Only if neccessary. The lack of high-end energy is basically just "natural" in most audio. High end doesn't use/produce much energy compared to lows. This is why a kick might be mixed in at -6dB and the cymbals and overheads might be mixed in at -18dB, yet they still sound related in volume.

If you run & listen to some tones (a very valuable experience if you've never done this) you'll find out that everything above 17kHz is pretty much dog whistle zone and that most speakers and microphones don't even really reproduce them. These frequencies will exist in recordings as the higher order harmonics above the much lower freq's that are the recordings fundamentals. The vast majority of music will exist a sloping down from the midrange to almost nothing around 20kHz so what you are seeing in your analysis of commercial CD's is typical of music and not showing any specific preparation. To me you don't really need to worry about whether freq's above 20kHz are rolled off by an LPF when preparing your mixes - this is resolved either by the AD when you record back in at 44.1kHz (which has anti-aliasing filters designed just for this issue) or by the Sample Rate Conversion algorithm the hardware or software you are using has when converting from a higher resolution sample rate.

I run a low pass filter only to solve a problem, i.e. when doing a transfer to vinyl master (where large amounts of ultra highs will only cause distortion) or to "warm up" a recording that was mixed with excessive amounts of high end. If your recording's high end sounds nice to you there is no reason to use an LPF just to make a spectrum analyzer's display "look right". We're making recordings to please the ear and not the eye after all!!

Best regards,
Steve Berson

Atlantis wrote: When checking commercial songs (in 44.1 kHz) with a spectrum analyser, I often see a sharp roll off in frequencies at around 20 kHz or so. Why exactly is this done? In order to make the fact that the frequencies are abrubtly cut off at 22050 Hz inaudible? So is this something I should be reproducing by using a low-pass filter with a sharp Q (I'm using 20504 Hz, Q of 5.79)?

I've usually have seen this on mp3's. a sharp cut at about 17k.
due to the compression. It's much more infrequent from a wave,aif or cda. file.

I also like to use 20k & use the bell curve to give more air without conflicting with to many things in the high range. That has all to do with what I'm hearing & what is needed & nothing to do with what I'm seeing or simulating the visuals of another cd.
Ed

These freq's don't translate well. If you're seeing a large amount of energy at 20k, it's going to translate very different from system to system. I usually see this from newer engineers, lots of energy at the extremes. Tons of subs, tons of 20k. The treble bass syndrom. eq'ing from the bottom or the top until you can hear it, it's a disaster waiting to happen. If you're monitors have trouble reproducing these freq's, you'll have to keep your eye on the spectrum analyzer to make sure you aren't going overboard in these area's.

TotalSonic wrote: If you run & listen to some tones (a very valuable experience if you've never done this) you'll find out that everything above 17kHz is pretty much dog whistle zone and that most speakers and microphones don't even really reproduce them. These frequencies will exist in recordings as the higher order harmonics above the much lower freq's that are the recordings fundamentals. The vast majority of music will exist a sloping down from the midrange to almost nothing around 20kHz so what you are seeing in your analysis of commercial CD's is typical of music and not showing any specific preparation. To me you don't really need to worry about whether freq's above 20kHz are rolled off by an LPF when preparing your mixes - this is resolved either by the AD when you record back in at 44.1kHz (which has anti-aliasing filters designed just for this issue) or by the Sample Rate Conversion algorithm the hardware or software you are using has when converting from a higher resolution sample rate.

Thanks, yes I've tried listening to a couple of tracks with a hipass filter with a very high frequency setting and sharp Q to only hear the air frequencies. Stopped at about 18 kHz, but it's damn harsh to hear.

when preparing your mixes - this is resolved either by the AD when you record back in at 44.1kHz

Didn't really get this part. I'm not doing any analog stuff here if that helps, just working with digital material and don't need to "record back in at 44.1 kHz" at all.

I run a low pass filter only to solve a problem, i.e. when doing a transfer to vinyl master (where large amounts of ultra highs will only cause distortion) or to "warm up" a recording that was mixed with excessive amounts of high end. If your recording's high end sounds nice to you there is no reason to use an LPF just to make a spectrum analyzer's display "look right". We're making recordings to please the ear and not the eye after all!!

