ORIGINAL: DonM

All:

I concur with much of what is being said, but I would add however.... The difference between 44.1k and 48k in capture frequency (thanks to the Nyquist theorum) correct me if I did the math wrong but about 1/5 of an octave.

The difference is in between having the filter going from 44100/2=22050 to 20Khz and 48000/2=24000 to 20Khz. Thats nearly twice the bandwidth.

UnderTow

Please stay strong. My father has had 3 heart attacks... Cancer, and kemo... Tripple bypass.... And we play 18 holes of golf every Tuesday morning.. Stay strong, and nothing can stop you

That's encouraging, Ognis. This is my third, too. And after doing everything I'm supposed to do: I dropped 85 lb, stopped smoking, exercise an hour a day, don't drink and I always eat my veggies. I'm just paying the price for a lifetime of being a rock and roll musician, I guess.

I want to add a sincere thanks to everyone who has wished me well, and thanks to everyone who does not wish me well but was kind enough to keep it to themselves.

Hmm... I think we're both wrong. It's not 1/5 an octave nor nearly twice the bandwidth. Here's my logic

Humans can hear approximately 20hz to 20,000hz which is divided across approximately 10 Octaves (see my chart below) Since doubling each frequency is in fact an octave you can see simply 20hz to 40hz is one octave, 40hz to 80hz is another octave and so on. Once we reach "higher" frequencies the number of cycles increases between octaves (obviously) That makes the difference between what is capturable under the nyquist for SR's of 44.1Khz and 48Khz of 1950hz. In that octave range (dividing the difference between the start and end of octave 11) by 12 semi-tones that makes the 11th octave divided into 1575 cycles and some change per semi-tone. That would make the octave capture difference between a 44.1khz and 48khz sample rate about 1 and less than half a semi-tone in the eleventh octave. Maybe I am misunderstanding your comments about nearly twice the bandwidth. Let me know where I've strayed.




-D

ORIGINAL: DonM

Hmm... I think we're both wrong. It's not 1/5 an octave nor nearly twice the bandwidth. Here's my logic

...

Maybe I am misunderstanding your comments about nearly twice the bandwidth. Let me know where I've strayed.

I just meant that the filter needs to let everything through at 20Khz but nothing at half Nyquist. For 44.1Khz that is 22.05khz - 20Khz = 2.05 Khz bandwidth. At 48 Khz that is 24Khz - 20Khz = 4Khz bandwidth.

UnderTow

Don, I think everyone is overthinking this issue.

Raising the sample rate also raises the cutoff frequency for the anti-aliasing filter, that is true.

However, it is also moot. That's because modern A/D converters' final anti-aliasing filters are digital linear-phase filters that can be very steep without smearing. Shifting the anti-aliasing filter's knee up isn't necessary, as it would if we were talking about a conventional reactive (analog) filter.

The reason for that is modern converters are oversampled, meaning we're actually sampling at much higher frequencies than 44.1 or 48 or even 192. It's more like 5.6MHz! That relaxes the filter requirements such that its knee is WAY beyond audible frequencies. It's only AFTER that step that we decimate down to 44.1/48/88.2/96/192 by basically tossing out all the extraneous samples. At this point we have a DIGITAL signal that can be DIGITALLY band-limited to 20KHz without artifacts.

The bottom line is that while you still want wide bandwidth in all your analog components, there is virtually no audible benefit to digitally recording at sample rates greater than 44.1KHz.

If you can hear a difference, it's due to inadequacies of your hardware. There should be no difference. To insist that there is a difference is to argue against Nyquist himself, whose theorem unequivocally states that any audible wave can be accurately reproduced -- not approximated, but accurately reproduced -- as long as the sample rate is slightly higher than double the highest frequency you need to record.

UT:

Right! Now I understand. And you can see from my info that the difference between 44.1 and 48 Khz sample rates is really not that much in octave info. I perform blind tests with my students to see where their octave perception is most acute. As Fletcher and Munson taught us over 65 years ago - that center is around 500 to 3000 hz. The ability to discern about one and 1/2 semi tones above 20Khz is possible but quite amazing. Most folks talk about the 'air' in that location. Which as an on-location orchestra recordist I have a bunch of 'airless" recordings I've done over the years. But I would say that I choose 96khz as a SR for some of my orchestra stuff based on info I've gained from DPA microphones and their thoughts about stereo image and their omni mics.

