Almost all my songs are audibly clipping, what am I missing?

[QVprod]

I'll add in here. It's not simply academic. The distinction really does matter. You can have clipping lights on the master but not hear audible clipping on the output. You will however hear that clipping upon export once floating point is no longer saving you from it.

I'm pretty sure the playback already isn't floating point.

I only found one interface that supports 32-bit float: https://zoomcorp.com/en/us/audio-interf ... s/uac-232/ and I'm not sure Reason would be able to take advantage of it.

The problem you’re having is you keep thinking the master and output are the same thing. What I said in the last post proves otherwise. Try it for yourself. It isn’t hard to get a clip light. You won’t hear it unless you’re really pushing it.

[antic604]

I'm pretty sure the playback already isn't floating point.

I only found one interface that supports 32-bit float: https://zoomcorp.com/en/us/audio-interf ... s/uac-232/ and I'm not sure Reason would be able to take advantage of it.

The problem you’re having is you keep thinking the master and output are the same thing. What I said in the last post proves otherwise. Try it for yourself. It isn’t hard to get a clip light. You won’t hear it unless you’re really pushing it.

I'm aware of the distinction (and resulting patching opportunities), but I fail to see how that's useful for anything other than room correction software that you don't want included when rendering your song. No one would put limiter between the master and audio interface, to prove they can clip the master but still don't hear it.

[crimsonwarlock]

I'm aware of the distinction (and resulting patching opportunities), but I fail to see how that's useful for anything other than room correction software that you don't want included when rendering your song. No one would put limiter between the master and audio interface, to prove they can clip the master but still don't hear it.

Actually, I have a brick-wall limiter patched directly in front of the audio interface, not to prove anything, but to protect my near-field monitors from damage by level spikes. The limiter is patched last, after my room correction software. I have that setup in all my templates.

[selig]

Not sure we’re disagreeing, except:
The truth is you can’t clip the master, because ‘floating point”.
That said…
You CAN clip the output and should thus always pay attention to levels at all stages.

Not sure either

I thought you're making a quantum physics theory-like argument, that the signal on the master doesn't clip while being there (which technically might be true), until you send it to master's output, i.e. to a file or speaker, because it needs to be converted from math to "reality"...

In that case we agree, although I see this argument as academic, because there's no point to have signal on master other than to hear it or export it to file. The peak & VU meters measure the output, not the signal "inside" of master.

So, similarly, I can appreciate the fact that when not measured we only know the distribution of probability for the position and speed of a particle (as illustrated by Schroedinger's cat being both dead and alive at once), but we'll know on of them for sure only when we'll "look" at it.

The peak/vu meters measure the samples, NOT the output. They literally DO measure the signal "inside". This is why they cannot show true peak, which requires simulating the D/A reconstruction filter to determine the likely analog equivalent peak levels.
The same thing happens with the waveform display, which shows digital samples and not the reconstructed analog signal.

This is a floating point conversation we are having, right? All digital signals inside of Reason are floating point. The ONLY place they are converted is when writing a fixed point file or sending audio to the D/A for conversion to analog. Thus, that is the only place clipping can occur.

It's like building a sailboat inside a big room with a little door. You can build it as big as the ceiling allows, but it is the DOOR you have to concern yourself with if you ever want to get that boat outside of the building. Trying to exit a boat bigger than the door means some "clipping" will occur but ONLY when you actually attempt to leave the space.

[selig]

I'll add in here. It's not simply academic. The distinction really does matter. You can have clipping lights on the master but not hear audible clipping on the output. You will however hear that clipping upon export once floating point is no longer saving you from it.

I'm pretty sure the playback already isn't floating point.

I only found one interface that supports 32-bit float: https://zoomcorp.com/en/us/audio-interf ... s/uac-232/ and I'm not sure Reason would be able to take advantage of it.

And I'm pretty sure that interface has a 24bit OUTPUT. What monitors could possibly handle 1500dB dynamic range for playback?

