Pre- or post-fader?
I use Cubase SX 3. I have a question regarding where one would use an effect like Ozone in the stereo bus chain. Should it be used pre-fader or post-fader. In Cubase there are eight slots on the stereo output bus. Six of the eight are pre-fader and the last two are post-fader. Also, now that I think about it, overall which effects should one use pre- and post-fader?
Ozone is a suite of mastering plug-ins, and, if you want to do things the conventional way, you would save the mastering until later and not have any mastering type plug-ins on your stereo mix buss.
Usually you master an entire album at once, trying to balance out the apparent loudness and overall EQ curve of each song mix with respect to each other so that the listener doesn’t feel like they have to adjust the volume or EQ controls from one song to the next. Unfortunately, these days, we have to contend with the loudness wars, and everyone trying to make the apparent loudness of their mixes as loud, or louder, than everyone else’s. I personally hate the whole loudness wars thing (if you are one of the few that haven’t heard of this, just Google the term “loudness wars”), but, it’s become a necessary evil that many clients simply insist upon no matter how much you argue against this.
If you wish to make your songs as loud as possible, that should be saved for a separate mastering stage, and you should NOT be trying to achieve those levels of loudness while mixing. Unfortunately, this whole loudness wars things makes it hard to use modern CDs as reference checks while you are mixing since they have already been mastered loud and bright, and will always sound MUCH louder and brighter than your mixes unless you are putting massive amounts of peak limiting on your stereo bus. One solution is to have a peak limiter / loudness maximizer plug-in on your stereo buss that you ONLY turn on when you want to compare to modern CD mixes to see if you are in the same ballpark. Another solution is to simply attenuate the output of the CD player to match the current levels of your mix without using a loudness maximizer on your mix.
Anyway, the point is, you really should NOT be using Ozone type plug-ins while you are mixing. You want a nice clean and full dynamic range mix file to work with that you can mangle to death later if you want, while still having the original untouched mix in case you want a professional to master it later, or if the world comes to it senses at some point and starts demanding that we put some dynamic range back into music.
If you insist on using some plug-ins while you are mixing, then you really should only be using some fairly gentle compression to kind of “glue” the mix together a bit, followed by possibly some EQ to correct for what the compression does. I don’t mix in the box, but I still mix digitally through an external digital mixer (feeding up to 56 channels digitally from the computer to the board), and so I’m not using any plug-ins at all on the master stereo bus. Since I have some really nice high-end analog stereo compressors, sometimes I’ll run out through high-end D/A converters, through the compressors, and then back in through the master quality A/D converters, to get some of that real analog compression and flavor. BUT, I do it very sparingly, AND I also simultaneously print a direct digital copy from the board that hasn’t gone through the conversions and compressor. That way, if I later decide I want a different kind of compression, or want a pro mastering engineer to do the mastering, I still have an un-touched master copy of the mix.
But, to answer your original question, if I was going to put some compression and EQ and other mastering type effects on the master stereo bus, I would want those to all go PRE-FADER so that the compressor is NOT affected by any master fader movements (such as during a fadeout at the end, where you still want the compressor to act the same, but you just want to fade out the overall volume).
However, as always, you need to watch your gain staging along the way, and be aware of the internal math inside of your plug-ins. You should always make sure you aren’t clipping the master buss during your mix, even before you put any plug-ins on it. Even though the mix engine of Cubase is floating point and you can’t cause digital distortion within it by going over zero, your D/A converters on your soundcard are still fixed point and they WILL clip and cause nasty digital distortion. Plus, if any of the plug-ins you are using are fixed point, then you’ll be clipping them internally and it won’t matter that Cubase is floating point… once the clipping occurs, the damage is done. You need to save headroom for mastering anyway and know that you can make it as loud as you want at that stage, so there is no point in trying to get the levels up to digital zero while mixing. To maximize the sound quality of your converters without overdriving their analog stages, Your average RMS levels while recording or mixing should really be around -18dBFS, with the highest peaks not hitting above -6dBFS. If you are doing everything at 24 bits, then you have PLENTY of dynamic range to work with and no need to drive things as hot as possible. You’ll find that most gear sounds MUCH better when you are NOT pegging the digital meters. Plus, if you keep those general levels in mind for each track, and the stereo bus, while recording and mixing, you’ll have room later in case you decide the snare or some other sound needs to be turned up a bit overall. If you had already maxed out the levels on each channel, then you have no room to turn it up anymore without clipping!
