|
Maximising signal to noise ratio?
|
View this Thread in Original format
| paulc_dj |
This may sound like a stoopid question to you guys, but I am really stuck on this:
I seem to have a problem when mastering, in that what I hear in Fruity when playing back and what I hear in the rendered file are two different things, volume wise (The rendered file is lower in volume). How do I set the master channel up so that what I hear is what I get, so to speak? I am using L2 Ultramaximiser, LinEQ Broadband and Fruity Multiband Compressor on the master channel BTW if you need to know.
Also when I open it up in Soundforge, it sits between the two middle lines on each channel, so I have to do like a 200% inc in volume to get it loud enough and then I get very slight distortion. I look at the Fruity DB meter I have on the master channel and it shows me that its continually registering 0db on each beat and I thought that you need to master to 0db, or is it meant to be more than that?
Help!
PC :conf: |
|
|
| Eldritch |
| You should master to -0.3 dB. Never over 0.0 dB, because that will cause clipping which I think is the reason for the low volume of the rendered output. (Fruity maybe tries to attenuate the signal or something?) |
|
|
| cybernetica |
I am not sure, but maybe the problem is that some program is using the ASIO drivers of your soundcard and the other one is not. Try to set all your programs to use the same audio out driver. If I produce in FL for example, the audio output is much louder than in any other program without ASIO drivers on.
Alternatively, to solve your problem with soundforge, you could just do a simple normalization on your audio file. |
|
|
| paulc_dj |
Cheers for the prompt replies! Just out of interest what does the phrase "maximising signal to noise ratio" actually mean in practice?
PC :tongue2
Edit: I have just checked in Soundforge and the signal only comes up to -6.0db (its in between that on the lines). How do I set that vol higher. I don't recall setting it to -6.0db anywhere. |
|
|
| /I\ |
from what i can gather, maximising signal to noise ratio is the old analogue realtime cousin of normalization.
With normalization your app will find the loudest peak of your sample and adjust the complete sample so the loudest peak is at zero making the whole sample at its loudest without digital (or even analogue) cliping.
Same applies to signal to noise ratio when mastering, meaning with each stage of your mastering process you want the inputs and outputs of each plugin to be as close to zero as possible without overloading or you will get digital clipping. Conversley in the old days of analogue a week signal would introduce hiss and other unwanted staff into your audio signal just like recording onto a tape deck with a weak signal
Just guessing here with your prob, maybe your eq plug is at a lower output and sending a weak signal to another plug in your mastering chain |
|
|
| Derivative |
Signal to Noise ratio is the 'absolute' dynamic range you can record it using your audio interface.
If you are recording using microphones:
1) Ambient noise will affect SNR. For instance, you want to record yourself singing a vocal hook but your PC is closeby and it is generating 35dB of fan noise. You have just reduced your SNR if that is captured in the recording. Because it will limit the amount of pre and post gain you can apply without hearing your PC whirring in the background.
2) To get the highest SNR possible you ideally want to record in *absolute* phucking silence, in an acoustically treated room that doesnt reflect or artificially amplify sound by way of its surfaces. You also want your PC or any other source of ambient noise source in another room if possible or far enough away that it isnt raising the ambient noise floor on your recording.
If you are recording straight into the inputs on your audio interface, then you can get the most dynamic range by:
1) Using balanced cable runs if applicable. You need equipment that can take balanced ins/outs though. Balancing reduces signal attenuation across the length of cable, eliminates cross talk where cables overlap or cross at right angles. It also eliminates ground loops which are the no.1 cause of audible buzzing in recordings. It also provides another 4 dB of headroom or thereabouts so you can run the input signal hotter without clipping.
All purpose:
1) Dont normalise. At the very least try to limit how much you do this. Normalising scales up the noisefloor in proportion with the peak signal. It will make all ambient noise on a recording more perceptible. |
|
|
| MrPit |
| quote: | Originally posted by Derivative
Signal to Noise ratio is the 'absolute' dynamic range you can record it using your audio interface.
If you are recording using microphones:
1) Ambient noise will affect SNR. For instance, you want to record yourself singing a vocal hook but your PC is closeby and it is generating 35dB of fan noise. You have just reduced your SNR if that is captured in the recording. Because it will limit the amount of pre and post gain you can apply without hearing your PC whirring in the background.
