Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Matt]

If you have DSD files in your library, please run Library Tools > Analyze Audio on them.

Then, report the highest peak level you see for your DSD files.  It should be a percentage, like 80%, 90%, etc.

I'm asking because I got into an interesting discussion where someone suggested it should be safe to boost the level when doing DSD to PCM by +6dB.  However, there's some squish in this number so it's not clear it's a universally good idea.

If we sample enough real world files, I think we'll have our answer.

Thanks for any help.

[Otello]

No, unfortunately it's NOT safe to boost +6dB all DSD files; it should be a standard, but it's not universally respected.

[6233638]

It's going to be some time until my library has finished scanning, as I have spent the last couple of weeks ripping a lot of discs, and have been skipping the analysis of DSD files since having the ability to bitstream them. (no ReplayGain)
Scanning DSD seems to take significantly longer than any other format on my system.

So far I have one album with peaks of 93%, 96%, and 100%.
Everything else is below 88% (−6 dB) with the highest being 82% so far, but there's a lot to go.


Is this why Bitstreaming DSD is louder than decoding to PCM? (a +6dB boost?)
I also noticed that I had a track with a peak of 115% in my library! (not DSD) Is this a worst-case scenario for intersample overs?


And on the subject of DSD, might I suggest either extending the 24kHz, 48dB/octave filter to 30kHz, or changing the 30kHz from 24dB/octave to 48dB/octave.
Since you introduced DSD to PCM conversion options, I have been doing some analysis of my library, and some discs do have information which extends beyond the limits of the 24kHz filter, but the 30kHz filter leaves a lot of ultrasonic noise.
If I use the 50kHz filter and then run a 30kHz/48dB low-pass DSP on it, that seems to pass the full signal range without any of the ultrasonic noise. (at least down to around -120dB)

[6233638]

Sorry for the delay, but that's everything scanned now. Aside from that one album (notorious for being poorly mastered) the highest peak in my DSD tracks is 82%.

[cmbe]

Hi,

After almost two days, my DSD files are finally analyzed: 252 Albums/1292 tracks.

The top 10 files are:

 (all from the same album)
 1 - 100%
 2 - 96%
 3 - 93%
 (other albums)
 4 - 82%
 5 - 81%
 6 - 79%
 7 - 79%
 8 - 78%
 9 - 77%
10 - 77%

[tiberiuspv]

105 SACD, 1140 tracks, all classical music.
Max peak level is 70%, min is 4%. Number of tracks in 5% buckets above 50% peak level:
Exactly 70%: 8
65%.. 69%: 22
60%.. 64%: 34
55%..59%: 51
50%.. 54%: 195
I don't use replay gain in general, as it does not necessarily do the right thing for classical music...

[Henglander]

Only 250 files Classical and Prog rock
Max Peak level 68%; Min Peak level 14%

125 Files < 50% Peak level

[Henglander]

I would be happy if I could simply edit the album replay gain

[Matt]

Thanks for the help everyone.

It looks like there's not really any safe amount of volume to add to the existing DSD-to-PCM conversion.

Several users have a peak of 70% (right about -3dB), but that could just be a common mastering level.  And since some users have files with peaks of 100%, it means there's no extra headroom.

Thanks again.

[bulldogger]

If you have DSD files in your library, please run Library Tools > Analyze Audio on them.

Then, report the highest peak level you see for your DSD files.  It should be a percentage, like 80%, 90%, etc.

I'm asking because I got into an interesting discussion where someone suggested it should be safe to boost the level when doing DSD to PCM by +6dB.  However, there's some squish in this number so it's not clear it's a universally good idea.

If we sample enough real world files, I think we'll have our answer.

Thanks for any help.

How do you analyze the entire library of DSD files? I see Library Tools>Analyze Audio but I can only do a file, at a time. I have about 550 iso files. Perhaps the ability to set a manual boost if all are your files are below a safe threshold would work? I guess that would mean that you would have to analyze each file each time you added a new one.

[JJJ]

If you have DSD files in your library, please run Library Tools > Analyze Audio on them.

