|
|
#46
05-08-2010, 08:22 PM
|
||||||
|
||||||
Quote:
Quote:
Quote:
I will route Ch6 directly to the master bus. I will route Ch5 to Ch3 (Aux) with Voxengo Audio Delay adding 50 ms latency. Ch3 is routed to the master bus so the combined total delay on Ch5 is now 100 ms, the same as Ch6. I will route Ch4 to Ch2 (Aux) with Voxengo Audio Delay adding 80 ms latency. Ch2 is routed to the master bus so the combined total delay on Ch4 is now 100 ms, the same as Ch6. Everything else gets routed to Ch1 (Aux), my submix channel which (you guessed it) adds 100 ms latency and is then routed to the master bus. This way every channel has exactly the same amount of latency so everything triggers in time (though delayed by 100 ms). Quote:
Go back to the earlier page where I set up a side chain so that the magnitude of the piano controls how much the compressor 'works' on the guitar. I did this in FL Studio and unfortunately explained it more in terms that were particular to FL Studio but the important bit in that example is that the piano was fed into an auxiliary channel and this channel goes into the auxiliary input of the compressor. The compressor's peak detector then uses the level of the piano to determine gain reduction on the guitar instead of the magnitude of the guitar signal determining gain reduction on the guitar. I hope this makes sense. Its one of those things that is difficult to explain in words but very easy to explain when you see it being done. Youtube may help in this regard since you can search for a 'how to sidechain' video in the DAW that you use. Quote:
For things like side chain compression. Again not strictly necessary but how else are you going to do it? You can manually ride the volume fader so that when the piano hits you slam down the volume fader on the guitar to compensate. This of course is not very accurate and not very elegant but it achieves the same thing (but not as accurately or elegantly of course). For CPU problems? Again not strictly necessary. You can put a reverb plugin on every channel that needs reverb but if its the same reverb with different amounts of wet/dry signal why add 5 reverb inserts and crash your computer when you can have 1 reverb send and use the cpu headroom on other things? Quote:
http://www.mediafire.com/?zdzkfyh0wy0 This post is big enough so I'll explain the trickery in the next post. 
|
|
#47
05-10-2010, 03:46 AM
|
|||
|
|||
|
Thanks for the link to your sample - sounded very cool. It has quite a bit of depth, and like how the melody comes in so subtley. I can hear effects, of course - but wouldn't know which ones would be done during the playing and which ones during the mixing. I really love the sound of it near the middle - has a great feeling to the sound.
I see what you mean about fixing the latency now. Will admit, reading it makes more sense than if I were to try something like that - it is logical reading how you set it up, but to come up with it myself would be something else altogether... Thanks for the youtube side-chain tip. My version of Magix samplitude doesn't have a side-chain compressor plug-in I now know, but saw how to set it up using a different version. I'm wondering if a free plug-in (saw one or two on-line) would work - or since this version doesn't offer one, it won't be able to handle that type of plug-in. I'll have to just down-load one and see. And I see how it's useful to do things like you say - aux and side-chain. Again, the side-chain compressor sounds very useful and in your example you did for me, I really like the subtle sound of the two instruments "interacting" in that way. And using the volume manually in my program is, as you say, not very accurate I've found out - also takes an awful lot of time, and I just end up pressing "un-do" anyway... Thanks also for the drum idea - grouping for certain things. A friend of mine is putting together some drums for a song of mine, and I'm looking foreward to mixing each drum part, and seeing how it will then all come together. I will keep that tip in mind as well. Thanks agian - slowly some things are making sense. It's really nice being able to ask some questions and get such good information. I am also reading a nice book by Roey Izhaki - I don't understand a lot of it, but am slowly learning what all the effects actually are and what the do (and why people use them, and how and why they were developed in the first place).
|
|
#48
05-10-2010, 09:27 AM
|
|||
|
|||
|
Yes I forgot to mention that you need to use a compressor (or any signal processor or generator) that has an auxiliary input. Some dont but there are other ways to achieve the same effect.
A compressor just detects the gain of whatever signal is running through it and if it goes over a certain threshold it automatically pulls back (suppresses) the gain. Its just a more precise way of riding the channel volume fader and waiting for the sound to get really loud and then you pull down the fader to compensate. But of course why do that manually and not very precisely when you can have a compressor do that for you? A side chain compressor does the same thing except you have 2 sounds on 2 different channels. When 1 sound gets loud you want to automatically lower (suppress) the gain of the other channel.
|
|
#49
05-10-2010, 10:47 AM
|
|||
|
|||
|
Well, I will try a little later to see if I can get a plug-in to work. I used to have a very nice (and much more expensive version of the program), but going to a newer computer it wouldn't work anymore (things were on floppy - new computers don't have them...). I'd still be using it if it worked - probably could have done things like that (for the price I payed, anyway).
