Topic: Sennheiser 421s - pros, cons?

[Church-Audio]

Yeah, when you're talking about a digitial filter, there are all sorts of audible artifacts that can be introduced if you're not careful, but you can also build a much more agressive filter.  It's just math.  If you use a real analog filter, then you generally won't get audible artifacts unless something breaks into non-linear operation (like a spurious oscillation), but you can't be as agressive in removing undesired frequency components.

So when an analog signal is digitized at 44.1khz it first goes through an analog lowpass filter, as I understand it.  What slope is typically used? or where does the rolloff start if the signal must be way down by 22khz?

And if a digital file is resampled to 44.1khz or lowpassed to remove everything over 22khz, then the steep slope required to keep a relatively flat response up to 20khz is likely to produce audible artifacts when using simple digital techniques. 

Do I have it right?

It's not common to use more than about an 8 pole filter for an anti-aliasing filter ahead of the A/D.  If it's flat to 20 kHz, then you're not going to be very far down at 22 kHz, so aliasing happens.  That's all the more reason to justify putting your band edge lower, perhaps at 16 kHz.  That way, you get more attenuation by 22 kHz.  For anti-aliasing, you HAVE to use a real filter.  Common digital filtering techniques will not work to avoid audible aliasing if you are using a 44.1 kHz sample rate.  However if you are using 88.2 kHz or 96 kHz, then your anti-aliasing filter only has to be sufficiently effective by the time it gets to 44.1 kHz or 48 kHz, respectively.  You should be able to get 50 or 60 dB of attenuation of products above the 1/2 sample rate frequency if you use one of the higher sample rates.  Then you can actually use digital techniques to remove inaudible portions of the signal between the cutoff frequency of your anti-aliasing filter and your 1/2 sample frequency.  Of course, it depends if you really want to do that.  Some people (like Chris Church) think it's a bad idea to remove the portion of the signal spectrum that is inaudible and that's really the essence of our disagreement.  Can you hear it if inaudible information is removed from a signal?  I claim that you can't.  Chris claims that it makes a difference because somehow the inaudible portion of the signal makes the audible portion more real sounding.  He could be right, but I seriously doubt it.  If someone comes up with a reasonable procedure to create files to compare without introducing audible artifacts, then I guess we'll see.  One file will need to have ultrasonic information in it.  The other file should be exactly the same, except missing the ultrasonic information or at least it should not have as much ultrasonic information in it.  I doubt that this will be an easy test to construct.

Here is a MLS sample flat from 20hz to 90khz with a 20hz LPF butterworth 7th order with a band pass ripple of 0.5

[Church-Audio]

Yeah, when you're talking about a digitial filter, there are all sorts of audible artifacts that can be introduced if you're not careful, but you can also build a much more agressive filter.  It's just math.  If you use a real analog filter, then you generally won't get audible artifacts unless something breaks into non-linear operation (like a spurious oscillation), but you can't be as agressive in removing undesired frequency components.

So when an analog signal is digitized at 44.1khz it first goes through an analog lowpass filter, as I understand it.  What slope is typically used? or where does the rolloff start if the signal must be way down by 22khz?

And if a digital file is resampled to 44.1khz or lowpassed to remove everything over 22khz, then the steep slope required to keep a relatively flat response up to 20khz is likely to produce audible artifacts when using simple digital techniques. 

Do I have it right?