Then, report the highest peak level you see for your DSD files.  It should be a percentage, like 80%, 90%, etc.

I'm asking because I got into an interesting discussion where someone suggested it should be safe to boost the level when doing DSD to PCM by +6dB.  However, there's some squish in this number so it's not clear it's a universally good idea.

If we sample enough real world files, I think we'll have our answer.

Thanks for any help.

A DSD->PCM conversion should apply a 6dB gain. In relatively rare occasions, poor mastering will mean doing so creates clipping.

It's funny but just last night I played SACD ISOs for the first time.  The lower volume versus my Oppo player was immediately noticeable and I had asked on another forum whether the standard 6dB gain was being applied by JRiver.  I was referred here.

Whenever I have done DSD->PCM conversions I have always used SoX to check there is sufficient headroom. Hasn't been an issue so far.

Couldn't a dynamic headroom check be built into DSP->Volume Control.  I was actually looking for it there last night.  One could Analyze SACD ISOs and JRiver evaluates and saves a gain setting for each ISO (6dB or less if applicable). It would be a case of analyzing only once (say automatically on import).

[ted_b]

Matt and gang,
I have over 1200 SACDs in my JRiver library.  I would recommend the following:
* do not analyze audio for an ISO as it will also scan the multichannel content which may have VERY different levels.  (Note:  all my 1200 SACDs are extracted to stereo DSF folders ...and multichannel DSF folders if applicable)
* do NOT apply 6db gain to DSD-to-PCM conversion automatically.  If I did that (and I don't any more cuz I have DSD-capable DACs) I would be lopping off as much as 3db off the peaks of 30% of my albums...i.e clipping like mad.   

You need to do what Audiogate now allows (which unfortunately adds 6db as the 0db default); first choose normalize (bad name, sounds like replaygain is added, which it isn't) then it analyze the peaks for the album (Audiogate calls it "normalize playlist") then automaticallyset the conversion to that (less 6b).  In most cases it will be like 2-3db under it (i.e it will read -2.9db in Audiogate).  In Saracon you have the option to add 6b as a setup default, but that is simply an option due to most folks finding it's too hot.

The net/net is that 6db was supposed the be the right number cuz DSD modulators are 6DB quieter to prevent overload, but..guess what...some albums are recorded too hot (go figure    ).

[JJJ]

Ted and I are broadly in agreement except I think JRiver should look at the stereo content of the ISO and determine and appropriate dB gain for that 'album' - 6dB if it doesn't produce clipping; less than 6dB if appropriate (in my experience it is relatively rare that 6dB is indeed too hot, especially in classical recordings) - and look at the multichannel content (if applicable) and determine an appropriate gain for that content. This could be done when a new ISO is added to the library and recorded.

(Audiogate finally caught up on the issue.  In Saracon one could manually enter a more appropriate gain level.)

[6233638]

Media Center can already do analysis of SACD ISO. (this is how the peak levels discussed in this topic were determined)

[Matt]

Since there's no amount of gain that can always be added, we'll continue with no change by default.

Note that this is independent of peak-level normalization and/or volume leveling -- both of which can be optionally applied at playback time when playing DSD as PCM with MC19.

[ted_b]

Media Center can already do analysis of SACD ISO. (this is how the peak levels discussed in this topic were determined)

Yes, but are they done independently (2 channel and multi) cuz "Bob" mastered the 2 channel and it's 2db hot; "Bill" mastered the multichannel and it's 3db soft.  Happens all the time.  Don't convert the multichannel with the hottest peaks on the ISO or it will still be too soft.

[JJJ]

Ted's right.

Since there's no amount of gain that can always be added, we'll continue with no change by default.

Note that this is independent of peak-level normalization and/or volume leveling -- both of which can be optionally applied at playback time when playing DSD as PCM with MC19.

Well there is an amount which can be added...it's just not necessarily constant.  In most circumstances the 6dB amount would be right.  In some circumstances it may be too much. A quick analysis will determine the right amount. Can a level of gain be stored by the user, for example, for the stereo album within an SACD ISO which would be applied solely to that album (and not the multichannel album in that ISO)?