Newbie Question - Best DSP settings?

[d2b]

Re: Newbie Question - Best DSP settings?

I don't consider Trim Silence before and after the payload to be destructive. We do desire to achieve "instant start" characteristics and to avoid useless silence at the end of a selection to allow for proper crossfades to the next composition.

I'm not aware that Volume Normalize is considered destructive. Is it not the same as the volume control? We have assumed that Volume Normalize was a linear function applied to the payload, but perhaps this is not the case. Otherwise, it would be non-linear compression, would it not?

i don't know enough about the Replay Gain algorithms to be sure of this, but it would seem that any process that dynamically affects the gain during playback to achieve a consistent perception of loudness would be a non-linear process. If so, I would consider that to be effectively more "destructive" than applying what I would hope to be a uniform linear adjustment of the gain, i.e. the volume.

Your experience and advice is appreciated. I would also like to hear from Spoon about this.

"d2b"

[garym]

Re: Newbie Question - Best DSP settings?

In windows I use foobar2000 player for my FLAC files with RG tags. My main home system is Squeezebox based and it handle RG tags properly as well. I think Sonos also handles RG tags.

Spoon will have to answer, but I always thought that Volume Normalize was destructive as it actually changes the underlying audio. This is what the description page says:
file:///C:/Program%20Files/dBpoweramp/Help/Codec/DSP/help.htm*Volume_Normalize

In the old days there was something similar for mp3 files called mp3gain. This program would adjust the actual underlying audio. However, it would write "undo" information in an APE tag so the user could at a later date "undo" the mp3gain volume adjustment. It doesn't appear that the DSP "Volume Normalize" has the "undo" option. The end result, it seems to me, is that your underlying audio is actually modified permanently to make the volume adjustment, so the audio would no longer be bit perfect back to the CD audio.

For example, I'm guessing that if one rips a CD and uses volume normalize, that if one later checks the AccurateRip results of that set of album files, that the AccurateRip would NOT provide a match (because the underlying audio has now been modified). Seems like the silence removal DSP would also cause a nonAR match (if enough silence was removed), but not sure about the latter as I don't know enough about how AR deals with silence in its checking algorithm.

EDIT: I just experimented with this. If I use the Volume Normalize (peak to peak) DSP, the resulting rip is no longer considered accurate in comparing to the other rips in the AR database.

I'm not aware that Volume Normalize is considered destructive. Is it not the same as the volume control? We have assumed that Volume Normalize was a linear function applied to the payload, but perhaps this is not the case. Otherwise, it would be non-linear compression, would it not?

i don't know enough about the Replay Gain algorithms to be sure of this, but it would seem that any process that dynamically affects the gain during playback to achieve a consistent perception of loudness would be a non-linear process. If so, I would consider that to be effectively more "destructive" than applying what I would hope to be a uniform linear adjustment of the gain, i.e. the volume.

Destructive to me means that the ripped audio files are no longer bit perfect compared back to the original CD (once decoded for playback). Thus, nondestructive means that the files would match other entries in the AR database. ReplayGain tags can't be destructive under that definition, as the audio of the file itself is not being changed. If used by the player, the RG tags are simply equivalent to turning the volume up or down on a particular track.

[mville]

Re: Newbie Question - Best DSP settings?

I don't consider Trim Silence before and after the payload to be destructive. We do desire to achieve "instant start" characteristics and to avoid useless silence at the end of a selection to allow for proper crossfades to the next composition.

I understand what these DSPs achieve and why you might need to use them, but as I said, most users of dBpoweramp are after bit-perfect rips and so, in these terms, those DSPs are destructive.

[d2b]

Re: Newbie Question - Best DSP settings?

I'm glad I posted in this forum. I'm beginning to agree entirely with all of you. After re-reviewing the EBU specs and testing the media players for their use of the RG tags, we may modify our procedures.

Clearly there's no harm in including the RG tags as long as a playback application applies them as a linear process, as seems to be the requirement upon further review. Maintaining bit-for-bit accuracy in the payload is logically a desirable goal. But, in our case, we have no need or desire to compare decoded post-rip files with linear PCM representations of the original files as long as the files remain effectively lossless in the sense of only a linear adjustment of volume level at the byte level. Of course, I may be wrong when assuming volume normalization is a just linearly-applied adjustment of the resulting decoded PCM files.

However, we have no problem securing favorable AccurateRip results when the albums we rip using our current procedures are already in the AccurateRip database. Why is that, if removing digtal zeros ahead of and after the principle payload and Volume Normalization are destructive processes? Is that judgement made prior to the application of the DSP algorithms?

Spoon, where are you when I need you?

"d2b"

[garym]

Re: Newbie Question - Best DSP settings?

However, we have no problem securing favorable AccurateRip results when the albums we rip using our current procedures are already in the AccurateRip database. Why is that, if removing digtal zeros ahead of and after the principle payload and Volume Normalization are destructive processes? Is that judgement made prior to the application of the DSP algorithms?

It may be that the AR match is reported based on ripping data just BEFORE the "volume normalize" DSP is applied. What happens if you try to take an album after the fact (after ripping) and test it again against the AR database (using, say, PerfectTunes, or CueTools, or foobar2000 utility)? Does it still match?

