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 25th June 2006
  #151
Lives for gear
 

Quote:
Originally Posted by Mike Caffrey
What an odd tone to take in a public discussion that so far has been civilized, interesting and informative.

It's also especially odd that you're putting words in my mouth when everyone else is seeing the same posts and there were none where I said "I have have no interest in trying.....".
Sorry man, this was getting very frustrating for me. I didn't say "you had no interest in trying it" I said "IF" you had no interest in trying it. Two different things.

Let me sum up this conversation....

THEM: Always record as close to 0 as possible.

ME: Recording at lower levels if fine, leave some headroom in your daw. You don't lose any audible detail. I, as a rule, record at -6/-10 and my mixes have improved since I started doing that. Not sure why.

THEM: Yes you do lose detail. Converter design is written in stone. The more bits you use the more detail.

ME: Uh... ok ... fine, I get that. So you've recorded using my method and noticed an audible lack in detail? I cannot hear it. Is it actually audible with modern 24 bit converters?

THEM: Well, no I haven't. (Note: This is a reasonable assumption because none of these guys ever said "yes I have' to that question even though I asked MANY times) But converter design is written in stone. And I'm a pro. You will lose detail.

ME: Well, I can't hear it. YOU may hear a loss of detail. I don't. Let me know after you actually try it. I''m a pro too. I've done it both ways. Can you give it a try?

THEM: (other converter stuff / theory...) ..many years ago I remember some low level stuff not sounding so good...

ME: ARGH! Can you just try it sometime in the present? It may not work for you. I'm just curious as to what might happen. Can you try it?

NOTE: This is the point where I say "IF" you (not you specifically but the people who keep claiming a quality loss) have no interest in actually trying it what's then point?

Not a single person who's supporting this "lost detail" scenario has come back and said... "Yes, I did an entire session tracking at -10/-12/-15 and YES it seemed to lose detail compared to my previous work."

I'm not talking about A/B tests here I'm takling about making the kind of subjective quality determinations we make in the studio every day... this plugin's comp is better etc. We don't scientifically A/B that stuff we go with what we hear during the course of working.

That's all I was asking, record some tracks, mix them and let me know if you lose any quality by leaving some headroom.

Anyway, here's two 24 bit wav short samples of drums recorded from a Korg X5, you tell me which was recorded at 0.01dbFS and which was recorded at 27dbFS? Don't zoom in look at them in a daw first, just listen to them and see how they sound.

www.theaudiocave.com/GSTest/File1.wav
www.theaudiocave.com/GSTest/File2.wav

On one of them I added (the maximum) gain of 24db that SX allows (without moving a fader) and the other I had to bring down a few db (again in that audio segment) so they would be close in apparent volume.

No scientific A/B just a general comparison... tell me which one you think lacks detail or resolution and I'll tell you which is which.

Playing them back in my studio they sound almost identical to me, with any difference being the un-calibrated level matching. I don't hear any additional noise or lack or detail or snap or fatness or whatever.

But of course this doesn't prove anything... which is why I've been asking people to try it for themselves in the real world and see what happens. As professionals you'd think we'd be more open to that kind of thing.

Many people are so busy "defending" their way of working that they miss an invitation to simply try something new. I never said tracking up to 0 is "wrong". I said it's wrong for me. I've been inviting you to try what I do and give me feedback.

I said I disagree with telling other people (universally) to track up as close to 0 as possible. I still disagree with that. Maybe an addendum? something like ...

Quote:
The way converters are designed tracking as close to 0 as possible gives you the best possible resolution in your audio. However other factors may come into play during real world tracking and mixing.

You may find, due to those other factors, that leaving lots of headroom in the daw is a better approach with no audible loss of quality. Or not. Try both and decide for youself.
As a professional, that's the advice I would give.

Lawrence
Old 25th June 2006
  #152
500 series nutjob
 
pan60's Avatar
 

so if i am on track so far,
1 track with some compression i you do not, i have not.
2 hit as close to 0 as possible.
3 bad the signal if ITB plug-ins are use, ( other then PT-LE ) or confirm how the plug-in handles the math.
question
dose lowering the slider on the ITB console actually pad the signal going to a plug-in or dose this have to be accomplished some other way?
as in are the faders post or pre, i would assume post.
i know dumb question but my brain hurts.
Old 25th June 2006
  #153
Lives for gear
 

Thanks for your long response and those sound clips! Currently I have no computer speakers available to test them on.

This is my real world experiences about the signal level/bit resolution thing. I use an RME Fireface 800 converter. The tricky thing about this unit is that the converter clip indicator and the digital software meter clip indicator (set at 3 samples) show different things. I asked RME about this and they didn't know which one was more correct. They didn't know if the op amps in the converter generated a lot of distortion close to the clipping point. So for safety reasons I've calibrated the digital software meter to 1 sample just to make it as sensitive as possible about clipping. I then record the signal as hot as possible and don't care about the clip indicator on the box, all I care about is the digital peak meter not going to red. I used to record at -6dB but since I noticed sound degradation slightly below that level and always used the digital peak meter calibrated to 3 samples, I decided to give full scale a second chance with the digital peak meter calibrated to only 1 sample. Looking back now I'm not sure if the converter turns really harsh close the clipping point, it might also just have been the digital signal that clipped regularly for one or 2 samples which caused the harshness when a lot of tracks were recorded. But when I'm recording like this I need to trim the signal depending on what plug-ins I use. It feels good to know that I have a full scale signal available when I need to.

But generally, when it comes to the input signal, for me it's not so much about the peak signal, that doesn't really describe what the dynamic range is like, since the material might be very unbalanced in terms of dynamics in different parts of the songs. I focus more on trying to get the peak to average good, in other words trying to record the signal so that it's as easy as possible to balance the whole mix and still have a good enough dynamic range. That might include for instance using a little compression and riding the trim knob during recording as well as doing a few takes of each thing and later choose what I prefer. So instead of focusing on the peak level I focus much more on tracking such that I know that it will be easy to balance the track in the mix. I focus on the balance. For me it's much more important to be able to have a good overall track balance than to have extreme mix volume and the transients hitting exactly just under the clipping point. But these days, as a general rule I track to full scale and make sure I can trust the peak meter.
Old 25th June 2006
  #154
Lawrence, great post.

To sum up what is said here YMMV but you don't know anything unless and until you try it for yourself.

If anyone came on to this board as said "I know this preamp or compressor or what have you sounds best because it has the best specs" they would be tossed on their ear. How is that different than saying the math says I should record as close to 0dbfs so that sounds better? Until you try it you are speaking "specs" not "sonics". Try it both ways then come back and report findings.

As I said before on this thread, for me recording lower levels just sounds better. If that works for you great, if not great.... but we spend so much time trying this micpre and this cable why is it bad to try a few tests with lower levels vs. higher levels. It is just common sense to me....

Quote:
Originally Posted by Lawrence
Sorry man, this was getting very frustrating for me. I didn't say "you had no interest in trying it" I said "IF" you had no interest in trying it. Two different things.

Let me sum up this conversation....

THEM: Always record as close to 0 as possible.

ME: Recording at lower levels if fine, leave some headroom in your daw. You don't lose any audible detail. I, as a rule, record at -6/-10 and my mixes have improved since I started doing that. Not sure why.

THEM: Yes you do lose detail. Converter design is written in stone. The more bits you use the more detail.

ME: Uh... ok ... fine, I get that. So you've recorded using my method and noticed an audible lack in detail? I cannot hear it. Is it actually audible with modern 24 bit converters?

THEM: Well, no I haven't. (Note: This is a reasonable assumption because none of these guys ever said "yes I have' to that question even though I asked MANY times) But converter design is written in stone. And I'm a pro. You will lose detail.

ME: Well, I can't hear it. YOU may hear a loss of detail. I don't. Let me know after you actually try it. I''m a pro too. I've done it both ways. Can you give it a try?

THEM: (other converter stuff / theory...) ..many years ago I remember some low level stuff not sounding so good...

ME: ARGH! Can you just try it sometime in the present? It may not work for you. I'm just curious as to what might happen. Can you try it?

NOTE: This is the point where I say "IF" you (not you specifically but the people who keep claiming a quality loss) have no interest in actually trying it what's then point?

Not a single person who's supporting this "lost detail" scenario has come back and said... "Yes, I did an entire session tracking at -10/-12/-15 and YES it seemed to lose detail compared to my previous work."

I'm not talking about A/B tests here I'm takling about making the kind of subjective quality determinations we make in the studio every day... this plugin's comp is better etc. We don't scientifically A/B that stuff we go with what we hear during the course of working.

That's all I was asking, record some tracks, mix them and let me know if you lose any quality by leaving some headroom.

Anyway, here's two 24 bit wav short samples of drums recorded from a Korg X5, you tell me which was recorded at 0.01dbFS and which was recorded at 27dbFS? Don't zoom in look at them in a daw first, just listen to them and see how they sound.

www.theaudiocave.com/GSTest/File1.wav
www.theaudiocave.com/GSTest/File2.wav

On one of them I added (the maximum) gain of 24db that SX allows (without moving a fader) and the other I had to bring down a few db (again in that audio segment) so they would be close in apparent volume.

No scientific A/B just a general comparison... tell me which one you think lacks detail or resolution and I'll tell you which is which.

Playing them back in my studio they sound almost identical to me, with any difference being the un-calibrated level matching. I don't hear any additional noise or lack or detail or snap or fatness or whatever.

But of course this doesn't prove anything... which is why I've been asking people to try it for themselves in the real world and see what happens. As professionals you'd think we'd be more open to that kind of thing.

Many people are so busy "defending" their way of working that they miss an invitation to simply try something new. I never said tracking up to 0 is "wrong". I said it's wrong for me. I've been inviting you to try what I do and give me feedback.

I said I disagree with telling other people (universally) to track up as close to 0 as possible. I still disagree with that. Maybe an addendum? something like ...



As a professional, that's the advice I would give.

Lawrence
Old 25th June 2006
  #155
Lives for gear
 
kurt's Avatar
I’m with you on every line you wrote Lawrence. First time during the mix when I tried to pull down all faders 20dB down except master & group masters & pulled up monitor level +20dB, the whole new word of cleanness opened to me. Since then I’m sold.

Well put Lawrence.
Old 25th June 2006
  #156
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Good Post Lawrence

At - 10 or -6 into a 24 bit converter, I don't suspect there are any real world signals that you would hear any difference in. I myself, never said you would hear loss of detail. Theoretically also, you should still have around 100dB S/N and .001%THD. This is incredible quality.

For one reason or the other, I have recorded many tracks more than -10dBFS into 16bit, and not really had anything remotely bother me. I did possibly lose 1.5bits of resolution but the source was probably more noisy than the 14 bits I recorded at. There's probably no way anyone would hear any loss, especially when its done and mixed. My original thread was about getting ideal performance out of converters, not what one actually heard, of course. It was a "theoretical" discussion, and theoretically, a good converter should get its "ideal" performance when signal peaks are recorded right up to 0dBFS. There shouldnt be any harshness or nastiness with this pure, full scale signal. What happens after conversion is again, up for grabs!

Do you use Protools HD perchance? ( I dont know what SX is).
Old 25th June 2006
  #157
Lives for gear
 

Quote:
Originally Posted by Dave Derr
Do you use Protools HD perchance? ( I dont know what SX is).

No HD. Cubase SX. Same thing (for most practical purposes) as Nuendo.

Lawrence
Old 25th June 2006
  #158
Lives for gear
 
midnightsun's Avatar
 

METER CHARACTERISTICS

Quote:
Originally Posted by Dave Derr

What happens after conversion is again, up for grabs!
Endless hours of trouble shooting has left me with the impression that different DAW meters are possessed with different poltergeists. Subjective case in point-- I used RADAR at 48/24 conversion, realtime out TDIF to TASCAM SX1. I became very comfortable and pleased tracking near zero per the SX1 meters. I take the same TDIF cables and plug them into a G5/Logic Pro and track using the same approach but watch the meters on logic. As I approacch zero things quickly go to hell in a hand basket.

My first thought is that the system isn't calibrated. I verify with test tones that the RADAR meters read exactly what Logic meters and SX1 meters read. I observe that the SX1 and RADAR meters give me the same feel as to how hot my signal is. I strangely observe that the RADAR and Logic meters give me a completely different feel. The Logic meter beacons me to run the signal hotter and I start getting garbage. The RADAR meters do not prompt me to run too hot. Again test tones reveal that I am getting identical signals to all meters.

Okay, I am a slow learner... but this is what I think is going on... I think that the meters are reacting at different rates to transients. When the meters start dancing during a "real tracking" situation they have different characteristics. Another possibility is that there is something happening to the digital audio signal after conversion that I just don't understand.

My solution is that I consider the meters to only be an extension of my ear. I use the meters as a tool to help guide me to getting a print that I am happy with. I don't feel married to the meters, bits, sample rate, or resolution. I realize that my approach lacks the science and depth of many who have contributed to this thread.
Old 26th June 2006
  #159
Motown legend
 
Bob Olhsson's Avatar
 

Quote:
Originally Posted by midnightsun
...Do you think that it follows that if a person is prone to record to close to 0 that an analog console might mitigate some of the mathematical gibberish?
Yes provided no DSP beyond the D to A filters is attempted and provided the console can handle the signal or you can trim or pad the DAC down to a level the console can handle. Some DACs also don't have enough analog or digital headroom above digital zero.
Old 26th June 2006
  #160
Lives for gear
 
midnightsun's Avatar
 

Quote:
Originally Posted by Bob Olhsson
Yes provided no DSP beyond the D to A filters is attempted and provided the console can handle the signal or you can trim or pad the DAC down to a level the console can handle. Some DACs also don't have enough analog or digital headroom above digital zero.

Gotcha, thanx
Old 26th June 2006
  #161
Quote:
Originally Posted by kurt
I’m with you on every line you wrote Lawrence. First time during the mix when I tried to pull down all faders 20dB down except master & group masters & pulled up monitor level +20dB, the whole new word of cleanness opened to me. Since then I’m sold.
do you mean you pulled up your analogue monitor level? if so (and disregard this if it's not so) depending on what you use as a monitor controller, what you may be hearing is your volume pot opening up
Old 26th June 2006
  #162
Lives for gear
 
kurt's Avatar
Err, not sure what you mean. I’m using a SPL 2380 Surround Monitor Controller to control my Dynaudio BM15 & Genelecs 1030. The system is calibrated to 83db.
The change I’m talking about is pulling potencjometer from 45% to 65% on it´s scale.
Why don´t you try for you self? You can use some of Bob Katz k-system recommendations. Just pull your Monitor level up & keep your faders where the listening level is at the usual listening level.
Old 26th June 2006
  #163
Quote:
Originally Posted by Miles Jackson
The above argument reflects a common misunderstanding about digital audio. Beneath the Nyquist frequency (sampling rate/2), more samples per sine wave do not create a more "accurate" reproduction of a sine wave. For instance, if you sample a 10 kHz sine wave at 44.1 kS/s, you'll only manage to get 4 samples per wave cycle; however, as Nyquist demonstrated, that is enough information to perfectly represent the 10 kHz sine wave. Sampling the 10 kHz sine wave at a higher sampling rate does not provide a more "accurate" representation of the sine wave. I know this is not intuitive, but that's the story. Blame Nyquist if you don't like it!

Some good resources to get a handle on this stuff: Dan Lavry's white papers at www.lavryengineering.com, Nika Aldrich's Digital Audio Explained for the Audio Engineer, Bob Katz's Mastering Audio.

Miles
And if you're not recording a sine wave? Most of what we record is far more complex than a sine wave.
Old 26th June 2006
  #164
In your paraphrasing, You've diluted you position and exagerated the other side.

You also haven't acknowledged that the two sides are really at the core of it arguing two separate points.

Quote:
Originally Posted by Lawrence
Not a single person who's supporting this "lost detail" scenario has come back and said... "Yes, I did an entire session tracking at -10/-12/-15 and YES it seemed to lose detail compared to my previous work."
There's no way I'd do that to a client unless I was nearly 100% certain that it was better.

Quote:
Originally Posted by Lawrence
I'm not talking about A/B tests here I'm takling about making the kind of subjective quality determinations we make in the studio every day... this plugin's comp is better etc. We don't scientifically A/B that stuff we go with what we hear during the course of working.
Where do you think my opinion came from? School? A book?

And I don't know about you, but I do A/B things. Like when I first got a 2 inch machine. I turned all the alignement screws while playing back so see what the actually did. I recorded guitars in 2dB increments from -20db to pinned. That's how I learned that I like the sound of tape, but not tape saturation.

I don't remember, and I'm not going to remember what the specific process was for making my decisions with digital. I was a long time ago.

So far you've said that recording as hot as possible wihtout crossing zero is fine but you prefer recording lower becuase you might run the signal though somehtign that doesn't work right and as a result it will sound bad, but the discussion wasn't about what the optimal level for recoding something that will be run through a plugin or even what's the optimal level to record for mixing (which is a reasonable assumption, but still an assumption). The discussion was a very micro discussion about recording/creating the file. That may be the source of some of your frustration.


Quote:
Originally Posted by Lawrence
That's all I was asking, record some tracks, mix them and let me know if you lose any quality by leaving some headroom.

Anyway, here's two 24 bit wav short samples of drums recorded from a Korg X5, you tell me which was recorded at 0.01dbFS and which was recorded at 27dbFS? Don't zoom in look at them in a daw first, just listen to them and see how they sound.

www.theaudiocave.com/GSTest/File1.wav
www.theaudiocave.com/GSTest/File2.wav

On one of them I added (the maximum) gain of 24db that SX allows (without moving a fader) and the other I had to bring down a few db (again in that audio segment) so they would be close in apparent volume.

No scientific A/B just a general comparison... tell me which one you think lacks detail or resolution and I'll tell you which is which.

Playing them back in my studio they sound almost identical to me, with any difference being the un-calibrated level matching. I don't hear any additional noise or lack or detail or snap or fatness or whatever.

But of course this doesn't prove anything... which is why I've been asking people to try it for themselves in the real world and see what happens. As professionals you'd think we'd be more open to that kind of thing.
No it doesn't, especially since you talking about a mix context, in which case you need multiple tracks to see what the cumulative effect is (which may do more towards making your point). And you have to comapre with multiple types of sources.


Quote:
Originally Posted by Lawrence
Many people are so busy "defending" their way of working that they miss an invitation to simply try something new. I never said tracking up to 0 is "wrong". I said it's wrong for me. I've been inviting you to try what I do and give me feedback.

I said I disagree with telling other people (universally) to track up as close to 0 as possible.
No one has been telling other people how to record. I don't remember the very original post, but the whole last part of this discussion came from someone asking Dave's opinion on the risk vs reward of coming as close to 0DBFS as possible.

Ok now, we're debate summaries.

Somone has started the discussion of metering. If you recording to a converter that gives you a pre conversion signal, you can't hear digital distortion. Now you're reltying on your meters. Are the fast enough? What if they're not perfectly accurate, maybe you're recording hotter than the meters lead you to believe (another agrument for recording with more headroom). Or, it could be lower.


Another thing I've heard only a little about was "peaks within peaks". It wasn't something explained adetuately, but it was an explaination for why to print mixes well below digitial zero. That might be interesting to hear about from someone who can explain it.

Until I'm convinced, through theory or repeated expereinces of hearing that reocrind with the most bits possible sounds worse than recording with fewer bits, I'm going to stick with the more bit method. And if listening to your file makes me think that there's no difference in sound, the I'll keep recording with the most bits possible because, why not? It doesn't sound any different.

When it comes to mixing, my gains staging is different so I have different considerations so the whole plugin head room discussion is irrelevant to me.
Old 26th June 2006
  #165
Lives for gear
 

I think I need to point out one thing in this context that you seem to forget about, the quality of the monitors. We are currently dealing with differences it takes a set of really great monitors to hear. I think it's easy to say that you should use 10 dB of peak headroom because you are using bad monitors (with soft clipping etc...). And with lower input signal you will automatically push a compressor with the right signal level to automatically create the most softness. In this case it's pretty obvious that you will think that the sound is better when it goes in for instance at -6dB, you might simply have forgot to calibrate the input signal for the effects when you were recording to full scale and that's why you think recording at full scale sounds worse. There is only one thing that I could come to think of would justify not recording at full scale, that would be if the op amps in the converter actually become really harsh near the clipping point. I can't comment about that, I've asked RME and they didn't know, they said "maybe".
Old 26th June 2006
  #166
Lives for gear
 

Quote:
Originally posted by Mike Caffrey:
The reason to care about bit rate is not so that we can have a song that goes from the acutal volume of a whisper to the acutal volume of a jet enginer from 6 inches away, the reason is that is sample is described more acurately with a longer digital word. If you could do a freeze frame on a single sample of audiio, a 64 bit description would sound far better than a 16 or 24 bit description.
I agree that there's no real practical value in recording something that would take full advantage of the dynamic range that an ideal 24-bit system would give us, but at the same time the accuracy that we get when our bit rate goes up is manifested entirely as an increase in dynamic range. All of those extra "steps" which are used to describe our signal in more detail describe detail that's lower in level. Each bit gives us 6 dB more in dynamic range, and that dynamic range goes from 0 dBFS downwards. If we're recording signals that are nowhere near our digital noise floor all that extra "resolution" is giving us is a lower noise floor...it doesn't change the accuracy of the signal we're capturing/reproducing at all. This isn't my opinion, it's the way digital audio works.

As for a 64-bit word sounding better than a 24-bit word, there's no reason that it would. Aside from the fact that taking a single sample is meaningless, we're not even able to capture a "true" 24 bits' worth of resolution, let alone 64. But even if we could, all it would do was describe stuff we can't hear in more detail anyhow.

Quote:
Originally posted by Mike Caffrey:
I think the term for what I'm describing is aliasing.
Aliasing is when you capture a signal that's above the Nyquist frequency, so when your D/A converter tries to reproduce it it winds up producing a frequency that was never there in the first place. Antialiasing filters in any good converter will prevent this from happening, and in any case it's related to sampling rate, not bit depth.

Quote:
Originally posted by Mike Caffrey:
If you sample a sine wave at fewer point in the wave, there are gaps. The more samples the fewer gaps and since the wave is curved, making it smoother makes it more round, but we can describe it as smoother or more accurate if you want.
As long as there are more than two points per wave, the sample taken will be just as accurate as if there were three, or ten, or a thousand, or whatever. If there are two or less, it just won't work. It's not intuitive, but that's the way digital audio works.

Quote:
Originally posted by Mike Caffrey:
Now, if your tlaking about wnat comes out and saying that the D/A converter inerpolates or extrapolates and fills in the gaps that that there aren't any gaps, fine. But which would you prefer gaps filled in by math or more an more samples to minimize the gaps even further?
Since the result is the same, I'll take either...and if it gives me more processing power and takes up less hard drive space, I'll take the lower sampling rate. But that's not really what we're discussing here.

Quote:
Originally posted by Dave Derr:
My original thread was about getting ideal performance out of converters, not what one actually heard, of course. It was a "theoretical" discussion, and theoretically, a good converter should get its "ideal" performance when signal peaks are recorded right up to 0dBFS. There shouldnt be any harshness or nastiness with this pure, full scale signal.
Actually, with a theoretically ideal converter, shouldn't performance be perfectly linear all the way from 0 dBFS down to the noise floor? Why would it be any difference? And aren't today's good converters consistent over most of their dynamic range?

I'd agree that there shouldn't be any harshness with a full scale signal, but there also shouldn't be any with a signal sampled anywhere within the "usable" dynamic range of a converter, and with a 24-bit converter that range should be fairly large. Again, there are plenty of reasons why you shouldn't record too close to either extreme, but as I see it one of the benefits of 24-bit audio is that you don't need to worry about getting as close to zero as possible...not because if you do it won't sound good, but if you do your chances of going over do increase significantly, in which case it will sound worse.

[quote]
Quote:
Originally posted by Mike Caffrey:
And if you're not recording a sine wave? Most of what we record is far more complex than a sine wave.
But none of what we're recording is more complex than a bunch of sine waves added together, which is why digital audio works.

-Duardo
Old 26th June 2006
  #167
Lives for gear
 
Nordenstam's Avatar
 

Hi!

Quote:
Originally Posted by Mike Caffrey
Another thing I've heard only a little about was "peaks within peaks". It wasn't something explained adetuately, but it was an explaination for why to print mixes well below digitial zero. That might be interesting to hear about from someone who can explain it.
There's a link to a great Nika Aldrich article in my previous post that wraps it up neatly.

The problem is two fold. Short summary:

The sample points always lands at some arbitary position on the waveform beings sampled. It's near impossible for the sample position be at the exact top or bottom of the waveform the instant it's sampled. That's fine, it's how digital work. But it does present a problem for metering. Looking at the level of each sample point does not tell much about the level of the actual waveform stored in those samples! A regular sample peak meter doesn't tell what's happening on the output of the reconstruction DAC. An oversampled peak meter, like RME implements in their interfaces, gives a true report of the peak level of the actual waveform stored in the samples.

Any signal with frequencies near simple ratios of the sample rate will show this in a striking manner. The attached image 'overshot5.gif' shows how the sample points varies with 3dB in the representation of a pure 11026Hz sinewave. I've snipped out 100 millisecs between those different looking parts of the sinewave, but it's the exact same wave being stored and reproduced on the output. A peak level meter would report a level oscillating 3dB in amplitude. An oversampled peak level meter would report the true constant level.


Taking another look at the overshot5.gif image leads to the next part of this problem. The left side looks like it's been squared off on the top and the right side looks like a triangle wave. Luckily, all harmonics are above half the sample rate! Lowpass filtering the signal at 20kHz will remove those harmonics, leaving a smooth sine wave as the end result. That's where it gets interesting.

Have a look at the attached image called overshot4.gif. The upper part is a square wave, the lower part is the exact same wave filtered down to the fundamental lone frequency, a sine wave. Look what happened to the peak! It's 3dB louder. Lowpass filtering will remove squared tops and replace them with smooth flowing waveforms. These will have higher peak level than the initial square wave. Squaring the tops of the waveform is common practice, primarily from clipping and limiting, but also from processes. Filtering on the output of the DAC will make those peaks go above the 0dB digital.

There's more to this than I remember right now. Please read the Nika article!

Oh, btw, this is a processing issue, not a recording issue. An ADC will not produce such illegal sample values on it's own.


Andreas Nordenstam
Attached Thumbnails
My Theory About Prosumer Audio-overshot4.gif   My Theory About Prosumer Audio-overshot5.gif  
Old 26th June 2006
  #168
The Distressor's "daddy"
 
Dave Derr's Avatar
 

Quote:
Originally Posted by Duardo
I'd agree that there shouldn't be any harshness with a full scale signal, but there also shouldn't be any with a signal sampled anywhere within the "usable" dynamic range of a converter, and with a 24-bit converter that range should be fairly large. Again, there are plenty of reasons why you shouldn't record too close to either extreme, but as I see it one of the benefits of 24-bit audio is that you don't need to worry about getting as close to zero as possible...not because if you do it won't sound good, but if you do your chances of going over do increase significantly, in which case it will sound worse.
Very true. Modern converters work great over a huge dynamic range, but you probably dont want to purposely record 15 or 20 dB down, because that is losing 3 - 4 bits of its performance, right off the bat. The thing thats scary is how unforgiving some DAWs and Filters are when processing signals that are close to Full scale. Im thinking we really need to take some people to task for not allowing plenty of headroom, while at the same time, not dropping useful bits and data. Just recently I noticed wierd clipping in an effects loop on my protools system. I ended up backing down the output of the effect even though I knew it was quite a ways from clipping itself. There is often no metering in some of our day to day DAW tasks.
Old 27th June 2006
  #169
Quote:
Originally Posted by kurt
Err, not sure what you mean. I’m using a SPL 2380 Surround Monitor Controller to control my Dynaudio BM15 & Genelecs 1030. The system is calibrated to 83db.
The change I’m talking about is pulling potencjometer from 45% to 65% on it´s scale.
Why don´t you try for you self? You can use some of Bob Katz k-system recommendations. Just pull your Monitor level up & keep your faders where the listening level is at the usual listening level.
all i'm saying is, it may be that what you're hearing is an improvement in the analogue gain staging of your monitoring chain. unless you have a stepped volume control, you're using a potentiognfijdfoiadnfmeter to attenuate the analogue signal, and the lower that pot is set, the more it's attenuating, and the worse it will sound, so turning up your monitor level will open up the pot, and therefore your mix will sound more open. i'm not sure if the spl has a stepped volume knob, but if not that may contribute to the "whole new world of cleanness" you're hearing. just a thought, that's all.
Old 27th June 2006
  #170
Quote:
Originally Posted by Duardo
I agree that there's no real practical value in recording something that would take full advantage of the dynamic range that an ideal 24-bit system would give us,
-Duardo
Very intertesting post, which I'm going to respond to separately. I want to digress for a moment about it's style.

You quoted a bunch of things I said, and one of them registered in a way that made me rethink everything. No conclusions yet, but there's an easy A/B I want to do now that will be very informative (for me at least). So thanks for that!

In the midst of all of that you dropped in a quote from Dave Derr and then followed that with another from me. When I Dave's quote, at first I though it was mine, but I don't remember thinking the though I was reading. so I searched and saw it came from Dave.

I'd say that's kind of a big deal, in a bad way. I don't think any harm was meant, so I'm not pissed or anything, but I do feel misrepresented and I'm sure I wasn't the only person who viewed Dave's quote as coming from me. So I'd say it's bad form to quote without indicating who it's from, especially when you're quoting multimple people in the same post.

Also, in addition to sort of putting words into my mouth, you also deny the author credit to their idea/contribution.

I hope I'm not starting another digression, but it's something you may want to think about, and I'd prefer more clarity if you're going to be quoting me.

I think it's a good habit to develop because little things like that can get blown way out of proportion, especially around here.
Old 27th June 2006
  #171
Quote:
Originally Posted by kurt
I’m with you on every line you wrote Lawrence. First time during the mix when I tried to pull down all faders 20dB down except master & group masters & pulled up monitor level +20dB, the whole new word of cleanness opened to me. Since then I’m sold.

Well put Lawrence.

That doesn't' really correlate to what Lawrence was talking about. He's talking about recording lower in the first place. You'll still be hitting you're plugins just as hard if they're on inserts pre-fader and when you pulled down all of those faders, you just made a big mathematical calculation. Digital faders aren't volume controls, they're math controls. Yes, you hear a difference in volume, but that fader is controlling a mathematical alteration to the number that represents you're audio.

Also, unless you were listening to each track one at a time soloed, you're also hearing a difference in how you hit the summing buss, either analog or digital.
Old 27th June 2006
  #172
If you record hot - all your DAW faders end up real low...

If you dont record so hot... your DAW faders end up being far closer to "zero" (up higher)
Old 27th June 2006
  #173
[QUOTE=Duardo]I agree that there's no real practical value in recording something that would take full advantage of the dynamic range that an ideal 24-bit system would give us, but at the same time the accuracy that we get when our bit rate goes up is manifested entirely as an increase in dynamic range. All of those extra "steps" which are used to describe our signal in more detail describe detail that's lower in level. Each bit gives us 6 dB more in dynamic range, and that dynamic range goes from 0 dBFS downwards. If we're recording signals that are nowhere near our digital noise floor all that extra "resolution" is giving us is a lower noise floor...it doesn't change the accuracy of the signal we're capturing/reproducing at all. This isn't my opinion, it's the way digital audio works.

As for a 64-bit word sounding better than a 24-bit word, there's no reason that it would. Aside from the fact that taking a single sample is meaningless, we're not even able to capture a "true" 24 bits' worth of resolution, let alone 64. But even if we could, all it would do was describe stuff we can't hear in more detail anyhow.
Quote:
Originally Posted by Duardo

So the true test of this is to set a level going into a converter with a dynamic rage acceptable for 16bit recording, record that signal at 16 bits, then change the session/project (ie protools/radar) and record the same thing at 24 bits.

Then A/B both though the same converters without altering either (don't convert one to fit in the other session).

It's kind of a difficult A/B for playback, but with what you're saying there should be no difference.


One thing that's happend a few times is a client will bring in some sessions they created at home. We'll track drums and what not to all of them and at the end of the day, one sounds somehow different than the others. Inevitably, when I check, It's 16bit and the others are 24.

I guess one possible explanation is that the dynamic range or the recorded content is more suitable for 24bit, but there's always a perceptable differnce in those cases.

Quote:
Originally Posted by Duardo

Aliasing is when you capture a signal that's above the Nyquist frequency, so when your D/A converter tries to reproduce it it winds up producing a frequency that was never there in the first place. Antialiasing filters in any good converter will prevent this from happening, and in any case it's related to sampling rate, not bit depth.


As long as there are more than two points per wave, the sample taken will be just as accurate as if there were three, or ten, or a thousand, or whatever. If there are two or less, it just won't work. It's not intuitive, but that's the way digital audio works.
I was translating from graphics, where sometimes the DPI is set too low and then edges are pixelated and jagged. Then when the rate is higher the line appears "smoother". I think anti-aliasing fills in the little corners and can acutally make a smooth edge. I don't know a lot about graphics, but it's something along those lines.

Quote:
Originally Posted by Duardo
But none of what we're recording is more complex than a bunch of sine waves added together, which is why digital audio works.

-Duardo
Maybe I'm wrong, but doesn't that comination of waves create a wave that's more complex than a single sine wave?
Old 27th June 2006
  #174
Lives for gear
 

Originally posted by Mike Caffrey:
Quote:
In the midst of all of that you dropped in a quote from Dave Derr and then followed that with another from me. When I Dave's quote, at first I though it was mine, but I don't remember thinking the though I was reading. so I searched and saw it came from Dave.

I'd say that's kind of a big deal, in a bad way. I don't think any harm was meant, so I'm not pissed or anything, but I do feel misrepresented and I'm sure I wasn't the only person who viewed Dave's quote as coming from me.
Sorry about that, there was certainly no harm intended. When I responded I just went ahead and responded to the statements I did in the order in which they were made. I've gone back to that post and indicated who said what.

Originally posted by Mike Caffrey:
Quote:
So I'd say it's bad form to quote without indicating who it's from, especially when you're quoting multimple people in the same post.

Also, in addition to sort of putting words into my mouth, you also deny the author credit to their idea/contribution.
Again, I apologize, and I certainly didn't mean to deny anyone anything. I just figured everyone who was reading had been reading along and would know who said what, but I see that it can be confusing. Sorry again.

Originally posted by Mike Caffrey:
Quote:
Maybe I'm wrong, but doesn't that comination of waves create a wave that's more complex than a single sine wave?
Yes and no...sure, a combination of sine waves is more complex than a single sine wave, but as far as Nyquist is concerned as long as the complex waveform contains nothing at or above half of the sampling rate, it is as simple as though it were a sine wave. All of the components of anything that's not (square wave, etc) will either be sine waves that are below the Nyquist frequency or will have been filtered out by anti-aliasing filters.

-Duardo
Old 27th June 2006
  #175
Quote:
Originally Posted by Duardo
Yes and no...sure, a combination of sine waves is more complex than a single sine wave, but as far as Nyquist is concerned as long as the complex waveform contains nothing at or above half of the sampling rate, it is as simple as though it were a sine wave. All of the components of anything that's not (square wave, etc) will either be sine waves that are below the Nyquist frequency or will have been filtered out by anti-aliasing filters.

-Duardo
Exactly. What he said.....

I am sure you know this but you might not have thought of it this way before.

Yes a combination of sine waves are more complex but think of how the converter sees it. It does not "see" the whole wave like we do in a DAW, it is really just seeing the snapshot of the wave compared to the wave snapshot that just preceded it.

If you take it down to that level it really does not matter how complicated the wave form is it just needs to have a voltage shift that is 1/2 of the sample rate or lower.
Old 27th June 2006
  #176
Lives for gear
 

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.

I think one of the reasons many so-called ITB mixes sound bad is for this reason. It's remarkable how often I work (as either a mastering engineer or a mixer) with someone who brings me a DAW session that is just shredding the input or the output of some plug-in or channel or buss. I'm not talking subtle distortion here. Im talking just bone-stupid clipping and harsh, spitty audio. For me, it's painful to even look at the meters, let alone hear the sound. Usually I'll say "Whoa!" and grab the digital fader or input and pull it down and the person will turn to me and say "Why did you do that?"

And I'll explain "Because you were absolutely searing the output of that buss. This is a DAW, you can't do that. It results in completely unmusical clipping and truncated resolution." This is usually followed by me giving about an hour long lecture on the rudimentary principles of gain structure and why it's important. Then the client says "Wow, I never thought about that."

Elegant gain management --- i.e. negotiating your way through the headroom available to you, either in voltage (analog) or data (digital) --- is the most fundamental aspect of recording.

I have a radical concept: what if Protools/Logic/Nuendo, etc. had a bult in alarm that would sound whenever the user overloaded the i/o of a channel, buss, or plug-in within the system? Or better yet: prevented them from doing it at all! Like maybe the fader just won't go as high as you want it to... or maybe at some point, turning the fader up on one channel simply results in all other channels being lowered?

Of course "real" engineers (either seasoned through experience or sometimes formal education) don't have this problem as much as musicians and home-hobbyist types.

The DAW is an unforgiving environment! I long for the days when the home hobbyist was using a Tascam 4-track cassette or maybe 8-track reel-to-reel and maybe a Tapco console. At least then, bad gainstaging at least resulted in something interesting: either analog distortion [too hot levels] or seas of hiss [too low levels].

Man, I'd take some hiss over clipping DAW levels any day!!

So what I want to know is this: is my DAW concept a bad idea? Or a stroke of genius. What if DAWs stopped the user from clipping?

Lemme know what you think.

Cheers,
Chad
If it's too easy for the amatures, IE. no incentive for them to "understand" the signal processing and flow in regards to level, then there's no room for us pros to come in and fix it for them (:
Old 27th June 2006
  #177
Gear Nut
 

Quote:
Originally Posted by Dave Derr
....The thing thats scary is how unforgiving some DAWs and Filters are when processing signals that are close to Full scale. Im thinking we really need to take some people to task for not allowing plenty of headroom,....
First, let me say I love your products and am a customer. Okay.....

I think you're being unfair in the above statement. Digital audio must have some limits by its nature. It was never meant to record conventional sources at "full scale." Many years ago, depending on location, it was designated (IIRC) that 0dBvu = -15 or -18 dBFS. That leaves plenty of headroom. That's where I always record. If I see single digit peaks, I back off a little. I can't hear any degradation either at 24 bits (and you didn't claim there was any, I know). Admittedly, there are a few plugs that depend on "drive" to get their sound, and then I have to raise the level with the plug's input gain. And, to concede to your comments somewhat, there are plugs that are sub par.

I was gratified to read Paul Frindle's comments on this forum and at PSW. They made me feel more secure in my conservative levels. Check out his comments if you haven't already.

Additionally, FWIW, I am in accord with Lawrence's comments in this thread regarding practices (with no negative insinuations towards you).

Thanks, Dave, and keep up the great work.

-Naren
Old 27th June 2006
  #178
Quote:
Originally Posted by not_so_new
Exactly. What he said.....

I am sure you know this but you might not have thought of it this way before.

Yes a combination of sine waves are more complex but think of how the converter sees it. It does not "see" the whole wave like we do in a DAW, it is really just seeing the snapshot of the wave compared to the wave snapshot that just preceded it.

If you take it down to that level it really does not matter how complicated the wave form is it just needs to have a voltage shift that is 1/2 of the sample rate or lower.
That makes sense. it's just looking a moments in time, it has no idea what's before or after.
Old 27th June 2006
  #179
Quote:
Originally Posted by Duardo
Originally posted by Mike Caffrey:

Sorry about that, there was certainly no harm intended. When I responded I just went ahead and responded to the statements I did in the order in which they were made. I've gone back to that post and indicated who said what.

Originally posted by Mike Caffrey:

Again, I apologize, and I certainly didn't mean to deny anyone anything. I just figured everyone who was reading had been reading along and would know who said what, but I see that it can be confusing. Sorry again.



-Duardo
And then I go and make a post that reverses our content!

**** happens.
Old 27th June 2006
  #180
The Distressor's "daddy"
 
Dave Derr's Avatar
 

DAWS AND FULL SCALE SIGNALS

Naren said: "Digital audio must have some limits by its nature. It was never meant to record conventional sources at "full scale."

It wasn't? Was there a law passed that I don't know about? ::wink::

And you'd better tell every mastering engineer in the world who puts the PEAKS on CDs right up against Full scale on every single record... at least within -.5dBFS.

Personally, I wouldn't even think of putting out a DAW that couldn't handle full scale signals... on every single friggin channel! By not handling full scale gracefully, DAW manufacturers are saying "Well you've got these great converters, but we won't let you get the best performance from them... and still work cleanly in our DAW system".

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