Help with normalization

I have a track in dsf format - sounds great. I ripped it to flac for comparison purposes (to 24/88.2). The track also sounds nice but the volume is much reduces. When I ran them both in Foobar's Dynamic Range Meter they both showed peak volumes of -7.12.
How is that possible when one track is clearly not as loud. Do I need different software to find the actual volume differences between the tracks?
Sorry if this is a silly question - I'm pretty new to this stuff and I am trying to see if I hear in degradation converting dsf to flac (24/88.2) for storage purposes and to have all tracks in flac. However, I do want them to be the same loudness so I'm trying to understand how the DR Meter works and what the values are telling me.

Thanks for any help you give

 

Digital music has a maximum instantaneous volume level that can't be exceeded. Within that envelope, music can have different digital peak volumes compared to the absolute loudness. This is one of the reasons why CD was a revelation, you now had all that dynamic range and headroom that consumers had never experienced before. The digital audio file isn't bad if it is quieter; in fact, that it is quieter probably means it is higher quality.

In the last 10 years, commercial recorded digital audio has been sonically compressed to remove the dynamics, removing the hit of the drum that would peak the meters, so the rest of the music can be turned up. Brickwall limiting is what we call it taken to the max, where the volume is cranked up unnecessarily just to sound louder than the next CD, audio quality be damned.

Here is is an analysis I recently did. A re-release on vinyl by a small label. The re-release was very hot, very loud sounding compared to the original vinyl, so using RMS loudness analysis, I set them both to the same apparent volume (the value highlighted in yellow). Then analyze.



The one on top is the original LP issue, the re-release on the bottom. The top one uses the digital to the fullest: loud parts are captured as well as the quiet. The original looks a lot louder because it has more dynamic range. The one on the bottom is the reissue; they used digital processing to make it louder, but in reality, they destroyed the dynamic range of the music, cutting out all the peaks. The originally quieter one on the top is the better one.

Anyway, let's proceed, as that probably wasn't much of an education.

First, how did you "rip it to FLAC"? There are just a few software DSD converters, and they have been reused in products like dbPoweramp and foobar. In conjunction with the DFS to WAV (flac) are tools to convert ISO to DFF to DSF etc. It is probable that the software toolchain of one converter produces a different output of another; the implementation of integration and software filtering will be different on each.

The DSD file is being converted to PCM in order for you to hear it on your computer in any case, whether in the player or by external utility. The decoder in one may recognize a gain flag in the file, or may be designed differently, producing different levels. One may be exceeding the maximum PCM volume, giving you digitally clipped but louder audio. Your player may have an EQ, volume control, or DSP you forgot about, or one only applied to DSD. Or they may have just decided in the player to increase the gain of DSD files because the converters are always low-output.

The PCM FLAC audio file can be increased in volume up to the maximum allowed by the digital format. This is called normalization. If you open the flac audio file in an audio editor (try free Audacity), you can examine if musical peaks ever reach 0dB. If it looks like the bottom waveform in the picture, the normalization feature can increase the volume, but if the audio file looks like the top, it likely is already at max, and normalization probably won't do much. Normalizing the volume will not give you audio benefit though, except to match the volume to the rest of your music library.

 

Not really enough info to say for sure but DSD file playback or conversion to PCM involves setting the PCM output level gain, default is usually 0dB. It may have been set to 0dB for the flac conversion but set higher for the DSF file playback, so the DSF playback would be louder even though files compare.

 

I played it back on a thumb drive to my oppo 103 so no gain added there. It must have happened at the conversion but what stumps me is why they came back with identical peaks in the dl meter (unless I don't understand how that works - a distinct possibility!)
Thanks for your thoughtful replies - looking forward to more opinions.
Thanks!

 

Well, assuming you aren't doing any processing in the Oppo, the playback paths are different for DSD and PCM, and gain could be different. If the PCM conversion and Dynamic Range check were both done in foobar, then they both use same PCM conversion (the dynamic range tool only works on PCM), so it is understandable that you get exactly the same readings on the files since, in effect, you are measuring the same file twice.

 

Ah, good to know (that makes sense). The problem then is that it involves some guess work to get the volumes the same level - probably close to a basic +6db bump to the flac file would do it in most cases?
Thanks

 

Yea, it's kinda guess work with foobar, you set the PCM gain and hope you don't go over, which will lead to clipping in the PCM file. +3dB is a more conservative setting, but depends completely on the DSF files, and ultimately, how much time you wanna spend on it