[citraian]

http://teribil-audio.com/2014/02/how-to-display-dynamic-range-rating-in-foobar2000/ Maybe this helps

 

[Dobrescu George]

I remember when I first started listening to my AHG Q701 headphones.
I found there were some CDs I never wanted to hear again on those 'phones, not because of the music, but because of the nasty clipping on some songs....

The thing is that with less revealing iems or headphones i do not get the same results... maybe you are totally right, and the albums were not that great to begin with...
 
Sad, because some have some really interesting parts within..
 
 

http://teribil-audio.com/2014/02/how-to-display-dynamic-range-rating-in-foobar2000/ Maybe this helps

Okay, i did installed, but i do not understand how to read the dynamic range. 
 
 

Name	Dynamic Range	Peak	RMS
End of the line	4	0	-4
Duality	4	-0.3	-6
World painted blood	4	0	-6
Seeds of gold	5	-0.2	-7

 
 
So, this means that only end of the line and world painted blood have real clipping, and the other two songs actually have guitars that clip recorded, but the finished songs to not clip themselves?
 
I am pretty interested in understanding more...^_^
 
After understanding how oversampling and upsampling works, i can understand why hugo would not show clippings, it could be creating points in the wave, lowering the time the clipping occurs, because it oversamples music a lot, so maybe, by this process, and it's own interpolation filter, it is able to repair all clippings, by just making clippings shorter. this makes a lot of sense, after me studying how to interpolate music. 
 
So, basically, songs which have recorded clipping instruments, but do not clip themselves, cannot be repaired, as the clipping cannot be un-done in this situation... 
 
 
If i understood wrongfully, please, advice me how to read the data!

 

[citraian]

Well, I can't help you that much. You will have to use google for more details on how to interpret the DR measurements.
One thing I can say for sure is that those recordings are far from great.
 
I'm still curious on why the Hugo isn't clipping

 

[JaZZ]

  The thing is that with less revealing iems or headphones i do not get the same results... maybe you are totally right, and the albums were not that great to begin with...
 
Sad, because some have some really interesting parts within..
 
Okay, i did installed, but i do not understand how to read the dynamic range. 
 

Name	Dynamic Range	Peak	RMS
End of the line	4	0	-4
Duality	4	-0.3	-6
World painted blood	4	0	-6
Seeds of gold	5	-0.2	-7

 
So, this means that only end of the line and world painted blood have real clipping, and the other two songs actually have guitars that clip recorded, but the finished songs to not clip themselves?
 
I am pretty interested in understanding more...^_^
 
After understanding how oversampling and upsampling works, i can understand why hugo would not show clippings, it could be creating points in the wave, lowering the time the clipping occurs, because it oversamples music a lot, so maybe, by this process, and it's own interpolation filter, it is able to repair all clippings, by just making clippings shorter. this makes a lot of sense, after me studying how to interpolate music. 
 
So, basically, songs which have recorded clipping instruments, but do not clip themselves, cannot be repaired, as the clipping cannot be un-done in this situation... 
 
If i understood wrongfully, please, advice me how to read the data!

 
The Hugo doesn't remove the clipping. It's not its task to correct the signal, and it can't differentiate between wanted and unwanted clipping to begin with. For «Seeds of Gold» a maximum amplitude of –0.2 is indicated, which hints to no clipping, but in fact the recording is full of clipped peaks, which could mean that it's been «normalized» to a lower level than originally, maybe during the obvious dynamic range compression.

 

[Dobrescu George]

  Well, I can't help you that much. You will have to use google for more details on how to interpret the DR measurements.
One thing I can say for sure is that those recordings are far from great.
 
I'm still curious on why the Hugo isn't clipping

this much i can explain. 
 