In general, any type of plug-ins that you would normally use as INSERTS on a channel would be PRE fader plug-ins. These are the types of things that are 100% wet type of processing… in other words, you wouldn’t be doing a blend of the original and the effected sound — it would be 100% effected (“wet”). These types of processes are things like EQ, Compression, and any other types of dynamic processing (gates, limiters, expanders, de-essers, transient modification, etc.). You want those pre-fader, again, for the same reason as I discussed above… you don’t want the compressor to change the way it is compressing the audio based on volume changes you make during the mix. You want the compressor to be responding to the same level all the way through… otherwise, as you rode the fader up and down during the mix, you’d have to automate the threshold level and other settings on the compressor, and the compressor would be fighting you against those volume changes anyway, doing it’s job to try to level the volume changes out.
Traditionally, time & delay based effects, such as reverb, delay/echo, chorus, flange, etc., are NOT used as inserts. You would normally set those up as send/return type of effects. In programs like Cubase, you add an FX channel, into which you “insert” the effects you want, then you use the effects sends on the regular audio or instrument tracks/channels, to buss parts of the signal to those effects. You are blending the amount of the reverb or delay or other effect in with the original sound through the use of the send levels and the FX channel’s return levels (or
fx channel volume fader in Cubase). This type of use came about because in the old analog studios, you maybe only had one reverb chamber, and maybe a few digital effects that you had to be able to use for as many tracks as needed. So, you set them up as effects sends and returns, and then you could use the effects sends on each channel to determine how much of each channel got sent to a specific reverb or delay or whatever. Thus, you could have a LOT of reverb on one sound, while just a little bit of the same reverb on another sound, and any combination you want for as many tracks as you want.
When you are using time/delay based effects on an FX track and using effects sends on channels to control the amount of each channel going to that effect, you want the effects sends to be set up POST fader (which is the default). That way, if you bring the level of the vocal down, for example, the amount of vocal sent to the reverb (for example) will also decrease proportionally, so that the balance between the dry signal and the reverb stays the same. Same with delays, chorus, flange, or any other time/delay based effect.
However, sometimes you may want to create a special effect where the dry signal goes away, but the reverb (or delay, etc.) remains. In this case, you’d want to change the effect send to be PRE fader, so that the level going to the effect remains constant, no matter what you do to the channel fader. One use of this is to gradually pull down the channel fader so that there is less and less dry signal, while the reverb remains the same, which makes it sound like that audio is slowly fading away into the distance. This type of thing isn’t done very often though. In general, keep your effects sends set at the default POST-FADER setting.
With modern computers and the endless amounts of processing power, and plug-ins that you can have on as many channels as you want, some people put reverbs and delay and other time based effects as INSERTS on channels, and then use the wet/dry controls of the effect to control the blend. Of course, in this business, there is no “right” or “wrong” way to do anything, so you can certainly work that way if you wish. In this case, most of the time you would want those effects to be PRE fader so that they are acting on the same relative level of the original audio, and also so that the level of the reverb/delay will change correctly as you raise or lower the channel fader. The default insert settings on channels is pre-fader anyway, so that’s the way you’d want to keep them most of the time.
Thanks for you answer to the last question. This brings me to a related question.
One of the effects on the Ozone is an spectral EQ analysis unit. Is it helpful to have that plugged in temporarily in the stereo output/prefader bus to ascertain if any of your signal is breaching the 0dB floor or as you state in your previous email at about -18dBFS ? Otherwise, how does one know if their signal is too hot, other than obviouse distortion? If one does not have a spectral analyzer how does one know if certain parts of the mix are too hot, barring obvious distortion?
So, for example, let’s say the spectral analyzer indicates that the bass drum is breaching the 0dB mark or let’s say even -6dBFS, I can then adjust the bass drum track to fall within whatever range is best.
I hope this makes sense. As always I appreciate your patience
The spectrum analyzer is simply going to break up the audio signal into lots of different frequency bands and give you a rough graphical representation of what’s happening with each. It doesn’t specifically show you what the kick drum is doing, for example, although you can kind of figure out the key frequencies of certain instruments and get an idea for how they are impacting your overall frequency curve of your mix. Really, though, a spectrum analyzer is more helpful for figuring out if you’ve got some weird EQ balance problems, such as way too much low/sub frequencies, especially if your monitors, or your room, are not accurately letting you hear what’s happening in certain ranges (usually, low bass and subs are hard to reproduce accurately in a home studio). One way to kind of use a spectrum analyzer is to check out what the curve looks like for similar songs from other artists that you think sound really good, and then check to see if your mix is in the same ballpark as far as the relative curve (but not necessarily the amplitudes of each, since the major releases will have been mastered already and will be much higher).
