I have been ripping in BrainDead profile with "--xlevel --ms 15" options using v1.14 . What other options would increase the quality theoretically with minimal additional bitrate addition? What exactly does "--minval 2" do? Do you recommend v1.15 for such bitrates?

Also how is the difference in quality with alpha SV8 encoders compared to v1.15? Assuming there will be a lossless SV7->SV8 transcoder, is the additional quality difference worth to wait?

I've just abxed an improvement over 1.15r in 1.95z67 : 7/8 (standard setting, I don't know if braindead can be ABXed. Maybe with the Amnesia sample).

NOOOOOO!!!!!!!!
Don't start tweaking commandlines if or you'll have undesired effects!! They're only for experts and aren't used at all.THAT'S EVIL!!
MPC is optimized for using with the regular switches:
Use --quality 8 --xlevel . Period.

Hmmm. That's surely a way to gain credibility. Why quality 8? Why not 10? Or 7.63?

People spend too much time worrying about quality switches when Klemm has already done the dirty work for you. --standard --xlevel is all you need for first-generation playback.

Almost all music has no problems with --standard and the few samples that have problems with --standard generally have problems in higher settings as well. There is a reason --braindead is called Braindead: for those braindead people who absolutely have to waste every single possible bit when there is no need to.

Hey it's my problem if I waste away my bits. The difference of 10% between insane and braindead profile is not much for me to spare. 300kbps on average is reasonable, it is still half the size of lossless being almost lossless. I just bought a 120GB drive for my music and it looks like all my cds could take only 40GB with 10% par2 redundancy included I have many CDs of organ and harpsichord music which lossy encoders have problems, I want to make sure of the quality.

"--ms 15" is a safe switch to use with Braindead and could bring higher quality. Do a search. I thought there might be other such switches that are safe to use with braindead profile and would only add 2-3% to space cost. It is worth for my peace of mind. And hey, there're people using lossless out there. It's a free world.

I respect you opinion , however you'll never reach absolutely transparent audio on absolutely all (100%) possible cases unless you use lossless.There will be always a rare sample that will not be encoded transparently , so if you really want to get transparent audio in all cases no matter how rare you have to choose lossless.
IMHO no lossy codec is "almost lossless" as you say , no matter how high quality the codec , even if you go at "braindead" levels , no matter how many switches you use to improve quality , there will be always a "problem sample" waiting to be encoded out there.
I'm no "braindead" audiophile so I basically choose between:

1)For general music I want to keep (I don't care if it's "absolutely transparent in all cases"): MPC standard
2)For important albums/favorite music: Lossless

In the end it's a highly subjective topic but my philosophy is something like "only waste storage space if transparency is a must , in that case go lossless".

Well as you said , it's a free world (Not really absolutely free in all cases )

Before experimenting with "expert" switches you can increase bloat by using a higher quality setting like --quality 9.12 instead of the profiles (braindead translates to quality 8). This is safer and will probably give you the highest quality per bitrate-increase ratio (at least in theory. You won't be able to hear the difference anyway.) If you think you can find better settings than Frank Klemm and co, take a look at mppenc --longhelp.

If your concern is quality, don't wait for SV8. Main focus of SV8 is not increasing quality (since quality allready is close to perfect) but to create a more robust and flexible container. The only major quality issue that will be fixed in SV8 is internal clipping. Some remaining artifacts are difficult to fix in the current stream version, which SV8 should change.

I thought people only did this crap with MP3. What's this world coming to???

Using "--quality 8 --xlevel" with "--ms 15" is SAFE and COULD give better quality! ms 13 is default in q8 otherwise. Read the thread carefully. I asked for such switches that are safe. Other than that I have no intention to screw up the presets.

atici:

I think the reason that you're finding a mostly negative response to your question is that, once again, this community is pretty much founded upon the basis of blind testing, etc. If you're admitted beforehand that you're tuning beyond the level of audibility and that the discussion is only "theoretical", then it's not surprising that most people aren't interested, or even that they are annoyed.

Honestly, I don't see much of a point in this discussion either. I think the whole idea of using switches which will provide no audible difference, and almost admitting as much, pretty much goes against the whole idea of this place. It gives people the wrong idea about what is actually necessary.

If you want to spark a discussion about better quality, find an existing problem beforehand and point it out to us.

@Dibrom

But we know some switches could do better in some samples, that's why we pick higher presets other than standard. That's why I am asking, to be on the safe side (yes quite safe please). And as I said there're people who use lossless for double the bitrate. But if it's the case, as some claimed, for MPC problem samples increasing the preset and the Mid/Side stereo mode would not solve the problem then I'd agree with you there's no point in discussing. On HA we mind : "It won't probably provide audible difference but what if it does", or "What if using this switch potentially could solve a problem that could otherwise be heard."

