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: lossyWAV 1.1.0 Development Thread. (Read 182058 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

lossyWAV 1.1.0 Development Thread.

Reply #75
V1.0.1g  -q 0 Vs. Original

foo_abx 1.3.3 report
foobar2000 v0.9.5.2
2008/05/22 13:00:52

File A: C:\Documents and Settings\b.lossy.wav
File B: C:\Documents and Settings\b.wav

13:00:52 : Test started.
13:01:09 : 01/01  50.0%
13:01:21 : 02/02  25.0%
13:01:32 : 03/03  12.5%
13:01:44 : 04/04  6.3%
13:02:02 : 05/05  3.1%
13:02:17 : 06/06  1.6%
13:02:32 : 07/07  0.8%
13:02:52 : 08/08  0.4%
13:03:04 : 09/09  0.2%
13:03:15 : 10/10  0.1%
13:03:28 : 11/11  0.0%
13:03:42 : 12/12  0.0%
13:04:01 : Test finished.

----------
Total: 12/12 (0.0%)

For me V1.0.1g  -q 0 = V1.0.1f -q 0.5 it's better but the artefact is  still obvious & easy to spot. (3 min to ABX)

lossyWAV 1.1.0 Development Thread.

Reply #76
For me V1.0.1g  -q 0 = V1.0.1f -q 0.5 it's better but the artefact is  still obvious & easy to spot. (3 min to ABX)
Thanks for that - taking into account what David said about reality checking and that the lowest preset *should* introduce noticable artifacts, I think that I will step back a bit from the settings I put in place for 1.0.1g.

If I keep the spreading_function string at '22222-22222-22223-12224-12234-12345' but reduce the minimum_bits_to_keep to 2.333 and 2.667 respectively for -q 0 and -q 1, the resultant bitrate for -q 0 is 312kbps/325kbps and 352kbps/367kbps (default / --linkchannels) for my 54 problem sample set (to which I must add Mardel's latest problem sample).

I feel that this is a reasonable compromise as the HF regions are given a little more influence in the result at a small change in bitrate.

lossyWAV beta 1.0.1h attached to post #1 in this thread.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV 1.1.0 Development Thread.

Reply #77
With this sample and -q 0 setting, it looks like level of added noise changes very frequently, and this modulation is more noticeable than the noise itself.

lossyWAV 1.1.0 Development Thread.

Reply #78
At present you would have to wavegain the file before processing with lossyWAV. What upper limit do you think would be reasonable, as it's an easy fix (1 value changed in the nparameters unit)?

Up till now, with dozens of tracks to go, the biggest album difference is +6.23 dB. Don't know what scale factor that is.

lossyWAV 1.1.0 Development Thread.

Reply #79
At present you would have to wavegain the file before processing with lossyWAV. What upper limit do you think would be reasonable, as it's an easy fix (1 value changed in the nparameters unit)?
Up till now, with dozens of tracks to go, the biggest album difference is +6.23 dB. Don't know what scale factor that is.
About 2.0488, I reckon.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV 1.1.0 Development Thread.

Reply #80
... I think it's very useful to have at least one non-transparent setting at the bottom of the range - it lets people hear what kind of artefacts lossyWAV is adding. It also confirm to the average user that the quality scale works, and stops people from saying "even q0 is transparent - what happens if you lower the bitrate still further". ...

This is a great argument IMO. We should really stop to expect -q 0 being transparent. Good quality as a rule should be enough.


I can ABX -q 2 too but it takes 23min with a pause in the middle ...

Thank you, sauvage 78. Your results show that -q 2 isn't transparent with this sample, but (please correct me if I'm wrong) seems to be a setting which is 'close enough to transparency' to be useful for very high quality DAP use. To a slightly minor degree this can be said also of -q 1.5.
Remains the question: What quality setting brings full transparency?

...  haven't heard any sound artifacts at -q 2.5 --lowpass 17000.

Thank you, Mardel.


If I keep the spreading_function string at '22222-22222-22223-12224-12234-12345' but reduce the minimum_bits_to_keep to 2.333 and 2.667 respectively for -q 0 and -q 1, ...

I am having trouble to keep up with the changes.
What's exactly the difference between 1.0.1h and 1.0.1f?
1.0.1h spf string = '22222-22222-22223-12224-12234-12345'? What was it with 1.0.1f?
1.0.1h minimum_bits_to_keep to 2.333 and 2.667 respectively for -q 0 and -q 1? What was it with 1.0.1f?
Any other changes?

Anyway thank you for your restless struggle for improvement.
lame3995o -Q1.7 --lowpass 17

lossyWAV 1.1.0 Development Thread.

Reply #81
If I keep the spreading_function string at '22222-22222-22223-12224-12234-12345' but reduce the minimum_bits_to_keep to 2.333 and 2.667 respectively for -q 0 and -q 1, ...
I am having trouble to keep up with the changes.
What's exactly the difference between 1.0.1h and 1.0.1f?
1.0.1h spf string = '22222-22222-22223-12224-12234-12345'? What was it with 1.0.1f?
1.0.1h minimum_bits_to_keep to 2.333 and 2.667 respectively for -q 0 and -q 1? What was it with 1.0.1f?
Any other changes?

Anyway thank you for your restless struggle for improvement.
1.0.1f had the existing --spf string, 1.0.1g & h have the modified one, although the minimum_bits_to_keep was:

2.000 and 2.500 for -q 0 and -q 1 respectively in 1.0.1f.
3.000 and 3.000 for -q 0 and -q 1 respectively in 1.0.1g;
2.333 and 2.667 for -q 0 and -q 1 respectively in 1.0.1h;

I increased the "resolution" of the threshold index so that fewer bits are potentially lost due to the requirement to floor the spreading_function result to read the corresponding bits_to_remove value from the threshold_index array. By widening the threshold_index array by a factor (now 64) the flooring of  (spreading_function result * 64) has less of an impact.

I will leave --lowpass implemented as it will give those with more acute hearing some scope to include frequencies in the analyses above the existing 16kHz limit.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV 1.1.0 Development Thread.

Reply #82
1.0.1f had the existing --spf string, 1.0.1g & h have the modified one, although the minimum_bits_to_keep was:

2.000 and 2.500 for -q 0 and -q 1 respectively in 1.0.1f.
3.000 and 3.000 for -q 0 and -q 1 respectively in 1.0.1g;
2.333 and 2.667 for -q 0 and -q 1 respectively in 1.0.1h;

I increased the "resolution" of the threshold index ...

Thank you, Nick. So we have 1.0.1f's spf string with 1.0.1h - I appreciate this very much.
The other things make lossyWAV more defensive without a remarkable bad impact on bitrate as I understand it. So that's fine.
lame3995o -Q1.7 --lowpass 17

lossyWAV 1.1.0 Development Thread.

Reply #83
With this sample and -q 0 setting, it looks like level of added noise changes very frequently, and this modulation is more noticeable than the noise itself.
I'll have a look at the bits-to-remove detail for the sample at -q 0 and see if it is large stepping in the bits-to-remove (i.e. added noise) - from what you say, it would seem to be.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV 1.1.0 Development Thread.

Reply #84
halb27:
you're right -q 2 was near transparent with 1.0.1f & it is even more with 1.0.1h.
I tried to quickly ABX 1.0.1h -q2 vs original but I gave up after 5min as it was a boring task, I think it is still ABXable but it will take me 40 min just to prove to others something I am already sure for myself & even if -q2 is ABXable the fact that I can hear a difference doesn't mean that this difference sounds ugly.