But, more specifically, regarding your question, if ANYTHING in your stereo mix is clipping, it’s going to show on the stereo bus meters. The stereo bus is the sum of all the channels, and the highest peaks in any frequency range are going to be represented by the stereo bus meters. The only possible exception is if you have inter-sample peaks that exceed 0dB. These won’t be represented by the typical peak meters in DAW software… but, you would really only have the possibility to get those if you are already pushing your levels up near clipping anyway. There is a new free plug-in from SSL that attempts to reconstruct the waveform from the digital data to try to determine if there is any inter-sample clipping happening. Do a search for that if you wish, and the page there also has a good explanation of what inter-sample peaks are all about and how they can cause clipping even if the typical peak meters don’t indicate any clipping. That type of plug-in requires a LOT of CPU power, though, so it’s not something you would typically leave on all the time during a mix.
Regarding individual sounds/tracks, though, the best way to determine if they are clipping is to simply look at the channel meters for each channel. Again, while you are recording those tracks, you should be watching the levels on your A/D converter meters (if you have them), or the input meters in Cubase, to make sure the average levels are somewhere around -18 to -14, with most peaks not hitting above -6 dBFS. That should give you plenty of headroom and keep you in the best sounding range of your converters. Then, when you are mixing, if you are using a lot of plug-ins on each channel that are adding gain (such as compressors where you crank up the output gain, or EQ when you are boosting frequencies), again watch the levels on the meters, or within the plug-in itself, and reduce the input or output gain of the plug-ins as needed to keep each channel’s signal level in a good range. If all your tracks are at the recommended levels with peaks not getting higher than -6 dBFS, then, in most cases, you should be able to leave your stereo master fader right at unity gain (0dB), and not have to worry about clipping… but, you should still watch that meter as well.
Another way to keep your levels in check when mixing, is to start with what are typically the loudest peak elements, such as kick drum and snare drum, and after you have them sounding the way you want (always checking the sound within the context of the mix), then setting the levels on those so that your stereo bus meters are not going above -6 dBFS at the loudest parts of the song. Then, if you set you levels for all your other tracks using your ears to what sounds right relative to the drums, you should be fine with your levels.
I worked furiously over the weekend and experimented a bit with mixing. I don’t know if this is typical or is a general rule in mixing but here is what I ascertained.
It seems that when one is mixing a track, it is best to keep the volume fairly low in the pre-fader chain. What I did was insert Ozone in one of the inserts spots and check to make sure that the signal was not running way above the 0dB line. I would then boost the the signal a bit in the post-fader-chain with some compression and maybe even boost it some more when I got to mastering.
I don’t know if that all even makes any sense but keeping the volume lower in the pre-fader-chain seemed to give the music more “space” and it did not sound as tin-like. Does this make any sense? As you can tell, I do not know squat about the technical aspects of sound. I don’t even know what the 0dB actually means.
Am I out on a limb?
No, you’re not out on a limb at all, although, I would still recommend putting your compression in a pre-fader insert slot so that the way the compressor is reacting doesn’t get changed when you move the fader up and down.
But, the general principle of not slamming your channel and stereo master levels all the way up to the top is the way to go, and this can indeed make the sound “open up” a bit. Why the sound seems to open up can be because of a few factors… first, and probably the most likely, is that you aren’t driving your D/A converters at maximum levels. Some converters, both A/D and D/A, can start to sound “pinched” and smaller when you drive them really hard/hot. That’s why most people recommend that when recording and mixing you keep your average levels somewhere in the -18 to -14 range, with peaks not going above -6 dBFS. Also, some software, especially fixed point software, could start to sound bad if you are always maxing out the digital levels. It all depends on how they do the math and how good the programmers were at implementing things.
As far as what 0dB means. By itself, it doesn’t mean anything. dB stands for decibel, which is a measure of signal level, but it is a ratio and has to be expressed relative to some reference level. In the days of analog, when you talked about dB’s, you were really just doing a short name for dBVU. There was a common reference level that everyone worked with in the analog world, and most VU meters were calibrated to that, so 0dBVU would be when your VU meters were at 0. Analog had some headroom built-in to it, and there is no digital clipping in Analog (but there still can be distortion if your levels are too high), so it was possible, and common, to have signals recorded above 0dBVU. VU meters were more of a measure of the average, or RMS, levels, and they didn’t really show peaks too well since there were slow acting meters. Many meters had little LED lights built into them that would light up when the faster peaks were above a certain level.
