The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Why most commercial masters clip.. Mas­ter­ing Plugins
Old 26th October 2017
  #91
Lives for gear
 

Quote:
Originally Posted by Jonathan Race View Post
What style of music?

I think I'd feel comfortable with a limiter there as a safety net, just catching the odd peak here and there but nothing heavy. I don't know if it could still be classed as "unlimited" then though
Agreed

My/our music is sort of somewhere between ambient and hip hop or electronic... some of it is well dynamic enough already I guess:
https://www.youtube.com/watch?v=etk1FoTeG8g

https://www.youtube.com/watch?v=1PDootaiwmI

Sorry to push my wares...
Old 27th October 2017
  #92
Motown legend
 
Bob Olhsson's Avatar
 

Verified Member
FWIW we didn't even have a limiter in the cutting room at Motown and we cut hotter vinyl than anybody today. I always turn it off for vinyl because all it does is distort. When an L2 is part of the sound, I ask the mixer to put one in just the bass and bass drum with no gain reduction.

As for clipped CDs, if we were seeing unprecedented sales it would be one thing but there is zero evidence that the clipping is helping and radio is seeing an unprecedented number of people not listening to entire songs. This is lemmings marching over a cliff in order to be fashionable.
Old 31st October 2017
  #93
Lives for gear
I was reading a little bit more about ISP's and came across some specs on a high end DAC stating it can handle true peaks of up to +5dB w/out analog clipping. Ok that's great - it's a high end DAC and it has some extra head room so ISPs aren't a big deal. What about your typical consumer DAC? Say on the iPhone? Is there any headroom there at all? There must be some no? I ask because I ran some scans in Audacity and using afclip on some music released in 2016, by a very popular band, mastered by a legend in the industry, and the ISPs measure in the thousands. These are 24-bit 44.1khz files made available by the band. When I listen to the tracks on my phone or in my car, they sound pretty damn good. When encoded to MP3, these ISP numbers go up into the tens of thousands (for a 4 minute track) with true peaks over +1dB, but the sound is still pretty damn good. Dropping the track in volume digitally 1.5dB removes all ISPs, but it sounds pretty much the same to my ears when volume is adjusted for.

I'm not advocating for clipping or not...just trying to wrap my head around the theory vs. what happens in practice. I get that a lot of it is "fear" based and I agree with that notion (even though it is stupid since high RMS/Loudness can be attained easily with peaks at -1dBFS). Is the the biggest reason to avoid even one ISP on a track what happens in the MP3/AAC conversion process? If so, wouldn't there be a pretty easy way to protect against that, not in the mastering process, but in the design of the compression algorithm?

I'd love to hear from folks that actually code the compression mechanisms used, and whether it's possible or not in the algorithm, to simply scan a master for true peak level, bring the track down by that amount, and then encode. Seems simple but I'm sure it's more complicated than that.
Old 31st October 2017
  #94
Gear Guru
 
UnderTow's Avatar
 

Verified Member
Quote:
Originally Posted by M Mastering View Post
I was reading a little bit more about ISP's and came across some specs on a high end DAC stating it can handle true peaks of up to +5dB w/out analog clipping. Ok that's great - it's a high end DAC and it has some extra head room so ISPs aren't a big deal. What about your typical consumer DAC? Say on the iPhone? Is there any headroom there at all? There must be some no? I ask because I ran some scans in Audacity and using afclip on some music released in 2016, by a very popular band, mastered by a legend in the industry, and the ISPs measure in the thousands. These are 24-bit 44.1khz files made available by the band. When I listen to the tracks on my phone or in my car, they sound pretty damn good. When encoded to MP3, these ISP numbers go up into the tens of thousands (for a 4 minute track) with true peaks over +1dB, but the sound is still pretty damn good. Dropping the track in volume digitally 1.5dB removes all ISPs, but it sounds pretty much the same to my ears when volume is adjusted for.
And now you know why people have been clipping converters (or other stuff) for decades. If it sounds right, it is right.


Btw, I don't know which iPhone you are listening to but the iPhone 6 DAC is pretty damned good!

Apple iPhone 6 Plus Review
iPhone 6S Plus Audio Quality & Measurements

"Apple has more smart people and more resources than any other audio company on the planet, so as we see when it comes to audio engineering, the iPhone easily outdoes many so-called "audiophile" products."