2) To get the highest SNR possible you ideally want to record in *absolute* phucking silence, in an acoustically treated room that doesnt reflect or artificially amplify sound by way of its surfaces. You also want your PC or any other source of ambient noise source in another room if possible or far enough away that it isnt raising the ambient noise floor on your recording.
If you are recording straight into the inputs on your audio interface, then you can get the most dynamic range by:
1) Using balanced cable runs if applicable. You need equipment that can take balanced ins/outs though. Balancing reduces signal attenuation across the length of cable, eliminates cross talk where cables overlap or cross at right angles. It also eliminates ground loops which are the no.1 cause of audible buzzing in recordings. It also provides another 4 dB of headroom or thereabouts so you can run the input signal hotter without clipping.
All purpose:
1) Dont normalise. At the very least try to limit how much you do this. Normalising scales up the noisefloor in proportion with the peak signal. It will make all ambient noise on a recording more perceptible. |
word! |
|
|
| /I\ |
| cheers, Im homer. really :stongue: |
|
|
| Derivative |
Oh yea I should add that it isnt always desireable to absolutely kill every source of ambient noise in a room or obsess over the whole issue. Sometimes having a bit of life going on in the background is what you want. Its one of those things you need to weigh up depending on what you are recording and what you want the end result to sound like.
Totally 100% squeaky clean isnt something I try to shoot for because I like a bit of dirt in the mix. Theres loads of us that share a similar opinion. |
|
|
| Atlantis-AR |
| quote: | | Originally posted by paulc_dj This may sound like a stoopid question to you guys, but I am really stuck on this: |
I know this topic is quite old, but there are a few things I still want to say in reply to your original post. I hope you can learn from them, because they're quite important issues you seem to have overlooked...
| quote: | | I seem to have a problem when mastering, in that what I hear in Fruity when playing back and what I hear in the rendered file are two different things, volume wise (The rendered file is lower in volume). |
I must say I have no idea how this is possible at all. Even using different sound drivers shouldn't generate a different output (I should hope not anyway). What matters is FL Studio's internal process, which, yes, is slightly different each time (randomness), but shouldn't generate a difference of 6.0 dB, so it might be worth finding out what you're doing wrong.
| quote: | | How do I set the master channel up so that what I hear is what I get, so to speak? I am using L2 Ultramaximiser, LinEQ Broadband and Fruity Multiband Compressor on the master channel BTW if you need to know. |
Are you using the effects in that order? You'll definitely want to rearrange them so the L2 comes last. A limiter (or, more correctly, the dither noise, which is included in the L2 anyway) should come at the end of the mastering chain, as it sets the amount of limiting and the final output level.
| quote: | | Also when I open it up in Soundforge, it sits between the two middle lines on each channel |
This could be because you're limiting the sound first, possibly by dragging both the threshold and ceiling down together, which could result in a level around -6.0 dB. Place the limiter at the end of the chain, and set the output ceiling to e.g. -0.2 dB. Either that, or you're not limiting enough in the first place.
| quote: | | I look at the Fruity DB meter I have on the master channel and it shows me that its continually registering 0db on each beat and I thought that you need to master to 0db, or is it meant to be more than that? |
It sounds to me like you're mixing wrong. First, the mix level shouldn't be registering over -3.0 dB, as you need some room to work while mastering the mix later (plus, you don't want to clip your mixdown). Or, do you mean the level hits 0.0 dB after mastering? In that case, maybe ignore the next part, although you'll still want to set the output ceiling on the L2 to e.g. -0.2 dB.
If you're using FL Studio, keep everything at default, as it was designed that way. If you load an initally 0.0 dB normalised sample onto the step sequencer, and route it to a mixer channel set to default volume, it will come out peaking at -8.2 dB (not -8.0 dB, because the sample volume defaults to 78% rather than 80%, i.e. 2.2 dB less than -6.0 dB, or half volume, but this doesn't really matter).
If you start out with your kick (which is usually the loudest instrument) peaking around -8.0 dB, you will have 5.0 dB of additional gain to work on for the rest of the instruments, which is usually enough given you've balanced, EQ'd, and compressed everything right. Your mix should now come out peaking no higher than -3.0 dB at its maximum or minimum point (top or bottom, of either channel), which should give you enough room to work on while mastering.