Playback specs are determined by your D/A hardware interface, which are all fixed point with no need for floating point's extreme dynamic range. So that's where the bottle neck is at present, and why things can only clip when converting from floating point to fixed point file formats (either upon export or conversion to analog).

[Marc64]

I'm no master but try to lower the volumes on the faders and the slap a maximizer on the master with soft clipping on

[o_imseng]

I'd still have a R11 Suite licences for sale if anyone is interested, just PM me.

[avasopht]

Not sure either

I thought you're making a quantum physics theory-like argument, that the signal on the master doesn't clip while being there (which technically might be true), until you send it to master's output, i.e. to a file or speaker, because it needs to be converted from math to "reality"...

In that case we agree, although I see this argument as academic, because there's no point to have signal on master other than to hear it or export it to file. The peak & VU meters measure the output, not the signal "inside" of master.

So, similarly, I can appreciate the fact that when not measured we only know the distribution of probability for the position and speed of a particle (as illustrated by Schroedinger's cat being both dead and alive at once), but we'll know on of them for sure only when we'll "look" at it.

The peak/vu meters measure the samples, NOT the output. They literally DO measure the signal "inside". This is why they cannot show true peak, which requires simulating the D/A reconstruction filter to determine the likely analog equivalent peak levels.
The same thing happens with the waveform display, which shows digital samples and not the reconstructed analog signal.

This is a floating point conversation we are having, right? All digital signals inside of Reason are floating point. The ONLY place they are converted is when writing a fixed point file or sending audio to the D/A for conversion to analog. Thus, that is the only place clipping can occur.

It's like building a sailboat inside a big room with a little door. You can build it as big as the ceiling allows, but it is the DOOR you have to concern yourself with if you ever want to get that boat outside of the building. Trying to exit a boat bigger than the door means some "clipping" will occur but ONLY when you actually attempt to leave the space.

The SSL uses 64-bit integer for summing (can't remember if it's also used for EQ & Compressor calculations).

[antic604]

I'm no master but try to lower the volumes on the faders and the slap a maximizer on the master with soft clipping on

You might be on to something here!

[antic604]

I'd still have a R11 Suite licences for sale if anyone is interested, just PM me.

knobcloud.com

[o_imseng]

Soorry about my post somehow landed in the wrong thread. My apologies.

[o_imseng]

I'd still have a R11 Suite licences for sale if anyone is interested, just PM me.

knobcloud.com

Thanks antic didn't know about this site. Will up it there as well.
Regards

[jklok]

I would say that a well balanced mix has all faders at zero and is only used for automation. I often find that it's a patch or insert effect setting that creates the overload. Nothing makes me happier than finding the culprit hidden somewhere deep inside a combiner. It's always best to catch it earlier in the chain and having the signal going to -10 usually means that the settings within the channel are correctly balanced to begin with.

[Unaudited]

I would say that a well balanced mix has all faders at zero and is only used for automation. I often find that it's a patch or insert effect setting that creates the overload. Nothing makes me happier than finding the culprit hidden somewhere deep inside a combiner. It's always best to catch it earlier in the chain and having the signal going to -10 usually means that the settings within the channel are correctly balanced to begin with.

How do you set the volume before the mixer? A gain plugin I presume? Personally I do it the opposite way, I use the free kHs Gain to automate volume and the fader I use to set the overall constant volume that's not automated. Also, if all signals are -10 then how is the mix balanced? Everything would be the same volume...

[Sengin]

You gotta start somewhere, right? -10 dB on the channels is a place to start. Set everything to -10 (this is a good starting point because when combining all channels you probably won't hit 0 on master). Organize your channels in a way the song requires (e.g. kick channel outputs to drums channel, snare outputs to drums channel, etc). Bump the volume on your *speakers* (you probably want to "mix loud" and you've just set everything to -10) then solo what you consider the most important channel (most likely drums or lead) and adjust fader volume on the mix channel until it's where you want it. Then bring in one channel at a time and adjust its volume relative to the volume you've already set. Do this for every channel and you've got a starting point.