Quote:
I'm not advocating for clipping or not...just trying to wrap my head around the theory vs. what happens in practice. I get that a lot of it is "fear" based and I agree with that notion (even though it is stupid since high RMS/Loudness can be attained easily with peaks at -1dBFS). Is the the biggest reason to avoid even one ISP on a track what happens in the MP3/AAC conversion process? If so, wouldn't there be a pretty easy way to protect against that, not in the mastering process, but in the design of the compression algorithm?

I'd love to hear from folks that actually code the compression mechanisms used, and whether it's possible or not in the algorithm, to simply scan a master for true peak level, bring the track down by that amount, and then encode. Seems simple but I'm sure it's more complicated than that.
I don't think a lossy encoding algorithm is the right place for this kind of thing. You can be guaranteed that it would cause endless emails and online complaints from people that are not happy that their MP3's sound less loud than the source they are made from. These tools should just take whatever is thrown at them and encode them without doing and level changes (or EQing or anything else).


Alistair
Old 31st October 2017
  #95
Lives for gear
 
bcgood's Avatar
 

Verified Member
Quote:
Originally Posted by UnderTow View Post
Apple iPhone 6 Plus Review
iPhone 6S Plus Audio Quality & Measurements

"Apple has more smart people and more resources than any other audio company on the planet, so as we see when it comes to audio engineering, the iPhone easily outdoes many so-called "audiophile" products."



Alistair


Wow, I always wondered about that. Sometimes I plug my nice headphones into my iPhone 6S and listen to music and it does sound good. Also, it probably helps that I mostly use wave files from CDs in my iTune collection.
Old 31st October 2017
  #96
Motown legend
 
Bob Olhsson's Avatar
 

Verified Member
The problem is that "sounding right" using one converter often doesn't "sound right" with others. We are always mastering for a moving target.
Old 31st October 2017
  #97
Lives for gear
 
bcgood's Avatar
 

Verified Member
Quote:
Originally Posted by Bob Olhsson View Post
The problem is that "sounding right" using one converter often doesn't "sound right" with others. We are always mastering for a moving target.
It definitely seems like it would be good if there where more specific standardization to optimize audio quality and consistency for everyone.
Old 1st November 2017
  #98
Lives for gear
 
Jerry Tubb's Avatar
 

Verified Member
Quote:
Originally Posted by UnderTow View Post
iPhone 6S Plus Audio Quality & Measurements

"Apple has more smart people and more resources than any other audio company on the planet, so as we see when it comes to audio engineering, the iPhone easily outdoes many so-called "audiophile" products."
good review on the the 6S Plus!

we intentionally bought that one instead of the 7 a couple years ago,

specifically because i liked the sound of the audio output jack!

best, jt
Old 1st November 2017
  #99
Lives for gear
 
Hermetech Mastering's Avatar
 

Verified Member
I have a 6S plus, didn't realise it was so hi-fi lauded. My original Fiio X5 still sounds way better though.
Old 2nd November 2017
  #100
Lives for gear
 

Quote:
Originally Posted by M Mastering View Post
I'd love to hear from folks that actually code the compression mechanisms used, and whether it's possible or not in the algorithm, to simply scan a master for true peak level, bring the track down by that amount, and then encode. Seems simple but I'm sure it's more complicated than that.
You can perform forensic magic in RX to some extent. Say you have 1 or 2 annoying ISPs detected in the whole track - you can zoom right into the individual sample levels and pull just the adjoining samples down which are causing the ISP between them. As far as I can tell this is completely transparent when working in 32-bit floating.

Which makes me wonder if some kind of automated/intelligent version of this technique will eventually happen.
Old 2nd November 2017
  #101
Motown legend
 
Bob Olhsson's Avatar
 

Verified Member
The Sonnox codec manager or the Nugen master check pro both tell the tale. I should add that distortion accumulates with each additional process. The clipped material also really pushed the volume down in FM broadcasts which isn't exactly what one hopes to gain from clipping.
Old 3rd November 2017
  #102
Gear Guru
 
UnderTow's Avatar
 

Verified Member
Quote:
Originally Posted by MogwaiBoy View Post
You can perform forensic magic in RX to some extent. Say you have 1 or 2 annoying ISPs detected in the whole track - you can zoom right into the individual sample levels and pull just the adjoining samples down which are causing the ISP between them. As far as I can tell this is completely transparent when working in 32-bit floating.
What is annoying about the ISP's? That they are audible or simply that they exist? I ask because your cure is probably worse than the problem itself.

The reason the editing is inaudible is probably because the ISP's themselves are inaudible due to being too short to be audible. Adjusting single samples this way is, inherently, not taking the actual signal into account, just the sample values.

