The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
My Theory About Prosumer Audio Audio Interfaces
Old 13th June 2006
  #61
11413
Guest
I don't like to jam every signal up to 0dB because i like to mix with the faders at 0... no need to glue that hihat track up to full scale... no need for most of it when you're gonna back the faders off anywhere from 6-18dB. at least when mixing ITB. plus there's a lot of voodoo going on in those plugs I like to leave some headroom for..

so Dave Derr is right, up to a point.. and that point gets real fuzzy when you mix ITB.
Old 13th June 2006
  #62
Lives for gear
 

Some more thoughts on this topic... I must admit that besides the op amp quality I think digital jitter (noise artifacts due to an inaccurate internal crystal reference clock in the converter) should be a central part of this discussion. The problem that I see is the accumulative quantization distortion that builds up to harshness sooner if there is a lot of jitter in the signal. And this is especially the case today when people are limiting with plug-ins ITB. So take for instance something like a 32-bit audio engine combined with a medium quality converter and compare that with a 48-bit audio engine with premium grade converters. If that won't make up for a difference, then I think it WILL after downsampling and dithering. The difference might not be huge, especially not if you were using very high quality plug-ins, but I think generally the harshness comes from a little signal loss here and there which in combination with hard limiting will make the sound end up noisy and that's a problem.

There are a few extra critical effects in recording and mixing that need to be of very high quality when you are aiming for professional quality. I think generally bad prosumer mixes lack this: Compression, Reverb. I think the mix is killed when you combine these two effects in plug-ins that are not using the signal very efficiently, especially when you apply them in the wrong way. Bob Katz has pointed out the importance of using analog gear as an alternative to critical digital processes and to me that makes perfect sense. He has also pointed out what happens to a mix when it is too much limited, it gets wimpy. Mike Caffrey pointed out "too much compression" as a reason for low quality mixes and I think that's not just a lie...

Besides this, the main problems are likely to be find earlier in the recording chain.

One more thing. We need to realise that it soon becomes a matter of the engineer's ears and taste as well. There is a big difference between trying to make a mix punchy and heavy and trying to make it light and soft. All these differences in sound properties that the engineers are aiming for could alone very well be the reason when we "judge" bad prosumer sound quality. For instance I prefer an ultra soft high end and low end, a mix that is quite open. So I generally want to use such mics and to back off a little on those sharper elements like cymbals and hi-hat and also on the kick drum and bass to make them softer. The sum of all these decisions will have a big impact when we judge sound quality, it's not just a matter of mix self noise.
Old 13th June 2006
  #63
Lives for gear
 

Quote:
Originally Posted by 11413
plus there's a lot of voodoo going on in those plugs I like to leave some headroom for..
Yep, I sometimes do the same, because I don't really know what's going on under the hood on some plug-ins. Some plug-ins can make the sound surprisingly harsh only by clipping very little. Then I start asking myself how soon that clipping actually kicks in... I must admit that during the last year I've found myself wanting to disable more and more plug-ins. I have now ended up with Waves plug-ins that I think perform quite well compared to others, but I am still cautious and try to avoid them if I can.

Some plug-ins are just killing the signal too much. Maybe we audiophiles should make a blacklist or something. *joking*
Old 13th June 2006
  #64
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Headroom when mixing ITB

No person should call himself a plug-in writer if he doesn't leave lots of internal headroom for the DSP processing. Thats not to say theres not some stupid crap out there as far as plug-ins. But one of the first things a DSP writer learns is to leave headroom for full scale signal processing.

Actually, this is another reason to use all the bits. Because leaving headroom means you often have to round out and move the bits downward, dropping off more of the last bits, you end up moving the GOOD BITS (the MSBs or most significant accurate bits) downward also. So if you had 16 bit quality conversion and you wish to leave 20dB of headroom, you have to move the data 3.5 bits down, possibly truncating the last bits and lowering resolution farther, depending on the size of the buss and precision of the math. Floating point math eases much of this problem, but still its important that your upper bits are "well used". Most or all modern workstations use floating point math these days which if used properly, should keep the user with nice resolution. Still, you eventually start to lose the lower resolution as you get rounded out, and if you dont start out with great resolution and full scale signals, you are only quickening the loss of quality.

AGAIN, record your tracks to as close to full scale digital signals as is practical and hope that the DSP engineers aren't doing anything really stupid!

Let me remind you once again that its very similar to analog tape in that if you record your tracks low, you have 24 tracks of more tape hiss and distortion. The tape hiss is analogous to the noisy lower bits, and the headroom you didnt use on the analog tape is like most important bits you didnt use in digital.
Old 13th June 2006
  #65
Lives for gear
 

Quote:
Originally Posted by Dave Derr
No person should call himself a plug-in writer if he doesn't leave lots of internal headroom for the DSP processing. Thats not to say theres not some stupid crap out there as far as plug-ins. But one of the first things a DSP writer learns is to leave headroom for full scale signal processing.

Actually, this is another reason to use all the bits. Because leaving headroom means you often have to round out and move the bits downward, dropping off more of the last bits, you end up moving the GOOD BITS (the MSBs or most significant accurate bits) downward also. So if you had 16 bit quality conversion and you wish to leave 20dB of headroom, you have to move the data 3.5 bits down, possibly truncating the last bits and lowering resolution farther, depending on the size of the buss and precision of the math. Floating point math eases much of this problem, but still its important that your upper bits are "well used". Most or all modern workstations use floating point math these days which if used properly, should keep the user with nice resolution. Still, you eventually start to lose the lower resolution as you get rounded out, and if you dont start out with great resolution and full scale signals, you are only quickening the loss of quality.

AGAIN, record your tracks to as close to full scale digital signals as is practical and hope that the DSP engineers aren't doing anything really stupid!

Let me remind you once again that its very similar to analog tape in that if you record your tracks low, you have 24 tracks of more tape hiss and distortion. The tape hiss is analogous to the noisy lower bits, and the headroom you didnt use on the analog tape is like most important bits you didnt use in digital.
^&%$^&%!!! now i am back to square one again.

i had really thought that people had started to come to the consensus that peaking at -5 on the A/D or even lower during tracking was recommended, not just "ok". then you do not have to pull the digital faders back nearly as much while mixing.

and so this approach described above seems to indicate that the preferred method is to track pretty hot with no clips and then pull the faders back to avoid clipping the mix bus.

r.
Old 13th June 2006
  #66
500 series nutjob
 
pan60's Avatar
 

i was also under the impression that the consensus was to peaking at -5 to -10 during tracking was recommended.
i guess this is why i am also confused.
i am missing something?
Old 13th June 2006
  #67
11413
Guest
Quote:
Originally Posted by Dave Derr
AGAIN, record your tracks to as close to full scale digital signals as is practical and hope that the DSP engineers aren't doing anything really stupid!
you're about 1000x more optimistic than i'll ever be...



I treat plug-ins like i treat my friends.. i never assume they're anything but a potential menace.
Old 14th June 2006
  #68
Lives for gear
 

I would like to see Dave Derr and Paul Findle get together in a room and thrash out their ideas.

Dave - you seem to advocate the "resolution" theory that Paul utterly despises and tries to correct at any opportunity. Poles apart.

I used to strongly support the resolution theory, thinking that every wasted bit is a loss in audio quality. Paul's ideas have definately changed my thinking - although i'm still not 100% convinced. But I am completely convinced that we should never aim to get close to 0dBFS - for the following very simple reason:

24 bits has a mathematical dynamic range of 144dB (24 x 6dB).
BUT - the best converters have around 114dB, and cheap ones only 100dB or less.

IF you insist on peaking at just under 0dBFS, to obtain that mathematical 144dB resolution on your screen - you must have forced your analog signal to grossly exceed the analog dynamic range of your converter. By as much as 30 to 44dB.

I believe this is why tracking too hot into a DAW causes constricted, crunchy sound (nothing to do with clipping - even amatuers learn to avoid clipping). I'm talking about saturating the analog circuit of your A/D, and it doesn't sound good.

An interesting experiment is to create a sine wave that peaks at 0dBFS. Listen carefully to this as you raise the master fader. Especially on cheaper converters, you can hear the pure sinewave get crunchier the closer you get to 0dBFS. Then, as soon as you exceed 0dBFS you hear the much harsher clipping, but it's the saturation effect before you get to clipping that is interesting. Maybe you like it - maybe you don't.

That's looking at D/A - but i'm suggesting that exactly the same thing happens when tracking with A/D. Get to close to 0dBFS, and your sound can turn to ****.

Is it worth it for the sake of extra "resolution"? I don't think so. Also - is "resolution" actually lost at all if you track lower and "lose some bits"????? I don't think so any more.

The important thing to remember is that digitised audio is NOT like a digitised picture. It's too easy to draw the analogy to digitised pictures, but that's misleading. With a picture, you see every dot, and each dot is significant.

Audio is different. It is a continous waveform - and you only get to hear digital audio after the CONVERTER has converted this sample data into a continuous waveform. It connects the dots - so therefore, regardless of how many samples are there or not, you always get a continuous voltage waveform.

Also highly worthy of notice is the fact that not all digital audio samples that we see on the screen are necessary - or even "legal". This I found to be very interesting and changed my notions of digital audio.

Think about the reconstructed waveform for a while. It can be very different from the sample data we see on the screen. Because the analog circuitry is doing it's best to reconstruct a continuous waveform that connects all the dots. But there can be exceptions ...

Sometimes, to best connect the dots, the analog waveform has to exceed the samples. That's why samples very close to 0dBFS can cause analog "overs" when reconstructed.

Also - there can be "illegal" samples that can't actually exist in a true continous waveform. These will be smoothed out and probably won't cause much damage, but it's important to realise that not all samples we can create and view on the screen are actually going to work as expected in the reconstructed waveform. An example would be digital white noise - where signals are randomly generated, which simply could not be reproduced literally by a speaker.

An important thing to remember is that a converter can covert impossible or "illegal" digital streams into "legal" continuous waveforms that are then capable of being reproduced by a speaker. This may be an important part of the ITB/OTB summing debate ...

I think a lot of fears about "losing bits" is about losing harmonics or important frequencies. That's simply not going to happen. For a start, don't confuse bit depth with sample rate. Sample rate, not bit depth, determines the highest frequency or harmonic that you can reproduce. To a certain extent, higher sample rates mean you can use less bit depth and still get an accurate waveform.

If anything, the squaring of a waveform by low resolution digital quantisation is going to add harmonics - but that doesn't happen because the filtering in the converter removes the harmonics. The converter is connecting the dots anyway - so what we see an screen as a square wave, gets converted to a smooth wave anyway.

If you consider that all legal analog waveforms can be considered to be a complex blend of sinewaves, we can think about how accurately "low bit depth resolution" samples can represent a sine wave. The fact is - you only need two samples per cycle to be able to accurately reconstruct a sine wave. More bits aren't actually required, because there is only one sine wave that could fit the two samples and work. You can add more samples, but it won't make the reconstructed wave any more accurate.

The resolution of the bit depth isn't as important as people make out. You can dramatically lower the volume of a mix with the master fader, and the overall quality of the audio doesn't really change. Certainly the volume changes, and our human hearing response changes with volume, so we probably prefer the louder sounds, but really, the quality doesn't change a whole lot - even though we are "throwing away bits" like crazy.

I think the truth lies somewhere between the extremes that Dave and Paul are supporting.

I think you can track too hot.
I also think you can track too low.

But I think that we should concern ourselves with the analog sweet spot and headroom of our analog gear. ESPECIALLY OUR DIGITAL CONVERTERS, which are first and foremost analog gear.

And plugins are often modeled on analog concepts. So although floating point digital maths has virtually unlimited headroom - plugins that are designed to model saturation or clipping are going to saturate or clip.

Rant over. I'd be happy to proved wrong on any point - because i'm struggling to understand these conflicting ideas too.
Old 14th June 2006
  #69
Lives for gear
 
John Suitcase's Avatar
 

I tend to think that overdriving the converters, your preamps, any compressors/limiters you use is just as dangerous, and possibly moreso, than recording too close to 0db. If you're overdriving something in the signal path (the mic, even), you won't see a digital over, and the signal to the DAW may look fine, but it's going to sound wrong. Maybe not obviously so at first, but when you start to eq, compress, etc, you'll see it. So leaving some headroom through the whole path is important.

Of course, too much headroom, and you're going to lose range, and get more noise (in the analogue part, anyway.) So, you need to leave a little room, and listen very carefully. Look at the waveform, if you have to. If it looks very thick, with few high peaks, you're probably clipping somewhere. If you have 5db of headroom in the DAW, but the waveform seems too smooth, it probably is.

It's a tricky business!
Old 14th June 2006
  #70
Good post Kiwiburger

thumbsup

This is what I was pointing to in an earlier post on this thread....

Quote:
Originally Posted by Kiwiburger
I believe this is why tracking too hot into a DAW causes constricted, crunchy sound (nothing to do with clipping - even amatuers learn to avoid clipping). I'm talking about saturating the analog circuit of your A/D, and it doesn't sound good.
... I would add that I do believe the same thing happens inside the DAW 2 buss as well. When pushed too hard or too close to 0dbfs (most?) all DAWs seem to fold up and get small on me. When tracking if I push the converter as close can I can get to 0bdfs I find myself with less room to move on the DAW faders and I have to back everything down anyway. Better to start lower in level on the analog side of the conversion process than in the box in my eyes.

Funny I was just listening to Slipperman on one of his recorded rants (very interesting stuff) and he pointed this same thing out as well.

Not that he or anyone has the corner on the recording world but it is interesting that he feels the same way.....

don't ever try to shoot for 0dbfs...

at least I think that is what he said...

HO HO HO....
Old 14th June 2006
  #71
Quote:
Originally Posted by Kiwiburger
I would like to see Dave Derr and Paul Findle get together in a room and thrash out their ideas.
Great post.

Quote:

24 bits has a mathematical dynamic range of 144dB (24 x 6dB).
BUT - the best converters have around 114dB, and cheap ones only 100dB or less.

IF you insist on peaking at just under 0dBFS, to obtain that mathematical 144dB resolution on your screen - you must have forced your analog signal to grossly exceed the analog dynamic range of your converter. By as much as 30 to 44dB.
I don't know enough of the specifics, but I think there's a flaw in the math here, or maybe you're not taking the operating level of the converter into consideration, but I'm pretty sure it's possible to pass OdBFS without distorting the analog circuitry.



Quote:
An interesting experiment is to create a sine wave that peaks at 0dBFS. Listen carefully to this as you raise the master fader. Especially on cheaper converters, you can hear the pure sinewave get crunchier the closer you get to 0dBFS. Then, as soon as you exceed 0dBFS you hear the much harsher clipping, but it's the saturation effect before you get to clipping that is interesting. Maybe you like it - maybe you don't.

That's looking at D/A - but i'm suggesting that exactly the same thing happens when tracking with A/D. Get to close to 0dBFS, and your sound can turn to ****.
You're talking about a digital master fader, right?

I wonder if what you're hearing is alterations to the accuracy of the source signal caused by inadequately precise math or innaccurate math.

Here's a question I never thought of until now. If we can hear the difference between 44.1 thousand, and 48 thousand and 96 or 192 thousand, how many decimal places do we need to go for the math to be of an euqal resolution? Is it ok to round down to 10, or 100, or even 1,000? Is that even getting close to the level of detail we're at when we're splitting a second up into 96 thousand parts?

You're not really hearing clipping from the master fader if it's digital. It's not an audio signal getting it's top flattened by an op amp crapping out, it's a DAW mathematically mangling the signal.

Quote:
I think a lot of fears about "losing bits" is about losing harmonics or important frequencies. That's simply not going to happen. For a start, don't confuse bit depth with sample rate. Sample rate, not bit depth, determines the highest frequency or harmonic that you can reproduce. To a certain extent, higher sample rates mean you can use less bit depth and still get an accurate waveform.
I think of it like film with sample rate corresponding to fram rate and bit depth corresponding to flim width. There are high speed/frame rate cameras the provide, in a sense, a higher quality picture. That's what they use for slow motion. But 35mm looks better than 16mm and 70mm is better than 30mm. I've always felt like bit depth has made a bigger differnece than sample rate.

Quote:
But I think that we should concern ourselves with the analog sweet spot and headroom of our analog gear. ESPECIALLY OUR DIGITAL CONVERTERS, which are first and foremost analog gear.

And plugins are often modeled on analog concepts. So although floating point digital maths has virtually unlimited headroom - plugins that are designed to model saturation or clipping are going to saturate or clip.

Rant over. I'd be happy to proved wrong on any point - because i'm struggling to understand these conflicting ideas too.
Another thing to add to the consideration on the analog side is any type of limiting that is bulit in. Besides the options "soft limiters" many converters have brickwall limiters, with their filters to protect the converter. In a sense it's just semantics, becuase what difference does it make if you lop off the transient with a limiter or by hitting the ceiling of the circuit. A flat top wave still has an unpleasant sound.
Old 14th June 2006
  #72
11413
Guest
I want to stress the point about not recording to 0dB by default again... because I think anything you do to audio inside the computer is "destructive"... and less is generally more. I strive to get 90% of what i want *before* it hits the A/D.... which is why i like keeping the faders as close to 0 as possible.... less math to worry about.

Digital is really funny.. especially plug-ins... it's very easy (at least for me) to get sucked into messing with myopic things inside plug-ins, thinking i'm making a huge difference in sound... and then when i check back with the first mix i've changed it maybe 2-5%...

HAHAHA...

meanwhile i've wasted 2-3 hours...it can be a real timesuck if you're not careful...

digital is a bitch goddess... she gives with one hand and takes with the other.
Old 14th June 2006
  #73
Lives for gear
 

Quote:
You're talking about a digital master fader, right?

I wonder if what you're hearing is alterations to the accuracy of the source signal caused by inadequately precise math or innaccurate math.

Here's a question I never thought of until now. If we can hear the difference between 44.1 thousand, and 48 thousand and 96 or 192 thousand, how many decimal places do we need to go for the math to be of an euqal resolution? Is it ok to round down to 10, or 100, or even 1,000? Is that even getting close to the level of detail we're at when we're splitting a second up into 96 thousand parts?

You're not really hearing clipping from the master fader if it's digital. It's not an audio signal getting it's top flattened by an op amp crapping out, it's a DAW mathematically mangling the signal.
Yes i'm talking about the digital master fader - bringing a sinewave up to (and over) 0dBFS. When you clip your 24 bit fixed converter, the square wave clipping is extreme and very obvious. But i'm talking about the "browning" of the sound as you get close to 0dBF - especially with cheap converters (like consumer playback devices will have). My DAC-1 is pretty clean right up to 0dBFS. But my M-Audio Audiophile 2496 gets really crunchy as you get close to 0dBFS with a sine wave.

It may or may not be the op amp - but i'm convinced it's in the analog side of the converter. I totally convinced there is nothing wrong with the digital audio. I have proved this often by creating 32 bit floating wave files of sine waves that grossly exceed 0dBFS. Re-import them, drop the master fader, and you have a perfect sine wave again.

It's this notion that digital audio gets "harmed by running the summing bus too hot" or other ideas that I can't agree with at all.

I'm sure the signal gets harmed - i'm just sure that it's usually in the analog realm mainly. Or in plugins designed to emulate analog.

You can prove that digital audio exceeding 0dBFS isn't actually harmed, but using a spectrum analyser and looking at the harmonics. Because it's all in the digital realm, you aren't harming the waveform and there are no added harmonics. But your D/A converters might be saturating or clipping.

Digital audio data, and the reconstructed waves are two totally different things - and I thinks that very helpful to think about.

Seems to me that tracking as hot as you possibly can is a recipe for crunchy distortion - not the clarity of resolution it is supposed to achieve.

I just want to know where and why this occurs, so I can avoid it.
Old 14th June 2006
  #74
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Dynamic Range In Converters

This is from another post I made about the dynamic range. The 24 bit conveter has a theoretical range of 144dB, but the actual is much less because the last bits are pretty much trash. You don't lose the 30dB on the top end of the conveter, its on the low end. So when you record full scale, you get the 114dB dynamic range or whatever they spec the converter at, but when you record lower, say -30dB, you have lost the best bits. If your converter gets crunchy with a full scale signal, theres a problem somewhere and you should trouble shoot, as I say below. A sine wave should sound creamy smooth at full scale. Perhaps something before or after the converter is clipping.

"Remember the first bits are your most accurate and important. Each bit gets less accurate. Once you reach the 19th bit on a current 24 bit converter, you are starting to be just noise and distortion. The last 5 bits are close to useless for anything except dither. So if you throw away 30dB of the most important bits, you end up with an "theoretical" 19 bit converter. BUT, since the last 5 bits are questionable like I said, NOW YOU HAVE TURNED YOUR 24 BIT CONVERTER INTO A GOOD 14 BIT CONVERTER. What a waste!

Record right up to Full Scale if you can. The digital converter is best there. If you hear distortion or some degradation there, you need to troubleshoot."
Old 14th June 2006
  #75
Lives for gear
 

Quote:
Originally Posted by Dave Derr
This is from another post I made about the dynamic range. The 24 bit conveter has a theoretical range of 144dB, but the actual is much less because the last bits are pretty much trash. You don't lose the 30dB on the top end of the conveter, its on the low end. So when you record full scale, you get the 114dB dynamic range or whatever they spec the converter at, but when you record lower, say -30dB, you have lost the best bits. If your converter gets crunchy with a full scale signal, theres a problem somewhere and you should trouble shoot, as I say below. A sine wave should sound creamy smooth at full scale. Perhaps something before or after the converter is clipping.

"Remember the first bits are your most accurate and important. Each bit gets less accurate. Once you reach the 19th bit on a current 24 bit converter, you are starting to be just noise and distortion. The last 5 bits are close to useless for anything except dither. So if you throw away 30dB of the most important bits, you end up with an "theoretical" 19 bit converter. BUT, since the last 5 bits are questionable like I said, NOW YOU HAVE TURNED YOUR 24 BIT CONVERTER INTO A GOOD 14 BIT CONVERTER. What a waste!

Record right up to Full Scale if you can. The digital converter is best there. If you hear distortion or some degradation there, you need to troubleshoot."
Interesting discussion! First of all, the input signal that is converted looks different depending on what converter you use, so it's not only a matter of how the signal is stored in the samples or the signal level that determines the quality. The pre stages before the actual conversion is very crucial. Also, the dynamic range is much determined by both the AD chip in itself and by the jitter of the crystal reference clock. There is a huge difference in quality between converting a signal at high sample rates with a good op amp, a good AD chip and a stable crystal reference clock compared to using a lower sample rate, a worse op amp, a slightly worse AD chip and a more inaccurate crystal reference clock. It's of course very crucial to be able to record the signal as hot as possible but so that the signal is damaged as little as possible when doing so, because we want the signal to end up as accurate as possible. Make sure that the signal integrity is high when you record hot and back off on the input signal level when that's not the case. Also make sure that the input peak meter you are using is registering the signal in an efficient manner so that there are no transients passing the clipping point without the peak meter registering them. Leaving a few dB of safe input signal headroom even though the signal is clean when it is input hot I think is a good practise in general. I think my RME Fireface 800 converter performs the best when the signal is peaking somewhere around -6dB on 44,[email protected], but I'm sure I could run the input signal even hotter with a better converter (maybe even to full scale) and end up with a much better signal. This is of course very crucial for the rest of the digital processing, since the quality of the first input determines the quality of the last final output.

Another thing that I think is important when it comes to signal is the compression. Letting all the compression be done by digital plug-ins I think is not a good idea. When we limit we also make sure that the instruments are not falling apart. The kick drum is one example, we want it quite loud so that the punch and depth is audible and the mix feels big, but if the mix should be quite loud as well there is a chance that the kick drum will be overcompressed, which in combination with a bad digital peak limiter can make the sound quality very bad, especially when we try to compensate by cutting the low bands on the peak limiter. I think this is a problem with many bad sounding mixes. The tracks that take up a lot of dB in the mix should be treated very gently in order for the mix signal to stay high. Compensating with reverb on the snare and kick drum due to a too dynamic input signal is a recipe for mix signal loss. We tend to add reverb quite late in the mixing process and need to add more reverb when the dynamic range is high in order to make it audible in the mix, much also because less of the natural room is present in the signal. For these reasons I think it's important to have each instruments dynamic range pretty much under control as early as possible, so that the dynamic range of the mix will not be very unbalanced and so that we don't set the effects too wet, lose signal and end up with a mix that sounds flat and unnatural in the end. We can of course control this by multi band peak limiters on an M/S mastering setup, but then it's already a question of compromise.

There are a lot of critical steps where we can lose a lot of signal. Knowing about these and controlling them is very crucial, especially when we compare our mixes with the best... The problem with bad sounding mixes is exactly this, lost signal.
Old 14th June 2006
  #76
Lives for gear
 
Nordenstam's Avatar
 

Hello!

Quote:
Originally Posted by Dave Derr
"Remember the first bits are your most accurate and important. Each bit gets less accurate. Once you reach the 19th bit on a current 24 bit converter, you are starting to be just noise and distortion. The last 5 bits are close to useless for anything except dither. So if you throw away 30dB of the most important bits, you end up with an "theoretical" 19 bit converter. BUT, since the last 5 bits are questionable like I said, NOW YOU HAVE TURNED YOUR 24 BIT CONVERTER INTO A GOOD 14 BIT CONVERTER. What a waste!

Record right up to Full Scale if you can. The digital converter is best there. If you hear distortion or some degradation there, you need to troubleshoot."
Can see perfectly where your train of thought is! Guess most people have had an idea that's how digital is working. Me too. Fortunately, digital isn't that bad. The bit depth sets the *noise floor*.

The 'resolution' doesn't go to half when loosing 6dB. Yes, the signal is less accurate, but you have to think about what that implies. The non-accuracy of quantization IS the noise floor!

Loosing a bit makes the noise floor come 6dB closer. As Paul Frindle likes to point out, there is no such thing as resolution in digital audio. There's frequency content from zero to samplerate/2. And noise, from truncation or dithering. The loss of accuracy when loosing bits does not affect the top, it's the noise floor that moves.

20 year old converters truly had an inherent noise problem. Todays crop of converters have better dynamic range than just about any signal that's possible to connect to them. Tracking at -10dB makes NO difference as long as the converter have 10dB more dynamic range than the source. If you have a 120dB range converter and a 100dB signal, tracking at -20 makes no difference. The only thing you're missing out on is accurate sampling of the source signal noise floor. That's of course just a waste of time and energy. There really is no reason to track above -10dB. Contrary, getting close to full scale have so many drawbacks that there is a lot of reasons to avoid it!

Rainbowstorm: jitter affects ADC, DAC and SRC, not buffered events like data transfers and processes like plug ins.


Andreas
Old 14th June 2006
  #77
Lives for gear
 

Quote:
Originally Posted by Lupo
Hello!
Rainbowstorm: jitter affects ADC, DAC and SRC, not buffered events like data transfers and processes like plug ins.
Hmm... My two cents. Jitter affects the "original" signal in the conversion procedure. A jittered signal calculated by plug-ins will of course produce a more distorted result. So there is an indirect relationship between jitter and plug-in performance. There's also a relationship between a) plug-in performance related to the implementation of plugins and b) the DAC. When the DAC+monitors+room is not reveiling the true signal there will be overcompensations that result in signal loss. These overcompensations can be things like cutting frequencies that the mastering engineer will boost, panning instruments that the mastering engineer will re-locate with his M/S setup, backing off instruments that distort, removing reverb that is set too wet, adding compression that should have been applied more isolated earlier and worst of all reducing track signal level that the mastering engineer compensates for...
Old 14th June 2006
  #78
Lives for gear
 
synthoid's Avatar
 

Quote:
Originally Posted by Kiwiburger
Yes i'm talking about the digital master fader - bringing a sinewave up to (and over) 0dBFS. When you clip your 24 bit fixed converter, the square wave clipping is extreme and very obvious. But i'm talking about the "browning" of the sound as you get close to 0dBF - especially with cheap converters (like consumer playback devices will have). My DAC-1 is pretty clean right up to 0dBFS. But my M-Audio Audiophile 2496 gets really crunchy as you get close to 0dBFS with a sine wave.

It may or may not be the op amp - but i'm convinced it's in the analog side of the converter.

...

Seems to me that tracking as hot as you possibly can is a recipe for crunchy distortion - not the clarity of resolution it is supposed to achieve.

I just want to know where and why this occurs, so I can avoid it.
Yes yes yes. This crunchy behavior on hot inputs corresponds to my experience as well. I wanted to ask whether I'm the only one who has noticed that irrespective of the math, there is a very audible improvement in some converters by tracking at say -12db.

It might well be that some converters do not perform as well near their maximum input voltage as they do at lower input voltages. In a way, you would hope that this is so, because as I tried to point out, it is not a nice thing to have the most acccurate (analog) input range of a converter right next to the point where it clips (on the digital side). It would be much better if the sweet spot of the coverter were say at -6dbFS or -12dbFS (that is, at the analog input that produced these values on the digital side -- it's hard to talk about this stuff without being excessively wordy), because then you really should look at the last 6db or 12db of the converter's output as headroom and not as part of its normal working range, IMO. Having no headroom on the analog side is a bad idea.

The argument that the converter is the most accurate at the input voltage that produces the number with the most bits ignores the fact that the converter is also an analog device, with analog properties.

-synthoid
Old 14th June 2006
  #79
Motown legend
 
Bob Olhsson's Avatar
 

Instead of jawing about theory, I'd recommend that people do some actual testing for themselves.

My epiphany came when I set up a songwriter friend with a digi 001 system and he turned up a few weeks later with bounced mix files that peaked at -25 because he couldn't figure out how to get any higher levels.

I was working on a mastering project that had been mixed by a great engineer on an SSL 9k using Lavry Blue converters. My friend's demos using the 001 with just an old U-87, a 414 and a 57 sounded bigger, ballsier and more like analog than 90% of what you hear today including the high-end project I was working on.

It's important to never think of digital audio in analog terms. It's numbers. Period. Digital clipping doesn't sound that bad provided you never try to calculate any additional signal processing math using those numbers. At that point you'd have more resolution with a 12 bit signal than the mathematical gibberish that digital clipping produces.
Old 14th June 2006
  #80
Lives for gear
 

Quote:
Originally Posted by Bob Olhsson
Instead of jawing about theory, I'd recommend that people do some actual testing for themselves.

My epiphany came when I set up a songwriter friend with a digi 001 system and he turned up a few weeks later with bounced mix files that peaked at -25 because he couldn't figure out how to get any higher levels.

I was working on a mastering project that had been mixed by a great engineer on an SSL 9k using Lavry Blue converters. My friend's demos using the 001 with just an old U-87, a 414 and a 57 sounded bigger, ballsier and more like analog than 90% of what you hear today including the high-end project I was working on.
seriously?

this sounds a LITTLE over the top. i remember using my old 001...no matter how hard or light i would hit the converters the sound was pretty lacking.....and now that i am using the lavry blues it sounds pretty damn nice hitting the converters hard OR light.

the whole mix peaked at -25? how hot were the individual tracks when they first hit the A/D?
Old 14th June 2006
  #81
Lives for gear
 
Nordenstam's Avatar
 

Back to topic:
Quote:
Originally Posted by Silver Sonya
I think I have come to the conclusion that very few people understand the concept of gain structure or headroom within DAW's. Furthermore, I think this is probably the number one problem happening in audio today.
IMHO, the problem is better seen as insensitivity to distortion. Both are different views of the same subject, but gain staging alone does not explain the seemingly total lack of care for overdriven nastyness. Be it digital or analogue.

People growing up today are used to intentional and unintentional signal overloads on most any kind of music. There's a reason they don't notice. It's expected to be there, a part of the music culture. Even concerts are smashed through auto-mastering gear in the racks and the movies and DVD's are following quickly. Where can a young one get non-distorted sounds without digging into boring adult music?

If people have half an ear and realize the damage, there's so much wrong or irrelevant information around that it's really hard to learn the basics of gain staging in the digital world. Even if people want to.

Everything at -10dB digital would be a simple guide that would work well in most scenarios. At least better than everything at zero! The software makers could try to enforce it, but you can bet people will refuse to use it unless they understand why.


Keep working on making this a less distorted world! =)


Cheers,

Andreas
Old 14th June 2006
  #82
Yes, Virginia, there is a "sweet spot"

I want to compliment everybody on managing to have a vigorous discussion without it devolving into flame wars. Obviously everyone here is trying to do what's best for the audio -- we're just trying to sort out what that is. I hope I can add something helpful, speaking as both a recordist and EE.

Back in the 16-bit days, I was in the same camp as Dave Derr: "Don't waste precious bits". I do mostly location recording, and I'd keep a peak-latching digital level meter on the ADC output during rehersals and log the maximum of every piece. Then I'd have recalculate the preamp settings to hit -6dBFS, which usually ended up being -3dBFS during the actual performance. I chewed my fingernails a lot in those days!

Now we have 24-bit converters (well ok, 19-21 bit converters!) and I worry a lot less about using every last bit. Partly that's because I now have more dynamic range than I can really use. Face it: there is no performance venue that has an acoustic noise floor 80dB below peak concert level. (And if you do find one, go back an measure it after the audience arrives!) Studios may be built to NC30, maybe even to NC20 if the budget was big enough, but the effective acoustic noise at the mix buss goes up by 3 dB every time you double the number of tracks. So in practical cases, the ADC's noise floor is not the real limit.

If you want to get the best performance out of a converter, the thing you need to understand is "spurious-free dynamic range". Run a perfect sine in and look for the highest output "spur" -- the highest frequency bin that shouldn't be there. The difference in amplitude between the signal bin and the next highest (distortion) bin is the spurious-free dynamic range (SFDR).

When you measure SFDR vs. input level (output level for DAC's), it turns out that converters really do have a sweet spot: a range of input (output) levels where they perform the best. There can be a whole variety of reasons for this, from the performance of the associated analog circuitry to fixed-point math limitations of the digital decimation (image suppression) filters. Different designs have different optimal input levels, but there's always a sweet spot. And guess what? It's not at 0 dBFS. Most manufacturers who publish SFDR specs do the measurements at -10dBFS. Can you guess why?

One common difference between state-of-the-art and typical "prosumer" converter sets is how wide the sweet spot is, or stated differently, how quickly things go to hell when you get the gain-staging wrong. There's a big difference between, say, the average MOTU box and something from Benchmark, even if both products happen to use the same converter chip. One thing that distinguishes designs using the same converter chip is the quality of the analog parts of the signal path. Usually, the difference is most apparent with really hot inputs. My LynxII cards have a mixer applet whose meters turn red at -16dBFS. Looking at the data sheet reveals that they want a nominal +4dBu input level, which leaves 16dB of headroom to the +20dBu specified max input. (Bob Katz would be proud -- it's K16!). Sometimes I run them a bit hotter than that, but they do not like to be spanked.

Yup, gain staging still matters. It's starting to matter a bit less in the digital domain as more applications move to floating point math. In Sequoia, it doesn't seem to matter much what I do, as long as I'm not using third-party plug-ins. But back when I was mixing on a Yamaha 02R, it mattered a heck of a lot. I used to finish a mix, then go back and laboriously optimize the gain through every stage and summing buss, entering fader offsets to compensate. That was a 32-bit fixed-point system and it was still worth the trouble to optimize the gain staging. So if you have a DAW or plug-in that uses fixed-point math, be careful! Be especially careful if you're using very high-Q boosts or cuts. It's easy to clip internal nodes of narrow-band equalizers, even when the input and output levels seem reasonable.

Before I shut up, let me suggest an experiment you might like to try. Take a fairly squashed mix on your workstation and burn CD tracks at 0 dBFS and -10 dBFS. Everything else should be identical. Then mark your monitor level knob so you can play back both tracks at the same level. I predict you'll hear a difference on some material, even if you mixed in a floating-point workstation. Wanna know why? Here's a hint: What kind of math is used in the DAC chips inside your CD player?

David L. Rick
Hach Company (the day job)
Seventh String Recording (sleepless nights)
Old 14th June 2006
  #83
Lives for gear
 

Quote:
The 24 bit conveter has a theoretical range of 144dB, but the actual is much less because the last bits are pretty much trash. You don't lose the 30dB on the top end of the conveter, its on the low end. So when you record full scale, you get the 114dB dynamic range or whatever they spec the converter at, but when you record lower, say -30dB, you have lost the best bits.
I don't think you're really losing anything unless you get down to the noise floor of the concerters. If you have 114 dB dynamic range availalble and are peaking at -30 dB you've still got 84 dB of usable dynamic range (or, as was just mentioned maybe 74 to so you're at least 10 dB above the noise floor of the converters). Chances are you're actually using closer to half of that, but even if your signal has, say, a dynamic range of 50 dB you're still not "losing" anything. The six bits (or so) that are used to "describe" that signal will be just as accurate whether they're the bits that cover 0 dBFS to -50 dBFS, or -30 dBFS to -80 dBFS, or -50 dbFS to -100 dBFS. As long as you're at a level where you can't hear the noise floor of the converters you're fine, and in most cases even the ambient noise will be high enough (relatively speaking) that it won't be an issue. All of those extra "steps" of resolution are just used to describe lower-level noise that's below the noise floor anyhow, so you're not losing anything.

Of course, there are plenty of other reasons not to peak at -30 dBFS or -50dBFS, but losing bits or resolution isn't one.

Quote:
So if you throw away 30dB of the most important bits, you end up with an "theoretical" 19 bit converter. BUT, since the last 5 bits are questionable like I said, NOW YOU HAVE TURNED YOUR 24 BIT CONVERTER INTO A GOOD 14 BIT CONVERTER. What a waste!
It's only a problem if your signal dips down to the levels of the "marketing bits"...if the converter's designed right all of the "good" bits should be just as good as each other.

-Duardo
Old 15th June 2006
  #84
500 series nutjob
 
pan60's Avatar
 

if folks are hearing what appears to be a better sound closer to 0 then is that sweet spot something that is being introduced the the analog circuitry and is it possibly coherent only to a particular converter.
so far i think i will stick with the -10 has a general rule, not out of knowledge but ignorance.
Old 16th June 2006
  #85
GREAT thread. this is seriously important stuff to know.
Quote:
Originally Posted by Bob Olhsson
he turned up a few weeks later with bounced mix files that peaked at -25 because he couldn't figure out how to get any higher levels...sounded bigger, ballsier and more like analog than 90% of what you hear today including the high-end project I was working on.
i've experienced this on a few occasions. once when making rough demos on my friend's cool edit pro system, and once when mixing a project in pro tools that had been transfered from tape. in both cases the waveforms looked like silence, but the sound was fantastic. why? i only wish i knew. could some converter designers chime in here? maybe the name of this thread should be changed to something that would get their attention? if we can get some definitive answers, this might be worthy of a sticky.
Old 21st June 2006
  #86
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Resolution Vs Audible Quality

I suspect that most people won't hear much (or any) quality difference between good 24 bit quality and good 14 bit quality in a final product, except in the very low level stuff and the noise floor. Many of you are thinking Im crazy, but it would be interesting to take a good mix and truncate it down to 14 bits and do a blind ABX test against the 16 bit version. My recurring points about recording close to full scale are about getting the most resolution out of a given converter. Whether one hears the extra resolution or not really depends on dozens of factors. But theory, measurements, and listening tests should all show the best dynamic range and lowest distortion will occur with signals recorded close to 0dBFS.

The original Mitsubishi and Sony converters worked more closely to 14 bit resolution than 16 bit resolution, and yet there are many great records made on them. Again, I think the big musical things like songs, performance, arrangements, Eq, compression, balances etc grossly overshadow converter technology or performance. Without noise reduction, the best analog tape decks only had a dynamic range of 70dB, which is less than 12 bit resolution. However, Id prefer the tape due to its "pretty" non-linearities. With Dolby SR, the dynamic range of tape goes up over 100dB if my memory serves me correctly. We once put dolby SR on a Sony F1 digital recorder and I remember it sounded awful nice.

With the new converters, I wouldnt worry so much about what the converter is doing as much as all the big musical things. A little EQ or compression will probably have much more affect on the sound than whether you record -10dBFS on this converter or that converter.

I remember laughing a few years ago at a famous musician, saying he would only record on high def digital, even though his guitar sounds and vocal abilities were some of the raunchiest in the rock n roll world. Most of his singing is out of time and out of tune, and his guitar still sounded like razor blades on the "high-def" recording. Worrying and Nit picking about the technology rarely gets anyone very far as near as I can tell.

PS. NO, Im not going to tell anyone what musician Im referring to!
Old 21st June 2006
  #87
Quote:
Originally Posted by Dave Derr
I suspect that most people won't hear much (or any) quality difference between good 24 bit quality and good 14 bit quality in a final product, except in the very low level stuff and the noise floor. Many of you are thinking Im crazy, but it would be interesting to take a good mix and truncate it down to 14 bits and do a blind ABX test against the 16 bit version. My recurring points about recording close to full scale are about getting the most resolution out of a given converter. Whether one hears the extra resolution or not really depends on dozens of factors. But theory, measurements, and listening tests should all show the best dynamic range and lowest distortion will occur with signals recorded close to 0dBFS.

The original Mitsubishi and Sony converters worked more closely to 14 bit resolution than 16 bit resolution, and yet there are many great records made on them. Again, I think the big musical things like songs, performance, arrangements, Eq, compression, balances etc grossly overshadow converter technology or performance. Without noise reduction, the best analog tape decks only had a dynamic range of 70dB, which is less than 12 bit resolution. However, Id prefer the tape due to its "pretty" non-linearities. With Dolby SR, the dynamic range of tape goes up over 100dB if my memory serves me correctly. We once put dolby SR on a Sony F1 digital recorder and I remember it sounded awful nice.

With the new converters, I wouldnt worry so much about what the converter is doing as much as all the big musical things. A little EQ or compression will probably have much more affect on the sound than whether you record -10dBFS on this converter or that converter.

I remember laughing a few years ago at a famous musician, saying he would only record on high def digital, even though his guitar sounds and vocal abilities were some of the raunchiest in the rock n roll world. Most of his singing is out of time and out of tune, and his guitar still sounded like razor blades on the "high-def" recording. Worrying and Nit picking about the technology rarely gets anyone very far as near as I can tell.

PS. NO, Im not going to tell anyone what musician Im referring to!
Hi Dave.

So I actually want to know the answer to this question, I am not calling you out or anything so it's all good.



How would you explain this from Bob O above then?

Quote:
Instead of jawing about theory, I'd recommend that people do some actual testing for themselves.

My epiphany came when I set up a songwriter friend with a digi 001 system and he turned up a few weeks later with bounced mix files that peaked at -25 because he couldn't figure out how to get any higher levels.

I was working on a mastering project that had been mixed by a great engineer on an SSL 9k using Lavry Blue converters. My friend's demos using the 001 with just an old U-87, a 414 and a 57 sounded bigger, ballsier and more like analog than 90% of what you hear today including the high-end project I was working on.
Slipperman feels about the same way and honestly I have reached pretty much this same conclusion in my own experience as well. Math is fine but my ears tell me that having lower levels with digital gives me a better finished product.
Old 21st June 2006
  #88
Lives for gear
 

I have to disagree with the guy who's here saying "tracking as close to digital 0 or full scale as you can" gives better sound. In my experience that is simply not true.

Clients will bring in tracks that peak at -1 and immediately on loading the session I can hear a "digital hardness" that's quite different from what I normally hear when I track. It makes mixing to a good result more difficult.

Bringing levels down in most digital daws when tracking and mixing simply sounds better. You can offer all of the math theory you want, I know what I hear. I know what my clients hear. I know that when I started doing that my overall studio sound quality jumped noticably.

Why do people spread misinformation like this based on "theory" when in actuality all you have to do is try it? If you try it and find it not true then we'll agree to disagree. I don't think that'll happen though. Unless you actually try it you'll never know.

Look... I still own a d8b. I don't mix on it anymore (use it to monitor), I mix in SX which is easier overall for me, but I _can_ produce a great mix on a digital console that many people will say "sounds like ****". I have tracked though it to -10 peak on the input meters... as Mackie recommended ITFM. It "softens" the console sound considerably and it portrays the third dimension much better. Assuming you do the same with the mix bus and don't shoot for 0.

Track and mix too hot on it? It sounds like ****... even with good outboard converters... so do most daws in my opinion... they sound worse with really hot levels.

The guy who mixed all the audio for the network show "24" used a d8b for the first two years... and won industry awards. Proper "gain staging" is key in digital. If you can't hear the difference (in the end result of a daw session and mix) between tracking and mixing at -1 vs. -8/-6 or whatever in a modern daw then (...respectfully...) I can't help you ... you need better monitors or better ears. I hear it clearly through my $300 Alesis Monitor Ones (which I've grown to love, know like the back of my hand and cannot part with heh ).

People with Adam's will hear it even more.

Point? Doing something that simple makes a digital console that many consider "unusable" quite usable and capable of great results. You should do the same with your daw, it can only sound better. Forget limiting on the way in to get to -1, leave plenty of headroom and track what you hear to a lower peak level.

I treat SX (or any daw) the same way... -10/-6 ... it's fine, you don't lose anything. Actually ... you do lose something, a little harshness and some headaches. I track using conservative levels and mix to peak at about -6, -4 at the max. It sounds great, plugins don't overload and distort, the mix bus doesn't fold up and my clients are very happy with the results. It has more depth and better 3d imaging.

I simply don't care if another studio's mixes are 8db louder than mine. I mix for "good" not for "loud". Let the ME make it loud or you can experiment with "loud" after it's done... before you print if you want. Your _first_ job is to make it "good". More conservative levels simply sounds better. Not an opinion, this has been demonstrated over and over again by many people who've tried it.

People continue to dispute that, many who never tried it for (say, ... a month) and then compared the results to previous work where they "shot for close to 0". It will leave you convinced.

If I want to "saturate" an analog device on the way in for "that" sound? I do that and then turn the output level of the last analog device (before the AD) down to get my digital input peaking at about -6. For me, that's proper gain staging in the real world of digital daw recording (and for others who can hear the real world difference ) ... not shooting for "as close to 0 as possible".

It sounds great at mixtime.

Lawrence
Old 21st June 2006
  #89
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Subject run into the ground!

Sheesh I guess I'm done here. If a converter's manual says to leave 10dB of headroom or it will clip, then they are doing something very strange. But Im pretty much done repeating myself over how to get all the bits of resolution out of your converter. That is simply how converters work. To get 20 bits of resolution, you must go all the way to full scale on a 20 bit converter. If people find some sound better at a lower level, thats fine, just realize that you ARE NOT GETTING THE RESOLUTION SPECIFIED BY THE CONVERTER. THAT'S AS UNDENIABLE AS THE LAW OF GRAVITY.

Ill also repeat that recording down 6 to 10dB below clipping will not ruin any normal recording made on a modern converter, but if the tonality changes going to full scale, there is something wrong somewhere. Perhaps some converter boxes are poorly designed and can't actually handle full scale cleanly, but I certainly wouldn't put my name on any converter like that!

Unlike analog tape, converters are digitally filtered with "non-dynamic" filters whose frequency and tonal response are constant at ALL AMPLITUDES. Lastly, I suppose its also possible that when a mastering engineer "normalizes" a low amplitude mix, the normalizing process changes the frequency response... but that is also REALLY SCARY TO ME!

Use your ears when you are recording, but trust me when I say that if a converter is working with -6dBFS (or even lower) peaks, it is not working at it's designated resolution. Like buckling up... IT"S THE LAW!
Old 21st June 2006
  #90
Lives for gear
 
John Suitcase's Avatar
 

I did a session yesterday using my mobile rig, which is pretty "lo-end," A Soundcraft F1 mixer, through my MOTU 828(the old one), into Cubase SX. I used considerably lower input levels, though in SX I was still peaking close to 0db, but well below the red, which doesn't show up until you hit +6db, I think. Anyway, I felt that the clarity I gained on drums was significant, although the OH and Room mics sounded a little grainy to me. Their levels were very low, however, and seemed to a little harder to get up to a usable level.

As far as drums, though, I would say that I got much better transient response and "air" from the snare (SM57), than I had in previous sessions with this band. The Bass (recorded DI through a GT Brick) also seemed to benefit, having a more even tone, and better bottom end. Vocals sounded good, although I did overdrive something in my analogue side (probably the cheap-o Berhhy compressor I was using.)

All in all, I'd say that it makes the biggest difference on things like drums, where the initial transient is almost surely getting overdriven at the A/D when you record at higher levels.

I intend to continue experimenting, I'll be ordering a Echo Audiofire 12 soon, should be an interesting change from the 828.
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
thejook / So much gear, so little time
15
Ziggy!! / So much gear, so little time
12
DeadPoet / So much gear, so little time
1

Forum Jump
Forum Jump