Bus Level too hot!

[streak]

I have a problem I've never had before...I'm down to mixing a song that sounds the way I want. I have a master fader with compression/limiting being fed by bus 5-6. None of my tracks are even close to red-lining but bus 5-6 is way in the red. Audibly so. Anybody know if there is a way to pull down the feed to bus 5-6 without going back and trimming every track?

[grimepoch]

You could put an aux bus in there for bus 5-6 to go through and trim it down there. Internally you can be in the red as much as you want, DP has TONS of dynamic range. (I think like 400dB or more). The only time you need to MAKE SURE you are not redlining is at any point where that specific data is going to be written to disk (like a bounce to disk or a freeze). So, you can bring it down with whatever you want and you are not going to lose any data, just bring it down in range you need.

[streak]

Yeah, I thought about the aux bus route, but I figured it would already be slamming when it got to the aux bus. I was not aware of the internal headroom thing...Although I can definitely her!e some fuzziness around the edges. I must be at 401db! Thanks for the info

[grimepoch]

No problem. We've had a LOT of discussion about DP and how it handles the internal data. If you put the trim in on the aux, do you still hear the fuzziness? I very much DOUBT you are anywhere near close to 401dB Trust me!

The way it works, DP converts your 24 (or 16 bit) data into 32 bits. Those extra 8 bits are what form the dynamic range. Think of it as an exponent. (well, that is what it is!) So, when you get ready to send data out or bounce, you need to get that range down so that the entirety of you audio is within those 24 (or 16) bits. (or out to your audio interface).

BUT if you have red on your bus 56 and it is going out to the master fader in the red, you audio interface will clip. But, you should hear NO issues if you bring it down before it gets to an output destination. If you hear fuzziness, it has to be something else. (or something is not working properly).

The reason I say, I did tests where I added like 24 audio sinewaves together and then summed them, then trimmed the result down and wrote out to disk and it was perfect (as compared to the source waveform). I was studying the dynamic range AND roundoff error in DP to see what types of artifacts and stuff I would get depending on the range of source material. I was VERY impressed by the quality I was getting out.

[OldTimey]

Yeah, I thought about the aux bus route, but I figured it would already be slamming when it got to the aux bus. I was not aware of the internal headroom thing...Although I can definitely her!e some fuzziness around the edges. I must be at 401db! Thanks for the info

download sonalksis' free meter plugin, FreeG. Unlike DP's meters, which show you when your 32-bit audio stream is out of 24-bit range, FreeG displays all internal headroom. (I don't think that the Trim plug does). If you are clipping the FreeG plug, then, yes you have something to worry about.

[streak]

That is truly amazing.I'll bet your speakers loved that test! I tried the aux bux trim to no audible effect. Now I'm checking out individual plugins to see if one of them is getting warm and fuzzy on me....

[simonr]

Just a question about the conversion to 32 bits .

Does DP simply add 8 bits to my 24 bit file at the MSB end - thereby leaving the room you mention for louder material.

Or does it re-evaluate the amplitude of the sample based on a 32 bit bit depth.

For example:

Incoming 24 bit sample value : 8388608 (decimal - 50 % of maximum)

Does it get converted to 8388608 (decimal - now only around 0.2% of 32 bit maximum) giving allowance for extremely high amplitiudes

or does it get converted to 2147483648 (decimal - 50% of 32 bit max) giving the same maximum amplitude, but with much better granularity.

Just trying to get it straght in my head. ?

[OldTimey]

Just a question about the conversion to 32 bits .

Does DP simply add 8 bits to my 24 bit file at the MSB end - thereby leaving the room you mention for louder material.

Or does it re-evaluate the amplitude of the sample based on a 32 bit bit depth.

For example:

Incoming 24 bit sample value : 8388608 (decimal - 50 % of maximum)

Does it get converted to 8388608 (decimal - now only around 0.2% of 32 bit maximum) giving allowance for extremely high amplitiudes

or does it get converted to 2147483648 (decimal - 50% of 32 bit max) giving the same maximum amplitude, but with much better granularity.

Just trying to get it straght in my head. ?

well it is going from 24-bit fixed point to 32-bit floating point. Grimepoch knows a lot about this stuff. with 32 bit, there is an exponent and a mantissa, im not sure how many bits each is. I imagine the math is a little bit different from either situation you suggested.

[grimepoch]

1) Yes, it is true that once you are in a plugin, you are at the mercy of the plugin as far as how it handles that data once it gets it. Some plugins, like Ozone3 are 64-bit internally, yet the data is handed in and out at 32-bit. It is possible one of the plugs you are using is just not handling the data as well.

2) AFAIK, DP leaves your 24-bit value exactly as it is, as it becomes that mantissa of the 32-bit value. Then, the exponent portion is just tacked on and now allows you to move up and down without sacrificing ANY of the detail of the mantissa. (So if you lower the volume, just the exponent changes and your detail remains!

3) Other systems like TDM from ProTools chose to do 48-bit Integer representation which has a much smaller dynamic range, but allows for more detail depth per sample. There is a lot of debate about what you gain from each method, which I will not go into here. Essentially, the bit depth gain is down in the noise floor of the ADC and DAC conversion. But, this is not my specialty so I do not profess to know one way or the other, I have listened to both and think they both sound good

[OldTimey]

3) Other systems like TDM from ProTools chose to do 48-bit Integer representation which has a much smaller dynamic range, but allows for more detail depth per sample. There is a lot of debate about what you gain from each method, which I will not go into here. Essentially, the bit depth gain is down in the noise floor of the ADC and DAC conversion. But, this is not my specialty so I do not profess to know one way or the other, I have listened to both and think they both sound good

my understanding is that double precision (48-bit) can cause more of a performance hit, but there is less room for programming error to muck things up. In 32-bit floating point, as long as the program developers don't cut corners with their code, it can sound very good too.

[grimepoch]

I believe running internally with the floating point unit of a processor that yes, I think there is a performance hit when using large integers. HOWEVER, it has been a long time since I looked closely at the FPU stats of anything recent. Since I see most high speed stuff use Floating Point, I am making an assumption that the FPU (which is Floating Point Unit) still maximize on silicon intended for FP only. With the TDM system, they have external hardware to deal with the integers and it is really NO hit on the CPU. I have seen G4s doing AMAZING things on a TDM system with no lag. BUT, it comes at the price of the hardware of course.

Libraries are written so well these days, programming really is a no-brainer. When dealing with math, I think using floating point makes less mistakes because most high level math at any sort of realistic use typically works in floating point. For instance, a cup of water = 1, so a half cup of water = 0.5. Yet, working with integers. Do you call a cup of water 10 and a half cup 5? But what happens when now you want 1/2 of a cup? Yes, crude example, but just illustrates how integer math can be more complicated.

I think DAWs use Floating Point numbers because they are the fastest to be processed by modern CPUs. If integers were faster, I bet all the DAWs would have used them for the performance gain!

[OldTimey]

thanks for sharing your knowledge on this stuff, gp.

[grimepoch]

np A lot of this is a combination of talking with people here, researching LOTS of articles on the web, and some of the DSP stuff I did in school (I have a masters in EE).

Many of the white papers and write ups from MOTU, Digidesign, Izotope, AudioEase, and probably ten or more industry professionals and such is where I pulled a lot of this together from as well.

Subject to change