Ok now to combine two threads condense and conclude!!

Ben,

no.  24 bit interfaces are 24 bit - that is what they deliver to your computer.  They aren't perfect, of course, and have a maximum dB of 100+ to 120.  But neither are 16 bit interfaces delivering a full 96 dB, unless they are 24 bit.

So the full 144 dB can't be reached, and the signal at a very low level dissappears into noise - which is just about inaudible.  If you are recording a gnat's fart you won't hear it, but for any practical sound at less than an ear splitting level, you won't hear anything and your recorded signal probably disappears into the low level noise of your system - including amp noise.  For all practical puposes, a good ADDA interface specs out better than your ears.  Maybe not yours, but mine for sure.  What is the lowest dB sound we can hear -- 20 dB or so?  I used to know, but it has got lost in the noise of my fading memory.  The first 4 bits (the noise) is around 24 dB.  And the loudest, I think, is around 130 dB or so, where the ear starts shutting down to protect itself.

Digital converters don't mimic analog gear - they contain analog gear before the converter chips.  You can saturate those, tho most average units don't respond well.  Some transformer-based converters do, like burl, UA and JFC.  I know professionals that swear by them and find the analog part mimics tape.  But they also cost about $1000 a channel.  Most of us use "clean" (and cheaper) transistor ADDA units not driven so hard, since it don't sound so good.  Feel free to, tho, since it is art.  Digital harmonic distortion isn't harmonic but simply noise where the signal is greater than the ability of the numbers to represent it.  It clips, like transisters.  Again, you may like it and call it art.

You can get a rich, not abrupt saturation, by using a nice (read expensive) front end unit before your realitively cheap converter.  And again coming out to your analog transducer.  That is what I call art.  Or at least good sound.

@

Awesome link thanks Ben.

That Paul Frindle guy coded some of the Sonnox plugins I believe.

Ben the article you are interested in was in Issue 80 April 2011 Audio Technology. The front cover has a title 'The Low End'   Very good article as to why most PA's have too much bottom end. Very interesting article! and I totall agree. Most live sound engineers go way over with the low end. They are deaf to it.

The article you are interested in is called 'Too Low for Zero' by Jan Muths. This is not Stav refuting his original idea back in 2003 in his book. That was an error in your assumption and I never said or meant that so sorry. But it is obviously a much later article and it generally proves that recording lower is very good. And remember it is 8 years later than Stav's original idea.

If you cannot find it let me know and I can scan it and email it to you. It is only three pages.

Back around 2000 I was using an Emagic Audiowerk 8 sound card and although the converters are 18 bit, the specs say digitally the card can only go into 16 bit mode. I think many sound cards were like that at the time or the 24 bit ones were certainly not common.
Stav was suggesting we push the levels a little hotter then. I have got his email address, I could ask him what he thinks now about the situation.

But now of course 24 bit is standard for any sound card these days which is a good thing of course. It means we can go into that mode whenever we want to. All we need to do is create a session at that bit depth. If you have got a digital mixer you need to be able to put that into that res too and the Yamaha can of course. (Latest model handles all the sampling rates too which is even cooler!)

Now this might BE controversial:
"I hope this helps - because the subject of float versus fixed has gone round and round for some while - because it's difficult to grasp.
The most 'controversial' conclusion for people from all of the above, is that for the most part if you are dealing with audio audio data of say 32bits wide (cos that's how wide your processing data buss is), you are significantly better off to do this in a fixed point representation than float."
That's Paul from page 10, post #297.
Have fun children...

T

DeeringAmps


Well Ben here is your "Smoking Gun"
pg 15 post #430
 The reason I ask is that I think I've heard someone say that hitting you AD converter hard can degrade the audio quality a little bit." 
A) "That depends entirely on the ADC design, quality and performance - and indeed whether it deliberately distorts to avoid clipping :-(.
 Whilst it is true that hitting the ADC hard will increase signal to noise ratio, there's also the risk that it may produce more harmonic distortion  

And to clarify further this specific piece - they're talking here about analog distortion produced on the analog side of the ADC before you reach digital clipping.

And as discussed before, increasing the S/N ratio isn't really relevant if the quantization noise from the ADC is more than a little below the noise already present in your signal - you just end up a raising the noise floor by a fraction of a dB.

weather you guys are trying to revial the truth or find the ever so elusive invediable,this thread is really good and educating,thanks for the read guys.

seriously,

very intersting stuff.

Jeff Evans


*     Something that is not considered is that every time you add a track to a multitrack session the digital noise floor is in fact rising slightly. For one  or two tracks it is not a biggie but when you get into 32 tracks for example in 16 bit mode the noise floor comes up.

