ReplayGain

Hi Mike,

Replay gain doesn't alter the audio data in the file, but it does usually alter the audio data sent to the DAC - i.e. it "controls volume in software" (see last point for the unlikely exception to this).  If Asset is implementing this in a UPnP environment, then one of three things must happen.  If Asset is applying the Replay gain then that indicates scenario 1 is being used.

1    If the renderer is not capable of applying replay gain, then Asset it will alter the data in the PCM or DSD audio stream. before putting the data onto the network.

2    If the renderer is capable of applying replay gain in it's own preprocessing software, then the renderer will scale (i.e. alter) the data before sending the processed data to the DAC (Asset doesn't need to know about this scenario).

3    If the renderer is capable of controlling it's own analogue volume AND supports control of this from the Replay Gain metadata value, then it can apply the replay gain in the analogue domain.  This is unlikely to be implemented (and in Naim space, the 272 doesn't do this).  (Again, Asset doesn't need to know about this scenario.)

Scenarios 1 & 2 inherently degrade sound quality to some degree (even if the effect could be fairly minor in scenario 2).

Hi Mike

Whilst waiting for my NDX2, I am using a BlueSound Node 2 streamer. In Settings it gives a choice of Replay Gain.

1. Disabled

2. Track Gain.

3. Album Gain.

4. Smart Gain

Does Asset have this? (See photo attached) I have tried the above, but couldn't really notice any difference in volume or SQ.

Hi Huge,  thanks for the feedback.  & yes before anyone says it,  I'm guilty of the typo in the title line with 'RelayGain'.   

I understand the ReplayGain process:  In my case it it will require dBpoweramp to attach the ReplayGain ID tags to the metadata & ReplayGain will be "applied" by Asset & it sends the RG changed data stream to the renderer.      I'm aware that apart from attaching the tags - which does not affect SQ,   the Asset UPnP "RG Apply" feature can be turned on & off & its that area that I'm trying to get any SQ change opinions on.      My question was,  apart from volume,  does ReplayGain change the SQ,  & apart from yes it must do so in theory as its no longer 'Bit Perfect',  is any change perceptible.  Illustrate (dBp & Asset) claim the audio is unaffected & other www chat is not conclusive.  So my question is to the bat eared fraternity on the forum that have used it.

Hi Meerkat,  yes all you list is included in dBpoweramp CD Ripper as a DSP sellectable program. Or for post ripping & applying to bought in downloads or to apply it to a whole library,  it's in dBp Convert - add RG ID Tags.    I note that you couldn't really notice any difference in volume or SQ.     The difference in volume will be heard between albums that play at different volumes, RG removes that difference.  

Mike, I've edited for you...

Muchas gracias Richard

Hi Mike,

I haven't tried Replay Gain in Minim, but editing the file to change the volume using Audacity did cause a noticeable degradation of SQ.

Mike,

I use replay gain as tag value with quite a few ripped albums, to avoid big volume differences between tracks on playback. I Use minimserver, not asset. It’s easy to set minimserver to ignore the tag, should I wish so (I believe so at least). 

For daily use I prefer the convenience of relatively similar output levels to a very slightly better SQ without RG (in my view). With a few albums I actually prefer the SQ including RG adjustment (reduced gain). 

Claus 

I've applied ReplayGain ID Tags to a few well known (to me) albums as a test,  ranging from raw rock & roll, acoustic folk, jazz & fusion, coral & full orchestra & all a mix of 16 & 24 bit WAV & DSD.   All I can say is I don't hear anything different other than a similar volume level on each album,  they are quiet (pianissimo) & loud when they're supposed to be & as intended in the recording,  leading edge & general dynamics seem unaffected.  If anything the presentation appears more full bodied,  maybe because the volume control position is around the clock a little more.  Whatever I have decided to attach ReplayGain ID tags to the whole library overnight,  I estimate it could take at least 8-hours if the 10x test albums were a measure.   I will keep ReplayGain running as a test for as long as it takes for me to come to a conclusion one way or other;  and that will be well into December as I'm heading to southern latitudes for a while.  And if I do decide not to keep it,  it's a simple touch of a 'check box' to turn it all off,  leaving the RG ID tags in the metadata cannot affect the audio.

Mike

Are you upscaling your 16 bit files to 24 bits when applying Replay Gain?

sjbabbey posted:

Mike

Are you upscaling your 16 bit files to 24 bits when applying Replay Gain?

No, haven't tried that yet.   I read a bit about it but something did not read right & I didn't really understand.   Tell me more:    What does it do to SQ ??   Is it activated in the check-box next to ReplayGain Apply in the Asset configuration screen & if its a good thing to do why not set it automatically when RG is applied ???  

