Volume errors and false peaks

[Quad]

my ears tell me the sound is fine, the master volume on Acid Pro 6 visually depicts that the sound is breaking the peak and is in the red, can anyone tell me if its full of s**t, cause i'd like to think it is. Considering my music is at par with how loud i want it, im all for just turn up the volume but jesus christ its gotta be false.

[Quad]

anyone? Would really be a big help to know, would it even matter if im going to take the tracks through Cd architect anyway?

[Nordenstam]

Turn down the master fader in acid by 10dB's and raise gain on your pre-amplifier by an equal amount! It'll be exactly the same loudness in your room, sans peaking/clipping.

Oh.. but your mastering your own tracks, right? Well, if you like the sound of digital clipping(...), go ahead. Problem is that it'll be entirely unpredictable as the clipped audio overloads the output stages of the converters of the end users to various degrees. There's no way to know how this distortion will sound on the various pieces of playback gear available. If you like distortion, try adding it in some other (more predictable) way if you want control of the process.

[ed littman]

Originally Posted by Quad

my ears tell me the sound is fine, the master volume on Acid Pro 6 visually depicts that the sound is breaking the peak and is in the red, can anyone tell me if its full of s**t, cause i'd like to think it is. Considering my music is at par with how loud i want it, im all for just turn up the volume but jesus christ its gotta be false.

I don't know acid pro, but most programs run at higher internal sample rates that allow your master fader to clip without artifacts(to a point). The key is to see how it sounds after bouncing this file to a 16/44.1 file.
Ed

[dcollins]

Originally Posted by ed littman

I don't know acid pro, but most programs run at higher internal sample rates that allow your master fader to clip without artifacts(to a point).

Is this correct?

DC

[Ged Leitch]

Originally Posted by dcollins

Is this correct?

DC

Hi Dave, hope your good mate,
I do this in wavelab and raise the master fader by about 3db till i get the level - pre limiter, overs usually no more than 1db.

But of course...I don't need to tell you that

Dunno about higher internal sample rate though...

[ed littman]

Originally Posted by dcollins

Is this correct?

DC

NO
I meant bit rate
With some mixes I can clip at least 4db over with minimal artifacts. I would account this to the higher internal bit rate.
please school me if I way off base....
Ed

[Quad]

Havent tried the sample rate yet, but its master fader/pre amp trick worked pretty much perfect, i dont exactly follow the loudness trends i jus want somthing that hit hard enough for me, now all i have to do is wrestle with the compressor.

[max cooper]

I don't pretend to know the math of it all, but higher internal bitrate is something I've never heard of. Wouldn't they use this as a selling point?

[Nordenstam]

Originally Posted by Quad

Havent tried the sample rate yet, but its master fader/pre amp trick worked pretty much perfect, i dont exactly follow the loudness trends i jus want somthing that hit hard enough for me, now all i have to do is wrestle with the compressor.

That's splendidly great! Sincere apologies for assuming you where hunting for loudness by looking at the peak level meter. Keep up the good ear work! =)


As for the digital word length("bitrate), more or less should not make any difference to the clipping.

A floating point architecture can not run out of level if used in any predictable way, but the analogue converters and final 16 bit (fixed/integer) storage format is not floating. In fixed/integer/regular digital words, there is no headroom what so ever. If it exceeds zero, it clips. 4 or 8 or 48 bits gives the same results above zero: clipping. Though there is no headroom, there can be plenty of footroom! It's needed to cope with all the math, most calculations tends to increase the word lenght. Hence double precision(48 bit) and 64 bit computing.

These long words means there is an easy way to gain more than enough headroom even in a fixed bit world. Shifting the digital audio level down and away from the absolute zero( ) gives both headroom and footroom in a fixed/integer computing environment. The footroom is way big enough so the level shift does not do anything negative to the lowest levels at all, but the abudance of headroom is very positive.


Andreas N


PS, Max Cooper: it IS a selling point! Watch out for the Double Precision buzzword.

[max cooper]

Originally Posted by Lupo

PS, Max Cooper: it IS a selling point! Watch out for the Double Precision buzzword.

OK, that makes some sense. So how do they achieve this?

[Nordenstam]

Most calculations increase the word lenght. Like 2.21*3.31=7.3151. Those extra digits have to go somewhere to be stored, or else they'll be truncated. Truncating the above result to two decimals would give 7.31 - rounding it in the usual fashion would give 7.32, both are still pretty much off the actual figure which is almost in the middle. To avoid the truncation error, more bits are needed.

Double precision refers to using two storage slots instead of one for each number. In a normal computer, each storage slot is 32 bits wide, so double precision is 64 bits. If the starting point is 24 bits, as in typical audio, the computing device would have 48 bits available if using double precision. That's a typical format for specialized DSP chips in hardware boxes. More bits can be added in other ways too, like adding 32 extra bits to the 24 to get to 64 - again suitable for PC use.

The actual way of adding bits is to add a string of zeroes, inactive bits, to the word. As the word passes through the computing operations, the word length increases and those inactive bits gets more or less filled with relevant data.