It's not common to use more than about an 8 pole filter for an anti-aliasing filter ahead of the A/D.  If it's flat to 20 kHz, then you're not going to be very far down at 22 kHz, so aliasing happens.  That's all the more reason to justify putting your band edge lower, perhaps at 16 kHz.  That way, you get more attenuation by 22 kHz.  For anti-aliasing, you HAVE to use a real filter.  Common digital filtering techniques will not work to avoid audible aliasing if you are using a 44.1 kHz sample rate.  However if you are using 88.2 kHz or 96 kHz, then your anti-aliasing filter only has to be sufficiently effective by the time it gets to 44.1 kHz or 48 kHz, respectively.  You should be able to get 50 or 60 dB of attenuation of products above the 1/2 sample rate frequency if you use one of the higher sample rates.  Then you can actually use digital techniques to remove inaudible portions of the signal between the cutoff frequency of your anti-aliasing filter and your 1/2 sample frequency.  Of course, it depends if you really want to do that.  Some people (like Chris Church) think it's a bad idea to remove the portion of the signal spectrum that is inaudible and that's really the essence of our disagreement.  Can you hear it if inaudible information is removed from a signal?  I claim that you can't.  Chris claims that it makes a difference because somehow the inaudible portion of the signal makes the audible portion more real sounding.  He could be right, but I seriously doubt it.  If someone comes up with a reasonable procedure to create files to compare without introducing audible artifacts, then I guess we'll see.  One file will need to have ultrasonic information in it.  The other file should be exactly the same, except missing the ultrasonic information or at least it should not have as much ultrasonic information in it.  I doubt that this will be an easy test to construct.

So if we believe what you’re saying then sympathetic resonation does not exist! I believe that ultra sonic information from say an instrument like a horn can go up to 40k, there are sympathetic resonations that are above 20k that will effect what we hear in the audible range. Why should we limit our selves to 20k? When it’s a fact that many instruments produce sound well above that point, how can you say that the information above 20k does not effect the information we hear below it?

[Chuck]

A justifiable reason to use higher sampling rates, say above 44.1kHz, is to have more natural high end extension due to the neccesity of having the anti-aliasing filter. I think that if equipment is engineered with high fidelity in mind, the limits imposed by the filtering can be overcome.

[Chuck]

...real world here...

[RebelRebel]

dan lavry says it best.

http://www.lavryengineering.com/documents/Sampling_Theory.pdf

[Gutbucket]

^^ Go Dan! That's a convincing argument.

Chris,
Your proposing that two ultasonic tones will amplitude modulate to create a 'beat' or third tone in the audible range?  Similar to the rythmic 'beat' between two close but not quite matching frequencies in the audible range; ie. the 'beat' sound that slows down as two guitar strings are brought into tune, extrapolated up until the 'beat' is in the audible range of frequencies?

I don't think anyone would argue the physical phenomenon doesn't exist.

I've heard of technolgy exploiting this that uses ultrasonic transducers to create a fixed ultrasonic tone and a modulated one so that their interaction formed a third signal in the audible range.  It was billed as a way of creating sound with tiny ultasonic transducers for things like laptops.  I think the downfall was that there were huge power requirements for the ultrasonic signals which had to be absolutely enourmous in amplitude to generate even a quite low amplitude audible signal.  Probably killed all the rodents and insects in the neighborhood of the lab.

If that is indeed the case, the audible range singnals induced by the modulation of the ultasonic components would be most likely too low in amplitude to be heard, and most certainly masked by the signals in the audible range.  I don't know much about the technical aspect of such modulation but I believe the carrier and modulating frequencies to do so would also have to be higher than several hundred khz.

Maybe not, I just found this white paper: http://64.227.81.118/hss/pdf/HSSWHTPAPERRevE.pdf
Non technical article version: http://www.elecdesign.com/Articles/ArticleID/5535/5535.html
That could be your ammo Chris, but I'm not sure it applies.
The military is using similar technology to beam auditory warnings to craft approaching US Naval vessels and to create non-leathal crowd dispersal techniques from afar in addition to the advertising uses of the manufacturer above.

Then again I also found this research by David Griesinger that argues no and address this specific issue:
Power point: http://world.std.com/~griesngr/intermod.ppt
or Google html version: http://72.14.205.104/search?q=cache:ymfUfDx8lnoJ:world.std.com/~griesngr/intermod.ppt+modulate+ultrasonics+to+make+audible+sound&hl=en&gl=us&ct=clnk&cd=12

-gone campin'

[SparkE!]