So yes -q2 was listenable for DAP with this specific sample, specially in a noisy environnement, but my personal interest for lossywav is not in DAP even if I like the flexibility of using any lossless format.

Nick.C:
I was thinking lossywav was just removing unearable random noise in the back that takes much space to losslessly encode, but obviously it also add its own noise. I was thinking lossywav was as a kind of "lossless audio cleaner" by optimizing wav before encoding to lossless, I was thinking lossywav was not introducing "quantization like noise", but I was wrong it seems that even without quantization it add lossy-like additionnal noise. So I haven't followed all the lossywav developments, but now I am interested in knowing if lossywav applies some encoding strategy at some settings & not at some other settings. Even  without understanding in deep the used techniques, just to understand the logical step within the scale.

Why is Q0 to Q4 for DAP, Q5 magically called "transparent", Q6 & Q7 as good as lossless & Q8 to Q10 for transcoding ??? My technical knowledge is very limited but I would like to know if these claims are rationnal or just arbitrary. Is it just me or lossywav have become more lossy than it was supposed to be in the beginning ? If there is no different techniques at different settings isn't the scale a little too wide ? Wouldn't it be logic to have only 2 recommended settings within the scale covering each specific use:

"-near lossy" for DAP, covering Q0 to Q5 use, targetting transparency.
"-near lossless" for archiving, covering Q6 to Q10 use, targetting transcodable archiving.

I don't mean that the actual scale is bad, I like it, freedom is always good. But it is missleading for newbies as I ended using Q0 for archive thinking lossywav was overkill anyway no matter the setting. I dunno but

"-near lossy" could be the same as let's say -q 3 (ABXing can help here)
"-near lossless" could be the same as ... well I don't know but the least agressive possible while still optimizing the lossless encoding. (ABXing is useless here, only technical knowledge so I let the codecs developpers decide together, not only you Nick.C ... Monty, Ivan, Gabriel may have an hintfull opinion here)

I know this is still very arbitrary but it would help newbies using lossywav as it should, & not lowering its reputation ...

for me for there is only 4 uses for hybrid:
100% lossless for Archive + "-near lossy" for DAP.
"-near lossless" for Archive + 100% lossy for DAP.
"-near lossless" for Archive + "-near lossy" for DAP.
hybrid lossless part for Archive + hybrid lossy part for listening.

so in the end two well tuned settings can cover it all.

lossyWAV 1.1.0 Development Thread.

Reply #85
halb27:
you're right -q 2 was near transparent with 1.0.1f & it is even more with 1.0.1h.
I tried to quickly ABX 1.0.1h -q2 vs original but I gave up after 5min as it was a boring task, I think it is still ABXable but it will take me 40 min just to prove to others something I am already sure for myself & even if -q2 is ABXable the fact that I can hear a difference doesn't mean that this difference sounds ugly.

So yes -q2 was listenable for DAP with this specific sample, specially in a noisy environnement, but my personal interest for lossywav is not in DAP even if I like the flexibility of using any lossless format.

Nick.C:
I was thinking lossywav was just removing unearable random noise in the back that takes much space to losslessly encode, but obviously it also add its own noise. I was thinking lossywav was as a kind of "lossless audio cleaner" by optimizing wav before encoding to lossless, I was thinking lossywav was not introducing "quantization like noise", but I was wrong it seems that even without quantization it add lossy-like additionnal noise. So I haven't followed all the lossywav developments, but now I am interested in knowing if lossywav applies some encoding strategy at some settings & not at some other settings. Even  without understanding in deep the used techniques, just to understand the logical step within the scale.

Why is Q0 to Q4 for DAP, Q5 magically called "transparent", Q6 & Q7 as good as lossless & Q8 to Q10 for transcoding ??? My technical knowledge is very limited but I would like to know if these claims are rationnal or just arbitrary. Is it just me or lossywav have become more lossy than it was supposed to be in the beginning ? If there is no different techniques at different settings isn't the scale is little too wide ? Wouldn't it be logic to have only 2 recommended settings covering each specific use:

-near lossy for DAP, covering Q0 to Q5, targetting transparency.
-near lossless for archiving, covering Q6 to Q10, targetting transcoding.

I don't mean that the actual scale is bad, I like it, freedom is always good. But it is missleading for newbies as I ended using Q0 for archive thinking lossywav was overkill anyway no matter the setting. I dunno but

-near lossy could be the same as -q 3 (ABXing can help here)
-near lossless could be the same as ... well I don't know but the least agressive possible while still optimizing the lossless encoding. (ABXing is useless here, only technical knowledge so let the codecs developpers decide together, not only Nick.C)

I know this is very arbitrary but it would help newbies using lossywav as it should, & not lowering its reputation ...
lossyWAV has always added noise to the output, see the original lossyFLAC post started by David (2Bdecided) - the method used is his.

The Delphi transcode of the Matlab original started as just that - an attempt to duplicate David's method in a form that more users could make use of - Matlab is after all rather expensive....

Once agreement between both versions was achieved (about 0.3.0 roughly from memory), other ideas blossomed and added to the method.

The preset scale started as -1, -2 and -3. -1 = best, -2 = transparent, -3 = more aggressive. This evolved over time into -1 to -5, -1 to -7 and finally -q 10 to -q 0. As it states in the help:
Code: [Select]
Quality Options:

-q, --quality <n>   quality preset (10=highest quality, 0=lowest bitrate;
                    -q 5 is generally accepted to be transparent)
                    default=-q 5.
On the subject of transparency, as many problem samples as have been pointed out to halb27 and me have been included in the refinement of the quality presets. v0.6.4 RC1 was thought to be good until the day after release Alex B found a problem sample that halb27 also ABX'ed.

Transparency at -q 5 is not an arbitrary claim - although not every possible sample has been tested, therefore it is still "only" a claim (it should be noted that a very small group of users has been involved in the developmental ABX'ing process).

However, unless the optional (and still beta) noise shaping is selected, the actual bit removal process only rounds off the least significant <bits-to-remove> bits of each sample in a codec-block, thus adding white noise. If noise shaping is selected, the added noise is modified and a lot of it is pushed much higher in the spectrum (thanks to SebastianG's noise shaping coefficients).

Another, more advanced, method of noise shaping is in the early stages of development at present - but don't hold your breath.

Personally, I only use lossyWAV & FLAC as a DAP transcode - my lossless FLAC archive will remain intact.

The bottom line on the quality presets: 10 = best, 0 = worst - use one which you are comfortable with.

Thanks for taking the time to post.

Nick.

