Can you or can you not hear the difference between 48k and 96k? (06-08-01)

[Nika]

Yes, I don't like the wording of the question that I THINK they're asking and that most people are answering.

Cheers!
Nika.

[treidm]

Getting off the subject for a moment, why do all the unregistered posts show up blocked on my screen? Is this a problem with my browser or an internal problem?
Reid

[guitarjoe]

OH Yeah but how come my egosystems Pci card with 24 bit converters sounds so much more crisp and full at 96 than at 44.1 You really could'nt miss the difference. i wanted to send the card back until I got my software right and it would work with 96 not 44.1.Maybe with the apogee and other expensive converters the difference is less dramatic????

Thanks

[xstatic]

Maybe the real answer here is that we can not HEAR the difference, but we can Perceive it. In my opinion the real myth that needs dethroning is that all 24 bit, 96 k etc is the same. I am so sick of people telling me how awesome there system is with their on board 24 bit 96k crap. Give me high quality 16 bit 44 k anyday instead of sticking with some cheap 24 bit 96 k converter. Just look at the old Radar systems and the old sony and Mitsubishi digi reels. some of those are under 16 bits and still sound amazing. To me the good stuff is all in the design, parts, and build engineering. Not in the bits and freeqs

[DAS]

Lots of converters sound better at one sample rate than another. There are many reasons for this, but it is generally considered a flaw in the converter design.

[guitarjoe]

you can definetly hear a difference at 96 and of course analog sounds better it is just to expensive and slow

[MidiMagic]

The top end phase is lost at 48K.

But since it will end up on a CD, it makes no difference.

[Red Prince]

If the ear can't hear above 20kHz and the Nyquist frequency is indeed correct then why do we need higher sampling frequencies?

Because Nyquist theorem states that a wave of any frequency below half the sampling rate can be properly reconstructed but it does not state that every single DA converter on the market will reconstruct it accurately. If we all had perfect equipment, we would not need higher sampling frequencies.

Anyone who has high quality gear and is only recording for himself may only need to go slightly above 40kHz (so, 44.1kHz may be enough for him). But as long as we are recording for others to listen and we have no control of the quality of their equipment, we are better off with higher sampling frequencies. Of course, we need to use high quality recording equipment, too.

So, it is a pragmatic thing: Nyquist tells us what can be done, but not necessarily what is done.

[DAS]

Because Nyquist theorem states that a wave of any frequency below half the sampling rate can be properly reconstructed but it does not state that every single DA converter on the market will reconstruct it accurately. If we all had perfect equipment, we would not need higher sampling frequencies.

Anyone who has high quality gear and is only recording for himself may only need to go slightly above 40kHz (so, 44.1kHz may be enough for him). But as long as we are recording for others to listen and we have no control of the quality of their equipment, we are better off with higher sampling frequencies. Of course, we need to use high quality recording equipment, too.

So, it is a pragmatic thing: Nyquist tells us what can be done, but not necessarily what is done.

Based on where you appear to be going here...in what way does recording higher frequency signals (above 20 kHz) make the signals we do hear (below 20 kHz) come out better? This makes no sense. If the D/A converter is no good and can't reconstruct a signal within the boundaries of Nyquist then recording higher frequencies isn't going to fix it.

[Red Prince]

It is not that a DA converter is either good or no good at all, they all are just not made equal. Nyquist theorem is about mathematics, but electronic devices are physical, not mathematical. Some are better at doing their job than others (or else we would all be using consumer equipment). A sample of a higher frequency will give a less-than-ideal DAC more to work with so it can reconstruct the signal better.

There also are other reasons for higher sampling rates. To prevent aliasing, the analog signal has to be filtered, so frequencies equal to or higher than the Nyquist frequency never even make it to the ADC. But, again, this is the physical world, not the world of pure mathematics, so a filter cannot just let through everything below its cutoff frequency unmodified, and everything above it fully removed. Rather, there is a region around the cutoff frequency. As the frequencies come nearer to the cutoff frequency, they are gradually attenuated by the filter.

To get everything up to the 20kHz that we are all supposed to hear, to get all that through unmodified, it is better to build a filter that lets everything through up to a much higher frequency and adjust the sampling rate accordingly. That way the digital samples contain everything we can hear sampled exactly, plus some extra stuff sampled not so perfectly, but as long as we cannot hear the imperfectly sampled frequencies, we are fine.

This is the same as, for example, getting 20W speakers and sending only 4W to them. No distortion of the signal. Or as building a 600 lb elevator to only break if filled with 1200 lb (or probably a lot more).