From a technical point of view, you are probably causing just as much damage if not more by editing this way than the ISPs themselves cause. You just don't hear it because the issues are too short (either way).

At least with ISPs you know that on DACs with headroom, the signal will be accurately reconstructed. With the type of editing you propose, it is impossible for the signal to be accurate.

Quote:
Which makes me wonder if some kind of automated/intelligent version of this technique will eventually happen.
Any True Peak limiter does something similar but better because they take the signal into account, not just the sample values.

Alistair
Old 3rd November 2017
  #103
Gear Addict
 
bonne's Avatar
 

Quite a few "probably"s in your rationale here, Alistair. Have you looked into this yourself or are you mostly thinking out loud and theorizing?

MogwaiBoy did not say "Adjusting single samples this way", but rather "Say you have 1 or 2 annoying ISPs detected in the whole track - you can zoom right into the individual sample levels and pull just the adjoining samples down which are causing the ISP between them".

RX does this with 8x oversampling, so you get a pretty good idea what is going on, and can double-check after your adjustment if you have improved the sound or not.

Not only do customers use DACs of varying quality and headroom to present their music, after masters leave your studio, but many also make lossy format copies of tracks you have worked on for various purposes. Your ISPs may come back stronger and cause unnecessary audio problems downstream.

JB
Old 3rd November 2017
  #104
Gear Guru
 
UnderTow's Avatar
 

Verified Member
Quote:
Originally Posted by bonne View Post
Quite a few "probably"s in your rationale here, Alistair.
Not really. If it isn't clear how I come to my conclusions, feel free to ask me for a clarification of any step in my thinking.

Quote:
Have you looked into this yourself or are you mostly thinking out loud and theorizing?
I am talking from the perspective of looking at how digital sampling works both in theory and in practise.

Quote:
MogwaiBoy did not say "Adjusting single samples this way", but rather "Say you have 1 or 2 annoying ISPs detected in the whole track - you can zoom right into the individual sample levels and pull just the adjoining samples down which are causing the ISP between them".
Any editing of samples this way does not take the signal into account. No theorizing. Fact.

Quote:
RX does this with 8x oversampling, so you get a pretty good idea what is going on, and can double-check after your adjustment if you have improved the sound or not.
Does what with 8x oversampling? MogwaiBoy is talking about editing individual samples with the mouse. What is getting oversampled?

As for double-checking adjustments, that is why I asked what was annoying about the ISPs in question. Are they annoying because they are audible or simply because they get flagged as such by some software? If the ISPs are not audible to start with, there is not much point in fixing them. At least not this way.

Quote:
Not only do customers use DACs of varying quality and headroom to present their music, after masters leave your studio, but many also make lossy format copies of tracks you have worked on for various purposes. Your ISPs may come back stronger and cause unnecessary audio problems downstream.
Sure but the exact same thing applies to any non-signal based sample editing as proposed by MogwaiBoy. Either you trust what you hear in your studio or you don't. And if you don't, at the very least stick to tools that work on the signal, not on the individual samples. The individual samples are not the audio.

Alistair
Old 3rd November 2017
  #105
Gear Addict
 
bonne's Avatar
 

Quote:
Originally Posted by UnderTow View Post
Does what with 8x oversampling? MogwaiBoy is talking about editing individual samples with the mouse. What is getting oversampled?

- - -

Sure but the exact same thing applies to any non-signal based sample editing as proposed by MogwaiBoy. Either you trust what you hear in your studio or you don't. And if you don't, at the very least stick to tools that work on the signal, not on the individual samples. The individual samples are not the audio.

Alistair
Of course individual samples are not the audio, Alistair. For us oldtimers that goes without saying - no?

I sounds like you are unfamiliar with iZotope's RX application - am I correct?

RX calculates sample peak values and true peak values with 9x oversampling in the Waveform Statistics window, to get close to an approximation of what a typical DAC will do to your audio when converted. This also gives a good idea of what happens with ISPs that might come back and turn your audio into a strident and harsh mess, when further converted to lossy formats for online upload and what not. Many examples of this in today's releases.

What MogwaiBoy has offered here is a way to counteract this by going to the Waveform Statistics window and from there very conveniently make the neccesary adjustments, not on a single sample as you claim - it can be anything from a handful of samples to a string of hundreds of sample when neccesary. It's all determined when double checking the processing.

For some of us it's not enough to trust what we hear in our studio - or not. When that audio leaves the studio it might be very fragile if exposed to further lossy processing, even if it sounded like a million dollars on your favourite speakers.

