Does frequency response ABOVE 20 kHz really matter?

[chessparov]

I've seen this issue discussed before elsewhere, and was curious how important you guys think extended frequency response in various types of gear like pre's, A/D convertors, recorders, etc. are.

Thanks
Chris

[Geoff_T]

Originally posted by chessparov
I've seen this issue discussed before elsewhere, and was curious how important you guys think extended frequency response in various types of gear like pre's, A/D convertors, recorders, etc. are.

Thanks
Chris

Hi

In as nutshell... Yes, because the harmonics of the frequencies within the audio bandwidth are lurking there. There's often talk of Geoff Emerick hearing the treble boost from a dry soldered joint on a Neve transformer compensation network and I hear the effect (even with my cloth ears) on Neve racks of later units (3114/5/31105, etc) where the rack guy has left off the 600 ohm terminations producing a big hump (around +10dB) around 35 to 40KHz.

Believe me, mess with the HF stuff and you will hear the effect lower down... note I say "effect", I'm not saying that you hear 40KHz.

The proof of the pudding is in the eating and why PT and others have gone up the 96KHz clocks (and beyond).

[Bob Olhsson]

To put this another way, it's very hard to lop off the high-end without introducing audible artifacts below 20kHz. and any RFI or instability up there will also affect the audible range.

[cdog]

Originally posted by Bob Olhsson
To put this another way, it's very hard to lop off the high-end without introducing audible artifacts below 20kHz.

Then why can I insert an cut off filter on my mix @ 20 khz in the middle of a session and have no one in the control room notice (except Bugsy the dog, maybe)?

There is stuff going on up there I can see on a scope, I just can't hear it with my ears. It doesn't seem to matter recording @ 44.1 or 192khz, when I slap that 20khz filter on when everyone has their eyes closed and is told to listen closely, they think I'm just playing mind games with them.

Please explain.

[fishtop_records]

Originally posted by cdog
I just can't hear it with my ears. It doesn't seem to matter recording @ 44.1 or 192khz, when I slap that 20khz filter on when everyone has their eyes closed and is told to listen closely, they think I'm just playing mind games with them.
Please explain.

Most male humans over 25 or so are
down 10 dB or more over 15kHz.
The "standard" 20 to 20kHz is
typical and round and traditional.
For decades, everyone designed things
to fall off over 20kHz, and often things
were down 5 to 10 dB at 20kHz.

Most speakers (monitors, etc.)
have a real hard time over 20kHz.

There are interharmonic modulations
that occur between any two signals.
You could have one 36kHz signal
and another at 38kHz, and
the difference would be 2kHz,
which is very audible.

Plus, cheap filters that chop off
at 12db/octave totally trash the
phase response, and even male
humans are very sensitive to phase.
Phase is how we tell where the
jaguar is, which is how we
avoid being lunch.

[David Herbert]

You can also avoid this by only recording jaguars in mono. David

Originally posted by cdog
Then why can I insert an cut off filter on my mix @ 20 khz in the middle of a session and have no one in the control room notice (except Bugsy the dog, maybe)?

There is stuff going on up there I can see on a scope, I just can't hear it with my ears. It doesn't seem to matter recording @ 44.1 or 192khz, when I slap that 20khz filter on when everyone has their eyes closed and is told to listen closely, they think I'm just playing mind games with them.

Please explain.

it would also depend quite a bit on the program material. lots of instruments dont even have a whole lot over 20 khz. maybe lots is an overstatement, but some...

[Bob Olhsson]

Originally posted by cdog
Then why can I insert an cut off filter on my mix @ 20 khz in the middle of a session and have no one in the control room notice ...

The same reason I once mixed an album with both tweeters blown in a pair of JBL 4311s! You almost never hear something that you aren't listening for. Worse, you'll swear you do hear anything you expect to hear such as that equalizer that wasn't patched in!

Objectivity is never easy...

[Robotnik]

I'm more interested in 20Hz and below.

[chessparov]

Thanks for all the responses.

Now I better understand any Mariah Carey
vocals.

Chris

[jbuntz]

Originally posted by Robotnik
I'm more interested in 20Hz and below.

I noticed you said you like to boost 31.5 hz by 12 dB on kick. Is there any concern with making real world power amplifiers crap out?

[Ted Nightshade]

Wow, I think this is easily the best thread I've seen on this topic, so far. The only thing I have to add to the excellent on-the-money points made already- I recently saw a couple graphs of the effect of various cable capacitances on a passive attenuator output. At one capacitance, the frequency response was -1/2 dB at 20khz. I can live with that! But for the same capacitance, the phase curve was off 20 degrees at 20khz, and a goodly bit further down where everybody is hearing too! That was an eye opener for me.

I'm not sure exactly how this works with pres etc., but maybe Geoff or somebody else knowledgable could speak to this issue?

I recently saw a couple graphs of the effect of various cable capacitances on a passive attenuator output. At one capacitance, the frequency response was -1/2 dB at 20khz. I can live with that! But for the same capacitance, the phase curve was off 20 degrees at 20khz, and a goodly bit further down where everybody is hearing too!

was this a stand alone attenuator? like one that is sold with a knob and cable inserts on it? or just the attenuator? a spec like that makes me really worry about thinking about buying a cheaper attenuator. how common and easy is it for this to happen in a design? does the equipment on either side affect anything in this regard?

