DeeringAmps


Ben,
IF you WANT a real education read the GS thread I liked to earlier.
Skip Burrows works WITH and ON SSL consoles EVERYDAY.
Paul Frindle was on the design team AT SSL.
They are ADAMANT that you keep your signal levels the same in the digital realm as you would in hardware.
Danny's input PEAKING at -6 is like pushing the VU all the way to the right. His AVERAGE signal will be right at Ovu.
At 0dBfs the needle will be PEGGED.
Nothing personal here Ben. You are free to record however "sounds" best to you.
You don't have to believe a single word I say, (and probably shouldn't, I'm just another internet "Know it all") BUT Skip and Paul have REAL credentials.

'nough said.

T

Hey Tom I'm not that disrespectful, people have me all wrong.
 
This console thing is a whole different argument though, I don't even know how companies like SSL can still keep afloat.
 
Look I know that Stavs theory works, maybe it has something to do with cheaper converters and cheaper convertors to get the best out of them requre this technique.
 
Do you know what the dynamic range of the converter's are on the SSL console because I am starting to think this is where the answer lies.
 
I know my peak is generally between -6db -3db that's what I aim for and as I say I believe to get the best clarity out of the bass instruments, this is where I like to sit.
 
I will have to look at RMS values next time I record.
 
Neb
 
Oh I know who Paul Frindle is, I have an interesting post of his on Gearsluts and I think it ties in with some of this stuff let me look it up, it was really cool!
 

BenMMusTech


Ok so what about, sorry Drew, I refer to you as the most technicaly minded of us all here.

So what about the notion of missing information because we are not using all the bit's avaliable to us eg: harmonics.

I hope that makes sense.

Neb

Remember what I said about imagining there's no noise or inadvertently moving the target?

We've already agreed that we want the quantization level to be a bit below the noise floor. The noise is causing the signal level that gets sampled to bounce around randomly at a magnitude greater than the quantization step size. How is measuring a randomly moving signal in smaller increments going to help?

With a dithered signal, do you understand that you can still hear stuff below the quantization level?

Ok I got that drew but if the harmonics are all in the area where the quantization is happening and or below and even though we can still hear them, because of dithering and the quantization errors is it not possible we are still getting the harmonics but at a reduced fidelity.

Does that make sense??

Neb

Just out of interest the SSL console I was referring to in one of my previous posts is a fully analog console only. All A to D and D to A in our case is handled by AVID HD 192 interfaces outside. (3 to be precise)

But you still need an rms ref level at some reference point and SSL were suggesting the ref levels I mentioned before eg -24 -22, -20, -18 etc. -20 is a good one because it matches the K level. The VU was calibrated at -20 on the particular console we have had installed at the factory.

Ben some of your ideas are not over wrong in that recording hot was a good thing to do in the past. (I spent many years working with EMU samplers from 8 bit into 12 bit and 14 bit so I know what you are talking about. You did have to sample hot back in those days otherwise it sounded just plain bad!) But now with both 24 bit recording and the fact that almost any converter you buy now is going to be pretty excellent in the scheme of things so now we can virtually relax levels wise now and just work with say a K ref level of -20 without any concerns that things are not being recorded well.

Even if we are at -20 for example and the music dropped down 30 db below that which is very very quiet compared to the original level, we are still only at -50 and at 24 bit there is still over 80 db down below -50 even before any of that distortion that you are rightly talking about would even come into effect.

Bass does not benefit from being recorded hot either. All frequencies will sound the same pretty well at any level.

Jeff Evans


Just out of interest the SSL console I was referring to in one of my previous posts is a fully analog console only. All A to D and D to A in our case is handled by AVID HD 192 interfaces outside. (3 to be precise)

But you still need an rms ref level at some reference point and SSL were suggesting the ref levels I mentioned before eg -24 -22, -20, -18 etc. -20 is a good one because it matches the K level. The VU was calibrated at -20 on the particular console we have had installed at the factory.

Ben some of your ideas are not over wrong in that recording hot was a good thing to do in the past. (I spent many years working with EMU samplers from 8 bit into 12 bit and 14 bit so I know what you are talking about. You did have to sample hot back in those days otherwise it sounded just plain bad!) But now with both 24 bit recording and the fact that almost any converter you buy now is going to be pretty excellent in the scheme of things so now we can virtually relax levels wise now and just work with say a K ref level of -20 without any concerns that things are not being recorded well.

Even if we are at -20 for example and the music dropped down 30 db below that which is very very quiet compared to the original level, we are still only at -50 and at 24 bit there is still over 80 db down below -50 even before any of that distortion that you are rightly talking about would even come into effect.