Yes, I see how the side-chain would be a whole lot easier and more accurate - and you have the interaction between the different channels. Between you and the books I'm reading, I'm finally getting the idea of how compression works. I don't kow - it's kind of hard to trust my ears with things like that. I can hear when it starts sounding "artificial", but never know quite why - is it threshhold, ratio, attack, release (or just a poor compressor in my program)? Do you generally set the threshhold quite low when you do things - so almost everything is treated? (read you get a more natural sound that way). Is it also an effect that you will get an overall increase in volume (but may lower the peaks)? I thought it should bring down all the volume, but seems like I always get (at least a little - sometimes a lot) increase in overall volume no matter how I set the threshhold and ratio. Is there a better way to lower peaks other than compression, if that's the main problem (like singers who really go from low to high in volume)? For most things (guitar, keyboard, even percussion), seems like eq and just volume helps, but for singing I started using compression and not sure it sounds very natural most times.
|
|
#50
05-10-2010, 01:18 PM
|
|||
|
|||
|
How you set the threshold and ratio depends entirely on the amplitude of the signal that you want to compress. Again, its easier to visualise it so heres an example:
 This is an excerpt of the waveform view of an acoustic guitar recording I made. Before you get into compression you have to be able to understand what this picture is telling you so lets break it down. The Y axis is measured in decibels and it tells you the amplitude of the wave. Note that it is 0 dBFS (fullscale) at the extremes and -infinity in the centre. This is because the recording of my guitar came from a mic which produced an analog AC (Alternating Current) signal so it has electrical polarity. Polarity means that the quantity has a steady state at rest around which it fluctuates between 2 extremes (poles). Why is this picture a representation of an AC signal? Because that AC signal is an analogue of the sound pressure waves produced by my guitar which made a membrane in my mic vibrate which moved an electromagnet in and out of a coil which induced a current that moved and changed with the sound pressure wave. Sound consists of alternating compression and rarefaction of equilibrium air pressure so hopefully you can see how the electrical signal is analogous to the sound wave. The X Axis in the picture is time. This can be measured in milliseconds/seconds or in this case samples because this waveform is actually a digital recording of an electrical signal which is an analogue of a sound wave. I recorded at 44.1khz so thats 44,100 samples in 1 second of audio. Yes thats alot of samples. You can tell the frequency of the sound by looking at very roughly how many 'compressions and rarefactions' there are in 1 second of audio. You wont be able to do this accurately so its only good for telling whats high frequency and whats low frequency and nothing else. An FFT (Fast Fourier Transform) is what we use to properly 'see' and measure frequency. Either way, the frequency is not important in this example. Later on you can exploit the sidechain input of a compressor with an EQ or filter to get frequency dependant compression but ignore that for now. Fire up a compressor and take note of these key variables: 1) Threshold 2) Ratio Now if you set the threshold to -12dBFS that means that the compressor starts working if the signal amplitude reaches this level or exceeds it. See the picture below:  The red regions represent everything above -12dB so this is the region that is going to get compressed. Everything below the threshold is left alone. How much the red region gets compressed depends on the ratio. So if you set the ratio to 3:1 that means that for every 3dB over the threshold the input signal is, the output will be reduced to 1dB over the threshold. Go to the tallest point on the Left channel. Its dead on -6dB so this part of the signal is 6dB over the -12dB threshold. The ratio is 3:1 so after compression this peak will be reduced from 6B over threshold to 2B over threshold (6/3 = 2). Therefore, after compression the new peak level is -10dB (which represents peak gain reduction of 4dB). Everything over the threshold gets continuously squashed depending on how much over the threshold it is. Now this example holds true for a hard knee compressor with 0ms attack and 0ms release. That is, the onset of gain reduction occurs instantly when the input signal rises above threshold and ceases instantly when it falls below threshold again. Soft knee compressors begin gain reduction before the input signal reaches threshold (about 10dB before threshold) and reaches peak gain reduction when the threshold is reached. This results in more gradual gain reduction and less artifacts from very extreme compressor settings (very low threshold and very high ratios). Once you get the hang of that I'll explain envelope attack and release and then makeup gain. Remember: the action of a compressor is entirely dependant on the amplitude of the signal going into it. If you use the settings in this example (-12dB threshold, 3:1 ratio) and the input signal does not rise above -12dB then the compressor will do nothing! This is why you should never use compressor presets. You should always rethink how much compression you need based on each individual input. It sometimes helps to visualise it but after a while you can simply follow a peak meter and do the arithmetic in your head. Some people do it by ear but if you do that then you either indulge minutiae (fiddle with a compressor and have doubts its actually doing anything because the effect is so small) or you compress enough that it becomes audible. Whether you want to hear gain reduction happening depends on what you want to do. For example, Benny Benassi almost built a style of dance music around a bassline that 'pumps' with gain reduction applying and releasing. Sometimes the effect is desirable, other times not. Last edited by Pokiehat; 05-10-2010 at 01:31 PM.
|
|
#51
05-11-2010, 11:04 AM
|
|||
|
|||
|
Wow - thanks again for the in depth explanation (and pictures!). I understand (well, the extreme basics) of what you mean about how the sound is made, but never quite learned the difference between AC and DC, or polarity. I've have broken things using AC adaptors in DC equiptment, though...