With this sample and -q 0 setting, it looks like level of added noise changes very frequently, and this modulation is more noticeable than the noise itself.
I'll have a look at the bits-to-remove detail for the sample at -q 0 and see if it is large stepping in the bits-to-remove (i.e. added noise) - from what you say, it would seem to be.
I had a look:
Code: [Select]
Z:\WAV\tmp>lossywav "..\_swav\01 - Ginnungagap - The Black Hole.original.wav" -q 0 --detail -f
lossyWAV beta 1.0.1h, Copyright © 2007,2008 Nick Currie. Copyleft.
This is free software under the GNU GPLv3+ license; There is NO WARRANTY, to
the extent permitted by law. <http://www.gnu.org/licenses/> for details.
%lossyWAV Warning% : Detailled output mode enabled
Processing : 01 - Ginnungagap - The Black Hole.original.wav
Format    : 44.10kHz; 2 ch.; 16 bit.
Progress  :
Block    Time  00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 Tot.
====================================================================
    0    0.00s.  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0
  16    0.19s.  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0
  32    0.37s.  5  7 10 10 10  9  9 10  9  9  9  8  7  7  9  8 136
  48    0.56s.  9 11 10 10 10 10  9 10  9  9  9  9  9  9  8  9 150
  64    0.74s.  7  9  8  8  6  7  6  8  7  7  7  7  7  7  8 10 119
  80    0.93s. 10 10 10  9  9  9  9  8  8  8  8  9  8  8  9 11 143
  96    1.11s. 10 11 11 11 11 11 11 11 11 10 11 10 10 10 10 10 169
  112    1.30s. 10 10 10  9 10 10  9  9 10 10 10 10  9  9  9 10 154
  128    1.49s.  9 10  9  9 10 10 10  9  9 10 10  9  8  9  9  8 148
  144    1.67s.  9  9  9  9  9  8  9  9  9  9  9  9  9  9  8  8 141
  160    1.86s.  8  9  8  9  9  9  9  9  9  8  9  9  9  9  9  8 140
  176    2.04s.  8  8  9  8  8  8  9  9  8  9  8  8  9  8  8  8 133
  192    2.23s.  7  8  8  8  8  8  8  8  8  8  8  8  8  8  8  8 127
  208    2.41s.  7  8  7  7  8  8  7  8  8  7  8  8  7  7  7  8 120
  224    2.60s.  7  8  7  7  6  8  7  8  7  7  7  7  7  7  7  6 113
  240    2.79s.  7  7  8  6  7  6  7  6  7  7  6  7  7  7  7  7 109
  256    2.97s.  7  7  7  7  7  7  6  7  7  7  7  6  6  7  7  7 109
  272    3.16s.  6  6  7  7  7  7  7  8  8  7  8  8  8  8  8  8 118
  288    3.34s.  7  8  7  8  7  7  8  7  7  7  7  7  7  7  7  7 115
  304    3.53s.  7  7  7  8  8  8  7  7  7  7  7  8  7  7  8  6 116
  320    3.72s.  7  7  7  7  7  7  7  7  7  7  7  7  7  6  6  7 110
  336    3.90s.  7  7  5  7  7  6  7  6  6  6  7  6  6  7  7  7 104
  352    4.09s.  7  7  7  7  7  7  6  7  6  6  7  7  6  7  7  7 108
  368    4.27s.  6  7  6  6  6  6  6  7  7  7  7  6  7  6  7  6 103
  384    4.46s.  6  7  6  6  6  7  6  6  6  6  6  6  6  6  7  6  99
  400    4.64s.  6  6  7  7  6  6  7  6  6  7  7  7  7  6  6  6 103
  416    4.83s.  6  6  5  6  6  6  6  6  6  6  6  6  6  5  6  5  93
  432    5.02s.  6  6  5  6  6  6  6  6  6  6  6  5  6  5  6  6  93
  448    5.20s.  5  6  6  6  6  6  6  6  5  5  5  5  6  5  5  6  89
  464    5.39s.  6  6  6  5  5  6  6  6  6  6  5  6  6  5  5  5  90
  480    5.57s.  5  6  5  5  6  6  6  6  6  6  5  5  6  6  6  6  91
  496    5.76s.  6  5  5  5  6  5  5  6  6  6  6  6  5  6  5  6  89
  512    5.94s.  6  5  5  5  5  5  6  6 10 10 10 10  9  9  9 10 120
  528    6.13s.  9  8  8  8  9  9  8 10 10 10  9 10 10  9  9  9 145
  544    6.32s.  9  9  8  8  8  7  8  7  8  8  8  7  7  7  7  7 123
  560    6.50s.  7  6  7  7  7 10 10 10 10 10  9  8  9  9  9  9 137
  576    6.69s.  8  8  8  8  8 11 11 11 11 11 10 11 10 11 10 11 158
  592    6.87s. 10 11 11 10 10 10 10 10 10 10 10  9 10 10  9  9 159
  608    7.06s.  9  9  9 10 10 10  9  9  9 10 10 10 10 10 10  9 153
  624    7.24s. 10  9  9  9  9  9  9  9 10 10  9  9 10  9  9  9 148
  640    7.43s.  9  9 10  9  9  9  7  9  9  9  9  9  8  9  9  8 141
  656    7.62s.  9  9  9  8  9  9  9  8  9  8  9  8  8  8  8  8 136
  672    7.80s.  9  8  8  8  8  8  7  7  8  9  9  8  8  9  8  7 129
  688    7.99s.  7  7  8  8  8  8  8  8  9  7  7  8  7  7  8  7 122
  704    8.17s.  8  8  8  8  7  7  7  7  8  6  7  7  6  7  8  6 115
  720    8.36s.  7  7  8  8  7  7  7  7  7  8  7  7  6  6  7  6 112
  736    8.54s.  7  7  6  7  5  6  7  7  6  6  5  7  6  7  7  7 103
  752    8.73s.  7  7  6  6  6  7  7  7  7  7  7  8  8  7  8  8 113
  768    8.92s.  8  8  8  8  7  7  7  7  7  7  7  6  6  7  8  7 115
  784    9.10s.  8  7  7  7  7  6  8  7  7  7  7  7  7  8  8  7 115
  800    9.29s.  7  7  7  7  7  7  7  8  8  7  7  7  7  7  7  7 114
  816    9.47s.  7  7  6  6  7  7  7  7  8  8  7  7  8  7  7  7 113
  832    9.66s.  7  7  7  7  7  7  6  7  7  8  7  7  7  6  6  7 110
  848    9.85s.  7  6  7  5  7  7  6  6  7  7  7  7  7  7  7  7 107
  864  10.03s.  6  6  6  7  6  7  7  6  7  5  6  6  4  6  7  7  99
  880  10.22s.  7  6  6  7  8  7  8  7  8  7  8  8  8  6  8  8 117
  896  10.40s.  8  8  8  7  7  7  8  7  8  7  7  7  7  8  7  7 118
  912  10.59s.  7  7  6  7  7  7  7  7  5  7  6  6  7  6  6  7 105
  928  10.77s.  7  7  7  6  6  7  7  7  7  7  7  7  7  7  8  7 111
  944  10.96s.  7  8  7  7  7  7  8  8  7  7  8  8  7  7  7  7 117
  960  11.15s.  7  7  7  7  7  7  7  7  7  7  7  7  8  7  7  7 113
  976  11.33s.  7  7  7  7  7  7  7  7  7  7  7  7  7  7  7  7 112
  992  11.52s.  7  7  7  6  6  7  7  7  7  6  7  7  7  6  8 11 113
 1008  11.70s. 10 10 10 10 10  9  9  9  8  8  8  8  8  9 10  9 145
 1024  11.89s. 10  9  9 10 10  9  9  9  9  8  8  9  8  8  9  7 141
 1040  12.07s.  7  7  8  8  7  8  7  7  7  6  7  8 10 10 10  9 126
 1056  12.26s. 10  9  9  9  9  8  8  8  8  8  8  8 11 11 11 11 146
 1072  12.45s. 11 11 10 11 10 11 11 10 11 10 11 10 10 10 10 10 167
 1088  12.63s. 10 10 10  9  9  9  7  8  9 10  9 10  9  9  9 10 147
 1104  12.82s. 10  8 10 10  9 10 10 10  9 10  9  9  9 10  9 10 152
 1120  13.00s.  7  9  9  8  9  9  8  9  9  9  9  9  9  8  9  9 139
 1136  13.19s.  8  9  9  9  9  9  9  9  7  9  9  9  9  9  8  7 138
 1152  13.37s.  9  7  8  8  8  8  8  8  8  8  8  8  8  8  8  8 128
 1168  13.56s.  7  8  8  7  8  5  8  8  8  8  8  7  7  7  8  8 120
 1184  13.75s.  7  8  8  8  8  8  8  8  8  8  7  8  7  7  8  7 123
 1200  13.93s.  7  7  8  7  7  7  7  7  6  8  8  7  7  7  7  6 113
 1216  14.12s.  6  7  6  7  7  7  7  6  7  8  6  6  7  7  7  7 108
 1232  14.30s.  7  7  7  7  7  7  6  7  7  7  7  7  7  7  7  7 111
 1248  14.49s.  7  7  7  8  8  8  8  8  7  8  6  8  7  6  6  7 116
 1264  14.68s.  8  7  8  7  6  8  7  8  8  8  7  8  6  7  8  7 118
 1280  14.86s.  7  8  8  7  7  7  7  7  7  7  7  7  6  6  7  6 111
 1296  15.05s.  7  7  7  7  7  6  7  7  6  6  7  7  7  7  7  7 109
 1312  15.23s.  7  6  7  7  7  7  7  7  7  7  6  6  7  6  7  6 107
 1328  15.42s.  7  7  6  6  7  7  6  7  7  7  7  7  7  7  7  7 109
 1344  15.60s.  7  5  6  6  6  7  7  6  6  6  6  7  7  6  6  5  99
 1360  15.79s.  6  7  6  7  6  6  6  6  6  6  6  6  6  7  7  7 101
 1376  15.98s.  7  6  7  5  5  6  6  6  6  6  6  6  6  6  6  6  96
 1392  16.16s.  7  6  7  5  6  6  5  7  6  5  5  6  6  5  6  6  94
 1408  16.35s.  6  6  6  6  6  6  6  5  6  6  6  6  6  6  5  5
Average    : 2.78MB; 7.4112 bits; 32.93x; 0.50s;