JB
Old 3rd November 2017
  #106
Gear Guru
 
UnderTow's Avatar
 

Verified Member
Quote:
Originally Posted by bonne View Post
Of course individual samples are not the audio, Alistair. For us oldtimers that goes without saying - no?
But you are arguing as though they are or you are not understanding MogwaiBoy's post. He is talking about changing the level of individual samples with the mouse. A.K.A. editing the samples without consideration for the actual signal.

Quote:
I sounds like you are unfamiliar with iZotope's RX application - am I correct?
Incorrect.

Quote:
RX calculates sample peak values and true peak values with 9x oversampling in the Waveform Statistics window, to get close to an approximation of what a typical DAC will do to your audio when converted.
This is just the visual feedback/metering. Non of this is of any relevance when dragging the level of individual samples with the mouse. The processing is not oversampled and the export is not oversampled.

Editing the individual samples like this can cause distortion and aliasing that will not be measurable. It won't show up as an ISP, sure, but you are still tampering with the data stream with no consideration for the audio signal encoded in the data stream.

Quote:
This also gives a good idea of what happens with ISPs that might come back and turn your audio into a strident and harsh mess, when further converted to lossy formats for online upload and what not. Many examples of this in today's releases.
Individual ISPs (MogwaiBoy mentioned "1 or 2 annoying ISPs detected") will not cause "a strident and harsh mess". They are unlikely to be audible at all.

Also, what you seem to keep missing is that I am not advocating that we create masters full of ISPs. I am advocating using tools that take the signal into account rather than editing the level of individual samples. So a True Peak limiter with ISP protection or lowering the entire level to avoid ISPs. (Whatever is appropriate for the job at hand).

Quote:
What MogwaiBoy has offered here is a way to counteract this by going to the Waveform Statistics window and from there very conveniently make the neccesary adjustments, not on a single sample as you claim -
Go and reread MogwaiBoy's post. It states very clearly: "you can zoom right into the individual sample levels and pull just the adjoining samples down which are causing the ISP between them."

And I maintain that this is likely to cause as many problems as it solves, especially if the 1 or 2 individual ISPs where not even audible to start with (which they probably weren't).

Quote:
For some of us it's not enough to trust what we hear in our studio - or not. When that audio leaves the studio it might be very fragile if exposed to further lossy processing, even if it sounded like a million dollars on your favourite speakers.
And you are still missing the point that you can not trust this approach to solve ISPs either. It does NOT take the signal into account. The actual signal or sound is just as likely to break-up on a cheaper DAC outside of your studio with this type of editing as leaving the ISPs in. This is simply not a valid solution and I even question whether there is a problem in the first place (which is why I asked if the ISPs were audible).

Alistair
Old 3rd November 2017
  #107
Lives for gear
Quote:
Individual ISPs (MogwaiBoy mentioned "1 or 2 annoying ISPs detected") will not cause "a strident and harsh mess". They are unlikely to be audible at all.
agreed. i can think of at least one record in my collection that occasionally hits +3dbfs on a true peak meter. you'd never know from listening.

i've done a bit of manual sample-level redrawing, in my experience it can work ok on higher frequencies but lower frequencies will distort in an obvious, unpleasant way. generally i think it's better just to let the clipper/limiter handle them.
Old 3rd November 2017
  #108
Lives for gear
 
SmoothTone's Avatar
 

Quote:
Originally Posted by scraggs View Post
agreed. i can think of at least one record in my collection that occasionally hits +3dbfs on a true peak meter. you'd never know from listening.

i've done a bit of manual sample-level redrawing, in my experience it can work ok on higher frequencies but lower frequencies will distort in an obvious, unpleasant way. generally i think it's better just to let the clipper/limiter handle them.
Thanks for explaining that in terms I can understand!

OT, but how's the room shaping up? Care to update us in the "roomz" thread?
Old 4th November 2017
  #109
Gear Addict
 
bonne's Avatar
 

Quote:
Originally Posted by UnderTow View Post
Individual ISPs (MogwaiBoy mentioned "1 or 2 annoying ISPs detected") will not cause "a strident and harsh mess". They are unlikely to be audible at all.

Alistair
Hi scaggs, I didn't say this, Alistair is putting words in my mouth again.

