PDA

View Full Version : You learn something new everyday



d6jg
2019-01-23, 12:24
Yesterday I recorded an album to FLAC using the same software that I always use - VinylStudio from Alpinesoft (itís the best I have used by far and I have tried many).
Anyway the album in question was Bob Marley, Legend on 180g Vinyl. When I got to the track splitting stage I was astonished at the waveform I was presented with. The volume level on the whole album was set really low with the result that the waveform was narrow if you know what I mean. This wasnít a case of not having the software set incorrectly as with a USB type ADC you set a particular level of digital gain once and then never alter it. No the album was pressed at an incredibly low volume.
After splitting and writing individual tracks I ran Foobar2000 over them to set Replay Gain. It required a 15db uplift. A level which I have never seen before.
Having rescanned I proceeded to test play on my Living Room SB3. Still sounded desperately quiet when compared to other tracks.
Checked the SB3 audio settings to ensure Smart Gain was on. Yes.
Then I realised.
The SB3 was set to output at 100%.
It appears Smart Gain does not work if this is the case.
Now I think about it I can see the logic as to why it wouldnít but also how you could be fooled into thinking that even at 100% it could / should work.
As I say you learn something new about LMS every day!

Apesbrain
2019-01-23, 16:27
If you don't want those big RG values, you can amplify the files. Just amplify each track by the same amount. 12dB should do it.

d6jg
2019-01-24, 02:23
If you don't want those big RG values, you can amplify the files. Just amplify each track by the same amount. 12dB should do it.

You can indeed but I adopted RG as opposed to Normalisation some while ago.
There is a Normalisation option in VinylStudio but my collection is about 2000 CDs to 650 LPs so RG was easier when I started out. That said Iím still re ripping stuff that I foolishly did at 320kbps MP3 at the outset!

bakker_be
2019-01-24, 02:40
You can indeed but I adopted RG as opposed to Normalisation some while ago.
There is a Normalisation option in VinylStudio but my collection is about 2000 CDs to 650 LPs so RG was easier when I started out. That said Iím still re ripping stuff that I foolishly did at 320kbps MP3 at the outset!

Normalization is not the same as boosting the overall volume. Normalization loses the dynamics as it's more akin to dynamic compression. Digitally boosting these tracks with 12dB as proposed will lift everything by these same 12dB, ending up at Ī-3dB absolute for the loudest peaks and (-3dB - the difference between loudest and quietest) for the quietest parts. Think of it as just turning the volume knob on your amplifier.
It does seem strange to me however that RG doesn't work when your volume is fixed at 100. I've been using RG on all my files for years and never once had this issue on any client with this setting in effect. I thought that this was done server-side befor sending out to the client ...

d6jg
2019-01-24, 02:44
Normalization is not the same as boosting the overall volume. Normalization loses the dynamics as it's more akin to dynamic compression. Digitally boosting these tracks with 12dB as proposed will lift everything by these same 12dB, ending up at Ī-3dB absolute for the loudest peaks and (-3dB - the difference between loudest and quietest) for the quietest parts. Think of it as just turning the volume knob on your amplifier.
It does seem strange to me however that RG doesn't work when your volume is fixed at 100. I've been using RG on all my files for years and never once had this issue on any client with this setting in effect. I thought that this was done server-side befor sending out to the client ...

What you say about 100% Fixed is interesting as I could have sworn it used to work. Where I noticed is an SB3 which is in a Group with an O2 Joggler as a Control Point / Now Playing display but I always play the group. Wonder if itís that?

bakker_be
2019-01-24, 02:56
What you say about 100% Fixed is interesting as I could have sworn it used to work. Where I noticed is an SB3 which is in a Group with an O2 Joggler as a Control Point / Now Playing display but I always play the group. Wonder if itís that?

It shouldn't. I've used the group player plugin as well, for a combination SBT + SqueezePlayer, both set to 100% fixed, and it worked as it should. I do seem to remember that RG refuses to lift the volume to the desired level if this would induce clipping.

cliveb
2019-01-25, 08:30
Normalization is not the same as boosting the overall volume. Normalization loses the dynamics as it's more akin to dynamic compression.
I'm sorry but this is completely wrong. Normalisation is the application of linear amplification so as to bring the peak level to a defined value (usually, but not necessarily, 0db).
It most definitely does not affect the dynamics.

My advice to the OP is to go ahead and normalise his Bob Marley needle drop, and then recalculate the replay gain values.

Mnyb
2019-01-26, 00:29
Yes replay gain tags only teaks the volume at playback .