My thinking is that Asset have included this option to try to mitigate against any possible artifacts or other negative effects which might be introduced by the replay gain being applied. I guess Mr Spoon would be able to give a definitive answer.

I have to admit to not having experimented with replay gain as the thought of altering the metadata on several thousands of individual tracks would fill me with dread.

sjbabbey posted:

I have to admit to not having experimented with replay gain as the thought of altering the metadata on several thousands of individual tracks would fill me with dread.

Yes it did take a time,  I have around 10,000 tracks & it took about 13 hours to complete.

I will keep digging & have an experiment with up sampling

  I've applied Album Gain as opposed to Track gain, this plays the individual tracks at the volume the mixing engineer intended. It does make a better presentation with albums that are intended as a whole rather than a series of individual tracks. 

  Set 'Increase to 24-bit', this I believe applies head room to allow enough bit depth for the 16-bit ReplayGain process. Can't say I hear anything I can attribute to this.

All sounds good, Album Gain works well, no audible upside in basic SQ but nothing detrimental either that I can hear.

Replay gain it’s simplest form can normaluse a track or increase the gain. This in its crudest sense can increase the value of 16 bit samples in the renderer which won’t be ideal. For higher sample values effectibevley providing greater resolution this can be done with less distortion more easily...

For true volume gain control a sort of compression is typically required where the average volume energy is increased, but very few media software renderer devices that I am aware of do this as this is more an involved DSP function, and doing crudely will not sound good.

Yes Simon,  I’m aware of all that but unlike something better, this DSP function is freely available & no mattter what you say, it does do more or less what it says on the box.  I’m trying it to hear for myself what it can do, good and/or bad.  So far it’s a mix of good & neutral, nothing so far obviously bad.   If eventually after due course I do decide not to use it,  it’s a simple ‘check-box’ switch off.  Nothing lost other than my time,  but some interesting testing & mostly nice listening in the process. 

Indeed, it’s all there to tinker with.. but if such things matter to you, you are very much departing from sample bit accuracy. Certainly compression can be powerful and be part of the creative process, I sometimes master some mixes for a friend’s band and I use some specialised software compressors and the effect they have when used correctly is phenomenal in adding clarity to a track such as allowing a bass line to be more noticeable in a mix, making vocals clearer, making the drums feel to time better etc... to achieve some of this you need to often use something could ‘sidechaining’.. where you modulate the compressor with a trigger channel... but back to media rendering and it’s DSP compressor if available to you.. the compression effects will be more like that used by radio stations etc.. and if it makes the track sound more enjoyable or accessible for you then it’s got to be a good thing.

Mike-B posted:

  I’m trying it to hear for myself what it can do, good and/or bad.  So far it’s a mix of good & neutral, nothing so far obviously bad.   If eventually after due course I do decide not to use it,  it’s a simple ‘check-box’ switch off.  Nothing lost other than my time,  but some interesting testing & mostly nice listening in the process. 

OK  all done.   Bottom line is I prefer it without ReplayGain for serious listening.   It does what it says on the box,  it brings down or raises volume so all albums have a similar level.  It does not affect dynamics, startling rim shots still startle, loud passages are still loud & quiet is still just as quiet.   It works well when playing a mix of PCM pre & post 'loudness wars' recordings & more so with DSD (all DSD have lower volume)  as it avoids having to adjust volume from album to album.   But running a few selected tracks, & going back & forth with A-B-A-B comparisons, there is a slight softening with top end detail, lightness & presence,  I hate to use the 'veil' word, but something subtle is lifted with it switched out.    The nice thing with this on Asset UPnP is that its so easy to switch it in & out,  so for play lists, mixing up albums & whenever I feel like it,  its available.      

Hi Mike compression does affect dynamics, but quality compressors usually have a timing threshold such that transients are not compressed and are passed through so it doesn’t sound compressed... there is also a timing threshold on release too... and for example in EDM these qualities are exaggerated and taken advantage when mix mastering  to give a groove to the sound or a pumping feel.

But yes this processing is mangling the signal to some extent.. so I can totally understand your preference for not using for serious replay SQ.

Thanks for the update Mike

Volume change in digital is a multiplication which is not that expensive but will result in truncation. It should be followed by adding dither which gives the effect of rounding instead of truncation. Normal triangular dither add about 1.5dB of noise.

To make this even more fun it is not uncommon in mastering to add weighted dither as the last stage (like Apogee UV22). That type of material should be left alone. Normally you use triangular (just plain noise).

Its been subject to much discussion how audible truncation of 24-bit material is. While mixing where many operations is done it varies. Today it is common to use double precision floting point (64-bit) for DSP and then you only dither when converting back to 24-bit for storage or between sections.

So ideally you would not want to mess with the bits from the CD/download (I never do). For high quality playback I would do volume change in analog unless there is some fancy Dsp-trick I dont know about.