I'd recommend sub-grouping mix channels and doing them piecewise. For example, drums. Do solo kick, bring to -10. Then bring in snare and adjust snare's level until its where you want it relative to the kick, combine both outputs into a single mix channel. Solo trumpet, bring to -10. Bring in trombone, adjust trombone volume until its where you like it relative to trumpet. Set both to output to 'brass' channel. Now you can bring in drums and adjust the relative level of brass to drums.

I'd also recommend setting the input volume on your mix channel if you end up adjusting the fader to +-6dB (e.g. set input to -6 and fader to -1 instead of setting fader to -7). That's because the fader is easier to adjust and in smaller increments, so it's easier to adjust later as you get further into your mixing process (as you never set your levels 'right' the first time as mixing is an iterative process).

Then put a limiter (e.g. maximizer) on master to get the levels to where you want them (including trading off perceived loudness vs clipping).

[selig]

You gotta start somewhere, right? -10 dB on the channels is a place to start. Set everything to -10 (this is a good starting point because when combining all channels you probably won't hit 0 on master). Organize your channels in a way the song requires (e.g. kick channel outputs to drums channel, snare outputs to drums channel, etc). Bump the volume on your *speakers* (you probably want to "mix loud" and you've just set everything to -10) then solo what you consider the most important channel (most likely drums or lead) and adjust fader volume on the mix channel until it's where you want it. Then bring in one channel at a time and adjust its volume relative to the volume you've already set. Do this for every channel and you've got a starting point.

I'd recommend sub-grouping mix channels and doing them piecewise. For example, drums. Do solo kick, bring to -10. Then bring in snare and adjust snare's level until its where you want it relative to the kick, combine both outputs into a single mix channel. Solo trumpet, bring to -10. Bring in trombone, adjust trombone volume until its where you like it relative to trumpet. Set both to output to 'brass' channel. Now you can bring in drums and adjust the relative level of brass to drums.

I'd also recommend setting the input volume on your mix channel if you end up adjusting the fader to +-6dB (e.g. set input to -6 and fader to -1 instead of setting fader to -7). That's because the fader is easier to adjust and in smaller increments, so it's easier to adjust later as you get further into your mixing process (as you never set your levels 'right' the first time as mixing is an iterative process).

-10dB WHAT on the channels? VU? Peak? PPM? Are you using VU offset? You can easily clip the output with VU levels that high (see below for an example)

If you use the channel meters you are seeing average level, not peak level. A signal at -10dBVU could easily be clipping the outputs since the peak level can be 10dB, 20dB, or more above the average. By the time you combine 3-4 all at -10dBVU you may already be clipping the outputs. Check out the default snare on Kong Kit (the default kit) - if you set it to -10dB on the channel you're already clipping the output by +2dBFS.

That said, if you use a consistent peak level for all signals, like would happen back in the day of digital tape machines, you will leave headroom for the mix bus to allow many channels to be combined.
I really wish they had the option to have the channel meters follow the main meters, I really like the VU + Peak mode and use it all the time on the Big Meter to judge crest factor (peak level minus average level, a good indicator of loudness IMO).

[Sengin]

I am indeed talking about the channel meters. My post was intended to be an extremely brief starting place (and one that you can also do very quickly, so that you can get on with your song), not intended for in-depth discussion to cover 100% of the cases. If you load up all your instruments, -10 is in my (albeit limited, heh) experience is a good place to start. If you are clipping at -10, it may be a good idea to see why. Or if you want such a high dynamic range, bring all down to -12 instead and start there (or go lower until you aren't clipping master). This is a starting place after all, and you will be doing many more changes following this and will need to constantly adjust levels. That is, -10 is not supposed to be a 'cover all' case, but the general advice is to pick a level to set all channels to that you aren't clipping master when combined, and go from there. -10 has just been that point for me (and it's easy to see as the channel meters have a -10 labeled).

[jklok]

Maybe it’s coming from the analog era that makes me work like this. But in most cases I find that a typical patch is designed to work within this range. So when everything is cool in most of my mixes, I have all channels going to -10 at max and the master at most just hits the peak without clipping before turning on the mastering units. I am not using gain stage at all, I always find that it is a setting within a patch that needs adjustment. It could even be down to a bad sound file that needs normalizing or clip removal.

