1

Topic: Competent adjustment of loudness on a sound card

Greetings, the people.
Actually this question for certain rose in many branches on sound cards and .
I suggest to generalize in this subject experience saved up in long and fruitful discussions as this subject is rather actual among beginners, and not only. Of what I on own example was convinced. Understood that apparently long time installed loudness not correctly. Probably.
In general - let's understand.
All began that I studied for the first time a label of characteristics from   to chip CS4398 more attentively.
Also saw there a curious inscription:
[img=700x513, 115.9Kb] http://i85.fastpic.ru/big/2016/1215/f6/ … 11e7f6.png [/img]
The vendor specifies in dependence of parameter THD+N on a level . And this dependence rather essential.
Earlier at forums said that it is impossible to expose  in 0db. I and itself always in mixer Windows put value Master (which directly sets  cards, on how many I understand) in position of 50 % and already on the amplifier installed comfortable level.
Under characteristics of chips, it turns out that the maximum productivity is reached only at level 0db, that is at all without  that as I understand now quite logically.
That is at adjustment source audio level reasonably to put on 0db and to regulate or in applications (foobar, the browser etc.) or on the amplifier.
Though. On the other hand.
We reduce loudness on 20 db and THD+N worsens only on 10 db (107-97) if to take the same CS4398.
Theoretically, at loudness lowering subjective resolution of a source, increases. That is - it is perfect on the contrary.
So where truth?
Whether correctly I argue?
What thoughts about it are at you?

2

Re: Competent adjustment of loudness on a sound card

MagistrAndrew
At adjustment source audio level reasonably to put on 0db and to regulate... On the amplifier.
-- So.

3

Re: Competent adjustment of loudness on a sound card

MagistrAndrew
The Vendor specifies in dependence of parameter THD +N from a level . And this dependence rather essential.
You forget about presence "+N" in parameter. For all samplings, except 24 ' full scale ' (0dB) - measurement is restricted by a noise floor of an output of the chip (for 24 bits) or a noise floor actually a source signal (for 16 bits).
[q=] Note 1. One-half LSB of triangular PDF dither is added to data. [/quote] Hands to tear off to figures! TPDF has unique correct amplitude - 2LSB. If same it is applied in the chip at  - in its oven!

4

Re: Competent adjustment of loudness on a sound card

The Vendor specifies in dependence of parameter THD+N on a level .
Yes not   ! This concept - only for analog.
-60  is "level" of a numeral signal, i.e. usage of a small part of dynes. Range . From here and impairment of parameters.
About level: actually numeral material - it is already made and does not overload a digital-to-analog coder (in it there is simply nothing to be overloaded), and here an output signal of a digital-to-analog coder after its transformation into analog - this _ to overload analog parts of the circuit. And it to us ? smile
Any muses-signals has a so-called peak-factor, therefore level in a regulator I would not advise average to install above-3 .
If is  the analyzer peak level -  a buzzing, and it is necessary to be guided by it.

5

Re: Competent adjustment of loudness on a sound card

Solder
In a regulator I would not advise middle tier to install above-3 
Here the question where is a regulator - after  in analog, or to - in digit. Judging by a chip block diagram -  in digit. And at such "" as in notes, any  - is counter-indicative categorically! If a signal "not " in the further structure of gain without a clip -  before a node, in which "not ".

6

Re: Competent adjustment of loudness on a sound card

On how many I know in the majority of sound cards and  level  is regulated in a numeral part, in the chip.
Unless, there is a class of devices with the attenuator carried out separately - there  probably already capacity of an analog signal is regulated.
So, while the general-purpose answer turns out such, whether that?
If on a source there is no analog regulator of loudness - in drivers level should be installed in 0db.
So?

7

Re: Competent adjustment of loudness on a sound card

MagistrAndrew
No. In drivers ZK ASUS level by default 76 % (hardly it is more-2 dB FS).
It is made that artifacts of oversampling and a numeral filtration did not lead "numeral ".

8

Re: Competent adjustment of loudness on a sound card

vitamir
Whether so really to work out the general-purpose strategy for exhibiting of audio level of sources?
In particular sound cards and USB .