 

   
The Hugo doesn't remove the clipping. It's not its task to correct the signal, and it can't differentiate between wanted and unwanted clipping to begin with. For «Seeds of Gold» a maximum amplitude of –0.2 is indicated, which hints to no clipping, but in fact the recording is full of clipped peaks, which could mean that it's been «normalized» to a lower level than originally, maybe during the obvious dynamic range compression.

Yes. 
 
After thinking about it, i expect the notes that clip with other equipment to clip with hugo too, in seeds of gold.
 
But in metallica - end of the line i expect that there is no clipping at all. thanks for all the help, it is helping me understand technology a lot faster. Out of curiosity, do you get clipping with end of the line?
 
This is why i found the clipping with x5 strange, it actually is normal for x5 to clip with all the above files, because they were recorded with clipping and distortions;
 
 
 
 
 
 
By oversampling, the stress of interpolation is taken off the algorithm itself, and the complexity of algorithm is used for a shorter period of time, so even it a wave stays for a duration of 4-5 samples in clipping position, hugo is able to take into account the samples before and so reduces the height of the clipping samples. This should make them smooth and curved.
 
 
The audio hobby becomes more and more interesting.
 
 
Now this is the part where i want to ask a question, to everyone:
 
If we are talking about waves, it is abnormal in real life sounds for there to exist real clipping, so sqare waves are out of the question, if we want to touch the audiophile dream, to recreate live performance, so even if it affects the original recording, an algorithm that makes sqare waves into curved waves, which is how they actually look like live. The questions is if the sound would be closer to live performance after applying such an algorithm, than it is with the original clipping?

 

[tjw321]

@skerry2006aj If oversampling masks the problem with the clipped recordings, why don't you upsample on your computer and create new upsampled flacs to put on the X5?

 

[Dobrescu George]

  @skerry2006aj If oversampling masks the problem with the clipped recordings, why don't you upsample on your computer and create new upsampled flacs to put on the X5?

because they take up much much more space, and the original algorithms that most comercial programs use, from wavelab to audacity, do not resolve the problem.
 
The problem lies within the algorithm for oversampling that would solve the problem, upsampling by computer did not solve the problem. I am working at a solution right now

, thanks for trying to help.
 
The fact is that most over sampling, upsampling, and reconstruction algorithms try to keep the data of what has been originally recorded, this is where they fail, if the data is broken, for example has clipping, the algorithm keeps the data that way. the challenge is to try to remove the clipping without loosing musical data. 
 
What @JaZZ said is the best ideea, to edit a wave, wo make all clipping edges round, but it would take lots of time for an editing song, there are 44100 points in a single second. What would be awesome is if there would be an algorithm that would be able to do this instead of me. I hope that i can get to make one or something.

 

[Chris J]

this much i can explain. 


Yes. 

After thinking about it, i expect the notes that clip with other equipment to clip with hugo too, in seeds of gold.

But in metallica - end of the line i expect that there is no clipping at all. thanks for all the help, it is helping me understand technology a lot faster. Out of curiosity, do you get clipping with end of the line?

This is why i found the clipping with x5 strange, it actually is normal for x5 to clip with all the above files, because they were recorded with clipping and distortions;






By oversampling, the stress of interpolation is taken off the algorithm itself, and the complexity of algorithm is used for a shorter period of time, so even it a wave stays for a duration of 4-5 samples in clipping position, hugo is able to take into account the samples before and so reduces the height of the clipping samples. This should make them smooth and curved.


The audio hobby becomes more and more interesting.


Now this is the part where i want to ask a question, to everyone:

If we are talking about waves, it is abnormal in real life sounds for there to exist real clipping, so sqare waves are out of the question, if we want to touch the audiophile dream, to recreate live performance, so even if it affects the original recording, an algorithm that makes sqare waves into curved waves, which is how they actually look like live. The questions is if the sound would be closer to live performance after applying such an algorithm, than it is with the original clipping?

Electric Guitar distortion is clipping.
Synthesizers can synthesize square waves.

 

[Chris J]

this much i can explain. 


Yes. 

