Not streaming - dedicated PC

[ssully]

I agree with the premise. If they are level matched they should sound the same. In my case or philosophy I would much rather leave the source format as is, which means leave the ISO as is and convert on the fly if necessary. Do you notice a difference in gain between say playing an SACD directly versus playing back the converted PCM files. In my setup I notice about an 8 dB difference. Are you compensating for some of this by applying gain in the conversion process ?

Yes, as needed. Depending on the SACD mastering choices, peak of DSD may be as much as -6dbFS in PCM (SACD spec recommends 6dB of headroom in DSD mastering so that if converted to PCM it's unlikely to overload. That's quite conservative so often the level of the PCM version can be increased several dB without overload). So when I use Foobar2Ks SACD plugin to convert to PCM, I start at +6 gain, and watch the conversion process proceed track by track using F2Ks console window , and if there's an overload I stop and restart conversion at +5. Repeat as needed, lowering the gain by 1dB, until there are no overloads during conversion. ('Thriller' appears not to meet spec; it overload briefly on some tracks even when there is 0dB gain increase during PCM conversion)

 

[ssully]

I use ReplayGain on all my FLACs to level match my MCH music as I tend to flick around various albums and there’s quite a difference between many albums. ReplyGain doesn’t alter the waveform it’s just metadata to tell your player to raise/lower volume by x dB to match the norm level.

That works too, if there is no conversion. But I don't use Replaygain because a certain amount of my hard drive content is 'raw' DTS or Dolby files, which have to be passed bit-perfectly (i.e., without any DSP applied) . Replaygain turns them into white noise.

 

[ssully]

I found this in the linn.co.uk user forum, but you can find similar explanations elsewhere:

"If a DSD file is converted to PCM with no gain adjustment it will sound 6dB quieter than the equivalant native PCM file. This gain difference is deliberate and is required in DSD to prevent overload in the sigma-delta modulator."

He goes on to say that you can't blindly bump the gain by 6dB because the DSD modulator may have peaks that exceed -6dB.

This is exactly true. After 'blind' conversion you have to check that there were no overloads (or monitor conversion in real time, as I do).

I read elsewhere that the DSD spec allows up to plus or minus 3.1dB variation from the -6dB standard; so you would probably be safe by blindly adding 3dB gain.

This is generally true, IME.

I usually measure the dff peaks directly with foobar2k dynamic range meter and then add gain to the conversion based on the highest peak reported by DRM.

That should work too! I guess I should install the DR meter plugin...

 

[LuvMyQuad]

Is it like an SBE?

Im not sure what that is. ????

 

[LuvMyQuad]

How is having the files on two computers underestimating the possibility of a crash?

I meant, if you had no backup system now with all files on your main PC.

 

[LuvMyQuad]

Yes, the thing that's always kept me from RAID is having both copies in one box, which provides little protection against water/fire damage, or theft.

My current "off-site storage" is portable USB drives, but I eventually want to go to a second NAS box, tucked away at the other end of the building, doing nightly automated file sync.

The idea is supposed to be this....

RAID backup takes care of the most common type of failure... a crashed drive. A failure within the system itself. With a RAID system there may never even be a service interruption due to a drive crash.

And you are right. RAID does not protect you against a catastrophic on site failure. And very possibly neither does keeping the same files on two machines in the same location. In the event of fire, flood, theft, etc both machines may well be affected.

The recommendation most often given is, in addition to the RAID, the final precaution is an additional off site backup. Most people use either cloud storage or physical backups on portable drives, The move the portables to another site.

 

[Jim the Oldbie]

...I don't use Replaygain because a certain amount of my hard drive content is 'raw' DTS or Dolby files, which have to be passed bit-perfectly (i.e., without any DSP applied) . Replaygain turns them into white noise.

If DTS44 .wav files are converted to FLAC without decoding (still in noise form), ReplayGain can then be calculated & applied as metadata only. These files can then be decoded & played with ReplayGain compensation in Foobar. (Haven't tried it with Dolby44 .wavs.)

[Edit] This trick may not have worked in earlier versions of Foobar, if I remember correctly; ReplayGain would just analyze the noise. But in current versions, DTS44 is apparently detected & decoded during ReplayGain analysis, yielding the same RG numbers as for the pre-decoded, multi-channel file.

 

[atrocity]

But I don't use Replaygain because a certain amount of my hard drive content is 'raw' DTS or Dolby files, which have to be passed bit-perfectly (i.e., without any DSP applied) . Replaygain turns them into white noise.

I use ReplayGain with Logitech Media Server. I just don't have RG values calculated or stored for DTS or AC-3 files, so it all works.

 

[atrocity]

I don't think there's a way to use ReplayGain on an SACD .iso (Someone please correct me if I'm wrong about this) which is the main reason I convert them to FLAC. I usually do 2 conversions. The first determines the album peak as described by sukothai above. I then adjust the decibel boost accordingly in Foobar's SACD preferences before the final conversion, for maximum level. (A bit obsessive, but there it is.)

I haven't found a way to apply RG to any type of DSD file at all.

If you're using RG on your DSD->FLAC conversions, is it really necessary to bake any volume boost into the FLAC file? Not trying to say you're wrong, just wondering if there's a benefit I'm missing.

 

[markshan]

Im not sure what that is. ????