(note: I agree that it doesn't really matter if you no longer have bit perfect, matchable rips after the fact for your purposes. But I'd also suggest that many users of dbpa expect in ripping a lossless file that they end up with a lossless file that would match up with a bit perfect rip in the future. In my own case, I like to use PerfectTunes to check match against AR database after I copy a large music database to a new backup harddrive. The check is even better than simply checking to see that the FLAC files are still valid, decodable files.

[d2b]

Re: Newbie Question - Best DSP settings?

Well, delving deeper into this question may not be worthwhile if we decide to begin using RG tags rather than doing the testing you suggest to try and justify what we've been doing. It sounds like we should be using RG instead. Of course, it depends upon our media players of choice, all of which are said to utilize RG. Do they do it correctly? I don't know. According to the Wikipedia article on Replay Gain, they do, but it's not obvious to us that they correctly implement album and track RG for FLAC. Cue 1.7, for example, has only meager documentation that describes the user interface. There may be specs somewhere, but they're elusive and the developer has dropped out of sight. Development on Cue stopped 8 or so years ago! JRiver Media Center is also said to support RG, but I've yet to research their site for anything describing it. I guess I'd better look around for details.

I agree that the AccurateRip comparisons are almost certainly done before the post-rip DSP is applied, but this would have to be done on fly as each track is ripped.

You might wonder why we use Cue 1.7. We have a wide variety of non-technical volunteer DJs, some of whom are Luddites. Not to demean them, but virtual CD players on a color screen are intuitive and appealing to computer beginners, whereas JRiver Media Center is rather complex in appearance by comparison. Not only that, but there's virtually no way for a user to "mess with the settings" when using Cue 1.7.

"d2b"

[schmidj]

Re: Newbie Question - Best DSP settings?

Given that storage is cheap now, you might consider future-proofing yourself by ripping to one directory with no "destructive" DSP's, then use the converter to make an "air" copy with the appropriate DSP's. Don't like the results on a particular file, or a bunch of them, you've got the source file, convert it again with new settings.

The "trim" settings, while necessary for you, can be dangerous. I was burnt by the default (-40dBFS?) silence setting. It chopped the fades off of some music. I reset to digital silence, safer. Now some day I have to find the offending CD's and rerip. However if you have an analog source, like a transfer of tape or vinyl, you'll never get digital silence. Frankly, the best for those is to use a waveform editor at the time you transfer them. Any automatic trim of analog material will fail one way or the other depending on its setting and the residual noise on the source.

My experience with players properly applying replay gain tags is marginal and inconclusive. I use the replay gain (tag, not apply) set to loudness of -18 LKFS, pretty conservative for radio, but I continue to be surprised at the number of CD tracks that end up softer because the dynamic range would have pushed the peaks into clipping. I have very good, expensive software that will set a loudness profile and apply pretty transparent limiting to any resulting "overs" (peaks exceeding whatever you set in the software.) I use that on the stuff I record and produce myself, but I don't nearly have the time to do it to all the ripped stuff. You can always do a test of your playback devices, record something really hot with replay gain calculated, something really soft with replay gain calculated and play them back and see what levels you get. You can also take some files and change the replay gain tags from the calculated values and see if the output levels change. For instance it doesn't appear that the replay gain tags on my m4a's are applied on my smartphone when played through bluetooth in my car. One of these days, I'll probably make another m4a profile and directory with the replay gain applied, and copy those files to the phone.

If you do go the dual library route, you would be safest from a playback standpoint if you used the replay gain apply DSP on the copy you use to air. Then it is "idiot proof", the file you air is already completely processed.

[Melapi]

Re: Newbie Question - Best DSP settings?

Hi d2b (AKA Dennis),

I haven't chosen a media player as such for the mac yet. I have been looking at the following - Tomahawk, Fidelia & Vox. But my first job is to get my media collection out of WMA Lossless and into something more friendly like FLAC.

As I don't want to be locked into Apple anymore either, well not for my music collection. I'm considering transferring all my FLAC files to a dedicated music server or possibly a NAS. Something like the Asustor 7004T so I can stream Hi Res Audio and stream Kodi TV from UK/USA. Otherwise maybe the Bluesound Vault system with wireless speakers.

I want to be able to stream Hi Res music in the house. Maybe Plex is the way to go? To be honest there's so much to take into consideration/research it's another project for once the files are converted. Any thoughts with your knowledge would be welcome.

Regards
Adrian (Melapi = Orangutan I sponsor at a rehab centra in Borneo)

[d2b]

Re: Newbie Question - Best DSP settings?

It's probably not worth our time to try and maintain two different libraries. We're ingesting dozens of CD a week as it is,and I'm pretty adamant about tagging them thoroughly. (What we mean by "thorough" is a topic for a whole different discussion some day. Just imagine a music world where NO physical media exists, i.e., a completely file-based system from digital master to home or professional server, where both the essential and desirable (but consistent) metadata is embedded in the files at the point of origin, i.e., the label or artist.) Managing a library with potentially hundreds of thousands of tracks from thousands of albums is already a daunting task.

Our motto needs to remain the same as it was when we started: "Rip Once, Rip Right!" In this definition, 'ripping' also means tagging thoroughly. Quality control is a paramount concern. That's why we're going to focus next on our DSP configuration. At least what we've done doesn't need to be discarded or redone. Performance on-air has been excellent.

My advice to Adrian's benefactor is to follow the motto above! Your sponsorship is appreciated here. I encourage your changeover to FLAC as well, for a huge number of reasons.

Dennis, aka "d2b"

Anyway, we're definitely going to look into RG.