Bass does not benefit from being recorded hot either. All frequencies will sound the same pretty well at any level.

Ok I am going to play nice and not be rude, I put my gun away and you put your's away.
 
Do we agree that 24 bit is not 24 bit, the best converter's use a dynamic range of around 118db not the full spectrum of a 24bit 144db.  As I have mentioned the US Lexicon series use converters that use only 101db of 144db spectrum.  So 5 db above 16 bit.
 
Therefore the theory still has validity when it comes to lesser interfaces and converters.
 
Now even Presonus which have some of the best cheap interfaces on the market, I think if I remember rightly their converters use 114db of the 144db 24 bit spectrum, because they are so cheap and even though they claim to have low jitter specs, I still see the theory being of use because of the jitter problem and the distortion that jitter introduces.
 
I know I am vering off a bit here but do you get what I mean??
 
I am going to look at that Paul Frindle gearslutz post because he had posted about digital myths and this may help clarify what we are talking about.
 
Neb

BenMMusTech


Ok I got that drew but if the harmonics are all in the area where the quantization is happening and or below and even though we can still hear them, because of dithering and the quantization errors is it not possible we are still getting the harmonics but at a reduced fidelity.

Does that make sense??

Neb

What you're talking about is stuff that's already below the noise floor before your signal gets quantized. Since we use a quantization level even lower than the noise floor, we end up with quantization error (that sounds like noise) at a level below the existing noise floor. And keep in mind the dithering we are talking about here is not something we're adding - it's the "self-dither" from the background noise in the room, analog signal chain, etc. 

So you're talking about stuff that you can hear below the existing noise floor, but that will be masked by a little teensie, tiny bit of additional (quantization) noise? In the real world that's just not likely to happen, because it can only happen with a combination of all the factors where everything is just exactly so.

Ben what you are talking about now is the useable dynamic range that any converter is to actually going to give you. But that might be hard to quantify because while one converter might only use 101 dB out of the available 144 dB of range another might be better and in fact achieve say 120 dB out of 144 dB. How do we measure it? It is a good point and warrants some research into the subject to see how much actual dynamic range is available. I have heard that most converters use about 6 dB less dynamic range that is fully or theory wise avalaible. eg 16 bit in reality is 90 db of range and 24 is supposed to be 138 dB

But even if it varies etc I still feel that it is probably still a lot of dynamic range that is available. Even if a converter only used 101 dB of range and you had a signal way down at -50 then there is still 51 dB of range below that before the noise is going to even become an issue. Don't forget the -50 db signal is already 30 db down (from -20) and that represents a very quiet sound in itself.

I do agree about something though and that is the fact the final medium we are listening to is say 16 bit and I think it is good to make sure your final 16 bit master is reasonably loud and that still holds. Even a K-12 level is pretty loud and hot and still allows for plenty of transients.

Ok I will put it to bed but I will keep doing things my way!!

We will have to agree to disagree.

Peace Neb

Jeff Evans


Ben what you are talking about now is the useable dynamic range that any converter is to actually going to give you. But that might be hard to quantify because while one converter might only use 101 dB out of the available 144 dB of range another might be better and in fact achieve say 120 dB out of 144 dB. How do we measure it? It is a good point and warrants some research into the subject to see how much actual dynamic range is available. I have heard that most converters use about 6 dB less dynamic range that is fully or theory wise avalaible. eg 16 bit in reality is 90 db of range and 24 is supposed to be 138 dB

But even if it varies etc I still feel that it is probably still a lot of dynamic range that is available. Even if a converter only used 101 dB of range and you had a signal way down at -50 then there is still 51 dB of range below that before the noise is going to even become an issue. Don't forget the -50 db signal is already 30 db down (from -20) and that represents a very quiet sound in itself.

I do agree about something though and that is the fact the final medium we are listening to is say 16 bit and I think it is good to make sure your final 16 bit master is reasonably loud and that still holds. Even a K-12 level is pretty loud and hot and still allows for plenty of transients.

Check your specs Jeff on whatever interface you are using, I can't find mine, MOTU are very secretive but I believe mine are around 110db or that's what they use of the 144db of dynamic range, so in reality it's more like 20 bit not 24 bit. My interface that is.
 
Neb

BenMMusTech


Now even Presonus which have some of the best cheap interfaces on the market, I think if I remember rightly their converters use 114db of the 144db 24 bit spectrum, because they are so cheap and even though they claim to have low jitter specs, I still see the theory being of use because of the jitter problem and the distortion that jitter introduces.
   

Hmm. I wonder if I should point out that the level of jitter distortion is relative to the signal level, so if you have a hotter signal you get proportionally hotter jitter distortion too?