[Ted Nightshade]

Originally posted by chikkenguy
was this a stand alone attenuator? like one that is sold with a knob and cable inserts on it? or just the attenuator? a spec like that makes me really worry about thinking about buying a cheaper attenuator. how common and easy is it for this to happen in a design? does the equipment on either side affect anything in this regard?

It's a stand alone attenuator. Totally passive little box with knob.

I don't think what's being fed from the attenuator matters very much (in my case it's the power amps to monitors), but what is driving the attenuator is significant. I'm told by hutch at Manley that a 5k load from the attenuator and cables is about right for most devices. After a passive attenuator like this, you want short cables (1 to maybe 6', depending on what you can get away with due to cable capacitance) and very low capacitance ones. So I sprung for 2 times 3 feet of trick ultra-low capacitance cables (you want about 10pf per foot or less) to make the cables from the passive attenuator box to the amps, or the next device.

As far as cheaper attenuators, the real hazard is the stereo imaging between the two sides of a stereo attenuator- it's quite rare for both channels of a stereo pot to track very similarly, and with fluctuations of as little as 1 or 2 dB (most times the differences between channels are a *lot* worse than that!) the image will be wandering left and right as you turn the attenuator up and down. One solution is a stepped attenuator where each setting is matched carefully to make sure that both channels are being attenuated by the same amount, within some tolerance, at each setting. That's what I went with. Wasn't cheap either- $300 just for the raw step attenuator (channel attenuation within .03 dB between channels, 24 steps), and I still have to spring for the housing and connectors and having it assembled. Probably somewhere you can get a less fanatically fussy one for cheaper, but at those prices the Coleman attenuator boxes and that start to seem very reasonably priced- there's a whole thread on that on gearslutz if you search for it.

[Robotnik]

Originally posted by jbuntz
I noticed you said you like to boost 31.5 hz by 12 dB on kick. Is there any concern with making real world power amplifiers crap out?

I was exaggerating.

[chessparov]

Robotnik's avatar is inherently distracting for those more interested in 50D and "below".

Chris

[Duardo]

In as nutshell... Yes, because the harmonics of the frequencies within the audio bandwidth are lurking there. There's often talk of Geoff Emerick hearing the treble boost from a dry soldered joint on a Neve transformer compensation network and I hear the effect (even with my cloth ears) on Neve racks of later units (3114/5/31105, etc) where the rack guy has left off the 600 ohm terminations producing a big hump (around +10dB) around 35 to 40KHz.

Right, it's important that analog gear be flat well beyond what we can hear...but do A/D converters really need to capture all of that as well? If Geoff could hear the probems with that 3 dB 54 kHz bump on that console, or you can hear the effects of the humps around 35 to 40 kHz, should we not be able to assume that a good 44.1 kHz converter could capture those audible artifacts as well? They'd still be there, wouldn't they, even if the center frequencies weren't? I know the effects of the various "air band"-type equalizers can easily be heard when used in a 44.1 kHz recording.

Believe me, mess with the HF stuff and you will hear the effect lower down... note I say "effect", I'm not saying that you hear 40KHz.

Exactly. So do we really need to capture that? I would assume that even the analog circuitry in a 44.1 khz converter would have to be flat well beyond the range of the frequencies that come out the other end of that converter.

The proof of the pudding is in the eating and why PT and others have gone up the 96KHz clocks (and beyond).

I was under the impression that PT went up to 96 kHz because the market demanded it...

-Duardo

[Ted Nightshade]

Originally posted by Duardo
Right, it's important that analog gear be flat well beyond what we can hear...but do A/D converters really need to capture all of that as well? If Geoff could hear the probems with that 3 dB 54 kHz bump on that console, or you can hear the effects of the humps around 35 to 40 kHz, should we not be able to assume that a good 44.1 kHz converter could capture those audible artifacts as well? They'd still be there, wouldn't they, even if the center frequencies weren't? I know the effects of the various "air band"-type equalizers can easily be heard when used in a 44.1 kHz recording.



Exactly. So do we really need to capture that? I would assume that even the analog circuitry in a 44.1 khz converter would have to be flat well beyond the range of the frequencies that come out the other end of that converter.



I was under the impression that PT went up to 96 kHz because the market demanded it...

-Duardo

Actually PT beat most manufacturers to 192! Many still aren't there...

The issue of the filter at the Nyquist cutoff comes in. At 44.1 kHz the filter, be it analog or digital, must be rather steep. Hard to do in a way that's artifact free. Nonetheless it can be done rather well, and probably even better in the future. I'm looking forward to experimenting with this SLAM! headed my way, with it's selectable filter frequencies, but using the HEDD I found that most times I really did prefer 44.1, but sometimes 96 was preferable- depended on the program.

[Duardo]

Yeah, that's true, although most manufacturers did beat them to 96 kHz...and I imagine they did 192 because it was already part of the DVD-Audio specification...who knows.

