Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Replay Gain questions (Read 11380 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Replay Gain questions

I play a mix of DTS (Flac encoded 5.1), dvd-A to flac 24/96 and good old mp3s -  What I find is that the surround channel as a whole are WAY softer than the .mp3.  Not all are like this which tells me its possible to have the gain raised without too much loss of dynamic range (or am I wrong about this?).

So I thought that replay gain could compensate for this so when going between an mp3 and a surround, I would not have to adjust the volume all the time.  I do not see this working unless I am doing something wrong - when I went to scan a track it told me that the track already had been set - but it sure did not sound that way.

So can someone set me straight on this via a link if that's the easiest way or s full explanation -and does RG work on multi-channel tracks?

Replay Gain questions

Reply #1
here is proposed calculation of RMS: http://replaygain.hydrogenaudio.org/rms_energy.html

I don't know what foobar scanner is doing for DTS 5.1, and I rarelly use them, as they are stored on DVD/CD, but here is one example result:

Code: [Select]
DTS 5.1 CD itself [RGAG: +0.24 db, RGAP: 0.250000]

RGTG; RGTP
+0.24 dB; 0.250000
+0.18 dB; 0.250000
+0.21 dB; 0.250000
+0.25 dB; 0.250000
+0.22 dB; 0.250000
+0.15 dB; 0.250000
+0.26 dB; 0.250000
+0.34 dB; 0.250000
+0.32 dB; 0.250000
+0.27 dB; 0.250000


Code: [Select]
DTS 5.1 CD converted to 6ch FLAC [RGAG: +2.65 dB, RGAP: 0.693634]

RGTG; RGTP
+4.93 dB; 0.434052
+4.75 dB; 0.672180
+3.43 dB; 0.665527
+0.27 dB; 0.693634
+6.48 dB; 0.475281
+2.66 dB; 0.678558
+5.47 dB; 0.580597
+5.73 dB; 0.638672
+5.41 dB; 0.467438
+2.32 dB; 0.666077


Code: [Select]
DTS 5.1 downmixed to stereo FLAC with foo_dsp_downmix [RGAG: -1.69 dB, RGAP: 1.000000]

RGTG; RGTP
+0.50 dB; 0.529785
+0.42 dB; 0.820160
-0.56 dB; 0.727051
-4.10 dB; 1.000000
+2.37 dB; 0.534760
-1.91 dB; 0.999969
+0.83 dB; 0.725220
+1.64 dB; 0.691956
+0.73 dB; 0.714508
-1.82 dB; 0.844971


I guess that in first result foobar is processing it as noise (as DTS without decoder), but for second and third example I don't know what's happening, and obviously the third example works best in reality

Replay Gain questions

Reply #2
I just started a new thread about "Multichannel Replay Gain" at the Scientific Discussion forum:
http://www.hydrogenaudio.org/forums/index....showtopic=77936.

Before starting it I searched HA for existing discussions, but found only this thread here.

I'd appreciate if someone could explain (perhaps in the linked thread) how foobar handles the channels when it analyzes e.g. a 5.1 file.

Replay Gain questions

Reply #3
Quote
DTS 5.1 downmixed to stereo FLAC with foo_dsp_downmix [RGAG: -1.69 dB, RGAP: 1.000000]

Looks like output was clipped because of converting to FLAC. I think it's better to use WavPack with 32-bit floats.

Also, can you test built-in "Convert 5.1 to stereo" DSP plugin?

Replay Gain questions

Reply #4
seems you are right

Code: [Select]
DTS 5.1 downmixed to 32bit stereo WV with foo_dsp_downmix [RGAG: -1.69 dB, RGAP: 1.647962]

RGTG;RGTP
+0.50 dB; 0.529791
+0.42 dB; 0.820164
-0.56 dB; 0.727040
-4.10 dB; 1.647962
+2.37 dB; 0.534769
-1.91 dB; 1.066528
+0.83 dB; 0.725208
+1.64 dB; 0.691947
+0.73 dB; 0.714511
-1.82 dB; 0.844982

just for the record:

Code: [Select]
DTS 5.1 downmixed to stereo FLAC with built in 5.1 to 2ch processor [RGAG: -1.40 dB, RGAP: 1.000000]

RGTG;RGTP
+1.61 dB; 0.502350
-0.04 dB; 0.900513
-0.41 dB; 0.681213
-3.61 dB; 1.000000
+2.41 dB; 0.583405
-1.60 dB; 0.999969
+1.27 dB; 0.724274
+1.75 dB; 0.723633
+0.79 dB; 0.653046
-1.87 dB; 0.938782

maybe there are too many different encoders, decoders, surround processors for some new RG code to be practical, but then maybe I'm wrong
also no one reports what foobar scanner is doing in case of multichannel audio

Replay Gain questions

Reply #5
also no one reports what foobar scanner is doing in case of multichannel audio
1.0, 2.0:
  obvious

5.1:
  rgL = L + (?2)/2 · C + (?2)/2 · LS + LFE
  rgR = R + (?2)/2 · C + (?2)/2 · RS + LFE

other channel count:
  simple mono average over all channels (?ch)/n

Well that was in 2003, but probably hasn't changed since.

Edit: Fixed the 5.1 formula.
Full-quoting makes you scroll past the same junk over and over.

Replay Gain questions

Reply #6
So, that 5.1 RG calc should be used as upmixing matrix (inverted) for stereo files, so that they'll be played without oscillations in 5.1 playback

[edit] when used together with native 5.1 tracks, of course

Replay Gain questions

Reply #7
...5.1:
  rgL = L + ?2 · C + ?2 · LS + LFE
  rgR = R + ?2 · C + ?2 · RS + LFE

other channel count:
  simple mono average over all channels (?ch)/n ...

Thanks for the info, but it doesn't make much sense to me.

Are you saying that in case of 5.1 foobar first creates "simulated" left and right channel values and then calculates the actual value by using the standard rule for stereo files?

I don't know what is the actual unit & scale that is used, but if for instance all channels result the same measured value (i.e. are equally loud) and that value is represented by the number 25, a stereo file would then result ?[(4² + 4²)/2] = 25 before the dB conversion (I hope I got this right).

How do I use the equation for a 5.1 file in case all 6 channels have the measured value 25?

I didn't calculate this, but I think that a direct unadjusted and dump calculation for six speakers should result a value of about 40.
- In other words, the resulting replay gain value difference from this dumb calculation should be about 4.77 dB (2 vs 6 equally loud speakers without taking the special nature of LFE into account).

Replay Gain questions

Reply #8
First let me correct my mistake, the coefficient for surround channels was not ?2 but (?2)/2, so the equations should be:
  rgL = L + (?2)/2 · C + (?2)/2 · LS + LFE
  rgR = R + (?2)/2 · C + (?2)/2 · RS + LFE

Are you saying that in case of 5.1 foobar first creates "simulated" left and right channel values and then calculates the actual value by using the standard rule for stereo files?
Yes, that seems to be the idea behind the formula being used. The later stereo processing instead of single channel might be just to avoid inaccurate results with L/R phase shifted signal.
Full-quoting makes you scroll past the same junk over and over.

Replay Gain questions

Reply #9
Maybe the equation is still not complete or I don't understand correctly how the results are passed to the two "virtual" channels before making the final calculation.

My example value "25" would produce approximately the following:

rgL = 25 + 17.7 + 17.7 + 25 = 85.4
rgR = 25 + 17.7 + 17.7 + 25 = 85.4

and the resulting value would then be

√[(85.4² + 85.4²)/2] = 85.4

I.e. more than twice the number I got when I estimated the effect of four directly added speakers. Of course I may have made a mistake, but that's how I understood the process of RMS energy calculation (at http://replaygain.hydrogenaudio.org/rms_energy.html).

Though, foobar's calculation may still produce practically useful results, because the surround channels are often quieter than the front channels and the LFE channel will probably never have a significant role because of the equal loudness filter.

Edit: fixed some comma vs dot irregularities (comma is the decimal separator in my locale settings).