9

Re: Competent adjustment of loudness on a sound card

MagistrAndrew
If on a source there is no analog regulator of loudness
Or normal (recalculation in the raised digit capacity with correct ) the numeral. If at you a digital-to-analog coder 120 SNR and recalculation is correct - that it is not necessary to be afraid of digit.
vitamir
That artifacts of oversampling and a numeral filtration did not lead "numeral "
If  values go to "overload", it already  - in records spoilage. And so - well the largest peaks which  in record - anybody and does not note will be cut.

10

Re: Competent adjustment of loudness on a sound card

MagistrAndrew
Scalability of strategy in a non-admission "numeral ".
For this purpose it is necessary to understand, what numeral handling of a flow that becomes a sound, fulfill DSP + the driver specific .
Also, it is necessary to consider singularities of numeral handling in a specific chip of a digital-to-analog coder
Adding from 12/16/2016 16:41:
igorzep :

If  values go to "overload", it already  - in records spoilage. And so - well the largest peaks which  in record - anybody and does not note will be cut.

is not present, not absolutely. These are artifacts of algorithms of the numeral filtration applied in DSP/the driver/player on "numeral" . Who in a reality does not want to lose  dynamic range.

11

Re: Competent adjustment of loudness on a sound card

vitamir
No, not absolutely. These are artifacts of algorithms of the numeral filtration applied in DSP/the driver/player on "numeral" .
Nonsense! The band is cut already off in record according to frequency of sampling. Recovery in the analog form - is unambiguous, and if it exceeds FS is a spoilage __. Filters at  just also are engaged in recovery. They create nothing, and any artifacts do not add.  *  = .
If record a curve, on frequencies of close to fs/2 - it is possible to create an overload though in tens times! But it is a record problem. If in a source code with the raised resolution (we tell, with fs*8) - there were no such samples - that and at recovery they from do not appear anywhere!

12

Re: Competent adjustment of loudness on a sound card

igorzep
All is true, but the digital-to-analog coder DSP/driver/microcircuit can fulfill additional flow handling. In , for example, not in all modes the disconnected. As well as in some Creatives in some modes. On it also do "store" in 2. 3  which ignoring leads to appearance of distortions in a regenerated signal. We take  from , we twist a mixer slider in the driver in a position "100 %" and we listen to result. And is better we measure.

13

Re: Competent adjustment of loudness on a sound card

vitamir
Additional handling of a flow
Bosh. It any more a digital-to-analog coder, and stuff what that, is not clear that the doing.
Not in all modes disconnected
In a garbage box the device which is engaged in amateur performance. Another here there is nothing to tell. And in black lists everywhere where only it is possible that the device - is broken, and the direct duties does not fulfill.

14

Re: Competent adjustment of loudness on a sound card

igorzep
All modern digital-to-analog coders contain onboard the numeral filter which is  for creation of the intermediate coordinates. Alternative are the ancient digital-to-analog coders named NOS (Non Over Sampling), in  circles it is most "the present multibits".
All PCM have a singularity of operation  - on CCIF on high frequencies is  for 0 dBFS
[img=620x369, 90.5Kb] http://reference-audio-analyzer.pro/rep … d_ccif.png [/img]
MagistrAndrew
The optimal maximum level as a first approximation shows schedule Total Harmonic Distortion vs Level
[img=620x369, 99.4Kb] http://reference-audio-analyzer.pro/rep … s_dBFS.png [/img]
For STU with regular  the signal needs to be lowered to a digital-to-analog coder on-15 ~-20 . But without fanaticism since the main contribution to distortions is imported by the second and third harmonic. These distortions by the way it is not so much from a digital-to-analog coder, how many from .