It just wonder how much the envelope will be pushed as far as filter design, etc with respect to 44.1 kHz in the future...may not make much sense from a marketing perspective, although since CD seems likely to be the standard delivery format for some time to come (along with lesser formats like MP3) seems like squeezing as much as possible out of 44.1 kHz is a good idea.

-Duardo

[Geoff_T]

Originally posted by Duardo
Exactly. So do we really need to capture that? I would assume that even the analog circuitry in a 44.1 khz converter would have to be flat well beyond the range of the frequencies that come out the other end of that converter.

-Duardo

Hi

I'm not an expert on things with digits but the point with 44.1KHz surely is that it can't handle an audio signal greater than half the clocked sampling rate. So there's a barn door filter of around 96dB/octave around the 20KHz mark.

The convertor's analogue circuitry's role is to chop all the harmonics off...

Most ancient analogue circuits could handle a 44KHz signal (albeit not a square wave) with a loss of around 3dB or better.

[Duardo]

Right, I should have clarified that I meant any circuitry aside from the filter...and as I understand it, even the analog filters in today's converters don't start rolling off until well after 20 kHz and do so very gently after that, and then the digital filters after the initial sampling stage do the brick wall thing.

-Duardo

[Ted Nightshade]

Originally posted by Duardo
Right, I should have clarified that I meant any circuitry aside from the filter...and as I understand it, even the analog filters in today's converters don't start rolling off until well after 20 kHz and do so very gently after that, and then the digital filters after the initial sampling stage do the brick wall thing.

-Duardo

Still, the "brick wall" thing. (I like "barn door"!)

I'm with you on hoping that the 44.1 filtering continues to be a priority. Some of us in the wealthy part of the world may be going to higher sampling rate delivery mediums, but there are a hell of a lot of CD players out there, and will be for a long time.

[Blinddot]

I rescue this thread in order to post this link which I´ve found very interesting. It´s a white paper from earthworks

http://www.earthworksaudio.com/f_wpa...yond20khz.html

I haven´t found it after doing a bit of work with the search feature, so I thoght It could be useful to have at hand here at gearslutz..

Just forgive me if it doesn´t sound interesting to you..


saludos

[lucey]

Originally posted by Geoff_T
There's often talk of Geoff Emerick hearing the treble boost from a dry soldered joint on a Neve transformer compensation network

please tell this story again Geoff ... it'll send a few heads spinning!

[Drumsound]

Without even reading the responses...


YES!

[Renie]

Originally posted by fishtop_records
Most male humans over 25 or so are
down 10 dB or more over 15kHz.
The "standard" 20 to 20kHz is
typical and round and traditional.
For decades, everyone designed things
to fall off over 20kHz, and often things
were down 5 to 10 dB at 20kHz.

Most speakers (monitors, etc.)
have a real hard time over 20kHz.

There are interharmonic modulations
that occur between any two signals.
You could have one 36kHz signal
and another at 38kHz, and
the difference would be 2kHz,
which is very audible.

Plus, cheap filters that chop off
at 12db/octave totally trash the
phase response, and even male
humans are very sensitive to phase.
Phase is how we tell where the
jaguar is, which is how we
avoid being lunch.

Pat

Have you any similar info on female hearing?

thanks

[Jules]

interesting article BlindDot!

[Blinddot]

Thanks Jules, It has been the most interesting article I´ve read lately since the Lavry´s white paper on Dan´s aproach to sampling theory.

I´ve found very interesting what David says about metal dome tweeters, I´ve always found these kind of tweteers very harsh(as he say he does) but from time to time some of them just don´t sound that harsh, and now I understand why I can enjoy that very good transient response from the metal dome tweeters on the Yamahas MSP5 and MSP10 studio, without much of a harsh sound (like I hear on genelecs for example). It must be thanks to their extended HF response even when the rest of the system doesn´t reach those frecuencies. (at least my system) I deduce after reading the article, that it must be all about material resonances and I´d like to see studio monitors manofacturers to take advantage on this kind of stuff soon.


just forgive my poor english

[Brad Lunde]

Originally posted by Duardo
Yeah, that's true, although most manufacturers did beat them to 96 kHz...and I imagine they did 192 because it was already part of the DVD-Audio specification...who knows.
<SNIP>
-Duardo

They hired Ed Mietner and gave him a budget. That's how it was done far as I can tell.
Brad

[toledo3]

To the person who commented about 10dB down at 12k-

My girlfriend is a B.A in Speech and Hearing Science, and I could have sworn that most people have something like 50-60dB down past 12k....

Is your stat that says 10dB down, in comparison to what "perfect human" hearing would be? If so, then it makes sense with what I have heard. It is my understanding that the 60dB figure I mention, is in relation to "perfect flat" frequency response.

To the comments about Emerick, etc....

If the 54k signal imbalance effects the sound of the range below 20k... then you aren't necessarily hearing 54k, are you? You are just hearing the audible effects below 20k. This may seem like I am splitting hairs, but I think it is important to note.

Thinking for a moment that he may have actually heard 54k, well... what speakers was he using that played that 54k back?

It is obvious that what was being heard was well within 20hz to 20k... The way that this story is usually repeated leaves a false impression.