What do I not understand???????????

  • Thread starter Thread starter VTgreen81
  • Start date Start date
VTgreen81

VTgreen81

Active member
Ok........we've learned that the sampling therom states that the sampling freq must not exceed S/2 when S = the sampling rate.

A low pass filter cuts out all above S/2

Range of human hearing ends at 22KHz

So to sample the highest audible freq you'd need a sample rate of 44KHz.

Thus the CD standard of 44.1 Khz

If 44.1 captures all the info to reproduce every audible freq, what's the point of having a sample rate of 96Khz?
 
Because your hearing doesn't stop at 22k. You perceive frequencies much higher (and lower than 30hz) and when they're missing, some folks can very much tell the difference.
 
Although that hasn't been proved...

With 96k you get more information and all the filters and stuff have more data to work on, resulting in a more accurate result. And also, the hardware gets easier to make, as you don't have to fiddle with oversamplomng and such.
 
Your logic is correct however the 44KHz sampling rate does not cover the perceived frequencies thus, the 96k sampling rate has become the generally accepted ideal rate.

However, since most of the buying public thinks "more is better" the industry has been able to bullshit people into believing that 192k is better when in fact the extra sampling means only bigger files and unnecessary processing. This introduces distortions and aliasing that you don't get at 44.1k, 48k and 96k and you actually end up loosing accuracy.

Sad thing is it doesn't end there and a billion dollar junk industry exists. You can purchase all kinds of unnecessary audio junk like directional speaker wires and demagnitizers for cd's so you have to be careful and research what you're getting.
 
lets see if I understand now

If I record a freq of 6KHz i'll also have harmonics at 12,18,24 & 30KHz

At a sample rate of 44.1 I'd lose the harmonic overtones above 22K, and although not audible, they add some type of coloration to the sound? More a feeling of the vibration rather than a recognizable sound?
 
regebro said:
Although that hasn't been proved...

With 96k you get more information and all the filters and stuff have more data to work on, resulting in a more accurate result. And also, the hardware gets easier to make, as you don't have to fiddle with oversamplomng and such.
Click to expand...


so the samples all get lined up and we play connect the dots, and the more dots the more accurate the representation of the recorded signal? Sounds more like we're dicussing bit depth.
:confused: :confused: :confused:

i knew I was too old to go back to school
:rolleyes: :rolleyes: :rolleyes:
 
Bit depth and sampling frequency both affect the data amount. Most 96 khz converters of course use 24 bit too, which helps.
 
A lesson on sample rate.

What happens if you draw a dot-to-dot picture using only a dozen dots or so? You have a very rough, “barely recognizable” picture. If you add more dots then the picture becomes a little clearer. Add still more dots and it takes on greater detail. Add tens of thousands of dots, or perhaps millions, and it can approach photo quality.

A similar thing is happening with audio sample rates. A low rate of say 44.1 KHz is barely adequate to be recognizable at the highest frequencies. Here’s why.

A sign wave is a graphical representation of a sound wave. When drawn, it looks like a simple curved line forming a mountain followed by a valley. Perhaps you’ve seen it. It has a high peak and a low peak. A sample is a simple measurement of how high or how low the line is at any given point in time as it moves past a certain point. So a digital sample is only a numerical value.

At 100 Hertz, we have 100 vibrations occurring in one second. (Or in digital music, 100 sign waves end to end.) But we also have 44,100 samples (measurements) being taken in one second. 44,100 divided by 100 equals 441. We get 441 samples to describe each sign wave.

In other words, at 100 Hertz, we can draw a dot-to-dot picture of the sign wave using 441 dots. That’s not bad. We can plot out a very clear picture of the exact curve and thus our recording should be crystal clear. However, the higher the number of vibrations, the less dots we get to describe each sign wave. At 22 KHz, we only get two dots to draw the curve. Because 44,100 divided by 22,000 equals two. The highest frequency of 22 KHz is represented using only two sample points.

As you can see, the high frequencies suffer from accurate reproduction. That is why cymbals, bells, and things that go hiss don’t always sound their best.

If you increase the sample rate, you also increase the resolution of your carbon copy, so-to-speak. It is true that higher sample rates make larger file sizes because they contain more samples, (or dots.) But that’s a good thing.
 
So at what sampling rate do people on this board record? I had never considered this before, so I checked and found that my CPA9 is set to a default 44.1 kHz. I thought this was the read rate of most CD players. So if I record a song at 96 kHz, what's the benefit if it's only going to be played at 44 kHz?

Maybe tomorrow I'll try recording a track at both frequencies, and try and see if I can pinpoint any contrast between the two...
 