After thinking about it, i expect the notes that clip with other equipment to clip with hugo too, in seeds of gold.

But in metallica - end of the line i expect that there is no clipping at all. thanks for all the help, it is helping me understand technology a lot faster. Out of curiosity, do you get clipping with end of the line?

This is why i found the clipping with x5 strange, it actually is normal for x5 to clip with all the above files, because they were recorded with clipping and distortions;






By oversampling, the stress of interpolation is taken off the algorithm itself, and the complexity of algorithm is used for a shorter period of time, so even it a wave stays for a duration of 4-5 samples in clipping position, hugo is able to take into account the samples before and so reduces the height of the clipping samples. This should make them smooth and curved.


The audio hobby becomes more and more interesting.


Now this is the part where i want to ask a question, to everyone:

If we are talking about waves, it is abnormal in real life sounds for there to exist real clipping, so sqare waves are out of the question, if we want to touch the audiophile dream, to recreate live performance, so even if it affects the original recording, an algorithm that makes sqare waves into curved waves, which is how they actually look like live. The questions is if the sound would be closer to live performance after applying such an algorithm, than it is with the original clipping?

Oversampling doesn't get rid of clipping, but it could be the Hugo handles clipping better than other DACs if a recording is recorded right to 0 dB FS.

 

[Dobrescu George]

Electric Guitar distortion is clipping.
Synthesizers can synthesize square waves.

I also thoguht about this. 
 
So basically, if i manage to create an algorithm for repairing clipping, it can actually destroy some parts of music, or at least, it should never be used with electronic music. I guess that if a rule for repairing, like a certain degree of angle means that the wave is to be left alone should work.
 
Like, making the program able to recognize what waves should be left alone and what not. It is not entirely impossible, but it takes much longer time to write, and i must include most situations in which the algorithm excludes waves.
 
This is not something i want to write in a few hours of work, but it could be worth if i manage to finish the code, because there are many albums that could be helped.

Oversampling doesn't get rid of clipping, but it could be the Hugo handles clipping better than other DACs if a recording is recorded right to 0 dB FS.

This could also be the situation. 
 
I am still learning, i am new to high end audio, and it is fun to learn!
 
Sorry if i am disturbing the thread too much, but i am still trying to understand where the problem lies, and what can i do to correct it.

 

[JaZZ]

  After thinking about it, i expect the notes that clip with other equipment to clip with hugo too, in seeds of gold.
 
But in metallica - end of the line i expect that there is no clipping at all. thanks for all the help, it is helping me understand technology a lot faster. Out of curiosity, do you get clipping with end of the line?

 
Yes, there's a lot of clipping, but here more clearly in the context of a fuzz guitar (and a drum beat at the same time).
 
 
 
However, it sounds like a heavy-metal tune has to sound: distorted and with massive dynamic compression / peak limiting.
 
With the Hugo I hear distortion more clearly than with the Opera, but it's easier to attribute it to the fuzz guitar, so the listening is more relaxed and less afflicted by doubts (about recording or component clipping).
 

The problem lies within the algorithm for oversampling that would solve the problem, upsampling by computer did not solve the problem. I am working at a solution right now

, thanks for trying to help.  
The fact is that most over sampling, upsampling, and reconstruction algorithms try to keep the data of what has been originally recorded, this is where they fail, if the data is broken, for example has clipping, the algorithm keeps the data that way. the challenge is to try to remove the clipping without loosing musical data.

 
Think about it: «what has been originally recorded» is what you hear now, minus the missing ultrasonics (from a bandwidth limitation to 22 kHz) – nothing else.
 

What @JaZZ said is the best ideea, to edit a wave, to make all clipping edges round, but it would take lots of time for an editing song, there are 44100 points in a single second. What would be awesome is if there would be an algorithm that would be able to do this instead of me. I hope that i can get to make one or something.