Andreas N

[Quad]

Originally Posted by Lupo

That's splendidly great! Sincere apologies for assuming you where hunting for loudness by looking at the peak level meter. Keep up the good ear work! =)

Indeed, im not a fan of the loudness war, i have a volume knob or remote to adjust the volume haha

[masteringhouse]

Originally Posted by ed littman

NO
I meant bit rate
With some mixes I can clip at least 4db over with minimal artifacts. I would account this to the higher internal bit rate.
please school me if I way off base....
Ed

Isn't it bit depth?

[Ronny]

Originally Posted by masteringhouse

Isn't it bit depth?

Yes, it's bit depth not rate, but in particular floating point that allows to process over -0dBFs. You still have to attenuate somewhere before the DAC back to -0dBFs though, or you can run into problems on various cd players. If you can play +4dBFs and can't hear any distortion after the DAC, it's probably because the peak transients are far enough apart and happen too fast on the timeline to be noticed as distortion. IOW, only a few samples over at each transient that exceeds -0dB, but never the less the signal is considered illegal when it exceeds -0dBFs at the DAC. The way to get more footroom is by increasing bit depth, not necessarily increasing sample rates.

Bit rate has to do with the speed of a transfer or a streaming speed, such as constant or variable kbps for mp3's and mp4's.

[masteringhouse]

Originally Posted by Ronny

Yes, it's bit depth not rate, but in particular floating point that allows to process over -0dBFs. You still have to attenuate somewhere before the DAC back to -0dBFs though, or you can run into problems on various cd players. If you can play +4dBFs and can't hear any distortion after the DAC, it's probably because the peak transients are far enough apart and happen too fast on the timeline to be noticed as distortion. IOW, only a few samples over at each transient that exceeds -0dB, but never the less the signal is considered illegal when it exceeds -0dBFs at the DAC. The way to get more footroom is by increasing bit depth, not necessarily increasing sample rates.

Bit rate has to do with the speed of a transfer or a streaming speed, such as constant or variable kbps for mp3's and mp4's.

I've always thought that 0 dBFs was the same regardless of bit depth? Larger wordlengths increase the footroom not headroom?. Or maybe we're just getting into semantics of how dynamic range is defined when converting from various wordlengths?

[bob katz]

Originally Posted by masteringhouse

I've always thought that 0 dBFs was the same regardless of bit depth? Larger wordlengths increase the footroom not headroom?. Or maybe we're just getting into semantics of how dynamic range is defined when converting from various wordlengths?

You are correct, Tom. 0 dBFS is a fixed level regardless of wordlength. So a 24 bit file can define additional levels BELOW the lowest level that's defined in 16 bit.

However, in floating point, you can define and use levels ABOVE 0 dBFS for calculation purposes only. They are just not useable in the real world. Floating point is an intermediate calculation format that you can store in a file, but as soon as it's played back it has to be converted to fixed. And any levels that are in the floating point file that are above 0 dBFS will result in an illegal and distorted level.

Hope this helps,


Bob

[masteringhouse]

Originally Posted by bob katz

You are correct, Tom. 0 dBFS is a fixed level regardless of wordlength. So a 24 bit file can define additional levels BELOW the lowest level that's defined in 16 bit.

However, in floating point, you can define and use levels ABOVE 0 dBFS for calculation purposes only. They are just not useable in the real world. Floating point is an intermediate calculation format that you can store in a file, but as soon as it's played back it has to be converted to fixed. And any levels that are in the floating point file that are above 0 dBFS will result in an illegal and distorted level.

Hope this helps,


Bob

Thanks Bob.

I was leaving that aspect of 32 float out of the equation assuming that we were only exercising 24 bits of the 32. A good practice considering your explanation, correct?

[Ronny]

Originally Posted by masteringhouse

I've always thought that 0 dBFs was the same regardless of bit depth? Larger wordlengths increase the footroom not headroom?. Or maybe we're just getting into semantics of how dynamic range is defined when converting from various wordlengths?

All fixed integer bit depths have a peak of -0dBFs. When you add bits you aren't increasing the peak, you are lowering the floor, so IMHO, footroom is more appropriate than saying going from 16 to 24 adds headroom, because you are actually extending the noise floor of the system down, not the ceiling above which is always -0dBFs at 16 or 24 fixed int. Increasing word can allow more headroom to be utilized for example in a live orch recording, because you can lower peak input due to more dynamic range at the bottom. This aids in the prevention of getting overs, but it's still the added footroom pusing the noise floor down that allows it. IOW, it doesn't add headroom above the legal limit of -0dBFs.

[masteringhouse]

Originally Posted by Ronny

All fixed integer bit depths have a peak of -0dBFs. When you add bits you aren't increasing the peak, you are lowering the floor, so IMHO, footroom is more appropriate than saying going from 16 to 24 adds headroom, ...

Thanks Ronny I misunderstood your initial post. In light of what you and Bob mentioned later it makes complete sense.

[ed littman]

Originally Posted by Ronny

Yes, it's bit depth not rate,

Bit rate has to do with the speed of a transfer or a streaming speed, such as constant or variable kbps for mp3's and mp4's.

Thanks,
I must be getting lazy .......

Ed

[bob katz]

Originally Posted by masteringhouse

Thanks Bob.

I was leaving that aspect of 32 float out of the equation assuming that we were only exercising 24 bits of the 32. A good practice considering your explanation, correct?

Usually yes!

BK