View Full Version : How flat are my ears?

03-11-2007, 10:18 PM
Do not fool yourself into thinking that the ear is linear! Do not believe for one second that the ear is perfect! Far from it - but its non-linearity is remarkably predictable.

As an example, the Equal Loudness Contours (http://en.wikipedia.org/wiki/Equal-loudness_contour) describe the long standing observation that the ear has a very particular way of perceiving loudness versus frequency. This can be demonstrated very easily: adjust your listening level so that the music (e.g. an orchestra playing a symphony or similar) sounds full bodied and normal. Then dramatically reduce the volume to one-tenth and what do you notice? You can still clearly hear the middle frequencies but the bass has disappeared. Increase the volume, and the bass appears to return. This characteristic of the human ear (independent of race, creed or colour) is well documented, but recently with a substantial research effort (especially from Japan) the ISO 226 (http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=34222) Equal Loudness Contours have been revised. The revision show that much more bass is necessary than previously assumed for replay at a sub-optimal level to sound natural. We can draw some hard and fast rules from these results:

RULE No. 1: 'When music is replayed at home at a lower sound pressure level than you would have heard it live (in the concert hall or studio) you perceive the sound as bass-light." This situation is likely to apply to home listening. We normally listen at home to wide-bandwidth music at a substantially lower level than we would hear live*

RULE NO. 2: "When music is replayed at a higher level than you would have experienced live, you perceive the sound as bass heavy". This situation is most unlikely to apply to home listning as live music is so much louder than typically replay at home.

Now, to us here, this is no surprise at all. Nor is it any surprise to those BBC speaker designers in previous generations who developed the speakers we show here (http://www.harbeth.co.uk/usergroup/showthread.php?p=2375#post2375). In every case, it was impossible for the recording engineer (a Studio Manager in BBC-speak) to listen at anything like the true volume because of the danger of his monitored sound bleeding into other studios via the studio's building structure or common air conditioning system. And that is why, understanding the customer's likely replay level is the very first criteria on my work list. And also why so many hifi speaker sound thin and hard. Much more on this subject to follow.

* We mentioned here (http://www.harbeth.co.uk/usergroup/showthread.php?p=2350#post2350) a magazine article which suggested that '750W of power is necessary to reproduce acoustic peaks'. Whilst this may be technically true, it is completely irrelevant for socially responsible listening at home and positively dangerous for the hearing and for speaker units. The entire concept of high fidelity reproduction takes as a given that listening at home is at sub-real life volume levels. The trick is to make it sound natural at the lower level.

18-12-2007, 06:23 AM
Mr Shaw,

Many thanks for the information regarding Loudness Contours - it explains many of the listening experiences I have had, positive and negative.

I understand that you may not wish to divulge the specifics but can you tell me whether all Harbeth models are designed using the same loudness contour or is there a range? I ask this having read reviews online (StereoTimes) where the same reviewer states that the SHL5 comes alive at a lower volume compared to the other small speakers (all except Mon 40) which, everything else being equal, implies the SHL5 uses a quieter loudness contour. As someone who will buy speakers next year specifically for use at low volume levels this is especially relevant to me.


Colin Downes

18-12-2007, 10:43 AM
... Loudness Contours - it explains many of the listening experiences I have had, positive and negative ... can you tell me whether all Harbeth models are designed using the same loudness contour or is there a range?Hello Colin - and welcome. You make a very interesting point and in fact, I have been collecting information on the subject of equal loudness in preparation for a new and more detailed thread here, with supporting graphs. I'll try and attend to that over the Christmas break.

As a visitor said to me yesterday in connection with journalism "Does paper refuse ink?". He means, of course, that it is very easy to commit opinion to paper without scrutinising the facts quite as closely as one would wish. I am not really aware of designing for different listening levels and hence (perhaps) considering different loudness contours because (as I'll show later) the difference between say 83dB and 87dB - a sensitivity range that covers all Harbeths - has negligible influence on loudness perception. The key point here is that 83-87dB is almost certainly far below the level you'd have experienced at the recording which could easily have been 100-120dB and that 20+dB difference does have an influence on loudness perception. So, I wouldn't be concerned at all about selecting a Harbeth based on considerations of how loud you will play - they're all going to work equally well.