Replay Gain questions

Reply #10
Wow!  A lot of data with not much results. 

Is it possible to just SET a simple DB+ value to a track so when that track plays, it automatically gets its "volume" raised by n dbs?  And if this value could then be applied to all channels and go above the max volume setting (add after the volume circuit)?

For me that would be the perfect solution since when I listen to those tracks all I am doing is turning up my recv volume to compensate for the lower overall levels.

Call it a track boost level control that can be normally at 0db and be slid up or down and remembered for each track played.

Then also if I could simply select multiple tracks and change that slider and they would all get that new setting.

Problem would then be solved.

Replay Gain questions

Reply #11
For bsw11, you can manually Edit the Replay Gain tag values by an appropriate number of dB in fb2k (if you enable that context menu item, normally hidden by default) but it needs to be adjusted manually on each file (or by means of some other software or perhaps an AutoIT script). I have no experience of using 5.1 in fb2k to guide you.

Referring to Alex B's post and Yirkha's formulae, a simple sum of powers seems sensible at first, but of course, trying to derive the square root of the sum of voltages-squared would eliminate negative voltages (negative sample values).

So it seems somewhat arbitrary and relies to some degree on making assumptions. I'd have thought that Centre and LFE would tend to be divided by √2 (since they appear in both rgL and rgR downmixed channels, and 1/√2=√2/2=-3dB) whilst Left and Left-Surround would be added straight to rgL, but there might be reasons I haven't considered, why Yirkha's recollection may be correct.

So rgL = √(L² + LS²+ (C² + LFE²)/2) would not work for generating suitable instantaneous sample values, thanks to the lack of negative values.
I'd have thought something like
rgL = L + LS + (C+LFE)/√2
would be fairly sensible.
(note that √2/2 = 1/√2 = 0.7071, so the factor (-3dB) is the same as Yirkha's edited scaling, just that it's applied to C+LFE, not to C+LS)

C and LFE could be summed in power or voltage to the same effect being essentially independent frequency components with little or no overlap. If you have a full-range centre speaker, I believe it's intended that LFE may indeed be added to C by simple summation and fed to that speaker.

I'd imagine that equal volume in each channel is unusual, so the apparent discrepancy from using +25 in each of 6 channels is rather an unusual case, and would be expected to differ from the usual situation. Probably, Centre+LFE and Left/Right would be the loudest, with LS and RS being significantly quieter most of the time, mainly for atmosphere and ambient echoes except in certain action sequences when what's off-screen is highly relevant to the movie (e.g. surprise explosion or attack from behind). Left and Right would also be much quieter than in 2.0 music because much of their role in stereo is taken by the Centre channel in 5.1.

5.1 to Stereo Downmixing overall scaling may well be done differently to what is fed into the Replay Gain function in order to reduce the likelihood of clipping in the downmixed signal (which isn't an issue to the floating-point RG function).

A good solution, I'd have thought, would be that the RG function could simply sum the inverse loudness-curve weighted powers (in linear units, not dB) of an arbitrary number of multiple channels during each time interval. The disadvantage as the number of channels increases, is increased processing which isn't a problem when feeding the data to a 2-channel RG function.

Potentially, an ideal solution may be to model a typical 5.1 setup and use the Head Related Transfer Function etc. to calculate the waveform incident on each of the listener's two ears, then make the equal loudness curve adjustments and sum the powers derived from that calculation to ascertain the overall experienced loudness (or use Dolby Headphone's equivalent). Potentially, a set of in-ear microphones (used for binaural recording) could directly measure or derive the impulse resonponses (transfer functions from speaker to each ear) from each speaker of a real home theatre set-up or from impulsive noises (e.g. balloon popping) in the positions of each speaker. This may be more computationally intensive than a weighted addition of sample values, but ought to be more accurate. (Likewise, for stereo playback over speakers, the same approach could be taken to model response from each speaker of a typical 2.0 system to each ear)