Here is a MLS sample flat from 20hz to 90khz with a 20hz LPF butterworth 7th order with a band pass ripple of 0.5

Chris, that's a digital implementation of a Butterworth filter.  You don't get skirts that steep with a real filter.  And remember you can't use a digital filter to avoid aliasing.  You can only use a real filter to avoid aliasing.  Notice how the stopband skirt levels out?  A real Butterworth filter's skirts keep going down as you increase frequency and it does so at a 6 db per pole per octave rate.

[SparkE!]

So if we believe what you’re saying then sympathetic resonation does not exist! I believe that ultra sonic information from say an instrument like a horn can go up to 40k, there are sympathetic resonations that are above 20k that will effect what we hear in the audible range. Why should we limit our selves to 20k? When it’s a fact that many instruments produce sound well above that point, how can you say that the information above 20k does not effect the information we hear below it?

No sympathetic resonance? I said no such thing.  Sympathetic resonance occurs at harmonics of the fundamental frequency of excitation.  By definition, those sympathetic resonances are at higher frequencies than the fundamental frequency that excited them.  If the sympathetic resonances are in the ultrasonic range, I claim that you can't hear them. You seem to believe that somehow they affect how you hear information in the audible range.  One of us is wrong.**  Let's find out which one of us is that person.  But let's make sure that the test itself is valid.  You can't allow your signal processing on ultrasonic signals to affect signals in the audible portion of the signal or the test will be invalid.

What we need are two signals.  Each should have identical signals within the audible portion of the audio spectrum (both in amplitude and in phase) and only one of them should have significant energy in the ultrasonic portion of the audio spectrum.  The trick is how to make those two signals without inducing changes in the audible portion of the audio spectrum.  You can't just go imposing some arbitrary amplitude shaping to the spectrum without also re-equalizing the phase in the affected portion of the spectrum. Otherwise, it's likely that you will inadvertently affect the audible portion of the audio signal too.

**Two men say they're Jesus.  One of them is wrong. -- Dire Straits

[Church-Audio]

So if we believe what you’re saying then sympathetic resonation does not exist! I believe that ultra sonic information from say an instrument like a horn can go up to 40k, there are sympathetic resonations that are above 20k that will effect what we hear in the audible range. Why should we limit our selves to 20k? When it’s a fact that many instruments produce sound well above that point, how can you say that the information above 20k does not effect the information we hear below it?

No sympathetic resonance? I said no such thing.  Sympathetic resonance occurs at harmonics of the fundamental frequency of excitation.  By definition, those sympathetic resonances are at higher frequencies than the fundamental frequency that excited them.  If the sympathetic resonances are in the ultrasonic range, I claim that you can't hear them. You seem to believe that somehow they affect how you hear information in the audible range.  One of us is wrong.**  Let's find out which one of us is that person.  But let's make sure that the test itself is valid.  You can't allow your signal processing on ultrasonic signals to affect signals in the audible portion of the signal or the test will be invalid.

What we need are two signals.  Each should have identical signals within the audible portion of the audio spectrum (both in amplitude and in phase) and only one of them should have significant energy in the ultrasonic portion of the audio spectrum.  The trick is how to make those two signals without inducing changes in the audible portion of the audio spectrum.  You can't just go imposing some arbitrary amplitude shaping to the spectrum without also re-equalizing the phase in the affected portion of the spectrum. Otherwise, it's likely that you will inadvertently affect the audible portion of the audio signal too.

**Two men say they're Jesus.  One of them is wrong. -- Dire Straits

Well I know I am correct so I win! 

[laptaper]

Okay, the recording that got me all excited about Senn 421's (remember those? ;-)) is now seeding at http://www.shnflac.net/details.php?id=836332f70b10bc45fbf711a399c2d4c20dabac68.  16 bit today, 24 bit tomorrow.  You can see for yourself if I'm right or if I need a hearing aid. :-)