I have a collection of about 11 000 carefully collected MP3's which we use for personal and DJing purposes.

I have made a seperate snapshot of the collection and run MP3Gain 1.0 on it in Radio mode (Took more than just a while) and set the level to 89db because this was the highest I could set it to without having too many songs clip. (I do not totally understand clipping, but from what you've written, it is not good, so I try to avoid having any clip. I can notice the change in Winamp, and in CDs written by Nero for my car. On our DJing equipment, the sound is totally consistent, and this has made our lives far easier, but I have a feeling that 89db may just be a little low. I notice this most in my car.

I have a pretty nice sound system in my car, and, while the clarity of 'your' MP3's is incredible, they cannot be pushed loud enough at times. Is 89db too
low. I have noticed the latest version of MP3Gain allows to set a maximum level that ensures no clipping. How does this work? Will this not leave me
with a whole set of MP3s at different volumes?

How has the Ogg Vorbis community responded to the Replay Gain proposal?

Does anyone use CDDA Xtractor (http://xtractor.sourceforge.net/)? Should I be using EAC? Is is better to write audio CDs with Nero, Burrrn or BAO - with quality in mind.

Thanks alot!

-- There are 10 types of people in the world. Those who understand binary, and those who don't.

if you simply can't turn it up loud enough then you need to adjust the gain controls on your amp because there shouldn't be any real reason you can't crank them up just because they've been replaygained to 89dB. I've ran all my mp3s through mp3gain too and I have no problem reaching the limit of my amps/speakers in my car. Yes, it is a little quieter than the original CD, but nothing that can't be made up for by simply turning up the volume. Just happens that in your case your amp gains are set too low to allow you to do the same.

Yes, Max No Clip Gain will cause the volume to vary if it's done on a track-by-track basis. It's no better than peak normalization, which doesn't do well at equalizing perceived volume. Read The ReplayGain website for more explanation.

If you have a weedy amplifier gain control, it's not really the fault of RG. Many older albums and classical recordings come out at around 89 dB before adjustment. Some very dynamic ones even come out at 83 dB, which is the standard for movies, so these wouldn't play well. You can set the RG target volume in MP3Gain or a player like Foobar2000 to a different value if you like. You may find that clipping is inaudible if it's only during rare transients and only clipped by 1.5 to 3.0 dB for a handful of samples duration, though I don't take that risk myself, having adequate volume at 89 dB.

The Ogg Vorbis community is varied. Those on HydrogenAudio tend to be keen on ReplayGain and use an RG-capable player (Album Gain is probably favoured among the audiophiles as it better matches the indended volume). Some of the Vorbis users elsewhere aren't so sound-conscious and/or use DSP plugins like Dynamic Compressors to give a phat sound and more consistent volume, rather than aiming for transparency. Some use software that doesn't support ReplayGain. In Foobar2000 users, ReplayGain is very widely used, as it is in Musepack (MPC) users. Given that RG values got built right into the Vorbis header tag format, I'd say it's well supported by the Vorbis developers.

EAC, when set up properly in secure mode (or burst mode Test & Copy with matching CRCs), is almost universally considered the absolute best method for guaranteeing that a rip reported as having "no errors occurred" is accurate. I don't want to discover pops or glitches in my CDs so I only use EAC secure mode for ripping.

Most burning software is decent (including EAC if your burner is supported) and there's little to choose in terms of burn quality. Read the appropriate forum on HA to find out more. The CD-R manufacturer (not just the brand) tends to be more important, and here Taiyo-Yuden is usually considered the best. (Guide to identifying CD-R manufacturer, with photos)

If you're burning from lossy-compressed files, like .MPC, .OGG (Vorbis), .MP3, rather than lossless duplicates of the original CD, like .FLAC/OggFLAC, Monkey's .APE, .WV Wavpack Lossless you may wish to decompress to WAV before burning rather than do it on the fly.

This allows more control over the decode, and although difficult to ABX, dithering to 16-bit will ensure that the subtlest detail during fadeouts and quiet sections is preserved.

Programs like Foobar2000 (FB2K) make diskwriting with dither a doddle, and for DJing, if you intend to pitch-shift, I agree with 2Bdecided's previous recommendation (different thread) to use "No Noise Shaping" dither. Strong ATH Noise Shaping is recommended only for normal playback at the proper speed/sampling rate and achieves lower perceived noise by concentrating the dither energy in the ultrasonic region. It's only worth using if no more processing is going to happen. If you pitch-shift down (or slow the playback down with an accompanying pitch shift) you can bring the slightly increased ultrasound noise into more audible frequencies where it might sound more hissy than you'd like, so for DJing unshaped dither is better.

I believe there's also now a plugin to make FB2K into an audio CD burner. The other great aspect of FB2K is that it includes ReplayGain at the core, so it can tag files (or store database info, even for CD tracks!) with both track and album gain and peak values and give you full control. For preparing audio CDs, it can also resample other sampling rates to 44100 S/s with optimum quality, convert mono to two-channel and output at 16-bit bit-depth (all required by the CD format) with your choice of dither and ReplayGain on or off.

Nice sig, by the way!

Take a look at the mp3ain guide in the mp3 faq in the mp3 forum.


You can change the target value upwards from 89dB and mp3gain will tell you if this will introduce clipping. From what you've written, I think you've already tried this, and even 89dB causes some of your tunes to clip.

You can't have all the songs the same loudness, and all the songs as loud as the (previously) loudest song, for reasons explained on the replaygain website: http://replaygain.hydrogenaudio.org//faq_quiet.html


There are two options if 89dB is too quiet, and some songs are already clipping: increase the level and accept more clipping, or reduce the level of the clipped songs only, so that they don't clip.

Find the song which mp3gain says clips at the lowest level, mp3gain-it to the (higher) level you want, and listen to it. You might think it sounds fine, in which case, you have your answer. Warning: some songs sound quite OK clipped, others (e.g. piano and choral music) sound really really terrible! If you find that your songs do sound terrible when clipped, then you'll just have to reduce the level to prevent clipping. Whether you do this on all songs (so keeping the same loudness on all of them) or on just the clipped ones (so making these sound relatively quieter) is up to you.


You might find that clipping isn't annoying, and/or you might find that having a few songs slightly quieter isn't annoying, so you may have two good options!


Hope this helps,

Cheers,
David.

Holy Monkey! DickD that was quite an answer. I will read it over a couple of times more and try to understand some of that pitch stuff better. No need to elaborate though, I appreciate the effort. Thanks also _Shorty and 2BDecided.

I had read the Replay Gain site quite extesively, and was just looking for confirmation. I will stick to MP3GAIN's Track (Radio) Gain across the board. It will also have the added advantage of allowing me to safely push my equipment to their peaks, and hopefully minimising the number of tweeters we pop.

I think I definitely need to take a look at foobar2000! Thanks for the replies. Very much appreciated!

Regards
Warwick Chapman

-- There are 10 types of people in the world. Those who understand binary, and those who don't.