Downmixing 5.1 to stereo can be "correctly" done as a 5.1 to binaural transfer (for headphone playback - as with Dolby Headphone) or as a 5.1 to 2.0 transfer for loudspeaker playback, the latter resulting in a loss of spatial imaging (except for VERY specific head-and-speaker alignments if the exact inverse transfer function is employed!). The overall scaling of a downmix is either to match perceptual loudness or is a matter of controlling clipping from the downmixing of typical 5.1 source material.
Dynamic – the artist formerly known as DickD


Replay Gain questions

Reply #13
Dynamic,

Thanks for your comments. That's the kind of reply I would have liked to see in the thread I opened at the Scienfitic Discussion forum. ( http://www.hydrogenaudio.org/forums/index....showtopic=77936 )

I was a bit afraid that the discussion would split to two different places when I posted here. However, I needed to know how and why foobar works as it works. I'd like to continue this discussion there after it is clear that foobar's system can be documented and possibly used as a reference.


In the meanwhile I did some experiments and realized that foobar's "5.1 equation" is used only for weighting the individual channels in a certain way. After that the result is apparently divided by the number of included channels (...or square roots of them, or something like that -- I didn't try to do the math).

Practically it can make a multichannel file on a multichannel system play as loud as a stereo file plays on a stereo system, assuming both systems are calibrated to play a mono test signal through each individual channel equally loud.

On a single multichannel system the resulting replay gain value is not compatible with the value in a standard stereo file.

Here are the steps to reproduce what I did:

1) Create a short sine tone with foobar  (e.g: File > Add location > tone://1000Hz,10seconds)

2) Convert the test tone to a regular 44.1 KHz 16-bit mono wave file with foobar's converter. Name it as "mono wav".

3) Create a stereo and a 6-channel file from the mono file with Sox ( http://sox.sourceforge.net/ ):
stereo: sox -M mono.wav mono.wav stereo.wav
multichannel: sox -M mono.wav mono.wav mono.wav mono.wav mono.wav mono.wav surround.wav

4) Unfortunately foobar doesn't like the header in the surround.wav file and won't play it, but that can be fixed by converting the file to flac by using the standard command line flac encoder with the --channel-map switch, for instance:
flac -8 --channel-map=none surround.wav
(the stereo file can be converted without the switch: flac -8 stereo.wav)

When these two files are analyzed both get the same replay gain value.

Obviously the system can work correctly only when two separate playback system are used. A stereo system for stereo files and a 5.1 system for multichannel files. When both systems are calibrated so that a file a that has exactly equal volume level in all channels (2 or 6) plays at the same volume level in the listening position, the replay gain values are compatible.

When Replay Gain corrected playback happens through a single directly wired multichannel system, which outputs the stereo file only through the left and right front speakers, the multichannel file is 4.77 dB louder. (This is a calculated value that includes the LFE channel. In reality the test signal would need to be e.g. 50 Hz for the LFE channel and all other channels would then need to be connected to full range speakers so that they could reproduce the low frequency signal.)

Replay Gain questions

Reply #14
AlexB, I replied in the other thread too. The Replaygain standard includes a calibration file of -20dBFS pink noise that should measure 83 dB SPL when played on one channel only of a calibrated movie theatre surround sound system.

Therefore, I should expect that file to indicate +6.0 dB Track Gain mapped onto any 1 channel only of a stereo file or a 5.1 surround file that is otherwise silent. I'm note sure how mono "ought" to be treated when played over 2 speakers equally, or how it is mapped in mono-to-stereo converter DSP.

BTW, I think Wavpack is supposed to support multichannel particularly well, if that helps with your experimentation.
Dynamic – the artist formerly known as DickD

Replay Gain questions

Reply #15
There's an old thread about this. I think I made a helpful contribution.


I don't know what fb2k does.


In terms of stereo vs multichannel "equality" - it depends on how you playback the stereo file on the surround system - just through the front L+R speakers, fully duplicated in the rear speakers too, or via some kind of (fake?) surround decoding.

As a starting point, I'd be tempted to assume the first. This implies there can be no further scaling of the front two channels in surround mode - otherwise you'll get different results for native stereo vs exactly the same content encoded to the front L+R speakers of 5.1.

