|
A Discussion: Dynamic
Compression
(Fletcher-Munson Curves
Re-Examined)
(These are the notes from an
address to the Audiophile Society of NSW on 24th June 2007. Please note they
contain generalities due to time constraint and also keeping the audience in
mind. This is not a complete treatise on the subject but intended to raise
questions as much as answer them.)
Most audiophiles are fairly intelligent people, and if
you make the effort to explain clearly, most will usually understand. So
please pay close attention as we shall try to clearly explain why Dynamic
Compression explains why various similar equipment ought to sound the same
and yet doesn’t.
For example: Have you ever wondered why two amplifiers (or indeed two
speakers) can measure flat down to the same frequency and yet they often
sound VERY different. I remember a Classē DR-9 amp
were compared with a really good valve/hybrid amp and one listener said the
100 Watt Class A amp had "no bass." The valve amp was in fact less
measurably flat than the Classē .
How can that be? Well, your ears are not lying. While we all understand what
frequency response is, here we are talking about something that goes beyond
frequency response. Keep this in mind; it is not the frequency response but
the dynamic behaviour that is the key.
To help us understand we shall look at the famous, or rather infamous,
Fletcher-Munson curves.
F-M CURVES
Fletcher-Munson curves were conceived in 1933 and
although there have been the later revisions (Dadson-Robinson) and the newer
ISO standard variation. These are essentially minor changes. Our discussion
remains the same.
Take a look at the F-M curves. They are not frequency responses but
EQUAL LOUDNESS curves. At lower volumes the bass needs to be boosted to be
equal in loudness. Because of this F-M curves became mis-used in
earlier Hi-Fi equipment and became the false reason for Tone Controls and
some of us will also remember the Loudness Control switch on amplifiers.
Those days are behind us.
Features:
We have dB-SPL on the Left. Phon in the centre is relative to 1KHz (a kind
of reference point).
The BOTTOM line is the Threshold of Hearing.
The TOP line is the Threshold of Pain.
These two extremes define the 120dB dynamic range of human hearing.
Now what is 120dB in real terms? If we were to take a sound at the Threshold
of Pain and divide that by the smallest sound we can hear and then converted
that into amplitude terms, the ratio would million to one. But since it
takes energy to move sound through air, the real ratio is much higher than
that. Would anyone like to guess? 1,000,000,000,000:1 or a Trillion to One.
_________________________________________________________________________________
We can see that above 1KHz the curves are evenly spaced and all very
similar/consistent. But below 1KHz we see something rather different. As we
lower the frequency the lines bunch up.
Note the large vacant area at the bottom left hand side (sub 1KHz), NOTHING
can be heard.
For example, at 30Hz we hear nothing below 70dB SPL. Yet that is the
level of normal speech. At normal speech volume there is NO BASS!
_________________________________________________________________________________
Level of loudness to we experience:
Silent room: 40dB-SPL (C weighting or 25-30dB A weighting).
Soft whisper is 50dB.
Speaking normally and clearly is 65-70dB
Very loud talk is 80dB (louder than this)... and real loud yell 90dB.
A full orchestra climax is typically 95dB.
So if 70dB SPL is normal speech is considered average
loudness, not loud really. We just do NOT hear real bass at that level.
_________________________________________________________________________________
Another example of Dynamic Range we find is our sight.
Why is it that when you walk into a dark room it seems totally black but
after a short while you start to see things. Also, if you walk into the
light after being in the dark can be a painful experience. We have the same
kind of thresholds, minimum and threshold of pain. The pupil acts as an
aperture that regulates the amount of light that enters into the eye.
The pupil acts in a similar way to a volume control, and AUTOMATIC volume
control. The technical term is AGC, Automatic Gain Control.
Does ear do the same? Does the ear have an inbuilt AGC? Yes!
Anyone know how it works? For example, in the Middle Ear, we three bones or
ossicles, Malleus, Incus and Stapes (hammer, anvil & stirrup). Various
nerves and tissues tension these and other factors vary the sensitivity of
the ear. Hence we have a mechanism to compress Dynamic Range.
_________________________________________________________________________________
***BUT***
They don't vary ALL frequencies the same. In fact the progressively do so
less as we go below 1KHz - as F-M curves indicate.
Question: Can we redraw F-M curves, not as equal loudness curves, but as
frequency responses? YES!