This is what I said:
"RX calculates sample peak values and true peak values with 9x oversampling in the Waveform Statistics window, to get close to an approximation of what a typical DAC will do to your audio when converted. This also gives a good idea of what happens with ISPs that might come back and turn your audio into a strident and harsh mess, when further converted to lossy formats for online upload and what not. Many examples of this in today's releases."

and

"What MogwaiBoy has offered here is a way to counteract this by going to the Waveform Statistics window and from there very conveniently make the neccesary adjustments, not on a single sample as you claim - it can be anything from a handful of samples to a string of hundreds of sample when neccesary. It's all determined when double checking the processing." (- Alistair conveniently left out the last two sentences in his quote)

When listening back and double checking we are indeed listening to the analog signal before making the decision on the processing. Didn't feel the need to spell that out, it goes without saying.

scaggs: "in my experience it can work ok on higher frequencies but lower frequencies will distort in an obvious, unpleasant way. generally i think it's better just to let the clipper/limiter handle them". I have not found this to be true in my years of carefully finetuning this technique. It all depends on how big a sample window you use when determining how to go about treating clips from lower frequencies, without leaving tell-tale distortions, and you will know right away when you go too far. Clippers and look-ahead limiters tend to leave more distortions IME.

JB
Old 4th November 2017
  #110
Gear Guru
 
UnderTow's Avatar
 

Verified Member
Quote:
Originally Posted by bonne View Post
Hi scaggs, I didn't say this, Alistair is putting words in my mouth again.
I am not putting words in anyone's mouth. I am still responding in light of MogwaiBoy's post that I originally responded to.

If you are not addressing that specific scenario then we are not talking about the same thing.


Quote:
This is what I said:
"RX calculates sample peak values and true peak values with 9x oversampling in the Waveform Statistics window, to get close to an approximation of what a typical DAC will do to your audio when converted. This also gives a good idea of what happens with ISPs that might come back and turn your audio into a strident and harsh mess, when further converted to lossy formats for online upload and what not. Many examples of this in today's releases."
I snipped part of the quote because it doesn't change my point. 1 or 2 isps will not cause "a strident and harsh mess", even " when further converted to lossy formats for online upload and what not".

My points in a nutshell:

- A few short ISPs are unlikely to be audible before or after lossy encoding.
- The metering in RX will not tell you whether an ISP becomes audible down the line or not. (A tool like Sonnox Encoder Tool can help though).
- Editing individual samples the way MogwaiBoy proposes does not take the signal into account and is likely to cause as many problems as the ISPs being corrected.
- If there are many ISPs in the signal, there are IMO better ways to address them than direct manipulation of individual sample values.

Quote:
and

"What MogwaiBoy has offered here is a way to counteract this by going to the Waveform Statistics window and from there very conveniently make the neccesary adjustments, not on a single sample as you claim - it can be anything from a handful of samples to a string of hundreds of sample when neccesary. It's all determined when double checking the processing." (- Alistair conveniently left out the last two sentences in his quote)
I left these sentences out because 1) That isn't what MogwaiBoy wrote so that is not what I am talking about. 2) Whether grabbing one, several, or even thousands of samples and adjusting their level by hand, you are adjusting the data stream rather than the encoded signal. (Regardless of whether the damage is audible or not). 3) If the original isps were not audible to start with, it makes no sense to do this type of editing as there is a good chance you are fixing a problem that doesn't exist while potentially introducing new problems.

Quote:
When listening back and double checking we are indeed listening to the analog signal before making the decision on the processing. Didn't feel the need to spell that out, it goes without saying.
And the same applies to listening to isps hence my question to MogwaiBoy whether he could hear the isps or is just adjusting the samples in response to the metering. My comments apply to the latter scenario.

Something else to keep in mind: In most cases, isps will clip a DAC in the digital domain, usually in the anti-imaging filter, unless there is digital attenuation Pre-reconstruction. In many (most?) cases I reckon the consumers will not have the volume of their players maxed out so they will not have any issues with ISPs even if their cheap consumer DACs have no extra headroom in the analogue domain as the volume adjustment in their players happens pre-DAC.

It depends on many factors. For instance if listening to an MP3 that gets decoded into a floating-point stream that gets attenuated pre-DAC, isps probably won't clip anything. (Assuming ISPs didn't clip during MP3 encoding). On the other hand, if the player is playing at full scale, the ISPs are likely to clip at some point in the signal chain.

This brings up an important consideration for monitoring: If one is using digital volume control and one is not listening at full volume, one might not hear issues that would become audible if the digital volume was set to full scale. This is an area where an analogue volume control is an advantage as it allows one to leave the digital signal at full scale thus better revealing any ISP issues.