I personally think the --braindead quality switch is overkill, as least it is for my hearing capabilities since the --standard switch is typically transparent to me, although not always transparent. I tend to go for the placebo effect and use --xtreme --xlevel, which seems "safer," although it is probably unnecessary as I doubt my hearing will improve.

atici, unproperly tuned presets are not the origin of artifacts in mpc; it's the psymodel that allows for such slipups. Unfortunately it is impossible to create a perfect psymodel. That means that there is no single magic switch or a combination of switches which could cure the remaining flaws in the current psymodel thus creating a perfect one. All you can do is create overall bloat, hoping that your (potential) artifact will disappear. As I said, this is best done (and most safely done) using the quality scale.

In the past people have been playing with

• the ltq settings - the most reasonable to use nowadays would be ltq_gain x. x<0 will increase bitrate (and theoretically quality).
  
• nmt/tmn - bigger is better. For ultimate quality use --nmt 99 --tmn 99. In the past some used nmt 16 and tmn 32. Their excuse was "archival quality".
  
• minSMR - bigger is better. In the past was used to disable full spectrum encoding in the insane profile (iirc).

These switches can be used to increase or decrease bitrate. If used with the correct values, none should have any negative effects besides increasing bitrate. Like --ms 15 all can lead to a potential quality increase. However all of these switches scale with the presets/quality in an optimized fashion, so you can get the most out of your bitrate. There is no reason why you should be experimenting with these on their own, unless you want to have higher bitrates than quality 10 offers. Really, someone who knew what he was doing figured out which combination would give you the best effect which resulted in the quality scale.

I remember the debate back then, when the quality scale was first introduced. People thought they were taken their freedom to use whatever settings they wanted. I thought so myself. But from what I know, nobody is using the advanced switches anymore because there simply is no need to do so. You want higher quality? Just increase one setting and you will get higher quality! Of course you could tweak nmt/tmn and all that other stuff for higher quality on your own, but if you knew what you were doing you would end up somewhere on the quality scale again.

When you are tweaking in a range where there is no hearable difference from one setting to another you risk to actually decrease quality. Sure, your files will still sound great, but the bloat has increased. There is no sense in tweaking when you cannot verify the results.

HAHAHAHAHA Thanks for the extra funny switches Gecko!!!! Check this out:

Dimmu Borgir - Puritanical Euphoric Misanthropia - 06 - Puritania.ofr - 22,591KB Dimmu Borgir - Puritanical Euphoric Misanthropia - 06 - Puritania.mpc - 23,856KB

LOSSY BIGGER THAN LOSSLESS!!!!!!!!!!!!!! HAHAHAHAHAHAHA!!!!!!!

Switches used: --quality 10 --ms 15 Itq_gain 0.1 --nmt 99 --tmn 99

Oh, no! You are using ltq_gain the wrong way! ltq_gain has to use negative values to increase bitrate! Quality 10 allready uses ltq_gain of -30 so you are actually decreasing bitrate! Try "--ltq_gain -80".

It is interesting to see that bitrates don't vary much at this level of quality. With the small clip I tested it was around 1250kbps +- 30kbps only for each frame. (Checked with mpcbits).

Interestingly enough it absolutely will not get any bigger than 23,865Kb no matter what I do:

--quality 10 --ms 15 Itq_gain 0.1 --nmt 99 --tmn 99 is the same as
--quality 10 --ms 15 Itq_gain -80 --nmt 99 --tmn 99 and even
--quality 10 Itq_gain 0.1 --nmt 99 --tmn 99

Is it possible that --nmt 99 and --tmn 99 override the other switches??

Hm, there may be two possible effects at work here, but I'm just guessing.
1) ~1250kbps is the upper limit of mpc.
2) When you set nmt or tnm very high or modify ltq-gain accordingly, your 16 bit source file doesn't have this dynamic range you require for the masking. If I understand the correctly, then nmt 17 (="noise masks tone") means that a tonal signal must be 17db lower than the noisy signal in order to be considered as masked. (This may very well be a false interpretation.) So if you want 99db masking and your input file only has 96db (disregarding dithering) dynamic range, you can never reach this value and mpc does the best job possible: something close to 96db.

I don't know enough about the workings of mpc to understand how these switches affect encoder performance. (That's also a reason why I don't mess with them and just use the quality scale). But I guess it boils down to you demanding higher masking than the noise floor of your input.

Seems to me the obvious solution here is to just go lossless. By your own measurements, your CDs would then occupy 80 GB (with the same 10% PAR2 data) and all quality concerns would be gone forever. You'd still have 40 GB left for your other stuff, and if that's not enough - how much does a cheap 80 GB drive cost anyway?

Cheers,
Uosdwis

Umm refer to the one of the latest threads about this quality discussion: this for instance. Lossless occupies 3.5 times more space for me. All my archive is around 70GB in size now in MPC "q8 --ms 15". A good 120GB drive costs around $100.

Sorry, I was just quoting your own figures. Based on them, lossless appeared to be the way to go, really. Tweak away then, if it makes you happy.

/ Uosdwis