15

Re: Competent adjustment of loudness on a sound card

romanrex
All modern digital-to-analog coders contain onboard the numeral filter...
You here to the sermon will read it to me? Abc-books opened, in course wink
All PCM have a singularity of operation  - on CCIF on high frequencies is  for 0 dBFS
There is no such "singularity". There is a wrong preparation of a test signal. Does not happen "0 dBFS". It is a limit, instead of the maximum value. To it to "aspire", but never reaches. The reason for that is step function linearization. In the people - . Amplitude 2LSB the minimum (for TPDF), at noise-shaping - is more. All signals should be prepared so that they did not exceed 0 dBFS, including  and  values. That after a filtration some points are selected only, does not mean that only these points cannot fall outside the limits. There can not be no points of an encodable signal. Otherwise is a spoilage of a signal, instead of "a singularity of operation ".
However, even if a signal  as you showed, it is far not fatality, and in muses-signals of such moments there will be one piece on a track. Audibility of it is hardly less, than any. If it is finite all record rigidly not  under one level. But then in her already non-linearity distortions much more, than from this "clip" as a result of mockeries at it the compressor.
Since the main contribution to distortions is imported by the second and third harmonic
And you will change this insignificance in the form of simple harmonics of the low order for inadequate recalculation , as they say "on get rid a tick to deliver in "?
. I for numeral . But only when it is implemented correctly, instead of rejection of bits...

16

Re: Competent adjustment of loudness on a sound card

igorzep
There is a wrong preparation of a test signal.
If the test signal was wrong  would be on all digital-to-analog coders and would be visible on an initial signal, and so it only at family of digital-to-analog coder PCM.
On inadequate recalculation 
Inadequate modes are well fixed by tests. In normal digital-to-analog coders of problems did not meet.

17

Re: Competent adjustment of loudness on a sound card

romanrex
Also it would be visible on an initial signal
Wrong premise. And it there even is visible. Simply the analyzer has no restriction for a signal in 0dBFS. Simply because it it not ...
At family of digital-to-analog coder PCM
That for "family PCM". I understand it as a digital-to-analog coder accepting on input PCM the data vs. DSD... You possibly meant ruler TI PCM1 ***?
Inadequate modes are well fixed by tests.
These are very nontrivial tests. I know only one person who did them... The Published measurements about it - at all did not see.
In normal digital-to-analog coders of problems did not meet.
Tests did not do! However problems there are. Or at once it is necessary to name good few  abnormal. wink

18

Re: Competent adjustment of loudness on a sound card

igorzep
The Published measurements about it - at all did not see.
More than 100 various digital-to-analog coders http://reference-audio-analyzer.pro/report/dac.php

19

Re: Competent adjustment of loudness on a sound card

romanrex
More than 100 various digital-to-analog coders
Yes though 100500 test on everyones THD and other... What actually internal resolution of filters TsApa all of them equally do not show.

20

Re: Competent adjustment of loudness on a sound card

igorzep
All of them equally do not show.
Who does not want, does not see. Who wants to expose an optimum level - exposes on the basis of schedules since measurements are led not separately digital-to-analog coder chips, and the finite device with .
And it will be interesting to someone for nothing  about the lost digit capacity without uniform thought as this lost digit capacity to hear or as it it is compensated by higher frequency of sampling. Or what  in the theory without practice the most correct...
Adding from 12/17/2016 23:22:
I will tell it is more, when the subject is considered from outside  nonsense - all schedules "show nothing", and  is exceptional "the theory from a ceiling".

21

Re: Competent adjustment of loudness on a sound card

romanrex
Who does not want, does not see.
As it is possible  what did not measure...
Who wants to expose an optimum level - exposes on the basis of schedules since measurements are led not separately digital-to-analog coder chips, and the finite device with .
Optimum level always - as much as possible full usage of dynamic range!
Without uniform thought as this lost digit capacity to hear
It is perfectly heard in systems with the big dynamic range. Unlike catching of fleas in harmonics more low-100...-120 that almost any made digital-to-analog coder provides.
Or what  in the theory without practice the most correct...
To select  "practice" - idiocy and the full nonprofessionalism! By practice it is proved that first two moments of correlation (distortions and noise modulation) - are audible and artificial, while noise - is not becomes audible much faster and anyway - is preferable (besides always naturally is present, unlike.)
  it is exceptional "the theory from a ceiling"
not "the theory from a ceiling". It is a pure mathematics in the first, and in the second - a subject for a long time and perfectly researched. Though some designers of chips of "quite known firms", for some reason are guided, as you noted -  nonsense, selecting from  coefficients from calculations.
Or as it is compensated by higher frequency of sampling
It is not compensated! Though as you increase frequency - graduated transmitting function from it does not become less graduated!