Quote:
scaggs: "in my experience it can work ok on higher frequencies but lower frequencies will distort in an obvious, unpleasant way. generally i think it's better just to let the clipper/limiter handle them". I have not found this to be true in my years of carefully finetuning this technique. It all depends on how big a sample window you use when determining how to go about treating clips from lower frequencies, without leaving tell-tale distortions, and you will know right away when you go too far. Clippers and look-ahead limiters tend to leave more distortions IME.

JB
Which limiter(s) are you using for this? They don't all work as well. And I agree that a clipper isn't the solution either. Clippers don't address isps at all. (At least none that I have tried). That said, as I wrote previously, there is always the option to simply lower the overall level to avoid ISPs altogether.

Alistair
Old 5th November 2017
  #111
Gear Addict
 
bonne's Avatar
 

Oh yes, you are putting words in my mouth, Alistair, but don't seem to realise that yourself. The sentence you have gotten so hung on up: "1 or 2 isps will cause "a strident and harsh mess" is your own creation. You wont find that sentence in my posts. Nowhere did I say or imply that "1 or 2 isps will cause a strident and harsh mess".

You frabricated this hybrid sentence from Mogway's: "Say you have 1 or 2 annoying ISPs detected in the whole track" and my own: "This also gives a good idea of what happens with ISPs that might come back and turn your audio into a strident and harsh mess, when further converted to lossy formats for online upload and what not. Many examples of this in today's releases." - and then pretended this is something I have said or something that represents my stance on the topic we are discussing. This is an underhanded debate technique that I don't much care for.

Why you would do that is anybody's guess. I don't know you and don't know what's behind that side of your personality.

I wont start the guessing, but I have noticed through the years on GS that you show a propensity to get into very heated and unpleasent exchanges with fellow slutz from time to time - where you go to town with massive multi-quoting, critiqueing and insulting members who disagree with you, much like you do here. Boring stuff, if you ask me and counterproductive to good and interesting discussions on GS.

If you would care to come down off your high horse and relate to other members here on an egalitarian basis, you might learn something, and I will continue the discussion with you. If you choose to continue in same manner we've seen in last couple of days I'm out of the discussion.

JB
Old 5th November 2017
  #112
Lives for gear
Quote:
Originally Posted by bonne View Post
Oh yes, you are putting words in my mouth, Alistair, but don't seem to realise that yourself. The sentence you have gotten so hung on up: "1 or 2 isps will cause "a strident and harsh mess" is your own creation. You wont find that sentence in my posts. Nowhere did I say or imply that "1 or 2 isps will cause a strident and harsh mess".

You frabricated this hybrid sentence from Mogway's: "Say you have 1 or 2 annoying ISPs detected in the whole track" and my own: "This also gives a good idea of what happens with ISPs that might come back and turn your audio into a strident and harsh mess, when further converted to lossy formats for online upload and what not. Many examples of this in today's releases." - and then pretended this is something I have said or something that represents my stance on the topic we are discussing. This is an underhanded debate technique that I don't much care for.

Why you would do that is anybody's guess. I don't know you and don't know what's behind that side of your personality.

I wont start the guessing, but I have noticed through the years on GS that you show a propensity to get into very heated and unpleasent exchanges with fellow slutz from time to time - where you go to town with massive multi-quoting, critiqueing and insulting members who disagree with you, much like you do here. Boring stuff, if you ask me and counterproductive to good and interesting discussions on GS.

If you would care to come down off your high horse and relate to other members here on an egalitarian basis, you might learn something, and I will continue the discussion with you. If you choose to continue in same manner we've seen in last couple of days I'm out of the discussion.

JB
I noticed the same thing. Don't mean to gang up on him, but Undertow always seems to ignore the obvious intentions of a post's message and selectively skim for small errors or technicalities that he can then magnify out of proportion in ways that bury it; often derailing the entire conversation, hijacking the attention of the OP, and diminishing the perceived credibility of the poster that made the reply he is dissecting.

The strange thing is that he usually does this with posts that actually agree with his opinions or are in concurrence with his facts. Which is exactly the situation here. You both are saying that 1 or 2 ISPs won't be audibly noticeable, yet for some reason he chose to "straw man" you as somebody who disagrees.

He's obviously very intelligent but perhaps too much so for his own good in social situations. Perfectionism in informal communication doesn't often breed harmonious social interactions. I think he forgets sometimes that he's talking to people, not editing a textbook for a publisher haha

Last edited by psykostx; 5th November 2017 at 09:53 AM..
Old 5th November 2017
  #113
Gear Guru
 
UnderTow's Avatar
 

Verified Member
Quote:
Originally Posted by bonne View Post
Oh yes, you are putting words in my mouth, Alistair, but don't seem to realise that yourself. The sentence you have gotten so hung on up: "1 or 2 isps will cause "a strident and harsh mess" is your own creation. You wont find that sentence in my posts. Nowhere did I say or imply that "1 or 2 isps will cause a strident and harsh mess".