Going a little off topic maybe, but it sounds like you're not working the most efficient way, and you may be clipping things along the way.
Also, about normalisation, not only does it scale up the noisefloor, but it also introduces quantisation distortion if not done right. If you're mastering in your sequencer, place a limiter at the end of the mastering chain (NOT while mixing your track), and set the dither to 16 bit before saving the file in 16 bits (and bypassing the sequencer's dither) as well.
Sorry, got a little carried away, but I hope you can learn something from this. |
|
|
| Derivative |
Fruity's dB meters are kind of buggy and it has been established that 0dB on Fruity's mixer is not actually 0dB but more like -6dB.
I regularly end up running mixer channels into the red in Fruity with no audible signs of clipping and it is disconcerting as hell because I lose track of what levels I am running many channels at.
I wouldn't use Waves L2. If you need a limiter, its probably a better idea to use a compressor with a compression ratio of 30:1 and lower the threshold until the clip disappears. You shouldn't be clipping on the master bus anyway so its debateable whether you should need a limiter at all if you mixed everything down right. Adjust attack/release and knee as desired if you must use it. I have no idea why you have a limiter and a compressor on the master bus though.
L2 to my mind sounds and it has several dithering and dither noise shaping tools bundled in with it. I have no idea why Waves did this. Turn them all off. Its probably best not to touch anything dither related until you know what you are actually doing to the audio by dithering. I never both with dither of any kind except for sound testing purposes. I never actually dither any of my tracks.
If you must dither it has to be the last process in the signal chain. Waves L2 doesnt apply dither in realtime, it only works when you sample/render down to a lower bitrate. So if you have some crazy dithering setup going, the rendered result will be different to what is playing in Fruity (the render will be dithered. Fruity's playback will be undithered until you render it). |
|
|
| Atlantis-AR |
| quote: | | Originally posted by Derivative Fruity's dB meters are kind of buggy and it has been established that 0dB on Fruity's mixer is not actually 0dB but more like -6dB. |
That's strange. I have noticed weird things going on sometimes, but then I realised I had a different channel selected, or an insert channel rather than the master channel. That's talking about the meter included on the mixer, anyway.
| quote: | | I wouldn't use Waves L2. If you need a limiter, its probably a better idea to use a compressor with a compression ratio of 30:1 and lower the threshold until the clip disappears. You shouldn't be clipping on the master bus anyway so its debateable whether you should need a limiter at all if you mixed everything down right. Adjust attack/release and knee as desired if you must use it. I have no idea why you have a limiter and a compressor on the master bus though. |
I agree, but I think what paulc_dj was saying was that he was mastering in FL Studio, and I may have confused that a bit. I wouldn't use a limiter, let alone a compressor, on the master bus at all. If you set your levels up correctly, there shouldn't be a need anyway.
| quote: | | L2 to my mind sounds and it has several dithering and dither noise shaping tools bundled in with it. I have no idea why Waves did this. Turn them all off. Its probably best not to touch anything dither related until you know what you are actually doing to the audio by dithering. I never both with dither of any kind except for sound testing purposes. I never actually dither any of my tracks. |
I've sometimes wondered that too, but I guess they thought people would more readily want to dither to 24 bits rather than keep the higher bit depth process. I believe 24 bits is all the TDM bus supports anyway, so that might be why. If you have to limit (i.e. dither) within the mix, just use 24 bits, although this might be subjective. The other thing of course is that limiting is more intended for mastering, where you often do want to dither down to 16 bit.
| quote: | | If you must dither it has to be the last process in the signal chain. Waves L2 doesnt apply dither in realtime, it only works when you sample/render down to a lower bitrate. So if you have some crazy dithering setup going, the rendered result will be different to what is playing in Fruity (the render will be dithered. Fruity's playback will be undithered until you render it). |
I've found otherwise in Sound Forge. Selecting 16 bit dither adds dither noise to the signal, making the LSB's all 0's that will in turn be truncated off when you save the file in 16 bits. I would say it works the same in a sequencer, or at least I don't see why not. |
|
|
|
|