Hope this helps,


18-12-2007, 10:08 PM

Many thanks for the (surprisingly) swift reply. It is good to know that I don't need to worry about this particular issue regarding the Harbeth range, especially as I may be in the less than optimal situation of buying without a home demo next year (I will be moving to Norway and I cannot see any dealer near Bergen on your site). I am especially pleased that the Mon 40 (or 40.1) does not appear to be different, so if funds permit that would be my preferred option. Of course, if they have such a large back order that may be a moot point as I have no intention of waiting years to listen to music through speakers again :)


Colin Downes

19-12-2007, 09:40 AM
Noted Colin. You just happened to catch us at the start of the day reviewing overnight activity of the HUG. We are just building the first batch of M40.1 crossovers and matching them to the Master Reference Crossover now in anticipation of building the first sellable M40.1 - production will start in earnest after we return in the New Year.

08-04-2008, 05:52 PM
We've touched on the subject of how amplifiers, CD players and such have virtually perfect frequency responses - but what about the human ear? How flat are they? Audiophiles have total confidence in their hearing, but are they aware of how non-flat their ears are? The attached picture is redrawn from ISO226:2003 prepared by the world's top researchers into human audiology. They've revisited and revised the standard textbook curves from early in the 20th Century. Be prepared for a shock if you place total trust in your ears! Take careful note that the vertical scale is in divisions of 10dB spl where 6dB a doubling of sound pressure.

I've made this example at 80dB listening loudness which is a low-to-moderate listening level, as you would experience at home perhaps in the evening. The horizontal scale, 10Hz to 100kHz covers much more than the entire human hearing range, which is only about 20Hz to 20kHz. On the vertical axis at 80dB, I've predicted the frequency response of a normal, inexpensive hi-fi amplifier in green; perfectly flat in the 20Hz to 20kHz range. Then I've shown in red the inverse of the human ear's response - that means, what sounds 'flat' to the human ear at 80dB listening level. In blue I've marked the curiosities of the ear: there are some fascinatingly non-flat characteristics.

We can say from studying this graph and applying some background reading ...

1. That the ear needs a huge boost in the loudness of low frequencies for those low frequencies to be perceived as equally loud as the middle frequencies - hence the term equal loudness contours (http://www.lindos.co.uk/cgi-bin/FlexiData.cgi?SOURCE=Articles&VIEW=full&id=17).

2. There are some strange characteristic of the ear around 800Hz, 1.5kHz, 3kHz and 10kHz. All known healthy human ears regardless of race, creed or colour are known to have these characteristics - proof that we are all drawn from the same ancestors and that what sounds right to one ethnic group should also sound right to another.

3. The ear needs a dramatic boost in high frequencies (but not quite as much as it needs at low frequencies) for high frequencies to sound flat.

4. We should be very cautious about using the ear as a precision reference instrument as it clearly is not flat and we would be foolish indeed to rely on only the ear to critique or design audio equipment. We should use the ear in combination with technical equipment which is completely flat across the audio band to reach solid, reliable conclusions about audio armed with knowledge of how the ear can fool us.

5. This is the measurable side of human listening. Add to this the emotional dimension of listening and the vagueness of the human memory.

6. The ear transmits electrical signals to the brain by passing an chemical-electrical signal from nerve cell to nerve cell as fire fighters douse a fire with a bucket of water passed from person to person. This is an extremely slow chemical process. So slow in fact that above about 1kHz or so, the ear cannot trace the incoming signal as it has ceased by the time the signal has reached the brain. At 10kHz it is electro-chemically impossible for the brain to be receiving anything other than a vague sampled impression of the soundwave entering the ear. Beyond a critical (middle) frequency there is another highly mysterious (chemical) process at work in the ear/brain that senses the overall soundwave 'package' rather than the individual frequencies in the wave signal. In effect, an audio compression system is at work*. This is wholly different to audio test equipment which senses every discrete frequency from 20Hz to 20kHz (and beyond) in real-time. So to pursue super-wide bandwidth (96kHz?) seems to ignore the way the ear works.

Some kitchen table maths: Speed of signal along a nerve (http://www.answers.com/topic/nerve-conduction-study?cat=health) say, 50 mtrs./Sec. Let's guess that from ear to brain, one half metre of nerve. So the signal would take one hundredth of a second to reach the brain - about the periodicity of a 100Hz note. Contrast that with the velocity of an electrical signal in the amp (300,000,000 m/S) or a bullet in flight (1000m/S) and you'll appreciate that the signal our brain receives from our ears is at a comparative snails pace. Trust with caution only.

* Actually, the MP3 encode-decode system was developed after a careful study of the ear and brain and works by taking full advantage of the ears limited frequency (temporal) resolution especially in the upper frequencies. The key to MP3 is masking i.e. fooling the ear and that it does very effectively and cheaply.