? Are you going to lose quality simply doing as the below images show by using the default amplify setting (so there is no clipping)…

original recording:
<img](http://www.divshare.com/download/21179691-60c%22%3E%3Cimg) src=“http://www.divshare.com/img/21179691-60c.JPG” border=“0” />

after applying amplify effect:
<img](http://www.divshare.com/download/21179697-306%22%3E%3Cimg) src=“http://www.divshare.com/img/21179697-306.JPG” border=“0” />

I also tried running some of the quiet tracks through winamp with the pre amp on and the output set to disk writer (the wav export setting - no conversion) and I cannot hear any loss of quality compared to the original (using good headphones), but is there and I just cannot hear it? I could understand if I’m saving in different bit rates or frequencies and there is some process like interpolating or dithering going but I am just saving them in the same 16-bit / 44100hz format I recorded them in. Hope someone can help me so I know whether or not to amplify these quiet tracks as I would prefer they are the max level before burning to cd!

The dithering is there because Audacity works internally at 32-bit floating, not 16 bit. So everything is a conversion. If you were reasonable with your capture settings, I doubt you would hear any problems. People who get into trouble are the ones with very serious volume problems and the ones trying to do Audio Forensics.

A word on the Audacity timeline. It’s in Percent, not Acoustics or Volume To Your Ear. It’s done that way because it makes editing a bit easier and more convenient since the most audible part of the show appears using the most screen real estate. However, 0.5 converts to only 6dB change out of the 65dB range that you can actually hear. Since your natural peaks on the sound meter should not get closer to 0 than about 3, that’s only about a 3dB change which most people can’t hear at all.

I’d probably leave the shows just where they are. I’s possible some shows will present louder than others not related to the blue waves. That’s sound processing and you can do some of that in Audacity, too.

Note you only run into show problems if you try mashups over many multiple albums covering years.

Koz

A peak level of 0.7 should be fine. There will be little if anything to be gained by re-processing the files compared with just turning up the playback volume a little.

Speaking of absolutes the answer would normally be no, but with some possible exceptions.
In practical terms, amplifying will produce negligible loss of sound quality, but I’m aware that that is not what you asked :wink:

The “problem” is that with dither off, if you amplify by an inexact number of “bits”, then the numerical values of each sample after amplifying will be rounded to the closes 16 bit value. This “rounding” will cause a tiny bit of distortion.

An exception to this case is if you amplify to exactly double the original level.
Obviously this should only be done if the original peak level was less than 0.5.
Double the original level is about -6.0206 dB, but you want “exactly” double the level, so rather than use the Audacity Amplify effect we would be better to do it exactly using the Nyquist Prompt effect.
This should be done with “high quality dither” set to “none”.
Apply this code to the audio track using the Nyquist Prompt effect:

(mult s 2)

Thanks a lot for the explanation and advice! …after reading how adc’s actually work I now understand what you are saying about rounding off the mulitplied numbers on an incremental scale…so as long as I increase the amplitude by a whole number there will be no rounding off right? Would I be right in thinking dithering is not required at all in this case provided I am saving the file in the same format? Still I think I’m screwed becuase many of the recordings are just over 0.5 on the scale so it would mean clipping…I will mess around with various dither settings and just use the default amplify effect to see if I can actually hear any difference with dither turned on and off, I will probably notice nothing at all I struggle to tell between 320kbps mp3’s and WAV files even with a decent set of headphones (too many years of loud music!!!) I guess it’s just the thought of the information having to be approximated again is enough to prevent me from doing it so I will take the advice and either record each track again setting the levels for each track indivudually (though a pain in the ass as I’ve already done nearly a hundred 7’'s) or just leave them as they are which is about half the volume they could be, I like to listen to the wavs on portable devices which don’t often go that loud so it’s not ideal but it’s either that or accepting the distortion, however small it is.

Right, but remember that we are talking about binary numbers and not decimal numbers and we are talking about “multiplication factors”, not “dB”.


Yes.
As an example, if we have the binary number:
00110011
and we double it, we get (exactly)
01100110
Every 16 bit value less than 0.5 will have an exact representation when doubled so dither is not required. (values over 0.5 will clip if doubled)

If they are over 0.5 peak then I’d leave them as they are and just turn up the amp a bit when playing them.

With audio that has been recorded and encoded properly with a modern high quality MP3 encoder, double blind tests have failed to find anyone that can reliably tell the difference. :astonished:

What format files do you use on portable devices? Does the player support “ReplayGain” or “Sound Check”? If so you could use that instead of modifying the data.

320kbps mp3’s and WAV files

320 bitrate is approaching the point where nobody can hear the difference except possibly very young girls.

Bitrate quality ratings get very slippery past about 256 or so. It’s not a linear change as the rate goes up. The difference between 32 and 64 on a stereo show is immediately clear to anybody with a pulse, but the doubling between 128 and 256 is just barely perceptible and the difference between 256 and 320 is zero or an impossibly low number. The file size does keep doubling, however, so it’s a decision where you want to stop. WAV does have a bitrate. It’s something north of 1400.

Would I be right in thinking dithering is not required at all in this case provided I am saving the file in the same format?

Any time you pull a music file into audacity, it converts to 32-bit floating. Since this is a much higher quality format, dithering is not needed. It’s when you go back down to 16-bit that some of the audio voltage levels don’t work out right. Some get missed and some double up causing sound distortion.

You might be a good match for a WAV editor that doesn’t try to resample and not Audacity.