%lossyWAV Warning% : 0.0112 bits not removed due to clipping.
%lossyWAV Warning% : 17 sample(s) clipped to limiting amplitude.
Code: [Select]
Z:\WAV\tmp>lossywav "..\_swav\01 - Ginnungagap - The Black Hole.original.wav" -q 0 --detail -f --minbits 3
lossyWAV beta 1.0.1h, Copyright © 2007,2008 Nick Currie. Copyleft.
This is free software under the GNU GPLv3+ license; There is NO WARRANTY, to
the extent permitted by law. <http://www.gnu.org/licenses/> for details.
%lossyWAV Warning% : Detailled output mode enabled
Processing : 01 - Ginnungagap - The Black Hole.original.wav
Format    : 44.10kHz; 2 ch.; 16 bit.
Progress  :
Block    Time  00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 Tot.
====================================================================
    0    0.00s.  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0
  16    0.19s.  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0
  32    0.37s.  5  7 10 10 10  9  9 10  9  9  9  8  7  7  9  8 136
  48    0.56s.  9 10 10 10 10 10  9 10  9  9  9  9  9  9  8  9 149
  64    0.74s.  7  9  8  8  6  7  6  8  7  7  7  7  7  7  8 10 119
  80    0.93s. 10 10 10  9  9  9  9  8  8  8  8  9  8  8  9 10 142
  96    1.11s. 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 160
  112    1.30s.  9 10 10  9 10  9  9  9 10 10 10 10  9  9  9 10 152
  128    1.49s.  9 10  9  9  9 10 10  9  9 10 10  9  8  9  9  8 147
  144    1.67s.  9  9  9  9  9  8  9  9  9  9  9  9  9  9  8  8 141
  160    1.86s.  8  9  8  9  9  9  9  9  9  8  9  9  9  9  9  8 140
  176    2.04s.  8  8  9  8  8  8  9  9  8  9  8  8  9  8  8  8 133
  192    2.23s.  7  8  8  8  8  8  8  8  8  8  8  8  8  8  8  8 127
  208    2.41s.  7  8  7  7  8  8  7  8  8  7  8  8  7  7  7  8 120
  224    2.60s.  7  8  7  7  6  8  7  8  7  7  7  7  7  7  7  6 113
  240    2.79s.  7  7  8  6  7  6  7  6  7  7  6  7  7  7  7  7 109
  256    2.97s.  7  7  7  7  7  7  6  7  7  7  7  6  6  7  7  7 109
  272    3.16s.  6  6  7  7  7  7  7  8  8  7  8  8  8  8  8  8 118
  288    3.34s.  7  8  7  8  7  7  8  7  7  7  7  7  7  7  7  7 115
  304    3.53s.  7  7  7  8  8  8  7  7  7  7  7  8  7  7  8  6 116
  320    3.72s.  7  7  7  7  7  7  7  7  7  7  7  7  7  6  6  7 110
  336    3.90s.  7  7  5  7  7  6  7  6  6  6  7  6  6  7  7  7 104
  352    4.09s.  7  7  7  7  7  7  6  7  6  6  7  7  6  7  7  7 108
  368    4.27s.  6  7  6  6  6  6  6  7  7  7  7  6  7  6  7  6 103
  384    4.46s.  6  7  6  6  6  7  6  6  6  6  6  6  6  6  7  6  99
  400    4.64s.  6  6  7  7  6  6  7  6  6  7  7  7  7  6  6  6 103
  416    4.83s.  6  6  5  6  6  6  6  6  6  6  6  6  6  5  6  5  93
  432    5.02s.  6  6  5  6  6  6  6  6  6  6  6  5  6  5  6  6  93
  448    5.20s.  5  6  6  6  6  6  6  6  5  5  5  5  6  5  5  6  89
  464    5.39s.  6  6  6  5  5  6  6  6  6  6  5  6  6  5  5  5  90
  480    5.57s.  5  6  5  5  6  6  6  6  6  6  5  5  6  6  6  6  91
  496    5.76s.  6  5  5  5  6  5  5  6  6  6  6  6  5  6  5  6  89
  512    5.94s.  6  5  5  5  5  5  6  6 10 10 10 10  9  9  9 10 120
  528    6.13s.  9  8  8  8  9  9  8 10 10 10  9 10 10  9  9  9 145
  544    6.32s.  9  9  8  8  8  7  8  7  8  8  8  7  7  7  7  7 123
  560    6.50s.  7  6  7  7  7 10 10 10 10 10  9  8  9  9  9  9 137
  576    6.69s.  8  8  8  8  8 11 11 11 10 10 10 10 10 10 10 10 153
  592    6.87s. 10 10 10 10 10 10 10 10 10 10 10  9 10 10  9  9 157
  608    7.06s.  9  9  9 10 10 10  9  9  9 10 10 10 10 10 10  9 153
  624    7.24s. 10  9  9  9  9  9  9  9 10  9  9  9  9  9  9  9 146
  640    7.43s.  9  9 10  9  9  9  7  9  9  9  9  9  8  9  9  8 141
  656    7.62s.  9  9  9  8  9  9  9  8  9  8  9  8  8  8  8  8 136
  672    7.80s.  9  8  8  8  8  8  7  7  8  9  9  8  8  9  8  7 129
  688    7.99s.  7  7  8  8  8  8  8  8  9  7  7  8  7  7  8  7 122
  704    8.17s.  8  8  8  8  7  7  7  7  8  6  7  7  6  7  8  6 115
  720    8.36s.  7  7  8  8  7  7  7  7  7  8  7  7  6  6  7  6 112
  736    8.54s.  7  7  6  7  5  6  7  7  6  6  5  7  6  7  7  7 103
  752    8.73s.  7  7  6  6  6  7  7  7  7  7  7  8  8  7  8  8 113
  768    8.92s.  8  8  8  8  7  7  7  7  7  7  7  6  6  7  8  7 115
  784    9.10s.  8  7  7  7  7  6  8  7  7  7  7  7  7  8  8  7 115
  800    9.29s.  7  7  7  7  7  7  7  8  8  7  7  7  7  7  7  7 114
  816    9.47s.  7  7  6  6  7  7  7  7  8  8  7  7  8  7  7  7 113
  832    9.66s.  7  7  7  7  7  7  6  7  7  8  7  7  7  6  6  7 110
  848    9.85s.  7  6  7  5  7  7  6  6  7  7  7  7  7  7  7  7 107
  864  10.03s.  6  6  6  7  6  7  7  6  7  5  6  6  4  6  7  7  99
  880  10.22s.  7  6  6  7  8  7  8  7  8  7  8  8  8  6  8  8 117
  896  10.40s.  8  8  8  7  7  7  8  7  8  7  7  7  7  8  7  7 118
  912  10.59s.  7  7  6  7  7  7  7  7  5  7  6  6  7  6  6  7 105
  928  10.77s.  7  7  7  6  6  7  7  7  7  7  7  7  7  7  8  7 111
  944  10.96s.  7  8  7  7  7  7  8  8  7  7  8  8  7  7  7  7 117
  960  11.15s.  7  7  7  7  7  7  7  7  7  7  7  7  8  7  7  7 113
  976  11.33s.  7  7  7  7  7  7  7  7  7  7  7  7  7  7  7  7 112
  992  11.52s.  7  7  7  6  6  7  7  7  7  6  7  7  7  6  8 11 113
 1008  11.70s. 10 10 10 10 10  9  9  9  8  8  8  8  8  9 10  9 145
 1024  11.89s. 10  9  9 10  9  9  9  9  9  8  8  9  8  8  9  7 140
 1040  12.07s.  7  7  8  8  7  8  7  7  7  6  7  8 10 10 10  9 126
 1056  12.26s. 10  9  9  9  9  8  8  8  8  8  8  8 11 11 11 10 145
 1072  12.45s. 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 160
 1088  12.63s. 10 10 10  9  9  9  7  8  9 10  9 10  9  9  9 10 147
 1104  12.82s. 10  8  9 10  9 10 10 10  9 10  9  9  9 10  9 10 151
 1120  13.00s.  7  9  9  8  9  9  8  9  9  9  9  9  9  8  9  9 139
 1136  13.19s.  8  9  9  9  9  9  9  9  7  9  9  9  9  9  8  7 138
 1152  13.37s.  9  7  8  8  8  8  8  8  8  8  8  8  8  8  8  8 128
 1168  13.56s.  7  8  8  7  8  5  8  8  8  8  8  7  7  7  8  8 120
 1184  13.75s.  7  8  8  8  8  8  8  8  8  8  7  8  7  7  8  7 123
 1200  13.93s.  7  7  8  7  7  7  7  7  6  8  8  7  7  7  7  6 113
 1216  14.12s.  6  7  6  7  7  7  7  6  7  8  6  6  7  7  7  7 108
 1232  14.30s.  7  7  7  7  7  7  6  7  7  7  7  7  7  7  7  7 111
 1248  14.49s.  7  7  7  8  8  8  8  8  7  8  6  8  7  6  6  7 116
 1264  14.68s.  8  7  8  7  6  8  7  8  8  8  7  8  6  7  8  7 118
 1280  14.86s.  7  8  8  7  7  7  7  7  7  7  7  7  6  6  7  6 111
 1296  15.05s.  7  7  7  7  7  6  7  7  6  6  7  7  7  7  7  7 109
 1312  15.23s.  7  6  7  7  7  7  7  7  7  7  6  6  7  6  7  6 107
 1328  15.42s.  7  7  6  6  7  7  6  7  7  7  7  7  7  7  7  7 109
 1344  15.60s.  7  5  6  6  6  7  7  6  6  6  6  7  7  6  6  5  99
 1360  15.79s.  6  7  6  7  6  6  6  6  6  6  6  6  6  7  7  7 101
 1376  15.98s.  7  6  7  5  5  6  6  6  6  6  6  6  6  6  6  6  96
 1392  16.16s.  7  6  7  5  6  6  5  7  6  5  5  6  6  5  6  6  94
 1408  16.35s.  6  6  6  6  6  6  6  5  6  6  6  6  6  6  5  5
Average    : 2.78MB; 7.3919 bits; 38.17x; 0.43s;
%lossyWAV Warning% : 0.0056 bits not removed due to clipping.
%lossyWAV Warning% : 3 sample(s) clipped to limiting amplitude.
Code: [Select]
Z:\WAV\tmp>lossywav "..\_swav\01 - Ginnungagap - The Black Hole.original.wav" -q 0 --detail -f --minbits 3.5