With 4 db of headroom 32 channels will create a noise floor of around -64 dB. (it gets 6 db worse as you double the tracks so for a 64 track mix the digital noise floor is -58dB) Good news is in 24 bit mode a 32 channel mix is creating a total digital noise floor of way down at -112 dB! MUCH better than in 16 bit mode. Where 24 bit excels is when you have large numbers of tracks.

 

This part sounds highly questionable (and that's being generous).

BenMMusTech


Jeff and Drew wanting to have the last word are muddying the water, what has come of this is two or three importent things:

1) Digital mediums and recording devices do have an optimum operating level, in this case both danny, myself and this paul frindle agree -6db, although I have said the odd peak of -3db is alright.  And I mean odd!!

As I said before (repeatedly), anywhere the quantization level is reasonably far below the noise floor and you aren't clipping is "optimal".

We can go through it all again if you'd like.

drewfx1

BenMMusTech


Jeff and Drew wanting to have the last word are muddying the water, what has come of this is two or three importent things:

1) Digital mediums and recording devices do have an optimum operating level, in this case both danny, myself and this paul frindle agree -6db, although I have said the odd peak of -3db is alright.  And I mean odd!!

As I said before (repeatedly), anywhere the quantization level is reasonably far below the noise floor and you aren't clipping is "optimal".

We can go through it all again if you'd like.

Then why this:
 
"If you can't test accurately what your ADC actually does (very difficult) then my advice is to aim for -6dB peak values out of your converter as a reasonable safety compromise (only losing 6dB SNR) - and then lose perhaps another 6dB at the head of your DAW channels - giving you total headroom of around 12dB for your mix and processes."
 
And what about the prospect of the mere posibility of harmonic distrotion in you black box!!! 
 
I think you might have the science correct but I think the science is still pissing in the wind, 29 years to figure out tape and after 35 years we are still trying to figure out digital and the possibility's.
 
The three points I have made have nothing to do with the maths, screw the maths, it has to do with three people's opinion that there is an optimum operating level for digtal and this may tie in with this idea of harmonic distortion.
 
So even if the harmonic distortion is before the converter, at -6db we are going to hopefully pick some of that extra harmonic distortion up.
 
Sorry drew I am going with my ears and heart on this one, you may get the maths but something else is going on.  That ramdom factor that even maths most of the time can't explain.
 
Neb 

Jeff Evans


drewfx1 I am only quoting the article in question and why would 32 tracks not add some noise like they used to in the analog days with each one contributing some level of hiss. Except in the digital world the noise or quantisation error noise is much lower but why would it not add up. In 16 bit mode 64 tracks could be considered noisy. (assuming they are all up aound unity and in the mix)

Analog does tend to have optimum levels so we need to be aware of them and set our levels there. Ben  is wrong however stating that digital has an optimum level, it does not.  As long as you are not down at some very low level eg -60 dB or so what I have said is simply this, you can record at quite a wide range of levels digitally and there will no change to the sound quality. (there will be variances in rms levels hence some tracks will sound louder than others though so it might be wise to concentrate on consistent rms instead of peak levels)

And when any of you say record up to -6db, record what at -6db! Are you referring to a peak. Because you certainly could not track at -6dB rms could you. So if a transient makes it all the way up to -6 db then the rms part of that signal could easily be down at -14 dB or even -20 dB rms.

As I have said before forget peaks and keep track of rms levels all the way through a production and keep them at a similar level. (in my case it is K -14 dB only because my Yamaha digital mixer is calibrated there) Worry about the rms levels and the peaks will all vary above any rms level and they tend to take care of themselves.

As long as you are a nice level eg -20 dB rms then you never really have to worry about peaks because rarely will a track have a peak that is greater than 20 dB above its rms level.

That is one of the problems of digital compared to analog. Analog does have optimum levels and reference levels and we have all worked with them but digital does not and it is all over the place level wise so we are trying to give them some sort of consistecy. K system does this well. Sound quality wise I say it again there is no real optimum level for digital, other than being so low we are down in the digital noise or so high we are smashing into 0dB FS. If you keep away from those extremes you will be fine.

Jeff of course we are talking about -6db peak not RMS do you think that I am that unskilled and mad!!!
 
-6db peak in Sonar equates to -20 to -16RMS, And how can you go against the tide of opinion, including Danny and Paul Frindle -6db peak seems to be the optimum operating level of digital, I of course push it a little harder, sometimes, also I use outboard gear and MOTU has compression and EQ built into every channel I use these devices to hit the optimum operating level.
 
Man stop muddying the water, Drew and Jeff, I think we have found three points, those of us who are less experienced can use.
 
Lets leave it at that!!
 
Neb

Jeff Evans


And when any of you say record up to -6db, record what at -6db! Are you referring to a peak. Because you certainly could not track at -6dB rms could you. So if a transient makes it all the way up to -6 db then the rms part of that signal could easily be down at -14 dB or even -20 dB rms. 