But never at a positive values because of clipping so there is a target volume letís say -10dB ( the exact value is probably something else ) so the rg tags is an offset to reach the target volume usually negative .

Really well recorded stuff can have a small positive values .

But I suppose a reallly large positive value trips up the algorithm so it does not dare to raise the volume to much as clipping may occur .
Or also likely the volume was so very low that you needed more than 100% gain 0db at playback to match the other tracks .
The player can not provide more than full volume .

bakker_be
2019-01-31, 04:01
I'm sorry but this is completely wrong. Normalisation is the application of linear amplification so as to bring the peak level to a defined value (usually, but not necessarily, 0db).
It most definitely does not affect the dynamics.

My advice to the OP is to go ahead and normalise his Bob Marley needle drop, and then recalculate the replay gain values.

Probably all depends on the tools used. I haven't gone back and checked since discovering ReplayGain, but at the very least the "volume normalization" feature in Windows Mediaplayer adversely affected dynamics and was irreversibly applied to the files ...
I agree that the "official" definition indeed is as you say.

cliveb
2019-02-03, 06:58
Probably all depends on the tools used. I haven't gone back and checked since discovering ReplayGain, but at the very least the "volume normalization" feature in Windows Mediaplayer adversely affected dynamics and was irreversibly applied to the files ...
I agree that the "official" definition indeed is as you say.
Wouldn't surprise me if WMP's "volume normalisation" is some sort of compressor that attempts to boost the perceived volume to match typical CDs. Just guessing, though.

JJZolx
2019-02-05, 13:59
Now I think about it I can see the logic as to why it wouldn’t but also how you could be fooled into thinking that even at 100% it could / should work.

Does this mean that ReplayGain is disabled by LMS when a player's volume is fixed at 100%? Is that actually in the code?

Someone explain to me why it cannot work at 100%.

slartibartfast
2019-02-05, 14:40
Does this mean that ReplayGain is disabled by LMS when a player's volume is fixed at 100%? Is that actually in the code?

Someone explain to me why it cannot work at 100%.I have not tried myself but this thread suggests replaygain works with volume fixed at 100%.

https://forums.slimdevices.com/showthread.php?93913-ReplayGain-Problems

Sent from my SM-G900F using Tapatalk

bakker_be
2019-02-06, 02:41
I have not tried myself but this thread suggests replaygain works with volume fixed at 100%.

https://forums.slimdevices.com/showthread.php?93913-ReplayGain-Problems

Sent from my SM-G900F using Tapatalk

I have never known it not to work. I believe this is just a coincidence with this particular album, where the extremely low recording level would induce clipping when boosted to the reference level, which is something that ReplayGain doesn't do by default. In Foobar2K this behaviour can be circumvented, don't rightly know about it in LMS.

d6jg
2019-02-06, 03:47
I have never known it not to work. I believe this is just a coincidence with this particular album, where the extremely low recording level would induce clipping when boosted to the reference level, which is something that ReplayGain doesn't do by default. In Foobar2K this behaviour can be circumvented, don't rightly know about it in LMS.

It was me that started this thread. The album in question was recorded at a very low level and "seemed" not to be obeying the RG tags when the player (SB3) was at fixed volume. However the player was part of a Group created through "Group Players" plugin where there was another player also involved but I didn't test it on its own.

I will set another single player with 100% Fixed Volume and run the offending album through it as part of a playlist and advise. The sound level of the recording is so low that it will be immediately obvious but I can't do it until later.

BJW
2019-03-19, 09:02
Things got confused in this thread imo. I use RG, I do not use normalization and wouldn't want to.

But the other issue doesn't make sense. Think of output level as the volume knob. If it's at 10, that's 100%. If it's at 5, that's 50%. In either case, RG should be applied and noticeable (vs not being applied) bc it's done in software, prior to reaching the output stage.

So while I never put anything past server and it's hardware implementation, I believe the contention that 100% output defeats RG is wrong.

d6jg
2019-03-19, 09:21
Things got confused in this thread imo. I use RG, I do not use normalization and wouldn't want to.

But the other issue doesn't make sense. Think of output level as the volume knob. If it's at 10, that's 100%. If it's at 5, that's 50%. In either case, RG should be applied and noticeable (vs not being applied) bc it's done in software, prior to reaching the output stage.

So while I never put anything past server and it's hardware implementation, I believe the contention that 100% output defeats RG is wrong.

I was wrong. RG does appear to work at 100%.