lossyWAV beta 1.0.1h, Copyright © 2007,2008 Nick Currie. Copyleft.
This is free software under the GNU GPLv3+ license; There is NO WARRANTY, to
the extent permitted by law. <http://www.gnu.org/licenses/> for details.
%lossyWAV Warning% : Detailled output mode enabled
Processing : 01 - Ginnungagap - The Black Hole.original.wav
Format    : 44.10kHz; 2 ch.; 16 bit.
Progress  :
Block    Time  00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 Tot.
====================================================================
    0    0.00s.  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0
  16    0.19s.  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0
  32    0.37s.  5  7 10 10 10  9  9  9  9  9  9  8  7  7  9  8 135
  48    0.56s.  9  9 10 10 10  9  9  9  9  9  9  9  9  9  8  9 146
  64    0.74s.  7  9  8  8  6  7  6  8  7  7  7  7  7  7  8 10 119
  80    0.93s. 10 10 10  9  9  9  9  8  8  8  8  9  8  8  9 10 142
  96    1.11s. 10 10 10  9 10  9  9 10  9 10  9  9  9  9 10  9 151
  112    1.30s.  9  9  9  9  9  9  9  9  9  9  9 10  9  9  9  9 145
  128    1.49s.  9  9  9  9  9  9  9  9  9  9  9  9  8  9  9  8 142
  144    1.67s.  9  9  9  9  9  8  9  9  9  9  9  9  9  9  8  8 141
  160    1.86s.  8  9  8  9  9  9  9  9  9  8  9  9  9  9  9  8 140
  176    2.04s.  8  8  9  8  8  8  9  9  8  9  8  8  9  8  8  8 133
  192    2.23s.  7  8  8  8  8  8  8  8  8  8  8  8  8  8  8  8 127
  208    2.41s.  7  8  7  7  8  8  7  8  8  7  8  8  7  7  7  8 120
  224    2.60s.  7  8  7  7  6  8  7  8  7  7  7  7  7  7  7  6 113
  240    2.79s.  7  7  8  6  7  6  7  6  7  7  6  7  7  7  7  7 109
  256    2.97s.  7  7  7  7  7  7  6  7  7  7  7  6  6  7  7  7 109
  272    3.16s.  6  6  7  7  7  7  7  8  8  7  8  8  8  8  8  8 118
  288    3.34s.  7  8  7  8  7  7  8  7  7  7  7  7  7  7  7  7 115
  304    3.53s.  7  7  7  8  8  8  7  7  7  7  7  8  7  7  8  6 116
  320    3.72s.  7  7  7  7  7  7  7  7  7  7  7  7  7  6  6  7 110
  336    3.90s.  7  7  5  7  7  6  7  6  6  6  7  6  6  7  7  7 104
  352    4.09s.  7  7  7  7  7  7  6  7  6  6  7  7  6  7  7  7 108
  368    4.27s.  6  7  6  6  6  6  6  7  7  7  7  6  7  6  7  6 103
  384    4.46s.  6  7  6  6  6  7  6  6  6  6  6  6  6  6  7  6  99
  400    4.64s.  6  6  7  7  6  6  7  6  6  7  7  7  7  6  6  6 103
  416    4.83s.  6  6  5  6  6  6  6  6  6  6  6  6  6  5  6  5  93
  432    5.02s.  6  6  5  6  6  6  6  6  6  6  6  5  6  5  6  6  93
  448    5.20s.  5  6  6  6  6  6  6  6  5  5  5  5  6  5  5  6  89
  464    5.39s.  6  6  6  5  5  6  6  6  6  6  5  6  6  5  5  5  90
  480    5.57s.  5  6  5  5  6  6  6  6  6  6  5  5  6  6  6  6  91
  496    5.76s.  6  5  5  5  6  5  5  6  6  6  6  6  5  6  5  6  89
  512    5.94s.  6  5  5  5  5  5  6  6 10 10 10 10  9  9  9  9 119
  528    6.13s.  9  8  8  8  9  9  8 10 10 10  9 10  9  9  9  9 144
  544    6.32s.  9  9  8  8  8  7  8  7  8  8  8  7  7  7  7  7 123
  560    6.50s.  7  6  7  7  7 10 10 10 10 10  9  8  9  9  9  9 137
  576    6.69s.  8  8  8  8  8 10 10 11 10 10  9 10  9 10 10 10 149
  592    6.87s. 10 10 10  9  9 10  9 10  9  9  9  9  9  9  9  9 149
  608    7.06s.  9  9  9  9 10  9  9  9  9  9  9  9  9  9  9  9 145
  624    7.24s.  9  9  9  9  9  9  9  9  9  9  9  9  9  9  9  9 144
  640    7.43s.  9  9  9  9  9  9  7  9  9  9  9  9  8  9  9  8 140
  656    7.62s.  9  9  9  8  9  9  9  8  9  8  9  8  8  8  8  8 136
  672    7.80s.  9  8  8  8  8  8  7  7  8  9  9  8  8  9  8  7 129
  688    7.99s.  7  7  8  8  8  8  8  8  8  7  7  8  7  7  8  7 121
  704    8.17s.  8  8  8  8  7  7  7  7  8  6  7  7  6  7  8  6 115
  720    8.36s.  7  7  8  8  7  7  7  7  7  8  7  7  6  6  7  6 112
  736    8.54s.  7  7  6  7  5  6  7  7  6  6  5  7  6  7  7  7 103
  752    8.73s.  7  7  6  6  6  7  7  7  7  7  7  8  8  7  8  8 113
  768    8.92s.  8  8  8  8  7  7  7  7  7  7  7  6  6  7  8  7 115
  784    9.10s.  8  7  7  7  7  6  8  7  7  7  7  7  7  8  8  7 115
  800    9.29s.  7  7  7  7  7  7  7  8  8  7  7  7  7  7  7  7 114
  816    9.47s.  7  7  6  6  7  7  7  7  8  8  7  7  8  7  7  7 113
  832    9.66s.  7  7  7  7  7  7  6  7  7  8  7  7  7  6  6  7 110
  848    9.85s.  7  6  7  5  7  7  6  6  7  7  7  7  7  7  7  7 107
  864  10.03s.  6  6  6  7  6  7  7  6  7  5  6  6  4  6  7  7  99
  880  10.22s.  7  6  6  7  8  7  8  7  8  7  8  8  8  6  8  8 117
  896  10.40s.  8  8  8  7  7  7  8  7  8  7  7  7  7  8  7  7 118
  912  10.59s.  7  7  6  7  7  7  7  7  5  7  6  6  7  6  6  7 105
  928  10.77s.  7  7  7  6  6  7  7  7  7  7  7  7  7  7  8  7 111
  944  10.96s.  7  8  7  7  7  7  8  8  7  7  8  8  7  7  7  7 117
  960  11.15s.  7  7  7  7  7  7  7  7  7  7  7  7  8  7  7  7 113
  976  11.33s.  7  7  7  7  7  7  7  7  7  7  7  7  7  7  7  7 112
  992  11.52s.  7  7  7  6  6  7  7  7  7  6  7  7  7  6  8 10 112
 1008  11.70s. 10 10 10  9 10  9  9  9  8  8  8  8  8  9 10  9 144
 1024  11.89s. 10  9  9  9  9  9  9  9  9  8  8  9  8  8  9  7 139
 1040  12.07s.  7  7  8  8  7  8  7  7  7  6  7  8 10 10 10  9 126
 1056  12.26s.  9  9  9  9  9  8  8  8  8  8  8  8 10 10 10 10 141
 1072  12.45s.  9 10  9 10  9 10 10  9  9  9  9  9  9  9  9  9 148
 1088  12.63s.  9 10  9  9  9  9  7  8  9  9  9  9  9  9  9  9 142
 1104  12.82s.  9  8  9  9  9  9  9  9  9  9  9  9  9  9  9  9 143
 1120  13.00s.  7  9  9  8  9  9  8  9  9  9  9  9  9  8  9  9 139
 1136  13.19s.  8  9  9  9  9  9  9  9  7  9  9  9  9  9  8  7 138
 1152  13.37s.  9  7  8  8  8  8  8  8  8  8  8  8  8  8  8  8 128
 1168  13.56s.  7  8  8  7  8  5  8  8  8  8  8  7  7  7  8  8 120
 1184  13.75s.  7  8  8  8  8  8  8  8  8  8  7  8  7  7  8  7 123
 1200  13.93s.  7  7  8  7  7  7  7  7  6  8  8  7  7  7  7  6 113
 1216  14.12s.  6  7  6  7  7  7  7  6  7  8  6  6  7  7  7  7 108
 1232  14.30s.  7  7  7  7  7  7  6  7  7  7  7  7  7  7  7  7 111
 1248  14.49s.  7  7  7  8  8  8  8  8  7  8  6  8  7  6  6  7 116
 1264  14.68s.  8  7  8  7  6  8  7  8  8  8  7  8  6  7  8  7 118
 1280  14.86s.  7  8  8  7  7  7  7  7  7  7  7  7  6  6  7  6 111
 1296  15.05s.  7  7  7  7  7  6  7  7  6  6  7  7  7  7  7  7 109
 1312  15.23s.  7  6  7  7  7  7  7  7  7  7  6  6  7  6  7  6 107
 1328  15.42s.  7  7  6  6  7  7  6  7  7  7  7  7  7  7  7  7 109
 1344  15.60s.  7  5  6  6  6  7  7  6  6  6  6  7  7  6  6  5  99
 1360  15.79s.  6  7  6  7  6  6  6  6  6  6  6  6  6  7  7  7 101
 1376  15.98s.  7  6  7  5  5  6  6  6  6  6  6  6  6  6  6  6  96
 1392  16.16s.  7  6  7  5  6  6  5  7  6  5  5  6  6  5  6  6  94
 1408  16.35s.  6  6  6  6  6  6  6  5  6  6  6  6  6  6  5  5
Average    : 2.78MB; 7.3322 bits; 32.50x; 0.51s;
%lossyWAV Warning% : 0.0014 bits not removed due to clipping.
The issue might be the 11 bits removed around 1.1 seconds. They are removed because the signal is considered by the algorithm to be strong enough to mask the noise - in this instance obviously not. More analysis of the spectral makeup of the sample around the 1.1 second mark is required.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV 1.1.0 Development Thread.