Cheers,
David.
P.S. If you're going to play stereo files on through 4 channels on a surround system, just drop them by a suitable amount (either 3dB or 6dB) to make them match.
P.P.S. I'm reticent to make hard recommendations on this - there's plenty of theory to call on, but with the original ReplayGain I found that listening to the result was as important as relying on theory - I don't have access to a 5.1 system at the moment, so can't listen to the results, so don't really want to comment!

Replay Gain questions

Reply #16
AlexB, I replied in the other thread too. The Replaygain standard includes a calibration file of -20dBFS pink noise that should measure 83 dB SPL when played on one channel only of a calibrated movie theatre surround sound system.

Therefore, I should expect that file to indicate +6.0 dB Track Gain mapped onto any 1 channel only of a stereo file or a 5.1 surround file that is otherwise silent.

If I understood correctly that is practically what I tried explain. Here's another example:
- Have two systems in the same room. A 6-channel system and a stereo system. Have a single channel test tone and calibrate each individual channel on both systems separately (one at a time) so that the volume levels are equal. After calibration, on playback a 6-channel file (e.g. 6 channels at 0 dBFS) on the 6-channel system will be 4.77 dB louder than a stereo file (e.g. two channels at 0  dBFS) on the stereo system when the volume level is measured with an SPL meter that is in the listening position. This is simply caused by the difference in the number of channels because each individual channel in both files contain the same data (for instance, a 0 dBFS test signal).

- Foobar works differently. It assumes that both systems are calibrated to produce the same final output volume level when the above mentioned test files are played, i.e. the individual channels in the multichannel system are set to a lower volume level that compensates the increased number of channels.

The latter is not what happens in reality when the user has only one multi-channel system and no channel mixing/mapping or other postprocessing is applied. The stereo file plays only through L and R and is 4.77 dB quieter than the 6-channel multichannel file that plays through all six channels.

Quote
I'm note sure how mono "ought" to be treated when played over 2 speakers equally, or how it is mapped in mono-to-stereo converter DSP.

That is included in the Replay Gain standard. Mono is assumed to be duplicated to both speakers (a+a): http://replaygain.hydrogenaudio.org/rms_energy.html

Replay Gain questions

Reply #17
There's an old thread about this. I think I made a helpful contribution.

Unfortunately my searches didn't find it.

Quote
I don't know what fb2k does.

Neither do I. That's why I am asking. 

Quote
In terms of stereo vs multichannel "equality" - it depends on how you playback the stereo file on the surround system - just through the front L+R speakers, fully duplicated in the rear speakers too, or via some kind of (fake?) surround decoding.

As a starting point, I'd be tempted to assume the first. This implies there can be no further scaling of the front two channels in surround mode - otherwise you'll get different results for native stereo vs exactly the same content encoded to the front L+R speakers of 5.1

As I tried to explain, there are two possible approaches: the systems are calibrated individually and the files are assumed to be played on the correct system (stereo on stereo and surround on surround) or the files are assumed to be played on a single system that outputs only through the included channels (stereo to L & R and surround to all channels).

Any channel mixing, bass management, fake surround decoding, etc is post-processing and the post-processor must be able to handle the possible volume level issues.

When Replay Gain is analyzed on the source file level the only reasonable approach is to assume that the playback system is "perfect" i.e. that it contains a directly connected full range speaker for each channel and no post-processing happens.

Quote
P.S. If you're going to play stereo files on through 4 channels on a surround system, just drop them by a suitable amount (either 3dB or 6dB) to make them match.

IMHO, that's post-processing and not related to the file analysis.

Quote
P.P.S. I'm reticent to make hard recommendations on this - there's plenty of theory to call on, but with the original ReplayGain I found that listening to the result was as important as relying on theory - I don't have access to a 5.1 system at the moment, so can't listen to the results, so don't really want to comment!

I think this can and should be solved by simple maths because this time we can't evaluate typical surround recordings. There are no such things. A surround recording can be anything. It can contain only some room echo in the surround channels or it can be a psychedelic mix that is intended to be heard as coming from various directions and sometimes from inside of your head.