Hi Jeff. I can't speak for everyone, but for me, I fire up my input levels and play the song I am recording. I don't play lead guitar riffs or riffs in different keys that aren't in this song I'm recording, I stick to the exact song and chords/progression I'll be playing.
 
I have my input signal set so that it peaks at -6dB. I used to have it dead on -6dB and of course other notes/chords or what have you, would go over that and hit anywhere from -4dB to -2dB peak. I never had any problems doing it that way, but I've switched things to a final of -6dB peak and never go any hotter.
 
Whether this is right or wrong, messes with the science or whatever....it's where I feel the most confident with my personal recordings as well as those I record here. I also always run a light outboard compressor on each track so that it helps me to stay at -6dB without over-compressing. The signal is just conditioned enough to stay where it needs to be.
 
As for the rest of this thread...and I mean no one any disrespect that has posted in it, but it blows me away that such a conversation could even take place over something as simple as a meter and the signal that is fed into it. I turn them on, get a good signal that sounds good to my ears and never look at the meter again unless I hear a clipping sound somewhere or when I export my audio. It baffles my mind that individuals that are 1000% more intelligent than me would even spend more than a minute looking into something like this.
 
It's like....(and this is why the science part of the audio world just doesn't do anything for me) let's say I'm working with a guitar tone. I play, I set my amp or my pre-amp to where this sound is as good as it can be, set the mic where it sounds the best or run a DI, run both...whatever works. I listen for things that stick out to me as possibly being problematic. Ok, in this song, I'm using a good amount of gain. I notice that when I ride or chug on an A chord, my meters ramp up.

I either compress a little more to keep that in check or I lower my signal so that the A doesn't ramp me up too much and I can compress it more later after it's been recorded. I make sure I'm not clipping and I make sure the sound doesn't sound too stressed or too weak in signal. When I like what I got, I press record. I'm not worrying about K-systems, crushing my signal near 0dB or having too weak of a signal. I just go with it and always wind up with something that is presentable. I just can't see how or why we'd even worry about the rest of this stuff...but that's just me.
 
-Danny

I am interested in the statement that a peak of -6dB rms equates to 20 or -16 dB rms. Where did you get that info from? Yes, if the peak value happens to be 14 db say above its rms value as could be the case with a heavily slapped funk bass etc. But what if a sound comes along that only has a little peak say 3 db above its rms value.? What then. It means that the peak is still showing -6db but now the rms component is only down at -9dB a far cry from -20 or -16 as you put it Ben.

That is what is wrong with continuously monitoring peak values and nothing else. It is the rms values that need to be down around our ref level and keep them all the similar value and some peaks might be 3 db above that rms level and another sound might have a peak value 15 db above its rms level.

That is what is totally wrong with your (and many others too) approach Ben. Your theory only works if all sounds have the same rms to peak ratio and they certainly do not. My concept works best because it is about keeping rms levels constant and the success of it is not dependent on the sound itself, it works because you are doing something else (keeping rms levels constant) that is not related to the sounds rms to peak ratio.

You are muddying the waters Ben. I am basing my concepts on proven approaches which do work and I have used in many productions over the years. That is what is great about it, applying a sort of analog approach to digital. Choosing a ref level and working there! There are many as I have pointed out from our SSL training anything from -24 to -18 are common ones too. (Pro Tools HD interfaces are set at -18 dB FS as the ref level but they can be tweaked at the rear) At least the -20 dB rms is a standard that is used in the film industry so it is one level we can take from that and apply it to music production. Our new SSL mixer came calibrated to -20 dB which is interesting too.

Do you agree Ben that we need to be parking rms levels at some point anywhere between -18 to say -24 in the SSL case. Because if you do (and I get the impression you are starting to see the light) that is far cry now from the concept that you started out with saying we have to slam everything as high as we can!

drew good point about bringing tracks down to compensate, of course you are right. But what about a situation where all the tracks are actually at unity gain and no one on any track plays a sound at the same and each musician has a very quiet instrument eg a Vietnamese zither being plucked very quietly. (This is a soft sound I can assure you!) Imagine 32 tracks of this but no one is overlapping anyone else. Would you hear all that 16 bit noise building up, maybe! (BTW this article also mentions that any analog gear in this path is going to have a worse noise floor than digital at 24 bit so it is in fact the analog stages noise that you will hear building up) But yes under normal conditions it would not be an issue because this type of situation is pretty rare. But if you had to do it you would at least choose 24 bit then wouldn't you.

Danny I have just read your post. Yes I agree man! The reason what you do works is simply because digital does sound great and records really well over a very wide variety of levels and this is in direct contrast to what Ben is saying, he is saying there is only one special place for digital levels which is rubbish. Your work very well proves it in itself. No you don't have to get all crazy about K system and if you are a good mixer then you will still always get a good mix! But K metering is good and it does work and I believe even if you applied it Danny you would find your track levels would be very consistent and buss and final mix levels would also be the case. It does work and it does help. It makes mastering easier and more consistent too.

I don't agree that digital recording introduces harmonic distortion, that is bollocks! And it is not random either because it simply not there to start with. Look at the harmonic distortion figures of a typical digital recording setup and you will see it is almost immeasurably low. Like .001% maybe not sure. Wow and fluttewr is low too, noise is also low. These are things that are good about digital. So Ben as long we are well away from the extrems of digital then harmonic distortion is going to be very low or non existent. Sure if you bury the signal right down at -90 or something like that in 16 bit mode then you may hear something or if we are smashing into 0dB FS then it may be present then too but we are not there are we.

I want to mention that I am always talking about no processing on the way in to digital recording in terms of peak/rms ratios etc. Danny if I measured all the rms levels of one of your muktitrack sessions I would probably find they vary but the peaks would be very consistent but with me it is the other way around. You are compressing too at times on the way in which would be bringing rms and peak levels up to near each other. Heavily distorted guitar cabs are very compressed anyway so peak and rms levels could almost be the same! Which means if you are tracking and hitting -6db some of your guitar tracks could be pretty hot but you would pull your track faders down obviously to compensate. But with me I would still track those slammin cabs at say K-14. My track faders would be higher to achieve the same mix and same sound in the end.

Both ways work and many people here use the peak concept more consistent. That is a sort of digital thing. I like to keep rms levels constant and this is an analog thing being brought over from an era past into a modern digital approach and it also seems to work too.

I hate to even contribute to this, but I think that what some folks are missing is that there are different issues involved. The AD/DA part of it is one thing, and just a reasonable level anywhere from -20 to -3 should probably be fine (peaks obviously), though -3 doesn't make much sense because you'd have to be so careful during recording and it wouldn't really make any sonic improvement to do so, so it's not something you'd want to do.

But the biggest issue in that huge Gearlutz thread is about *digital processing* on those tracks recorded at high levels, not about the actual in and out levels themselves. If you run your tracks (and buses) overly close to 0dBFS and you have plugins on those tracks, you can have issues in some cases, because they can raise the level in unexpected ways, and it can happen between multiple plugins on the same track so you might not see it on the actual track meters because a subsequent plugin could bring the levels back down.

In a floating point world, it might not make much difference since there's no clipping involved going over 0 (though there might be in the actual algorithms of any given plugin even still), and you can't assume necessarily that all of your plugins are processing in floating point format internally I guess. Though if you do go over 0 in the floating point world you can be needlessly throwing away resolution bits. But in the 64 bit floating point world it might not make that much difference even there.

Yeh, you can subsequently bring down the track levels via the trim, but again you aren't gaining anything sonically by tracking them higher, so what's the point? You are making your tracking more succeptible to overs and then just turning it down anyway before it gets into the DAW's tracks. So it's kind of silly.

On the commonly discussed issue of 'more resolution' when you record hotter, the thing that Paul pointed out to me is that this isn't true. More resolution comes from having finer gradations. 24 bit is 24 bit. The gradations are the same across the whole possible range. It's the same resolution at the bottom as at the top. So you aren't getting more resolution by using more bits.

The real issue with resolution comes later during the actual mixing. But if you are on a 64 bit floating point DAW, and that's the case for all of us here on SONAR, then there are vast number of values between every 24 bit captured value, and those are used during the actual processing, until it's time to spit it back out to the D/A.
 
The whole issue of harmonic distortion on A/D is just silly. Who would want to do that? If you want (good, harmonious harmonic) distortion, get it on the way in from your front end. This is a non-issue that it's not worth wasting air arguing about.
 

Ben I am also saying that when digital systems are low in harmonic distortion I am assuming that any analog signal path is of the highest quality and is also introducing the least harmonic distortion possible. If I were to measure it I would be patching the oscillator direct to the sound card nothing else! And I would be putting the distortion analyser on the output of the sound card too!

But what you are saying is also correct in that harmonic distortion could be introduced in a variety of places prior to the A to D converters. And of course even if a pristine signal has got into a DAW now we have plugins that can introduce tons of harmonic distortion but that is on purpose.

Analog tape recorders are very high in distortion. Did you know of the order of several percent! Have you ever seen the square wave response of a reel to reel tape recorder, it is pretty bad overall. Only the finest machines can even achieve what looks like a decent square wave and even then it is not great. Digital can do this while walking in the park!

I try to avoid loud guitar cabs too I think they are too loud! I think we can leave that stuff to Danny. I love recording guitars direct and getting stuck into them inside the DAW. All be at a very low volume at 2am in the morning! I think we have really progressed in this area for sure. Plenty of distortion there but it's good distortion though!