INVERTED F-M CURVES
They have only been redrawn below 1KHz, as above 1KHz
there is little change. By redrawing in this way we can now examine the
dynamic behaviour of the ear at frequencies below 1KHz.
"DYNAMIC" behaviour, what do we mean? "Dynamic" is the opposite Static.
Static means constant and dynamic means constantly changing.
A bit like stationary versus moving.
For example, measurements of audio equipment are STATIC in nature; sine wave
based frequency response, distortion figures, Noise etc., they are all
STATIC measurements.
But music is not static in nature. It is DYNAMIC. Constantly changing in
level and frequency.
So is it any surprise that our hearing is DYNAMIC (just as our eyes are
dynamic)?
BTW, all our five senses, hearing, sight, touch, smell and taste, are
dynamic. Therein lies the spice of life.
Basically at mid and high frequencies our hearing is compressed. Yes, we
have dynamic compression built into our ears and for good reasons. Here
compression is not necessarily bad. Of course the last thing we want is our
equipment to have compressed dynamics. But look very carefully and you will
realise that dynamics gradually become less compressed below 1KHz and by 30
Hertz there is little or no compression. Note the even 10dB spacing at 30
Hertz.
Here we come to the main punch line of our discussion. We can make TWO
statements:
1) OUR HEARING IS SENSITIVE TO DISTORTION AT MID AND
HIGH FREQUENCIES AND LESS SENSITIVE TO DYNAMICS AS THE ARE COMPRESSED.
2) AT LOWER FREQUENCIES WE ARE LESS SENSITIVE TO
DISTORTION (we can tolerate a little more) BUT HIGHLY SENSITIVE TO DYNAMIC
CHANGES IN LEVEL, BECAUSE DYNAMICS ARE NOT COMPRESSED.
_________________________________________________________________________________
In that case, what initial conclusions can we draw from those two
statements?
1. If 70dB SPL (or 70 Phon) were not
compressed by 70dB - speech would be virtually inaudible, because 70dB would
become the new threshold of hearing. Regular day speech would largely be
impossible. We would constantly have to scream at each other.
2. Deep Bass is virtually inaudible below
60-70dB (SPL) because of lack of compression.
_________________________________________________________________________________
Example: High Voltage Lines
Let's change subject slightly and ask: Why are high voltage lines so thin?
Because the higher the voltage, the less current is required. There are
fewer losses at higher voltages and cables also will sag less and run much
cooler. Typically HT lines can be as high as 500KV, half a million Volts.
True or false, voltage does not compress, current does? Yes! The current is
the problem.
Is this why many of us love valve amps? Does valve
amps have superior and more natural dynamics because they are high voltage
and low current devices.
Believe me when I say that is much easier to mess up power supplies in
transistor amps that it is in tube amps. Typically voltages in tube amps are
up to ten times higher than transistor amps and that means up to one-tenth
the current.
Also, because AC is converted to DC, in typical power
supplies the peak current drawn out of the transformer can be 7-10 times
higher than the current required by the amplifier. So if the speaker
requires 5 Amps, then the peak current drawn in your power transformer can
be 30 to 50 Amps. AC has gaps in the power delivery, the power supply's
reservoir capacitors in-rush current has to compensate. Yes, current
is the problem.
Current equals dynamic compression unless very close attention is paid to
it. Tubes are less prone to compression.
_________________________________________________________________________________
Why is compression so more audible at lower
frequencies?
Let’s say we drop the volume at 1KHz by 10dB while listening at a healthy
90dB SPL. The ear now corrects that so we only hear it as a 3-4dB drop. Drop
volume down to 70dB SPL (-20dB total) and the perceived drop has barely
increased to -5dB. Yes we can hear that. But at 30 Hertz we hear the total
20dB drop because there is no compression.
At mid frequencies 2dB drop is noticeable, but 1dB
much less so and 0.5dB would hardly be noticed if at all. But these small
changes would be FAR MORE noticeable at low frequencies where the changes
are dramatically expanded.
Specifications show measured distortion, freq
responses etc, but nothing about the equipments dynamic behaviour. In all
cases it is because the better component has lower compression and superior
dynamics while still remaining acceptably low distortion. Now we know why
Classē DR-9 had "no bass" even though specifications disagreed. We heard an
amplifier deficient in dynamic behaviour that specifications cannot reveal.