Reply #86
...So yes -q2 was listenable for DAP with this specific sample, specially in a noisy environnement, but my personal interest for lossywav is not in DAP even if I like the flexibility of using any lossless format. ...

If keeping bitrate very low isn't your primary concern but quality is (like for me) you're better off using a higher quality setting like -q 4 or higher.

Nick C. may go into more detail, but I think I can answer some questions you addressed at Nick (hope you don't mind, Nick - oh, you were faster anyway).
lossyWAV definitely adds noise by rounding the wave samples in such a way that the least significant bits become zero, and a series of lossless codecs can make good use of this. Roughly speaking this is not done with quiet spots in the music but it is done when the music is loud. The added noise is controlled - it should be rarely audible with the lowest quality settings, it should be very rarely audible and - if it's audible - not annoying with the low quality settings, it should never be audible with the medium quality settings (though the focus here is also on bitrate and we don't allow for a safety margin), and it should be always inaudible with an ever increasing safety margin at the high quality settings.
Nick's description addresses this with emphasis on lossyWAV applications like DAP use or archiving/transcoding.
-q 5 is special as it takes exact care of the original lossyWAV idea originating from 2Bdecided: make some frequency analyses and use the number of bits to remove (zero out) depending on the signal's frequency region with lowest volume.
From experience so far the idea is correct and -q 5 is really transparent.
Below -q 5 2Bdecided's requirements are more and more softened, but NickC.'s other methods of controlling noise obviously do a very good job in keeping noise low. They can't work miracles however: the noise control of -q 0 has so low requirements that it's more remarkable that quality usually is still good than that problem samples exist.
Above -q 5 all quality requirements are fulfilled with an ever increasing added safety margin.
Originally there were just 3 quality settings: apart from the -q 5 related quality setting a setting with an additional safety margin for archiving quality, and another one with a softened quality requirement for DAP use.
This is clearer than the current settings. Movement came in when a certain demand came up for very low bitrate usage. It's all a matter of taste - I think it's a good thing to have a fine grained quality scale. Below -q 5 we're exactly in the same situation as the usual lossy codecs like mp3 for which everybody likes a quality scale to use according to personal quality and bitrate demands. Above -q 5 we have no analogue to the usual lossy codecs, but for the safety margin it's pretty much the same thing as for the lowering of the  quality demands: I think it's good to be able to choose the degree of added security.
What has happened with these many quality settings is that the 'standard quality' has gotten a bit out of focus, though for good reason as I think -q 5 is overkill for the most of us, but obviously the feeling has gone that the more we lower the quality requirements compared to -q 5, the more natural is an increasing chance for audible noise.
lame3995o -Q1.7 --lowpass 17