Well yes, I've noticed some rendered software output (Reason in particular) causes a lot of frequency buildup around 20 kHz (I assume because of some aliasing reasons, though I'm still very new to all this), in which case such a sharp LPF would be a good way to remove this?

And you're absolutely right about doing it for the ear, but since I've only been mastering for about two months now, and need a bit of a visual aid still.

Ed Littman wrote: [quote=Atlantis]When checking commercial songs (in 44.1 kHz) with a spectrum analyser, I often see a sharp roll off in frequencies at around 20 kHz or so. Why exactly is this done? In order to make the fact that the frequencies are abrubtly cut off at 22050 Hz inaudible? So is this something I should be reproducing by using a low-pass filter with a sharp Q (I'm using 20504 Hz, Q of 5.79)?

I've usually have seen this on mp3's. a sharp cut at about 17k.
due to the compression. It's much more infrequent from a wave,aif or cda. file.

I also like to use 20k & use the bell curve to give more air without conflicting with to many things in the high range. That has all to do with what I'm hearing & what is needed & nothing to do with what I'm seeing or simulating the visuals of another cd.
Ed

Yes, but this is not what I mean. I'm talking about 20 kHz here, which MP3s don't even get near to.

As an example, here's a screenshot of the frequency spectrum of Sarah McLachlan - Plenty, taken from a CD so no MP3:

http://atlantis.plastiqueweb.com/temp/screen.png

I can see the natural roll off in freqs above 12 kHz, but at 20 kHz there's clearly a deliberate LPF used.

Looks normal. at 22khz it falls off which it should.

Atlantis wrote: [quote=Ed Littman][quote=Atlantis]When checking commercial songs (in 44.1 kHz) with a spectrum analyser, I often see a sharp roll off in frequencies at around 20 kHz or so. Why exactly is this done? In order to make the fact that the frequencies are abrubtly cut off at 22050 Hz inaudible? So is this something I should be reproducing by using a low-pass filter with a sharp Q (I'm using 20504 Hz, Q of 5.79)?

I've usually have seen this on mp3's. a sharp cut at about 17k.
due to the compression. It's much more infrequent from a wave,aif or cda. file.

I also like to use 20k & use the bell curve to give more air without conflicting with to many things in the high range. That has all to do with what I'm hearing & what is needed & nothing to do with what I'm seeing or simulating the visuals of another cd.
Ed

Yes, but this is not what I mean. I'm talking about 20 kHz here, which MP3s don't even get near to.

As an example, here's a screenshot of the frequency spectrum of Sarah McLachlan - Plenty, taken from a CD so no MP3:

http://atlantis.plastiqueweb.com/temp/screen.png

I can see the natural roll off in freqs above 12 kHz, but at 20 kHz there's clearly a deliberate LPF used.

How was it posible for you to make the CD play tones over 20Khz?

I've been gone for a while, but I was hoping to bring this topic back to life as this is still something I don't fully understand yet.

Ammitsboel, to answer your question, I used a CD ripping program (CDex) to extract the CD track to a WAV file, and so it's only natural for frequencies as high as 22,050 Hz to be reproduced due to the 44.1 kHz sampling rate of CD's.

Some show the effect much more severely due to the dithering type used. Waves type1-like dither features more high frequency noise that tends to fil the resulting ~21 kHz gap up to around -105 dB, as in the second track, for example, whereas Waves type2-like dither features a more even distribution of low-level distortion through the entire spectrum. The bottom line, though, is that all have an unnatural drop in level around ~20 kHz.

Now the question is whether or not I should also do this when processing masters, by using a low-pass filter with a sharp Q before using limiting and dithering. Is there a specific reason this is done professionally, as in these commercial releases above?

It's really a natural occurance. between instruments, microphones, and gear, they will naturally drop off at these freqs. It also doesn't sound natural for anything to have a flat response at 20k. the only way this can happen is if you crank it in or you are using a synth sound that has too much top end. use your ears to decide what sounds best.

Michael Fossenkemper wrote: use your ears to decide what sounds best.

sounds like that's the bottom line for -any- thread in this forum ! :wink:

Thanks for all the help. I only thought it was done deliberately for some reason, but I only roll the extreme top end off ever so slightly if needed now.