 
You wouldn't want to hear the same recording without all the guitar distortion. Yes, it's certainly possible to create an algorithm that rounds all clipped peaks, but it would take the life out of the recording and change its characteristic drastically.
 
Your understanding of the Hugo's upsampling algorithm is wrong. It's all about «recreating» a possible waveform on the basis of 26,368 taps as a substitute for the missing ultrasonics – for the sake of creating a proper signal within the audio band. Also note: it's no 26,368 times oversampling; what's actually going on is described here.

 

[Dobrescu George]

   
Yes, there's a lot of clipping, but here more clearly in the context of a fuzz guitar (and a drum beat at the same time).
 
 
 
However, it sounds like a heavy-metal tune has to sound: distorted and with massive dynamic compression / peak limiting.
 
With the Hugo I hear distortion more clearly than with the Opera, but it's easier to attribute it to the fuzz guitar, so the listening is more relaxed and less afflicted by doubts (about recording or component clipping).
 
 
Think about it: «what has been originally recorded» is what you hear now, minus the missing ultrasonics (from a bandwidth limitation to 22 kHz) – nothing else.
 
 
You wouldn't want to hear the same recording without all the guitar distortion. Yes, it's certainly possible to create an algorithm that rounds all clipped peaks, but it would take the life out of the recording and change its characteristic drastically.
 
Your understanding of the Hugo's upsampling algorithm is wrong. It's all about «recreating» a possible waveform on the basis of 26,368 taps as a substitute for the missing ultrasonics – for the sake of creating a proper signal within the audio band. Also note: it's no 26,368 times oversampling; what's actually going on is described here.

Thanks again for pointing me to my mistakes.
 
After trying different methods, i had found one what works. 
 
You are absolutley right.
 
The reason i hear a lot of clipping with x5 and not with hugo is because if other degrees of separation between instruments in hugo. With x5, if the drum beat and fuzz guitar note are done in the same time, i perceive a distorted note, not two different notes comeing from two different instruments.
 
It seems that recovering transients can actually help even metallica - end of the line.
 
 
Moral of the day: X5 can handle clipping very very well, it is not clipping that really made the distortions, but some notes overlapping one over another, and creating a distorted perceived note. Hugo presents the three notes as distinct notes, and this is why i heared no clipping or distortion.
 
Or i guess that this is the explanation..

 

[JaZZ]

The reason why a sophisticated DAC algorithm (as with the Hugo) can have such a major effect is the need for an anti-aliasing filter – because of the sampling rate of 44.1 kHz, thus the steep low-pass filter at 22 kHz at the latest. The filter is needed to prevent aliasing – intermodulation of frequencies higher than the Niquist frequency (½ the sampling rate) with the sampling frequency, leading to mirror images. Sine waves below the Nyquist frequency would look like this without the low-pass filter (mentally replace the triangle shapes with orthogonal steps). In fact there are some filterless DACs around, and all of them suffer from a ~3.5 dB drop-off at 20 kHz, a consequence of the visible amplitude modulation. In turn a typical anti-aliasing filter leads to the infamous «ringing», which may be audible.

 

[514077]

  Well, I can't help you that much. You will have to use google for more details on how to interpret the DR measurements.
One thing I can say for sure is that those recordings are far from great.
 
I'm still curious on why the Hugo isn't clipping

I posted on the Hugo thread that I was listening to a radio drama ripped at 128kb/s mp3.  It seemed to smoothen out the rough edges of the playback that I noticed on other devices.  Not saying it's a miracle worker, but I noticed something positive in the listening.
Guess yull  just have to break down and buy Hugo, Skerry.  Who needs food anyway?  Should listen to Stan Freeburg on HiFi.

 

[514077]

On another note:  I don't recall if I posted this question on another X5 thread, but, there's a little iten of the 2.5 update allowing the use of X-FAT formatted cards.  Is there an advantage to this?  Will the data be stored more efficiantly on the card than with FAT32?