lossyWAV 1.1.0 Development Thread.

Reply #87
I may look dumb, but with my little knowledge it sounds like lossywav is cleverly removing hard to compress inaudible noise & then cleverly dithering easy to compress inaudible noise ? no.
so the problem I heard was not a "quantization like distortion" but more a audible "bad dithering" ?

also where does low, medium & high setting starts & ends ... that's a big problem for a noob, with the old 3 settings best/transparent/more agressive it was easier not to make mistake, I think both a scale for freedoom & recommended settings for noobs are usefull & complementary.

not everyone is willing to annoy the dev with stupid questions or to read plenty of technical threads just to know how to use correctly the codec. I know where to find the original thread & wiki help, the wiki were too general, the original discussion too technical. (& I was too lazy  )

finally I don't get the use of any quality setting above Q5, if the original idea was proven correct so far what will  a "more transparent than transparent" bring ? if it doesn't bring anything for transparency then it must bring at last some added safety for transcoding, is it technically the case at last on paper ? Is there 11 settings just to have 5 in the middle ? -q0 & -q1 were proven ABXable, you could as well remove -q9 & -q10 & have a 7 settings scale with -q3 in the middle being the actual -q5 ... it would be clearer. Anyway it's just my thoughts, I know I don't understand all the logic behind & I am not the dev afterall, in the end hybrid audio is a real headhache

Thks for taking the time to answer.

lossyWAV 1.1.0 Development Thread.

Reply #88
I may look dumb, but with my little knowledge it sounds like lossywav is cleverly removing hard to compress inaudible noise & then cleverly dithering easy to compress inaudible noise ? no.
so the problem I heard was not a "quantization like distortion" but more a audible "bad dithering" ?
No question is dumb - ignoring the answer would be .  By default lossyWAV does not dither at all and does not know if any noise is easily compressed or not.

also where does low, medium & high setting starts & ends ... that's a big problem for a noob, with the old 3 settings best/transparent/more agressive it was easier not to make mistake, I think both a scale for freedoom & recommended settings for noobs are usefull & complementary.
Why not just use the default value - it is included just this reason.

not everyone is willing to annoy the dev with stupid questions or to read plenty of technical threads just to know how to use correctly the codec. I know where to find the original thread & wiki help, the wiki were too general, the original discussion too technical. (& I was too lazy  )
Maybe the wiki needs more work, however I thought it was clear enough....

Thks for taking the time to answer.
No problem.

Nick.

[edit]
finally I don't get the use of any quality setting above Q5, if the original idea was proven correct so far what will a "more transparent than transparent" bring ? if it doesn't bring anything for transparency then it must bring at last some added safety for transcoding, is it technically the case at last on paper ? Is there 11 settings just to have 5 in the middle ? -q0 & -q1 were proven ABXable, you could as well remove -q9 & -q10 & have a 6 settings scale with -q3 being the actual -q5 ... it would be clearer. Anyway it's just my thoughts, I know I don't understand all the logic behind & I am not the dev afterall, in the end hybrid audio is a real headhache
You are correct, the higher the quality settings give less likelihood of problems in transcoding. I explained already the evolution of the numerous quality presets - different people have commented on the project at different times with with different aims - this has pulled the quality presets wider and wider. I also wanted to have a highest quality level which still gave some bitrate reduction compared to fully lossless (-q 10 gives at least a 20% bitrate saving on average). I cannot tell you which quality preset to use - that's up to your own evaluation and compromise between perceived quality and resulting bitrate when subsequently encoded.
[/edit]
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV 1.1.0 Development Thread.