When people report that they can hear a difference between something sampled at 48kHz and 96kHz, this is not because they are deceiving themselves as some seem to have implied in this thread, it is because the analog filters can do their job better (the filters are used both before the AD conversion and after the DA conversion). It is not about sampling/reconstructing the frequencies above 20kHz, it is about sampling/reconstructing those below it more precisely.

[DAS]

Yes, there is something to your points, but I think people are often fairly misguided about it, and I think that the way you state it is potentially misleading. Many people have deep misunderstandings about the way digital audio works (both at a theoretical and practical level), and this misunderstanding provides room for erroneous assumptions (many of which may seem like common sense) to creep in and take hold, which then allows for poor decisions in production techniques and equipment purchases.

For instance...

A sample of a higher frequency will give a less-than-ideal DAC more to work with so it can reconstruct the signal better.

This is vague enough to have several possible implications. One could interpret it to mean that higher frequencies (below Nyquist) are captured better by higher sample rates. In other words, with a sample rate of 44.1k my 15kHz sine wave will not be reproduced as accurately as it would with a 48k sample rate. Or that my 44.1k system reproduces my 10kHz audio signal more accurately than my 15kHz signal. This is one of those common sense conclusions some people jump to that is just dead wrong. I have no idea whether your statement was intended to apply that way or not, but someone could take it that way.

Or...

This is the same as, for example, getting 20W speakers and sending only 4W to them. No distortion of the signal. Or as building a 600 lb elevator to only break if filled with 1200 lb (or probably a lot more).

Analogies like this one are not that helpful and prey on the phenomenon I describe above. This stuff is to some extent common sense, but in digital audio these types of caparisons do not always apply. Actually they don't exactly apply in the examples you gave either. For instance, if you underpower a large speaker too much you can actually get more distortion.

But yes, filtering is a huge issue in digital audio, and it certainly is easier and better to build filters with gentle slopes that allow higher frequency signals to pass. And this would seem to require higher sample rates...however, with the development of high quality FIR filters (in the digital domain) and oversampling this is MUCH less of a concern than it once was. We now have the technology to do good filtering of signals that ultimately need to be band limited to just over 20 kHz. That said, I do agree that what people are often hearing when they report better/different sound at higher sampling rates are the different characters of different filter sets, but not the result of the acquisition of higher audio frequencies (that's really more of a side-effect).

However, contrary to your view, I strongly believe that some people do sometimes deceive themselves and hear what they want to hear. I've found that in rigorous objective listening tests people are not so able to hear the things they claim to hear. The trick is putting together a test that truly removes ALL of the other variables - extremely hard to do. And in some respects pointless, because....whatever makes anyone happy and whatever they think sounds good is absolutely what should be used. If you feel good about the gear you are using you will probably produce better music, which is what really counts in the end anyway. We sell audio gear. Feel free to buy more expensive equipment for high sampling rates: larger and more hard drives to store your high bandwidth audio, faster computers, more RAM, etc.... (just kidding) But seriously, this is all so much minutiae. Most of the mid-priced (and up) audio equipment sold these days is more than good enough to produce professional sounding results in the hands of someone who knows how to use it.

But....what if 384 kHz sampling rates are introduced tomorrow? Should we feel compelled to run out and upgrade? Based on the "more is better" statements I see people make in this thread and elsewhere it sounds like we should. Many of those statements are based on a very dubious set of assumptions. Of course, there are hi-fi and pro audio enthusiasts who will gladly spend $800 for a power cord because they believe it makes their system sound better. Whether or not it "really" does improve it is almost beside the point. Free will.

[Red Prince]

But....what if 384 kHz sampling rates are introduced tomorrow? Should we feel compelled to run out and upgrade?

I would not. I really see no point of going over 96kHz. That addresses the gentle slope issue to my personal satisfaction. I have actully been using 16-bit 48kHz but am moving up to 24-bit 96kHz with my next project.

As for needing bigger hard drives for that. Nope. Whenever mine fills up, I move the old data to a (data) DVD. Plus, the size of my audio files pales in comparison with my video files.

[analoguepilot]

im not sure... is the difference really all in our heads? when we see bigger numbers and words like clarity and definition, do our brains automatically percieve it to sound better? and since most people are now listening on ipods which are basically encoded mp4 files and mp3 players- why are we striving and using our last dollars to encrease the sonic depth of a song that will ultimately be compressed? and if a recording engineer mixes down a track in 96 and no one is there to hear it, does it really matter?

[cdanddvdpublisher]

I can't but I know people who claim to be able to.