I noticed that to keep everything properly balanced, you need to keep the frequency spectrum moving using the EQ/Comp and sidechain settings, as well as keep a close eye on the lo-end. I have the LPF/HPF setting enabled on all inputs and combine the Pump RE with the SSL compressors to avoid conflicts between different instruments. This is where I appreciate the flexibility of SSL. It seems that a standard patch most of the time has a level going to -10 on the channel anyway.

[SynthGang]

I would say that a well balanced mix has all faders at zero and is only used for automation. I often find that it's a patch or insert effect setting that creates the overload. Nothing makes me happier than finding the culprit hidden somewhere deep inside a combiner. It's always best to catch it earlier in the chain and having the signal going to -10 usually means that the settings within the channel are correctly balanced to begin with.

I would say that a good starting point for a mix would certainly have the faders at zero.. but beyond that I have a hard time understanding this fascination with refusing to touch... a control? lol

Don't get me wrong, I use that as a starting point for sure (clip gain is one of my best friends).. but once things get moving in terms of the mix then you bet I'm using those controls

[SynthGang]

Not sure either

I thought you're making a quantum physics theory-like argument, that the signal on the master doesn't clip while being there (which technically might be true), until you send it to master's output, i.e. to a file or speaker, because it needs to be converted from math to "reality"...

In that case we agree, although I see this argument as academic, because there's no point to have signal on master other than to hear it or export it to file. The peak & VU meters measure the output, not the signal "inside" of master.

So, similarly, I can appreciate the fact that when not measured we only know the distribution of probability for the position and speed of a particle (as illustrated by Schroedinger's cat being both dead and alive at once), but we'll know on of them for sure only when we'll "look" at it.

Thus, that is the only place clipping can occur.

Hey Selig,

What you explained has always been my understanding not to mention that, according to the manual, there's 64-bit summing in the mix bus in the Main Mixer Master Section.

It's also my understanding that this gives you (virtually?) limitless headroom, but I was able to easily (and audibly) clip the master - I'm now wondering where in my mind is the breakdown?

What I did was simple:

• Loaded a single instance of Subtractor, reset the device, set the waveform to sine, turned level up to max, amp envelope ADR set to 0 and Sustain set to 127

• Inserted an instance of Selig Gain, set fader to max (+24dB for those that don't have it)

Basically as soon as I bring up the Master Fader high enough to where my Audio Out Clip indicators come on, I'm hearing audible distortion. Using Voicemeeter Potato, I used the Virtual Aux channel of Voicemeeter, which I then routed into Audacity and recorded at 32-bit - waveforms are very heavily clipped.

I was just hoping you could shed some light on what I'm misunderstanding here if possible? My thinking is that I'm not "exporting" and recording the signal being played back by Reason in 32 bits. Is it just that I'm feeding it more than enough signal or am I missing something?

Here's what the audio sounds like (exported in 32-bit WAV and clip gain attenuated by -10dB):

Would really appreciate your insights!

Clipping.jpg (208.89 KiB) Viewed 5081 times

[jam-s]

As soon as the audio is passed from Reason to the driver of the audio interface it has to be converted into (24bit) fixed point sample data. That's the point where going higher than 0 dBFS causes the clipping.

[selig]

The problem you’re having is you keep thinking the master and output are the same thing. What I said in the last post proves otherwise. Try it for yourself. It isn’t hard to get a clip light. You won’t hear it unless you’re really pushing it.

I'm aware of the distinction (and resulting patching opportunities), but I fail to see how that's useful for anything other than room correction software that you don't want included when rendering your song. No one would put limiter between the master and audio interface, to prove they can clip the master but still don't hear it.

Not sure what distinction you’re talking about, everything you patch after the Master Outputs IS rendered with your song. Which is exactly why I put my limiter between the master and the audio interface, AND I still hear it if I clip the outputs. I do this to replicate how I’ve used mastering over the years, which is AFTER the mix stage. Patching it this way makes sense to me, but that’s probably just because it’s how I’m used to working.