You frabricated this hybrid sentence from Mogway's: "Say you have 1 or 2 annoying ISPs detected in the whole track" and my own: "This also gives a good idea of what happens with ISPs that might come back and turn your audio into a strident and harsh mess, when further converted to lossy formats for online upload and what not. Many examples of this in today's releases." - and then pretended this is something I have said or something that represents my stance on the topic we are discussing. This is an underhanded debate technique that I don't much care for.
I am not fabricating anything. As I pointed out multiple times, I am responding in light of MogwaiBoy's post. I am addressing what is implied by the context of the conversation. In case you forgot, you responded to MY post. If you are not talking about the same thing as me, you had no reason to respond to my post in the first place.

As a reminder, this is what I was responding to: "Say you have 1 or 2 annoying ISPs detected in the whole track - you can zoom right into the individual sample levels and pull just the adjoining samples down which are causing the ISP between them."

If you are responding to my post then we are talking about editing out (a few), probably inaudible, ISPs by hand. In THAT context, what you wrote is either about a few ISPs including whether something turns into a strident harsh mess or not or your post had nothing to do with anything I wrote.

If there really are thousands or tens of thousands of ISPs in a file, enough to cause "a strident and harsh mess" after encoding/decoding, hand editing every offending ISP is hardly the most efficient solution. Either way, there are better ways to address ISPs IMO which has been my point from my first post that you responded to.

Lastly, this is another important point, there is no guarantee that the presence or absence of ISPs as measured pre lossy encoding, has any bearing on whether there will or won't be (audible) ISPs or clipping during lossy encoding or decoding. The signal is being changed so much that the one does not reliably predict the other. There are ways to check this but pre-encoding metering is not a reliable indication.

None of the above means you can't edit ISPs by hand or do whatever you want in your own studio for your own clients. If you are happy with the results you get, then by all means continue but from a technical point of view, AFAIAA, what I write is accurate for all the reasons I have given (and more). If not, or if anything is unclear, please address the factual errors or technical flaws in my reasoning and leave out the ad-hominem attacks.

Quote:
I wont start the guessing, but I have noticed through the years on GS that you show a propensity to get into very heated and unpleasent exchanges with fellow slutz from time to time - where you go to town with massive multi-quoting, critiqueing and insulting members who disagree with you, much like you do here.
Says the guy attacking me instead of dealing with the topic at hand.

Quote:
Boring stuff, if you ask me and counterproductive to good and interesting discussions on GS.
Counter productive? Unlike you, I am bringing up relevant stuff in my posts and trying to deal with all the different aspects of ISPs and when they do or do not occur and when they are or are not a problem. Starting with the important distinction between what the metering says and what is actually audible followed by the important distinction between the samples in a data stream and the actual signal they represent. I also bring up the following:

- Where in the signal chain do ISPs actually cause problems? (In the digital Anti-imaging filters usually).
- What are the implications of this for different monitoring volume control options?
- How often are ISPs really a problem for the end user? (Maybe not that often due to pre-dac digital gain reduction in many modern scenarios).
- ISP metering pre-encoding does not reliably predict what will happen after lossy encoding. You need to actually encode/decode the signal to know what will happen with each codec.
- I also asked which limiter you used to come to the conclusion that it causes more distortion than sample level editing. Care to elaborate?

It is in light of all these and a few other considerations that I responded to MogwaiBoy's post.

If you are genuinely interested in furthering the conversation, you might want to address these points or bring up new ones but please stop going around in circles.

Alistair
Old 5th November 2017
  #114
Motown legend
 
Bob Olhsson's Avatar
 

Verified Member
Quote:
Originally Posted by bcgood View Post
It definitely seems like it would be good if there were more specific standardization to optimize audio quality and consistency for everyone.
This is precisely why avoiding intersample peaks is important.
Old 18th January 2018
  #115
Lives for gear
 
Masterer's Avatar
 

Verified Member
Quote:
Originally Posted by Greg Reierson View Post
Are you listening to low bitrates codecs that are common for streaming services?