Sector Boundary Error. It causes clicks on Redbook CDs on some players.

 

[atrocity]

Sector Boundary Error. It causes clicks on Redbook CDs on some players.

I think the problem has more to do with the burning software than the player, but I could be wrong. At least some burning software will (or so I've been told) automatically fill in the end of a track with an SBE using the data from the next track.

That live Devo CD/DVD from several years back actually has SBEs, which is incredibly annoying since a live show by definition needs to be completely uninterrupted.

http://forums.stevehoffman.tv/threads/sector-boundary-errors-and-redbook.258773/

 

[markshan]

A boundary error will only cause an interruption if it is corrected by padding the end of the file with zeros. There is a free program called Traders Little Helper which will correct SBE by padding or by shifting.

What I meant by "some players" is not that the player causes the SBE, but rather that some players don't care about it and play right through, while others introduce the click where the end of the boundary should be. If you play a disc with SBEs in Foobar it does not click, which is one reason some traders pass along discs with errors unknowingly.

 

[atrocity]

A boundary error will only cause an interruption if it is corrected by padding the end of the file with zeros. There is a free program called Traders Little Helper which will correct SBE by padding or by shifting.

What I meant by "some players" is not that the player causes the SBE, but rather that some players don't care about it and play right through, while others introduce the click where the end of the boundary should be. If you play a disc with SBEs in Foobar it does not click, which is one reason some traders pass along discs with errors unknowingly.

Wow, I didn't know that about Foobar. I'm trying to wrap my head around how that would work since I would expect an SBE to be indistinguishable from whatever other reason a file may have one or more digital zeroes. I wonder if there's a way to leverage that to correct a bad disc.

 

[markshan]

Wow, I didn't know that about Foobar. I'm trying to wrap my head around how that would work since I would expect an SBE to be indistinguishable from whatever other reason a file may have one or more digital zeroes. I wonder if there's a way to leverage that to correct a bad disc.

Because the Boundary is a function of Redbook CD. It expects the data to come in these "sectors". It does all of its processing based on them. Once the file is taken out of the realm of CDA, it no longer "needs" to conform to that standard.

The SBE doesn't add zeros, the correction of the SBE adds zeros if you (or your software) choose to correct it in that manner. That's why Traders Little Helper has the "pad" option to fix SBEs, or the "shift boundary" option. Think of the entire file as being a foot on a ruler and the sectors as being inches. If you draw new lines at 15/16ths spots, that doesn't change the length of the ruler, just the marking of that section.

Once zero padding has been introduced, there isn't any way I know of to remove it except old fashioned splicing. I've done that a couple times on shows that I was remastering/correcting.

That guy "lucpac" gives a pretty good explanation on that link you provided above.

 

[Jim the Oldbie]

...If you're using RG on your DSD->FLAC conversions, is it really necessary to bake any volume boost into the FLAC file? Not trying to say you're wrong, just wondering if there's a benefit I'm missing.

No, you're not missing anything at all. In fact, the more I think about this, the more inclined I am to just leave Foobar's SACD gain set to 0 dB for conversions to eliminate any possibility of clipping, and let ReplayGain do its job. Hell, the entire gain adjustment range is only 6 dB, and 6 dB is only one bit...

 

[sukothai]

I may be over obsessive, but I don't like anything messing with my audio bits before they get to the DAC. So, RG is not an option for me.

 

[atrocity]

Because the Boundary is a function of Redbook CD. It expects the data to come in these "sectors". It does all of its processing based on them. Once the file is taken out of the realm of CDA, it no longer "needs" to conform to that standard.

I completely misunderstood your point about Foobar, sorry. I somehow got it in my head that you were talking about Foobar playing CDs (which it does) and gracefully dealing with disc-based SBEs.

 

[markshan]

I completely misunderstood your point about Foobar, sorry. I somehow got it in my head that you were talking about Foobar playing CDs (which it does) and gracefully dealing with disc-based SBEs.

That was one of my points. Foobar won't make a click sound when a CD has SBEs.

 

[atrocity]

That was one of my points. Foobar won't make a click sound when a CD has SBEs.

Now I'm confused again. Are you saying that if Foobar2000 is playing a CD currently spinning in a drive and that CD has SBEs, those SBEs will no longer cause audible problems?

If so, I'm back to wondering how it knows the difference between an SBE and a normal reason for digital zeroes. I guess you could code something along the lines of "If it's the end of the track and there are 587 or fewer zero samples..."

I think the source of my confusion is that we're talking about files vs. CD-based but I'm missing a clear dividing line between them. I know that SBEs are by definition irrelevant to file-based playback, but if Foobar is spinning a disc then whatever issues exist on the disc would still exist in the output stream unless there's some technique to suppress them.

Apologies if I'm just being dense.

 

[markshan]

I think the source of your confusion is that you are still talking about SBEs and files corrected with padded zeros interchangably. An SBE is simply a marker placed in the wrong spot. It has nothing to do with extra zeros. Most early CD players will create a click at the SBE because they are expecting more data to be contained in that sector. Foobar doesn't work on sectors, because it is reading data rather than interpreting strictly by the Redbook standard.

If someone has incorrectly repaired the SBE of a live recording by adding zeros, Foobar certainly will play that silent gap. That gap has now become part of the data.