Here are some more over-looked examples:
Dynamic (conventional) speakers are inherently compressors of sound. Is that
why some like panel speakers like Quads. Why do others love horns? Answer is
obvious. But why do some like Lowthers? Because they generate less heat in
the voice coils - heat causes compression. More insensitive speakers are
more prone to dynamic compression.
(I note that the point was made during the
discussion that putting a horn in front of a driver did not reduce heat in
the voice coil. Quite correct. But if you only require a quarter of the
power for your desired volume level, does that not mean your voice coil only
have to dissipate a quarter of the same heat? And is then the opposite not
also true? If a speaker is less sensitive you have to put more power into
it, more heat and more compression.)*
Why are some claiming some magnets in speakers sound better than others,
like Alnico magnets? Could it be that they are
more stable under dynamic conditions? And on it goes!!!
_________________________________________________________________________________
Some other observations that reveal dynamic
compression:
Have you noticed that people resort to loudness to compensate for poor
dynamics? Loudness is not the same as dynamics. Just look at in-car stereo
systems. They seem to only sound dynamic when the volume is wound up
excessively.
Separation of musical instruments? We all know what that means and like to
hear it. Separation of musical instruments is related to dynamic behaviour,
it should be effortless. True dynamics are never FORCED! If your system is
not able to separate instruments, you have poor dynamics.
Masking effect:
IF a loud sound exists at the same time as a soft sound, the ear will adjust
the dynamic range to accommodate the loud and tend to mask the softer sound.
But this is LESS true of low frequencies. Because there is little or no
compression, we can hear more distinct level of bass with virtually zero
masking. We can hear quite loud bass and a delicate mid or treble sound
clearly at the same time. Some composers unconsciously take advantage of
that.
Dynamics are as much DOWN as it is up:
The best equipment seems to defy at times the masking
effect; Allen Wright has dubbed this as DDR, Downwards Dynamic Range.
Tympani sound fast when they stop fast. Bass overhang - after the event -
slows down dynamics. Dynamics perception is enhanced by sudden stops.
Have you heard some reviewers referring to some pieces of equipment having a
“black background”? That indicates low compression as well as
low noise.
_______________________________________________________________________________
We have only scratched the surface. There is much more.
One final example:
For dynamics to be effective at mid frequencies, it must have a sharper
leading transient edge. Compression has a response time. Our hearing is more
susceptible to damage at mid frequencies. Our low frequency hearing is more
robust. At mids and highs the response time is quite fast (protective, but
don't over do it) but at low frequencies, which have lower rise time, much
less so.
This may indicate that we are even more discernable to dynamic
behaviour than even F-M curves indicate.
Give it some thought.
Joe Rasmussen
At The Cross Roads.
Postscript:
Why does heat cause
compression in coventional cone driver (also called 'dynamic driver' but not
to be confused with our topic)? Efficiency ηo
is defined by the following equation
:
ηo is
defined by three Thiele-Small Parameters as shown, Fs (Free Air Resonance),
Vas (Compliance Value) and finally Qes. It is the last one that is of
interest here as it is related to the driving force. The higher the Qes the
lower the efficiency. But heat modifies Qes by the same ratio as the
increase of the DC resistance of the voice coil. Heat increases DC
resistance (DCR), so if DCR is increased by 10% then Qes also goes up by
10%. This will likely cause about 1dB loss. It is not difficult to
understand that a voice coil heat rises on the crest of a musical transient
and hence the transient is being slowed down, not rising to its full level
(a softer transient immediately afterwards is also effected as now the level
has changed). We have a classic mechanism causing compression.
Flat
membrane speakers do have an inherent advantage. In the case of ribbons
(both real and pseudo) we have a large area acting as a natural heatsink and
likewise electrostatics
have no enclosed voice coils.
But conventional speakers have voice coils trapped in a confined gap and has
to radiate heat into the magnetic/motor structure. That is why some drivers
have hollowed out pole pieces and/or raised spiders (suspension). These
methods help reducing heat and keep things more thermally stable. This also
shows the designers as being aware of the problem. BTW, metal cone drivers
must also have some advantage (but as all other things are not equal, all
drivers will have other problems as well, unless they are perfect - hah hah). |