However, for some basic testing you can easily create a test rig if you have 5.1 outputs on your sound device and three PC speaker pairs (or actually any amp/speaker combinations with input connectors - ghetto blasters, old home stereos - you name it). The system doesn't need to be "Hi-fi" when test signals are used. It is enough that you can adjust the channel levels individually in order to calibrate them.

Replay Gain questions

Reply #18
Because I fail at interpretation of stuff I don't fully understand anyway, just look yourself - that's all I have. If it doesn't correspond with results obtained by experimentation, note it's OLD.
Full-quoting makes you scroll past the same junk over and over.

Replay Gain questions

Reply #19
From your link:
Code: [Select]
...
  if (!check_tempbuf_size(samples*2)) return false;
                rg_float * left = tempbuf, *right = tempbuf + samples;
                unsigned n;
                const rg_input * src = data;
                for(n=0;n<samples;n++)
                {//L/R/C/LF/BL/BR
                        *(left++)   =  (rg_float)(src[0] + src[2]*SQRTHALF + src[4]*SQRTHALF + src[3]);
                        *(right++)  =  (rg_float)(src[1] + src[2]*SQRTHALF + src[5]*SQRTHALF + src[3]);
                        src+=6;
                }
                return AnalyzeSamples(tempbuf,tempbuf+samples,samples,2);
...

I'm not saying that I understand the code any better, but it certainly looks like your interpretation is correct, except that the equation is used only for weighting the individual channels. The final result is adjusted so that 1ch + 1ch + 1ch, etc = 1ch

I indeed did some further experiments and apparently there are later changes. When I reduced the LFE volume in the surround file the replay gain value didn't change at all. So it appears that LFE is removed from the equation.

Changing the volume level in one of the surround channels had an effect.

If the code really doesn't include LFE anymore and is now:

rgL = L + (√2)/2 · C + (√2)/2 · LS
rgR = R + (√2)/2 · C + (√2)/2 · RS

then I think that it would be more proper to include the surround channels as they are and include LFE as splitted to two channels:

rgL = L + (√2)/2 · C + (√2)/2 · LFE + LS
rgR = R + (√2)/2 · C + (√2)/2 · LFE + RS

If the LFE channel contains content that is loud enough to have a significant effect after the equal loudness filter then it certainly also effects the subjectively experienced loudness in the listening position. For instance, a recording may contain a passage that contains only low frequency bass tones and the sound engineer may have decided to mix more or less of that content to the LFE channel.

Replay Gain questions

Reply #20
I did not mean the surround channels are low - I really meant that the surround sources are low - all channels.

I did a bit of playing around last night and found a slightly different overall solution to this issue:

First off, I do play both stereo and surround tracks intermixed through the same system - I use an Auzentech Meridian 7.1 for a sound card which does a pretty good job of making 2 channel into 5.1. 

When I am playing 2 channel, I use the Izotope  Ozone DSP to clean up and bump up the mp3 sound - it automatically senses mulit-channel and turns off.  It turns out it also has both input and output level controls which I carefully decreased with the overall result is a much more balanced sound between the 2 sources!

(On a related issue if anyone knows the developed for foo_dsp_winamp, it needs a bit of tweaking as Izotope looses it's interface controls and needs to be restarted which causes crashes)

I am glad I started this discussion as even though this helps, there is a wide range of levels in 5.1 content that still needs to be adjusted with RG.

Replay Gain questions

Reply #21
There's an old thread about this. I think I made a helpful contribution.

Unfortunately my searches didn't find it.
No, it's not easily found, and when I did find it, it wasn't that good! Might still be right though...
http://www.hydrogenaudio.org/forums/index....st&p=198348

Quote
As I tried to explain, there are two possible approaches: the systems are calibrated individually and the files are assumed to be played on the correct system (stereo on stereo and surround on surround) or the files are assumed to be played on a single system that outputs only through the included channels (stereo to L & R and surround to all channels).
I don't see how those two specific approaches are different. For a stereo file, exactly the same audio reaches the listener from a calibrated stereo system as from a calibrated surround system replaying stereo through front L+R only.