But the vibrations of sound causing different pressure (which vibrate the membrane and so) I think I follow - like old telephones? Why did you choose to use -12 as the threshhold, actually? Is that just preference, or experience, or something you get from looking at the waves? So if you set it to 3:1, I see you get a 2db reduction (following along with your explanation). But none of it will ever go below the threshhold, or does it? What happens to something at (say) -9 in the original? -9 would be 3db over threshhold, so would it be lowered by 1db - which would put it on -10? Slightly confusing - but you devide the difference between the threshhold and peak (in question) by the ratio? Your 6/3=2 example: the 6 you are using is the difference between 6db and 12db, not the 6db peak? So I don't divide the -9db by 3, but the difference: 12-9=3, so it would be 3/3=1? I'm probably not saying it clearly, but I think I see how it works. The ratios in my program go between .5 to 5. This is the first number of the ratio I assume? I should see it as 1/2:1 going up to 5:1? So the lower the first number, the more effect you'll get? If you set your ratio to 1/2:1, you'd have a reduction of 12 on the -6 peak? This make take a little bit to start being logical - I admit it took quite a bit of thinking to get it a little straight in my head. But again, thanks a lot for all the time and the great explanations.
|
|
#52
05-11-2010, 01:12 PM
|
|||||
|
|||||
|
Quote:
Quote:
Quote:
* We know the peak level is -6dB which is 3dB above the threshold. * For every 3dB the input is above the threshold, the output is suppressed to 1dB over the threshold. Since we are 3dB over the output is suppressed to 1dB over. * Therefore the new peak level is the threshold (-9dB) + 1 dB = -8dB. This is 2dB of peak gain reduction which is half the amount of gain reduction as the previous example. ---- So heres the next bit. What do you compress and when? Theres no straight forward answer but if you are just compressing one sound in isolation but the most obvious use of compression in this example is for sound shaping. Basically, you are softening transients and making the volume of the entire passage more homogenous (less disparity between loud and quiet). It is common to compress a drum kit for instance where the drummer might hit the snare very hard a couple of times and the loudness is jarring. The traditional use for a compressor is in radio broadcasting where the person with the mic is prone to move around and raise their voice. In order to stop very sudden increases in volume which would annoy listeners, the compressor was devised to suppress very sudden and unexpected increases in volume. Now theres also the compressor's envelope to consider. The envelope has an attack and release phase. attack = the time it takes for the compressor to apply gain reduction after the input signal passes the threshold. release = the time it takes for the compressor to cease gain reduction after the input signal dips below the threshold again. Remember that with a hard knee compressor and 0 ms attack and release, gain reduction is instantaneous so it squashes transients (large sudden spikes in amplitude). If you have 50 ms attack then the compressor detects a peak over threshold and 50 ms later, the gain reduction occurs so high attack values mean you let transients through and then the gain reduction applies which makes the initial transient seem even louder. This is useful if you want to make sounds snappier, especially ones which are supposed to be snappy but become indistinct and muffled in a dense mix. Long release times mean that the compressor holds gain reduction for longer. This is useful if you want to avoid 'pumping', the audible effect of gain reduction being suddenly applied and released. Long release times are useful to 'glue' a mix together. You will set the threshold depending on what you want to do. So look at the left channel again, the peak part of the signal is half way on that picture and its -6dB. You may want to keep that snappy plucking part of the sound so you set the compressors attack to 50ms. So on the X axis measure along 50 ms from that peak and the tallest bit of the signal at that point is the minimum threshold value that the compressor has to have in order to have any effect. Set it much lower than that for more drastic gain reduction but be aware of 'pumping'. You can get peak volume information from meters so you don't always need to have a visual representation like in soundforge although it helps to see how it works. Now for some homework. If you have a program that can give you a visual representation like the above (Soundforge, Audition, Audacity etc) I want you to load it up and import a brief sound clip. Now I want you to find the peak volume of that sound file and I want you to set a threshold below this value and a ratio of 2:1. Now I want you to calculate what the new peak level is going to be when the compressor is hard knee, 0ms attack, 0ms release, 2:1 ratio and whatever threshold you choose to set. Then plug those numbers into the compressor and ok it. If your math is not terrible your calculation will be correct. After a while you won't need a visual representation or to read off X/Y axis values and you can do it solely on peak level meters. Furthermore, you won't need to rely on doing the math as much since you begin to strongly correlate degrees of gain reduction relative to peak just on the sound of it so it does become very intuitive. You may go through a phase of compressing everything to a very large degree and then you will naturally ease off. It helps to explore the extremes and try out massive gain reduction so you know what it looks and sounds like. This way you know whether that much gain reduction is necessary or whether its going to far for the sound you are trying to get. Hope this helps. Quote:
Also note that 1:1 ratio means that for ever 1dB over the threshold the input is, the output is suppressed to 1dB over the threshold. Yeah, you guessed it, 1:1 ratio gives zero gain reduction. You can set the threshold to anything you want and the compressor will do nothing. Quote:
For every 0.5dB over the threshold the input is, the output is 1dB over the threshold. A compressor with a ratio less than 1:1 has the opposite effect in the sense that you aren't reducing dynamic range anymore. You are increasing it. So go back to the example I made with -12dB threshold only instead of 3:1 ratio its now 0.5:1. The peak signal is -6dB so the peak level is 6B above threshold. For every 0.5dB in we get 1dB out above the threshold of course. Break it down: -12dB = threshold -11.5 = +1dB -11 = +1 dB -10.5 = +1dB -10 = +1 dB -9.5 = +1 dB -9 = +1 dB -8.5 +1 dB -8 = +1 dB -7.5 = +1 dB -7 = +1 dB -6.5 = +1 dB - 6 = peak input and another +1 dB Add all that up and we get +12dB over threshold. So for a ratio of 0.5:1 our new peak signal is no longer -6dB. Its 0dB. Last edited by Pokiehat; 05-11-2010 at 01:27 PM.
|
|
#53
05-12-2010, 03:01 AM
|
|||
|
|||
|
I'm going to have to read this again more - I see I'm not getting the hang of the ratio. But will try to go through it again today and tonight - so will reply a little later to the new (and old) stuff you explain.
Really - thanks a lot, it's more confusing than I thought, but would really like to try and understand this!
|
|
#54
05-12-2010, 03:21 AM
|
|||
|
|||
|
It is not confusing. Its actually very very simple and the math doesn't get more complicated than what a child would learn in primary school. If it seems confusing its because I'm not explaining it properly so I must apologize for that (I'm not a qualified teacher). Let me try to illustrate it in simpler terms:
3:1 ratio means that if input volume is 3dB over the threshold then the output volume is suppressed to 1dB over the threshold. So: If input is 3dB over, output is suppressed to 1dB over. If input is 6dB over, output is suppressed to 2dB over. If input is 9dB over, output is suppressed to 3dB over. If input is 10dB over, output is suppressed to 3.3333333333333 dB over. If input is 12dB over, output is suppressed to 4dB over. Practical examples: If: Threshold = -3dB Peak input signal = 0dB Ratio = 3:1 Then: Peak input is 3dB above threshold which is suppressed to 1dB above threshold. Threshold = -3dB then peak output signal = -3dB + 1dB = -2dB --- If: Threshold = -6dB Peak input signal = 0dB Ratio = 3:1 Then: Peak input is 6dB above threshold which is suppressed to 2dB above threshold. Threshold = -6db so the peak output signal is -6dB + 2dB = -4dB --- If: Threshold = -12dB Peak input signal = 0dB Ratio = 3:1 Then: Peak input is 12dB above threshold which is suppressed to 4dB above threshold. Threshold = -12db so the peak output signal is -12dB + 2dB = -8dB Last edited by Pokiehat; 05-12-2010 at 03:33 AM.
|
|
#55
05-12-2010, 11:49 AM
|
|||
|
|||
|
No, I think you are explaining it fine - I think I just have the problem! I'm usually not bad at math, but this is a little confusing.
In your original explantaion - yes, I see now you get a -4db change. I think I'm slightly confused because the numbers go backwards (or something...). I see now it goes from -6db to -10db, which is 4 (but misunderstood and went from -12db to -10db, thinking the difference was 2). Which it is, of course, but the reduction from -6 to -10 is 4, I think I understand now. It's also tricky to remember that it is supressed by "x" over the threshhold, not reduced by that amount (if I understand right). I'm going to print out that last post of yours - I understand it when looking at it, but know it will take time to sink in to my head. Seeing it written out like that is very helpful - thanks! I still have to read your attack/release more before I can see if I'm getting it - sorry to be a bit slow about this - I think I understand the concepts (what attack/release actually do), but I have to read that post a few more times and try out the things you say. I wil try and work it out and see if I got a bit more of it in a day or two. Again, thanks!
|
|
#56
05-12-2010, 12:48 PM
|
|||
|
|||
|
Quote:
Once it clicks and you get that eureka moment you won't need to think about it that much. You just kind of do it and you can roughly guesstimate how much gain reduction you are going to get at a glance just by seeing how much signal is over threshold.
|