I'd say that if I were doing only demos and in-the-box mixing... I'd be at 44.1khz... I know that may be shocking to some, but here's my logic. I did say DEMO - typically that's for the band members to learn a tune (wherever they may be in the world) and to get ready to go to real studio (whatever that means eh?) So all of my work is for broadcast or CDA or Film - so I work in the format needed by those delivery mediums - BTW FM broadcast is really about 128 kbs MP3!

As a classical recordist I must say the DSD and DXD stuff is interesting - but again for the 'right' reason. I believe stable stereo image is key (in fact I do part of my final mastering check on Acoustat Electrostatic panels to perform the highest resolution image test I can).

All in all SR matters as everything else does in the grand scheme. Thanks for your input.

-D

bitflipper, I agree with everything you say but with a caveat: It depends on the quality of the equipment used. Alot of budget converters don't have very good decimation filters (amongst other things). With these converters, it actually does help to go beyond 44.1Khz for recording. You will often get better results recording at 88.2Khz and SRCing the recorded files with a tool like R8Brain although this seems like a convoluted process. And of course this is only relevant when everything else in the signal chain is top notch and when the microphones are well placed on good instruments/cabs etc in good rooms with good performers...

In general, even budget euqipment has gotten so good these days that it makes more sense to worry about other things than sampling rate for most people. That doesn't mean it isn't interesting material and knowing one's tools can only benefit the end result. Then again, a little knowlege...

UnderTow

Yup, you're right. That why I added the caveat "If you can hear a difference, it's due to inadequacies of your hardware."

ORIGINAL: DonM

UT:

Right! Now I understand.

And you can see from my info that the difference between 44.1 and 48 Khz sample rates is really not that much in octave info. I perform blind tests with my students to see where their octave perception is most acute. As Fletcher and Munson taught us over 65 years ago - that center is around 500 to 3000 hz. The ability to discern about one and 1/2 semi tones above 20Khz is possible but quite amazing. Most folks talk about the 'air' in that location. Which as an on-location orchestra recordist I have a bunch of 'airless" recordings I've done over the years. But I would say that I choose 96khz as a SR for some of my orchestra stuff based on info I've gained from DPA microphones and their thoughts about stereo image and their omni mics.

I read once that phase accuracy (aka timing) of 44.1Khz modern samplers is arround 2 pico seconds. (With a good explanation of how that was calculated). That is way beyond what our ears or brains can perceive. Note that this is not the same thing as the fact that there is 0.023 microseconds between samples at 44.1Khz. The phase accuracy is actually much higher. If not, we have to start worrying about jitter which is probably a bigger problem and more audible.

So this would mean that good clocks and jitter rejection are more important than sampling rates.

I'd say that if I were doing only demos and in-the-box mixing... I'd be at 44.1khz... I know that may be shocking to some, but here's my logic. I did say DEMO - typically that's for the band members to learn a tune (wherever they may be in the world) and to get ready to go to real studio (whatever that means eh?) So all of my work is for broadcast or CDA or Film - so I work in the format needed by those delivery mediums - BTW FM broadcast is really about 128 kbs MP3!

Are you talking analogue FM or digital FM? But yes, always cater to the delivery format.

As a classical recordist I must say the DSD and DXD stuff is interesting - but again for the 'right' reason. I believe stable stereo image is key (in fact I do part of my final mastering check on Acoustat Electrostatic panels to perform the highest resolution image test I can).

SACD and DVD-A seem to have completely failed to catch the consumer's imagination. Maybe they are not fooled by the emperor's new clothes...

UnderTow

That is way beyond what our ears or brains can perceive.

Not 'way', although how conscious we can be of it is probably a matter of debate, but the fastest switch our body has is in the eye, involving 11-cis-retinal switching to another state, and is <5 picoseconds. Reacts to a single photon even, IIRC.

Never underestimate our hardware.