In the digital world, you have a certain number of bits in your digital data to work with. When all of those bits are “on”, or set to 1, there is no higher value you can possibly represent. So, if your signal level goes past the point where all the bits are on, the data will be “clipped”, and all the bits will still be at 1. If this happens for more than a few samples in a row, you get a nice squared off top to your waveform, which creates a sort of square wave, and causes some very nasty digital distortion. In short, digital distortion is nasty and should almost always be avoided, unless you are using it as some kind of special effect. So, when we talk in digital terms, our meters are calibrated with respect to the “full scale” value, where all the bits are on. Thus, we have dBFS (decibels full scale). Floating point digital data can avoid clipping since they have a certain number of bits that represent the data, and then a couple of bits or so (depending on the implementation) for the exponent (or multiplier) for the data. So, if you go above full scale with your bits, the multiplier in floating point math simply changes to the next multiplier value, so you can still represent the data accurately without clipping. Most modern DAW software programs use floating point math for their internal mix engines to avoid clipping, and most (BUT NOT ALL) plug-ins also use floating point math. HOWEVER, your soundcard converters (both A/D and D/A) are still fixed point. There is no such thing as floating point converters. So, even if you are recording with your audio software set to record in floating point format, you still can’t go above 0dBFS when recording because you will clip your A/D converters on your soundcard. Same thing when mixing… you are monitoring the analog output of your soundcard’s D/A converters, and if you go above 0dBFS, you’ll get clipping and digital distortion from the converters (even if the software is not distorting if it is floating point).
Hopefully that makes some sense. I tried to simplify it the best I could, but can still be hard to understand without some one-on-one time and graphs and charts, etc.
Silly question –
Does headroom apply to EQ?
As we’ve discussed, you want to leave headroom on your track volume-wise. Does one also want to leave headroom on EQ?
For example, if I slap a visual EQ device on the pre-fader slot on the master stereo channel, would I want to keep the eq levels below the 0dB threshold?
Thanks for your patience.
Headroom applies to anything in your signal chain. You want to make sure you have proper gain staging throughout any signal chain so that you aren’t overloading the input or output stages of any device, or plug-in, in your chain.
However, with respect to plug-in EQs, that doesn’t mean that you can’t boost the EQ levels in any band. You can boost and cut EQ as much as you want, as long as you aren’t clipping the output, or clipping the internal signal of the EQ if it’s a fixed point processing plug-in. In other words, just watch the output meters on the EQ itself (if it has meters), and keep them below the clipping point. If your output signal is getting too hot, then reduce the gain knob within the plug-in EQ itself so that you can still do the EQ you want, while keeping the output at a reasonable level. If the EQ itself doesn’t have any kind of meter within it, then just set the meters on your channel to show the pre-fader levels, and use the channel meters themselves to see if the EQ you are applying is causing your signal to clip or not.
Sorry I’m being a bit dense about this.
I use a synth plug in (Minimonsta) on one particular track. As I manipulate the filter resonance, the mid-range frequencies spike and go above the 0dB demarcation on my EQ which is on the pre-fader sterero output buss. I have included a link to the spectral EQ I use right here:
So if I have this EQ slapped on the pre-fader position of my stereo output buss and I see the synth part peaking the midrange area above the 0dB readout on this EQ, does that impact the headroom and cause distortion? Or does one raise and lower EQ according to EQ “taste”?
Again, I feel silly for asking this and thanks for your patience.
No problems… the only silly questions are those that you never ask!
However, I’m not sure I totally understand what you are asking in the above post. Are you using the EQ to try to correct the big peaks that are caused when you crank up the resonance of the synth’s filter? Or, are you just using the EQ to shape the overall sound?
If you are trying to tame those large peaks, the best way is to simply not crank the resonance so much, unless you are really going for that type of sound. Those types of filters can really get out of control quickly if you crank up the resonance. If that’s the sound you are going for, though, then you’ll just want to lower the output or master volume of the Minimonsta synth so it doesn’t cause clipping during those large resonant peaks.
Whether or not it causes distortion, again, depends on the system and the plug-ins. If the system is 32 bit floating point, and the plug-in EQ and the VST instrument both also use floating point math, then you won’t clip or cause any digital distortion as long as at some point you lower an output or fader somewhere to keep from clipping your D/A converters on your soundcard (which are fixed point, and will clip and distort).
You could certainly also use the EQ to try to tame those peaks if you don’t want to back off of the resonance control, and you could automate the EQ to really pull down the offending frequencies during the loudest peaks, and then just have it set more gently the rest of the time.
You are either using the EQ to shape the sound to “taste”, or you are using it to try to correct a particular audio problem you are having… or, it can be a combination of both. There is no right or wrong way to do things, and it all comes down to what sounds best to you and for the song!