22

Re: Competent adjustment of loudness on a sound card

igorzep
To you all truly write - the signal should be  in avoidance . You thus do not lose any dynamic range. Aliasing it in DSD, in PSM peaks of waves. Higher frequency of sampling shifts some peaks for 20 Kgts, but then it is necessary to put the filter since the UHF directs noises on a heard range. Whether sampling increase - normally always will be better to tell worse and only in DSD it leads to improving. Music needs to be played back in that format in which this it is written down, or with smaller frequency of sampling, saving thus reference frequency - i.e. 44/88/176/352 and 48/96/192/384 and never to mix, since it is normal just as a result of such blendings and artifacts of indefinite frequency get out.
Earlier there were multibits with dynamic range 60 - all was with them apprx. If in these 60 0 % of noise. Analog equipment too not  dynamics of a range - but on quality more abruptly digits. So chase a range less and look tonal quality of this or that  more. In this case Tserroz the standard - and to it all , "very popular"  a digital-to-analog coder - at Korga like were attempts it to recover but as that not so quitted. Probably chip such that with it to make nothing. Now in the presence of the oscilliscope, the analyzer of a spectrum, soldering station, a multimeter, hands and a head you can easily solder to yourselves a box on any . Well truth in an ideal it is necessary still  BIOS. The iron with the laser printer and chloride iron I think already life trifles, well 3-d the printer it already if absolutely that that on sale to do.

23

Re: Competent adjustment of loudness on a sound card

vanpetr
Forgive, on such amount of the senseless text, the casual words made in sentences, well matched to the generator of pseudoscientific delirium, I at all do not know as to answer... Therefore will not spend the and your time.

24

Re: Competent adjustment of loudness on a sound card

igorzep
As it is possible  what did not measure...
What exactly is not measured? Dynamic range? Here it (example) http://reference-audio-analyzer.pro/rep … 7.php#rw20
If it is a question the friend the detail information is necessary more THAT it would be desirable to measure and AS.
These are very nontrivial tests. I know only one person who did them... The Published measurements about it - at all did not see.
At present it looks as "not clear test not clearly whom of not clearly that".
It is perfectly heard in systems with the big dynamic range.
1. Included more loudly in 0dBFS - sounds accurately. Reduced loudness on 6  - sounds unsharply - equal loudness contours did not consider.
2. Included more loudly in 0dBFS - sounds accurately. Reduced loudness on 6  and lifted in  on 6  - sounds differently. The question, how many gave distortions before?
Question what was a digital-to-analog coder, if it something from a multibit series there quality decreases if the delta - to it in most cases all the same.
About  is that there is unique true algorithm, no them a lot of different under different tasks.
To worry about digit capacity of sense is not present, all the same on an output of a digital-to-analog coder after the modulator it turns out one or the several one-bit DSD flows, from coordinate system in PCM remains nothing. As and  from analog selects DSD a flow, and then translates in PCM. Some do the "impulse" digital-to-analog coders on logical high-speed chips, like Chord and NAD.

25

Re: Competent adjustment of loudness on a sound card

vitamir :
MagistrAndrew
No. In drivers ZK ASUS level by default 76 % (hardly it is more-2 dB FS).
It is made that artifacts of oversampling and a numeral filtration did not lead "numeral ".

About TI  - I like such behavior did not note, in CS4382 such behavior is watched on an one-acoustical signal.
To that I have been madly surprised, watching it on H-Faj Titanium.
It was necessary to mix an ultralow-frequency signal (hertz in 5) with a minimum level (-96 if I am not mistaken less,  is not able) as all became on the place.
Also is  mismatch of levels of record/vosprizvedenija at many cards.
At Asuseev, for example, an input standard 2V RMS, the output - is much more what virtually to expand SNR, therefore and there are certain nuances.