That’s two of us told you to leave it alone.

Koz

Thanks for the clarification! Before I go recording more vinyl, am I better off leaving the line-in sampling format as 16-bit / 44100hz and the same setting on Audacity (not 32-bit float) with no dithering because I will not be amplifying or saving them in a different format. I basically want to end up with files in audio CD format and I read somewhere if you do not need to apply effects and just want files ready to burn as an audio CD it is best to just set everything to 16-bit / 44100hz with no dithering so there is no ‘converting’ from the original sampling format and therefore no further loss or distortion. I tend to record about 20 songs at a time then just extract selections when I am saving each track, would 32-bit float just slow things down and offer zero benefit over recording in 16-bit in my case? Koz said to just use another WAV editor as Audacity will resample but as long as everything is set to 16-bit / 44100hz with no dithering and I am not applying any effects it will not be resampling will it? The only loss I want occurring is the original A>D process of the sound cards line-in which will be as good as my hardware allows, I don’t want to introduce any further software based processing / loss if that makes sense.

That will not necessarily have the effect that you intend.
On modern computers it is not like a simple wire connecting the ADC (Analogue to Digital Converter) of the sound card to the recording program. There is a complex “system” that involves device drivers and other software built into the operating system. The audio data may be resampled and/or scaled and/or dithered before it reaches Audacity. Audacity has little control over what the audio hardware, drivers and OS sound system do before the data reaches Audacity.

I’m not an expert on the Windows sound system but there is some useful information in this post: https://forum.audacityteam.org/t/mme-or-windows-directsound/17434/3


32 bit float format is generally a bit quicker than 16 bit (integer) format because computers use dedicated hardware for handling floating point data fast.
Disk operations are likely to take a little longer with 32 bit float because there is more data. Overall you will probably not notice much difference in performance.

32 bit float is categorically better if you are processing the audio because it is faster and higher quality than 16 bit, even allowing for losses due to the final conversion from 32-bit to 16 bit.


Yes that makes perfect sense. We all want the best quality that our system allows.
The problem is that we don’t really know what happens between the sound card input and the digital data reaching Audacity. We can optimise based on educated guesswork but without detailed knowledge of the audio hardware, device drivers, sound system API and so on we will never know for sure exactly what goes on in there.

On Vista/7/8 it is a good guess to use DirectSound with “exclusive mode” enabled in the sound system settings.
The sample rate set throughout (Audacity and the OS) to the final rate that you want (44100)
Audacity set to 32 bit float (this will not damage 16 bit data at all and is better if you decide to do any processing, amplifying, mixing, fading…)
IF you do no processing at all then there may be a marginal benefit in turning off dither, but in all other cases dither should be enabled.

Ok thanks again Steve, I am looking at this in a very simplistic way as you probably realize I have very limited understanding of the whole process and just want a simple answer but I should know it is never that simple, however going on what you said and what I read elsewhere I think I will just continue to record the vinyl with the current settings as I won’t be applying any effects and just want to end up with 16-bit 44100khz files. I recently got a new turntable (1210) and the sound improvement is huge but now I have this new equipment I just want to get the best recordings possible that my current setup will allow using which ever software and settings, I also have Nero wave editor on here I don’t know if that will do I better job for what I want but at the moment I am very pleased with the sound of these recordings using Audacity!

On that thread it says ‘‘However Direct Sound on Vista/7 does allow Audacity to use the new “exclusive mode” to take exclusive control of the sound device if this is set at “Sound” in the Windows Control Panel. The benefit of “exclusive control” is that no resampling is done in Windows. So if you choose a project rate bottom left of Audacity that is compatible with your sound device, you shouldn’t get resampling distortion or speed problems that can happen if you use MME.’’

…So if i set the sound device to allow Audacity to take exclusive control and set to direct sound in Audacity there should be no resampling or processing of the sound between the sound device and audacity? I am using win7 so I have these options, I had the exclusive control setting ticked in the sound device but it was set to MME in audacity, should I change this to direct sound now?

That’s the theory. Try it out and see if it works. (I can’t try this as I don’t have Win7)

Will give it a go mate, I doubt I will hear any difference but will try it with a few different types of music and will come back and post if I notice any difference!

Ohh now I have another question! Now that I have changed it to Windows DirectSound, for the input device I have 2 options:‘’ Primary Sound Capture Driver’’ or ‘‘Line In (High Definition Audio Device)’’ …both are recording when I test it but which one should I be using to get the best sound quality or are they essentially the same thing?

As far as I’m aware they are the same thing. Primary Sound Capture Driver’’ being an alias (AKA) for ‘‘Line In (High Definition Audio Device)’’

…well I’ve just recorded the same track in 16-bit 44100khz with the 4 different settings I have available: MME with Microsoft Sound Mapper, MME with Line In (High Definition Audio Device), WDS with Primary Sound Capture and WDS with Line In (High Definition Audio Device)… after sitting here listening to each recording repeatedly using some decent headphones I cannot tell ANY difference between the 4 recordings! My ears are thankfully not so bad I can’t tell between a 192kbps mp3 and a WAV file as I just tried it with these recordings and the difference is quite clear but between the 4 wav files I cannot hear anything different at all, the volume is exactly the same on all 4 tracks that’s without changing the levels and they all sound equally crystal clear! I guess that is a testament to the quality of the processing that is occurring (if any) when I change the settings, or maybe as you pointed out whatever you do some processing is going occur between the input device and Audacity all I know is my ears are definitely not good enough to tell the difference!