Reply #89
The issue might be the 11 bits removed around 1.1 seconds. They are removed because the signal is considered by the algorithm to be strong enough to mask the noise - in this instance obviously not. More analysis of the spectral makeup of the sample around the 1.1 second mark is required.

Well, the issue I hear is (I think) quick changes in noise intensity. They're highlighted in this picture:
[a href="http://img149.imageshack.us/my.php?image=32639078hq9.png" target="_blank"] )

lossyWAV 1.1.0 Development Thread.

Reply #90
The issue might be the 11 bits removed around 1.1 seconds. They are removed because the signal is considered by the algorithm to be strong enough to mask the noise - in this instance obviously not. More analysis of the spectral makeup of the sample around the 1.1 second mark is required.
Well, the issue I hear is (I think) quick changes in noise intensity. They're highlighted in this picture:
[a href="http://img149.imageshack.us/my.php?image=32639078hq9.png" target="_blank"] )
How long are the steps in the signal - if it's a multiple of 11.6msec (512 samples) then the steps are caused by different bits-to-remove being applied to the output and adding different intensities of white noise.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV 1.1.0 Development Thread.

Reply #91
How long are the steps in the signal - if it's a multiple of 11.6msec (512 samples) then the steps are caused by different bits-to-remove being applied to the output and adding different intensities of white noise.

Yes, they are.
But I can't see direct relationship between amplitude of noise and numbers that lossyWAV prints with --detail switch.  I can't find pattern
Code: [Select]
...
   80    0.93s. 10 10 10  9  9  9  9  8  8  8  8  9  8  8  9 11 143
   96    1.11s. 10 11 11 11 11 11 11 11 11 10 11 10 10 10 10 10 169
  112    1.30s. 10 10 10  9 10 10  9  9 10 10 10 10  9  9  9 10 154
...

lossyWAV 1.1.0 Development Thread.

Reply #92
Maye we should add a sticky somewhere. Inexperienced users should not touch any quality setting and will use the defaulted Q5 ??

lossyWAV 1.1.0 Development Thread.

Reply #93
Maye we should add a sticky somewhere. Inexperienced users should not touch any quality setting and will use the defaulted Q5 ??
That sounds like an eminently sensible suggestion - it's now on my to do list for the wiki / --help text.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

 

lossyWAV 1.1.0 Development Thread.

Reply #94
I had another idea: create a few 'presets' that are mapped to common Q settings. Maybe even hide the Q scale from the normal screen and document it with --longhelp or similar.

-normal [Default] = Q5
-medium = Q3
-Portable = Q2

- Extreme / Archiving / transcoding = Q6..10

lossyWAV 1.1.0 Development Thread.

Reply #95
I had another idea: create a few 'presets' that are mapped to common Q settings;

-normal [Default] = Q5
-medium = Q3
-Portable = Q2

- Archiving / transcoding = Q7
Good suggestion - also added to to-do-list. Suggested names and associated -q value:

--highest := -q 10;
--high := -q 7.5
--archive := -q 7.5
--default := -q 5;
--normal := -q 5;
--portable := -q 2.5.
--nasty := -q 0

When I make this modification, I will de-emphasize the -q <n> settings in the --help text.

This is a sensible compromise position which retains the current flexibility for advanced users while giving the less experienced user a big hint as to what setting to use. In light of recent conversation on the topic it is a required modification and will be posted today.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV 1.1.0 Development Thread.

Reply #96
I had another idea: create a few 'presets' that are mapped to common Q settings;

-normal [Default] = Q5
-medium = Q3
-Portable = Q2

-extreme = Q6..10
Good suggestion - also added to to-do-list. Suggested names and associated -q value:

--highest := -q 10;
--high := -q 7.5
--default := -q 5;
--normal := -q 5;
--portable := -q 2.5.
--nasty := -q 0

....


Yeah, I like this very much: shadowking's approach and your incarnation cause it looks like -q 2.5/5/7.5/10 are really those settings which IMO are of most practical use (7.5 or 10 for the main part as a substitute for lossless archivíng).
I'd just prefer one '-', so '-normal' instead of '--normal', I would drop '-default' as a synonym for '-normal' (of course '-normal' should be defaulted), and we shouldn't use a nasty mode IMO.
The '-q' parameter can remain as an advanced '--q' setting for the lovers of quality fine tuning.

The solution is so wonderful to me as we have seen a growing discussion on quality levels since we started having a finer quality granularity. This is covered by this new self-explaining quality granularity while still keeping up fine tuning abilities for those who like them.
lame3995o -Q1.7 --lowpass 17

lossyWAV 1.1.0 Development Thread.

Reply #97
That's exactly what I was suggesting, except that as I think -q 0 & -q 10 are not really recommandable I would not add them as recommended setting, & now that I understund the symbolic importance of -q 5 I would add it.

so ideally it would look like this for me

-q 2.5 = -portable (my "-near lossy")
-q 5 = -default
-q 7.5 = -archive (my "-near lossless")

-q0 & -q10 looks too much like flac -fast & -best which makes sense for flac but are sub-obtimal here.

I think that after some transcoding tests I would then either use the -default or -archive preset ...

Edit:
I agree with halb27 too, one - is enough, the easier the better.

If you really can't do without adding -q 0 & -q 10 to the presets then I would use maybe:
-q 0 = -noisy or -this-is-trash-metal
-q 10 = -overkill or -music-for-the-deaf
names that naturally sounds bad, but my english is too weak here.

lossyWAV 1.1.0 Development Thread.

Reply #98
...
-q 2.5 = -portable (my "-near lossy")
-q 5 = -default
-q 7.5 = -archive (my "-near lossless")
...

This brings us back to the 3 quality levels approach, but with the advantage of a proper naming.
Having thougt about it I also think it's better to stay with three levels.

I personally have dropped my lossless archive and encoded everything to -q 7 (for the most part), but on more or less rare occasion - depending on the meaning the track under consideration has to me - also used a higher -q setting up to -q 10 (and in very rare cases I encoded to lossless wavPack which I can use on my DAP). But it is really enough to use -archive as a rule, and I can still use a higher -q setting if available as an advanced option on those rare occasion where I want it.

So I suppoort this IMO better differentiation standard quality options / advanced quality options.

As a detail I suggest to use --shaping 0.5 with the -archive quality (more clearly in terms of the advanced options: with every -q setting from -q 7 up.
It makes the noise so much less audible (often totally inaudible when listening only to the noise containing correction file with the same volume setting as with the real music), and IMO is safe to use with these very high quality settings.
lame3995o -Q1.7 --lowpass 17

lossyWAV 1.1.0 Development Thread.

Reply #99
...
If you really can't do without adding -q 0 & -q 10 to the presets then I would use maybe:
-q 0 = -noisy or -this-is-trash-metal
-q 10 = -overkill or -music-for-the-deaf
names that naturally sounds bad, but my english is too weak here.

I really wouldn't like -q 0 among the standard options.
-q 7.5 should be overkill already, so it's hard to find a simple name for -q 10 (maybe -paranoid hits it best - but it's a pretty negative word. However if most members would appreciate a -nasty mode among the standard options a -paranoid mode would be a good counterweight).
Another argument for having just three standard quality options.
lame3995o -Q1.7 --lowpass 17