Quote
Any channel mixing, bass management, fake surround decoding, etc is post-processing and the post-processor must be able to handle the possible volume level issues.

When Replay Gain is analyzed on the source file level the only reasonable approach is to assume that the playback system is "perfect" i.e. that it contains a directly connected full range speaker for each channel and no post-processing happens.
Yes, agreed.

Quote
Quote
P.S. If you're going to play stereo files on through 4 channels on a surround system, just drop them by a suitable amount (either 3dB or 6dB) to make them match.

IMHO, that's post-processing and not related to the file analysis.
Exactly.

Quote
Quote
P.P.S. I'm reticent to make hard recommendations on this - there's plenty of theory to call on, but with the original ReplayGain I found that listening to the result was as important as relying on theory - I don't have access to a 5.1 system at the moment, so can't listen to the results, so don't really want to comment!
I think this can and should be solved by simple maths because this time we can't evaluate typical surround recordings. There are no such things. A surround recording can be anything. It can contain only some room echo in the surround channels or it can be a psychedelic mix that is intended to be heard as coming from various directions and sometimes from inside of your head.
...but there is no appropriate "simple maths" - even for stereo signals vs mono signals, it depends on channel correlation etc as to how loud the independent signals will sound when reproduced together, never mind electrically added together. It should be possible to get close enough though.

Quote
However, for some basic testing you can easily create a test rig if you have 5.1 outputs on your sound device and three PC speaker pairs (or actually any amp/speaker combinations with input connectors - ghetto blasters, old home stereos - you name it). The system doesn't need to be "Hi-fi" when test signals are used. It is enough that you can adjust the channel levels individually in order to calibrate them.
Oh, I have plenty of that - I have enough amplifiers and speakers  - nowhere to conveniently set them up while moving house though.

Cheers,
David.

Replay Gain questions

Reply #22
David,

I tired manually setting a replay gain and am now more confused then before - here are my questions..

1 - Does RG work on 5.1 Flac encoded and/or wrapped tracks?  (I tired it and could not hear a difference)

2 - When I open the RG Settings window I have it set to Track - apply gain - is this correct

3 - Why are there 2 Preamp levels - ie what does it mean With or Without RG Info?

4 - It seems to me that it is not saving anything per track, whatever level I setup on one track shows up when I open another track - please explain.

5 - If it actually works, if I select a set of tracks and open settings - will it apply my changes to all tracks I selected?

6 - Does the change take place right away or do I have to click OK first?

7 - How do I set a value for just one track?

8 - Why do I have to "Apply to mp3 tags" - shouldn't this just be automatic?

Thanks!

Barry

Replay Gain questions

Reply #23
I see a little confusion of two separate things here.


To change per-track ReplayGain information in files:
1/ Select one or more tracks.
2/ Hold the shift key and click on them with right mouse button to pull the context menu.
3/ Choose ReplayGain > Edit ReplayGain info.
4/ Enter/change the values *.
5/ Press OK to update the tags in the files.

*) It might be easier to use one of the ReplayGain > Scan selection as... commands first to have a reference, then just change the values a bit here and there.


To change how the player deals with ReplayGain during playback.
1/ Open Preferences >  Playback.
2/ Adjust the options.


For more information how it all works together, see for example [a href='index.php?act=findpost&pid=663183']this post[/a].
The settings for files without RG information is for the files with no ReplayGain tags.


The "Apply to MP3 data" command changes the MP3 file content itself (in contrary to just adding several RG tags), so that its loudness is changed even in players which don't know what RG is and how to handle it.
Full-quoting makes you scroll past the same junk over and over.

Replay Gain questions

Reply #24
The "Apply to MP3 data" command changes the MP3 file content itself (in contrary to just adding several RG tags), so that its loudness is changed even in players which don't know what RG is and how to handle it.

So does that mean MP3Gain is now obsolete?  Does foobar2000 work the same way, and is lossless in that the data can be restored to its original state?