BTW, you cannot clip the master section, thought we had cleared that up already?

Clearly there is still confusion here about how floating point audio works (and it’s not just Reason users that deal with this issue), so it’s probably a good thing we’re clearing it up for others as well.

[selig]

I am indeed talking about the channel meters. My post was intended to be an extremely brief starting place (and one that you can also do very quickly, so that you can get on with your song), not intended for in-depth discussion to cover 100% of the cases. If you load up all your instruments, -10 is in my (albeit limited, heh) experience is a good place to start. If you are clipping at -10, it may be a good idea to see why. Or if you want such a high dynamic range, bring all down to -12 instead and start there (or go lower until you aren't clipping master). This is a starting place after all, and you will be doing many more changes following this and will need to constantly adjust levels. That is, -10 is not supposed to be a 'cover all' case, but the general advice is to pick a level to set all channels to that you aren't clipping master when combined, and go from there. -10 has just been that point for me (and it's easy to see as the channel meters have a -10 labeled).

If you are clipping, there is only one reason (so you don’t need to see “why”) - it is because you are exceeding the headroom of the output or (of a fixed point file format during export). I’ve explained how you can have a channel reading -10dBVU and still be clipping the outputs (Kong Kits snare is a perfect example). That suggests your starting point is not low enough to leave headroom for even ONE track (let alone a mix of many tracks). How can you know how much headroom you have and how close to clipping you are if you don’t look at peak levels?

I 100% agree about using a consistent level for all tracks as long as it’s a PEAK level, because if your goal is to leave headroom on the master to avoid clipping the output then you simply cannot rely on VU metering for this job (right tool for the job issue IMO). In fact I made a video showing how this works, but maybe need to make another showing why it does NOT work if not using peak levels for judging headroom.

This is why every digital recorder (tape or disk) I’ve ever seen uses peak metering to set levels.

[selig]

Maybe it’s coming from the analog era that makes me work like this. But in most cases I find that a typical patch is designed to work within this range. So when everything is cool in most of my mixes, I have all channels going to -10 at max and the master at most just hits the peak without clipping before turning on the mastering units. I am not using gain stage at all, I always find that it is a setting within a patch that needs adjustment. It could even be down to a bad sound file that needs normalizing or clip removal.

I noticed that to keep everything properly balanced, you need to keep the frequency spectrum moving using the EQ/Comp and sidechain settings, as well as keep a close eye on the lo-end. I have the LPF/HPF setting enabled on all inputs and combine the Pump RE with the SSL compressors to avoid conflicts between different instruments. This is where I appreciate the flexibility of SSL. It seems that a standard patch most of the time has a level going to -10 on the channel anyway.

Standard patch level in Reason is -12dBFS, NOT -10dBVU (assuming VU offset = 12dB). I know this because of doing sound design for the Factory Sound Bank for Reason 6.5 as well as several individual devices since then. Some earlier patches exceed this occasionally, so it’s not quite as much a standard as I’d like to see. But as long as a single patch doesn’t clip (many do in the VST world, amazingly enough) I’m happy!

BTW, using filters on everything is probably overkill in most cases. Many times a low cut (high pass) filter will add level due to the phase shift, so it’s not always an effective problem solver. I’m of the school there is no free lunch. Meaning for every processor you engage in the signal path you have some positive and some negative consequences. The goal is for the positive to outweigh the negatives in all cases. For me to add ANYTHING to a signal path I must be absolutely convinced it is first necessary, and then second “does no harm” (the positives outweigh the negatives). If I’m not REMOVING some of the things things I try but that don’t work, I’m probably not listening well enough…

[crimsonwarlock]

...everything you patch after the Master Outputs IS rendered with your song.

With one important distinction: everything you patch in front of the main stereo outputs, being outputs 1 and 2 on the audio output panel in Reason. The signal that ends up at the first stereo output is what is being rendered out. You can test this simply by disconnecting these outputs and render out to file; all your mixer meters will show signals, but you end up with a file that has only silence.