Yes.
Old 18th January 2018
  #116
Lives for gear
 
Jerry Tubb's Avatar
 

Verified Member
Quote:
Originally Posted by Masterer View Post
Yes.
Good to see you here Masterer!

jt
Old 18th January 2018
  #117
Gear Nut
 

Quote:
Originally Posted by Bob Olhsson View Post
The clipped material also really pushed the volume down in FM broadcasts which isn't exactly what one hopes to gain from clipping.
+1. Same for vinyl.

I can hear hard-clipped distortion on FM from across the room even at very low casual listening levels. FM's preemphasis combined with transmitter/antenna system synchronous AM noise and receivers on the threshold of limiting due to low signal level or multipath provide a brutal reminder of how bad clipped masters sound.

Edit: I should add that in FM the difference channel is very fragile compared to Mid. In the US, it sits out on a 38 kHz AM-modulated subcarrier with a deviation significantly lower than Mid. Side gets the crap beaten out of it due to transmission impairments. Same as vinyl's vertical modulation.

HF clipped transients hammer the low frequencies on FM and mixes of some really great songs become a sea of intermodulated HF splattering grit. It's not like I live and drive in a fringe area either. Most times I'm in visual sight of the antenna farm or under it.

I stopped buying anything I hear that's hyper-saturated because I know I'll be disappointed and fatigued by listening to it. I enjoy it the best I can on the radio (or the broadcasters stream) but just won't open my wallet for it.

This non comm station isn't highly-processed and good-sounding masters sound clean. Clipped ones sound like dog poo. You'll hear a lot of new Indie material here: Listen Live | KXT 91.7
Old 18th January 2018
  #118
Gear Nut
 

BØRNS vs. Toad the Wet Sprocket.

90% of Americans listen to the radio weekly.

This is an actual recorded segue broadcast this morning on KXT-FM. Compare the hyper-saturated BØRNS track to Toad the Wet Sprocket's open mix. The transition from distorted dog poo to music shows its not the station's processing: It's the product.

The first clip (60s) is in the Left/Right domain. The second clip is M-S/Sum-Difference with Side in the right channel. Side is always revealing...

The reception conditions are below-average and typical of what you would hear in a car in the presence of terrain shielding. The receiver is being forced into mono - more so with the BORNS track - resulting in the Side channel muting. Stereo muting, or variable blend, are common. The Mid LF modulation is punching holes in Side. On the BORNS track I could see the Stereo indicator flicker.

BØRNS vs. Toad the Wet Sprocket: http://www.proaudiodesignforum.com/c...st_Example.mp3

BØRNS vs. Toad the Wet Sprocket in MS: http://www.proaudiodesignforum.com/c...n_Mid_Side.mp3

This is how your stuff sounds in the wild folks. Some of it ain't pretty.

Which is more listenable?
Old 18th January 2018
  #119
Motown legend
 
Bob Olhsson's Avatar
 

Verified Member
Actually, almost all U.S. radio listening is in automobiles.
Old 18th January 2018
  #120
Gear Nut
 

One thing that I discovered is that clipped saturated peaks boomerang back at you once they undergo phase rotation from pre-emphasis such as RIAA, FM and possibly some codecs.

You can think of saturation as a non-linear transfer curve or you can think of it as modulation of the signal by itself. A zero-attack zero-release compressor with full-wave sidechain and no averaging is a good analog of a modulator/soft-clipper. The source gets modulated by even harmonics to produce compressive odd-order distortion.

The harmonics created, besides creating a lot of intermodulated ear burn, fold the peaks of waveform back onto itself to subtractively reduce peaks, lower crest factor and bump up RMS levels.

When the phase relationship of the odd-order harmonics that caused the clipping to begin with are rotated relative to the fundamental, the peaks sum back and start to resume their original shape. The crest factor goes back to what it was and the "benefit" of saturation is lost.

Unfortunately phase rotation is not an "undue" for the distortion produced by hypersaturation because all the IM is still in there. If the phase rotation has a complimentary phase rotation, e.g. pre-emphasis/de-emphasis, then things go back to where they were. The problem is in the intermediate processes between encode/decode pre-emphasis. Take vinyl...

I did an experiment with clipped tracks and RIAA encode/decode. What I found was that the more the source became clipped the more level had to be reduced to maintain the same maximum peak level post-RIAA-encode. Clipping actually reduced the final RMS level that could be cut to disk. The material would have been better off and louder if it hadn't been clipped at all.

The real tragedy is that the music gets mangled and the overall goal of increasing loudness, by lowering crest factor is not only not realized but - in some circumstances - actually works against you.

Saturated peaks are like a bad penny when used excessively: They often come back.
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump