What levels do you record at in digital?

[Bob Savage]

How do you guys feel about compressing on the way in? Worth it/Not Worth it/Too risky?

Take this for what it is, the opinion of a hobbiest... but it depends on what instrument. Vocals seem to need it sometimes, especially with singers who have a wide dynamic range but little control over it. Clean guitar tracks seem to call for it at times too, depending on the amp. The clean channel on both of my Bogners (Uberschall and Shiva) have a huge dynamic range, and it's really easy to clip on the way in, so if I set the level where my highest peak isn't clipping, it leaves the rest of the playing too low.

 

[Benny H]

Are there 'hot to tape' sort of emulating compressors around these days? Or is that sort of gentle effect something you can do on any reasonable compressor?

What I'm talking about is just how you could afford a few hot peaks on tape cause they compressed 'naturally'. It would be nice if you had this safety all the time with digital.

 

[A Gruesome Discovery]

How do you guys feel about compressing on the way in? Worth it/Not Worth it/Too risky?

Definitely not- not if you're going for audio fidelity. Clamping down on dynamic range will always result in fewer bits used, and if your mission is to use as many bits as you can, you just failed (as I said though, I still think bits used means nothing in terms of fidelity). Any make-up gain also raises the noisefloor quite a lot.
As an effect or something, though, of course it's OK to use, but if your whole point is that you think
loudness = fidelity, and
compression = more loudness = more fidelity,
you'd be very wrong.

 

[EtherForBreakfast]

This is how I've always done it... For input level on mic pre, I will purposefully bash my strings very aggressively and play much harder than usual. I make sure I dont see any overs. This way, if I happen to go ape at a certain point of tracking, it will not clip.

To set the mic pre output, I watch the armed track meters in Sonar and make sure my hardest playing never gives me a digital 0dB or an over. What this always ends up with for me is having very fat, big, healthy drawn sine waves, with the highest peak ever going hitting -.1dB I am all for going in as hot as possible.

Nitro - That is crazy if you never touch volume levels on your tracks. Are you afraid to do so? WiLD!

 

[Bob Savage]

Definitely not- not if you're going for audio fidelity. Clamping down on dynamic range will always result in fewer bits used, and if your mission is to use as many bits as you can, you just failed (as I said though, I still think bits used means nothing in terms of fidelity). Any make-up gain also raises the noisefloor quite a lot.
As an effect or something, though, of course it's OK to use, but if your whole point is that you think
loudness = fidelity, and
compression = more loudness = more fidelity,
you'd be very wrong.

If I were to run a clean guitar tone through a compressor so that some of the highest peaks were leveled out, thus allowing the overall guitar level to be hotter, how is that using fewer bits? It would seem to me that I would be using fewer bits without the compressor, overall. I might lose some dynamic range, but the range on the clean I'm talking about is so wide that there will still be plenty of range.

 

[Bob Savage]

Nitro - That is crazy if you never touch volume levels on your tracks. Are you afraid to do so? WiLD!

Channel faders are scary, S C A R Y things and should be touched only by those who can harness their power.

 

[A Gruesome Discovery]

If I were to run a clean guitar tone through a compressor so that some of the highest peaks were leveled out, thus allowing the overall guitar level to be hotter, how is that using fewer bits? It would seem to me that I would be using fewer bits without the compressor, overall. I might lose some dynamic range, but the range on the clean I'm talking about is so wide that there will still be plenty of range.

Because hotter does not mean more bits. Greater dynamics means more bits. A compressed signal requires fewer bits to describe than the same signal uncompressed, generally speaking.

 

[Bob Savage]

Because hotter does not mean more bits. Greater dynamics means more bits. A compressed signal requires fewer bits to describe than the same signal uncompressed, generally speaking.

This is not the way I understand it. A -1db signal at 24 bits is going to be using the same amount of bits whether the sound has been compressed or not.

Hopefully the pros will chime in here soon to set one of us straight on this.

 

[A Gruesome Discovery]

This is not the way I understand it. A -1db signal at 24 bits is going to be using the same amount of bits whether the sound has been compressed or not.

Hopefully the pros will chime in here soon to set one of us straight on this.

Yes, that's true, but forget about the peaks for a moment; what's going on besides the peaks? If the rest of the audio data is much lower, let's say -18dB, then more bits will be required to describe this data, because there's 17dB of dynamic range there (i know, I'm oversimpifying for this example). If the rest of the audio is also loud, though, let's say -6dB, it will require fewer bits to describe; then the dynamic range is only 5dB. That's why compressing- reducing the difference between the "loudest stuff" and the "quietest stuff"- will cut down on the number of bits necessary to describe it.
I have to say again, though- it has nothing to do with sound quality. I love compression, but I have no idea what a "bit" sounds like.

 

[prowlergrig]

listen to gruesome. more dynamic range = more bits

not that anyone should listen to me, but i'm against using compression while recording, at least in the digital domain, when the compression effect can easily be added afterwards with the same (or better? ) effect.

 

[Bob Savage]

Yes, that's true, but forget about the peaks for a moment; what's going on besides the peaks? If the rest of the audio data is much lower, let's say -18dB, then more bits will be required to describe this data, because there's 17dB of dynamic range there (i know, I'm oversimpifying for this example). If the rest of the audio is also loud, though, let's say -6dB, it will require fewer bits to describe; then the dynamic range is only 5dB. That's why compressing- reducing the difference between the "loudest stuff" and the "quietest stuff"- will cut down on the number of bits necessary to describe it.
I have to say again, though- it has nothing to do with sound quality. I love compression, but I have no idea what a "bit" sounds like.

Alright, where are the pros!!!!???!! My understanding is different than yours. The way I understand it, the -18db signal will use less bits. Yes, the dynamic range will be big, so you could say there's a big difference in the number of bits being used between signals, but that's a different subject related to dynamic range only, not number of bits used.

I need to understand this... Andy, James, Bueller???????

 

[Bob Savage]

listen to gruesome. more dynamic range = more bits

not that anyone should listen to me, but i'm against using compression while recording, at least in the digital domain, when the compression effect can easily be added afterwards with the same (or better? ) effect.

I have personally never used compression on the way in, however, I've wanted to on clean guitars, but I don't have the outboard gear to do so.

In any event, I don't mind being wrong, but I'm having a tough time understanding how more bits are being used on the lower level signal.

 

[Bob Savage]

Alright Andy, I see you up there browsing the forum!!!!

 

[~BURNY~]

Yep, you probably won't be able to hear a difference between an heavily compressed sound @ 16bits and @24bits.
But I think limiting to prevent clipping on very dynamic sound (let's say a snare drum) makes sense if you make sure that the signal doesn't hit the threshold constantly.

 

[James Murphy]

i have already stated that Gruesome is technically right about this.. .waay back on page one. you've missed my point... and i'll quote myself:

"the "fixed resolution" is what is available...not what is actually used by the program material (edit: this refers to the music you actually want to record, as opposed to the noise floor you do not. see my "digital photo" analogy, which i will also quote below). each bit adds six db more amplitude range which in turn increases the dynamic range that can be recorded by a given system... and that's what dynamic range is, functionally (edit: this is key.... "functionally"), in this discussion: the difference between the the quietest and loudest amplitude that a given system is capable of recording. higher bit depths yield a bigger dynamic range ... whether or not that dynamic range is used is dependent on the engineer and the source sound. the results of not using as much of the dynamic range as possible is quantization errors that sound like distortion. Quantization errors are when a sample falls between bits in the encoding process and can't be resolved properly... so it is simply cut off... and for each bit in a system the audibility of quantization errors (distortion) decreases by six db.... so this brings us to noise floor. not the noise floor of the system, but of the room your mic is in or the output of the device you have directly connected. yes, you will record all this and if your actual signal you intend to record is not hot enough then you are more likely to hear it. And, more to my point, if your signal is too low level in a 24 bit recording, then when you dither to 16bits you will be far more likey to end up with audible quantization distortion.

that's the long and the short of it... .but am i saying everything should be squashed to hell and recorded all the way to zero at all times? absolutely not... i'm saying what i said to start with: record with your peak levels as hot as you can without clipping."

and the digital pic analogy: "let's analogize this with digital photography...

gruesome seems to be saying that it doesn't matter if the subject you want to photograph is only using 30% of the frame because it's still a 300dpi photo, and that's true enough, but what happens when you zoom in to just the 30% that you intended to shoot? you will see the pixels, the shapes will seem distorted, or "jagged"... same difference or even worse than if you shot only the subject filling the frame but only captured an only 72dpi image.

this is just an analogy of course.. but it's a pretty good one."

yes gruesome is right... but is the actual result of not recording your peaks as high as possible without clipping a desirable practice? nope.

 

[James Murphy]

every single time i've ever recorded, or gotten tracks recorded with a not very healthy peak level it has never failed to improve the sound quality to re-record with a more healthy signal... this is what is most important to me (next to performance)... the empirical result, not the esoteric (to most people) theory behind it. and to that end i've been speaking very loosely in regards to the technical details... more in a way to make myself understood in regards to the result i look for... not the theory as an absolute.... yes, a signal recorded with low peak levels in a 24 bit recording will still be 24 bits, but the lower those peak levels are, the more likely you are to have a "less than it could be" quality to the audio you recorded... and more things in that signal that you don't want. like, in most home studios, things like HVAC noise and the like. etc.

we actually agree on the theory totally... but i simplified in order to make my position more clear... which is that recording peak levels as high as possible in a digital system is good practice. i overstated the quantization noise problem a bit, in trying to further my point, and for that i'm sorry... it happens when dithering down from 24 to 16 if your signal level is VERY low to start with...and it's very subtle, not sounding so much like actual distortion but more of a slight "grainy-ness". you'll notice this even more if you apply a destructive gain increase (like with and audiosuite plug) to the audio (like zooming in to a low res digital photo or to a relatively small portion of a hi res digital photo, to be more specific). bah, i don't want to sound like a scientist here.. i'm not, but i know by how levels peak levels should be set... and i'm definitely NOT condoning squeezing the life out of one's tracks to get a higher average level at all times, quite the contrary. having said that, i won't hesitate to compress a vocalist or clean guitar, or anything else for that matter, on the way in to some degree if i think it's called for.

 

[Bob Savage]

Thanks James. I'm still struggling with the lower signal actually using more bits to capture the data, but the bottom line is, I guess I really don't care. I've always tried for the hottest signal without clipping, and what my ears tell me is what counts.

 

[James Murphy]

Thanks James. I'm still struggling with the lower signal actually using more bits to capture the data, but the bottom line is, I guess I really don't care. I've always tried for the hottest signal without clipping, and what my ears tell me is what counts.

yeah.. for all intents and purposes, lower peak levels = less resolution in my mind as well and always will... that's just paying attention to my ears, with theory aside. I'm not an expert in the science behind digital audio, but i think i've a reasonably sound grasp of the basics, and i know when my ear says that one recording level doesn't sound as good as a higher one. just don't squash everything with compressors, set healthy peak levels that don't disort, and use your faders for what they were meant for wave: hi Nitro! )....and you'll do fine.

 

[Nitronium Blood]

use your faders for what they were meant for wave: hi Nitro! )

 

[A Gruesome Discovery]

I should mention that I'm speaking purely in theoretical and observational terms not solely pertinent to audio applications, but any digital signal or system where there is at least one stage of analog conversion. I believe James and I are in complete agreement, however I think there's some theoretical properties of digital signals that I'm hinting at, and the only thing we're not totally eye-to-eye on is the undefined quality known as "too low" with regards to level. Theoretically I believe that any loss in resolution occurs at a much lower level than James does, but this is a factor of particular hardware anyway, so it's a moot point. I believe you can record lower without artifacts or losses, but in practice there's no point in doing so. So we are in agreement; your peaks should be as hot as possible, because there's no reason they shouldn't be. There's just some extra-nerdy theory that I've brought to the table, but it shouldn't affect any real-world applications in any way, I just find the subject fascinating, and the odd technical aspects worthy of note.
I thought this was a really good thread, I must say. This is the kind of stuff I love seeing here.