If you are only recordin it to normalize it and burn to CD there will be no benefit.

If you are using the computer as a mult-track recorder and so, then the additional data density will be beneficial, but suck up processing power.
 
I've noticed in a few tech articles lately that engineers are even upsampling 44.1khz projects for processing and digital mixdowns. It used to be you would only do that for greater bit depths but now they say the same benefits apply to higher sample rates which is fewer artifacts and greater preservation of the original quality.

In the analog world you usually want to use better formats as you go down the chain to preserve the original quality so maybe the same rules should apply to digital audio.
 
With you guys at my back we be pullin' an " A " fer sure.

Technical reading is such "gobbledy goop" but you guys really help to illucidate the concepts.

Thanks to all of you I've aced every weekly quiz and am one of the classes top contributers. Mid terms are coming up so I'll have lots more Q's.

Thanks again
Dan
 
Now I don't understand...

how will the extra bit depth be beneficial in any respect if the end result is a CD? Is the sampling rate different when CD's are pressed as opposed to burned off of a computer? I don't understand the pressing process one bit, and I suspect it's VERY different from what transpires in my computer.

Should I have been recording for my album in 96 kHz all this time?
 
The bit depth helps you keep the details when compressing. It seems like it also helps to keep clarity and smoothness with filters and reverbs although that is more controversial theories...

What you "should" do is all up to you. There is no wrong way to do anything as long as the result sounds good.
 
Bit Depth is how big each sample is (or how much info is in each "dot"), Sampling Frequency/Bit Rate is how many dots per second. They both play a part in the final result. I record at 24-bit, 44.1kHz, and dither my final stereo mix down to 16-bit before burning a CD.
(It's never too late to go back to school!)

VTgreen81 said:
so the samples all get lined up and we play connect the dots, and the more dots the more accurate the representation of the recorded signal? Sounds more like we're dicussing bit depth.
:confused: :confused: :confused:

i knew I was too old to go back to school
:rolleyes: :rolleyes: :rolleyes:
Click to expand...
 
Moving on

Okay so Sampling rate gives us the number of samples in a given period, bit word length tells us how accurately those samples are quantized. Quantization error will be no more than half LSB (least significant bit).

Now dither is a process that will correct what quantization error does exist, supposedly the digital equivalent to analog's Bias, but I'm kinda foggy on how it works.

Pulse modulation and averaging............... HEADACHE
 
Last edited:
Do I have to straighten you guys out again? Okay, listen up class.

If Sample Rate can be thought of as the horizontal resolution of the sign wave, then bit-depth can be thought of as the vertical resolution of the wave. Though, bit-depth does work slightly different than sample rate.

It defines the scale that is used to measure amplitude. More specifically it defines how many increments are on the vertical scale. NOT how many measurements are taken per second, but how many marks are on the ruler that we use.

Here’s an example.
If you used a scale with only 3 measuring marks on it, then the volume of each note recorded could only be set to one of those 3 values. The playback would have to make every note one of the following volumes; Loud, Medium, or Soft. (You only get three, so pick one.) However, if you increased the number to say 30, then there would be more positions to describe where the volume fell. The notes can be plotted to any of 30 different volumes. Again, it relates to the dot-to-dot story I told earlier.

8-bits, (poor), gives us a possible 256 marks on the ruler. 16-bit gives us 65,536 marks, (CD quality), and 24-bit gives us approximately 16.7 million marks.

Amplitude should be kept at high resolutions just as frequency should while recording and processing. Because when we process digital audio for effects, dynamics, EQ, or anything, the signal is subjected to intense mathematical formulas. We will no doubt lose some audio quality through the rounding of decimals. Higher rates are kept more accurate, lower rates are not. Once all processing is done, then we lower the rates back to industry standards.

So, always record as high as your hardware will allow.
 
Thanks Homerecor. I had forgotten most of this. LOL!

Time to get a new HD for all my 24-bit 96K recordings. :D
 
But how does pulse modulation work? I'm sure it won't be on a test, the teacher can't even explain it, but inquiring minds and all that........ :confused:

Also, my notes are incomplete :( , could someone remind me what Fourier analysis refers to?

Had something to do with complex vs. sine waves and then used to round off square or step waves. :confused:
 
Pulse modulation is a way of making D/A conversion. Essentially, what you do is generate high frequency pulses, loads of them if you want a high value, few of them if you want a low value. Then you slap a filter on it to remove the high frequency contents, and you got yourself an audio signal. :)
 
You must log in or register to reply here.
Back
Top