This is one of those explanations I am going to blog about
that is written more to someone who is a novice, not an electrical engineer,
and is curious about why things work the way they do. I have been tweaking the new tweeter in my
speakers trying to balance the loudness between the individual drivers and wanted
to share with you something that can help you understand one of the reasons why
speakers that may even look the same sound different. When designing a speaker from scratch,
understand that each driver used in any speaker is either a little louder or a
little softer than the other. When one
of the drivers is a lot louder, then that driver must be somehow made
quieter. The way driver loudness is
matched is to use what is called an attenuator: a quieting device connected
between the driver and the components of the crossover network.
Attenuation is necessary to balance out all drivers to the
same levels where woofers are typically the least loud and the other drivers quieted
to match the woofer’s level. High-end
manufacturers take great pains to match the driver levels precisely to produce
the highest quality sound possible. How
this is usually achieved in a high-end speaker is through an impedance matching
circuit called a T-pad. Unlike variable-resistance
L-pads found in mid-fi speakers (those knobs on the back that say plus or
minus, or something like that), T-pads precisely match the impedance the amplifier
is expecting to see to the actual driver impedance (the 8 ohm load amplifier to
a driver that may not be close to 8 ohms).
But why is this so important?
Let’s see how another example in a different technology is applicable to
audio attenuators.
If you have ever owned or known someone with a Citizens Band
or HAM radio, one of the things done to assure optimum transmission range and
best signal reception is to tune something called the standing wave ratio (SWR)
between the transceiver and the antenna.
This is done either by adjusting the length of antenna cable or by
tuning a portion of the circuit with passive components (some sort of
adjustment at the base of the antenna). If
the SWR is poor, the CB will not send out a strong signal as opposed to when
with SWR is very good (tuned in). (Search
for SWR
tuning to understand more about what this phenomenon does and think more
about how it relates to your stereo.) Connecting
an amplifier to an audio attenuator of a speaker encounters a similar issue that
a CB radio has when connecting to its antenna.
AND connecting a driver to this same audio attenuator encounters
additional problems as we will soon see.
Issues are encountered between the amplifier and the driver
because of the selected values for the components to make the attenuator (see this
link to understand how changing optimum values to available values
encounter these issues). If the
attenuator does not precisely match what the amplifier is most efficient at
driving (called its input impedance)
AND the driver does not match what the attenuator is most efficient at driving (called
its output impedance), issues develop
at each end of the attenuator.
Do this simple experiment.
Use the following three values in the white boxes at the top of the page
in the link
and then click on CALCULATE:
- Input Impedance=8
- Output Impedance=7
- Required Attenuation=5
The calculator shows the optimum resistor values required to
achieve this level of attenuation in the three orange boxes (3.091, 1.066, and
12.309). Now no one makes off-the-shelf resistors
of these exact values so a compromise must be made and herein lies the issue. There are values available of 3.0, 1.0, and
12.0 ohms, a tolerance error of less than 1% for each resistor. Plug these values (3, 1, and 12 respectively)
into the orange boxes and click on CALCULATE once again. This will show you the effects of minor 1% changes
in these resistance values has on the optimum performance of this
attenuator. You will notice that everything
changes. The closer the values used in
making the attenuator are to the optimum values recommended in this calculator,
the better the driver will sound. BTW,
desired values are: VSWR=1,
Reflection Coefficient=0, and Return
Loss=[higher is better].
Are you with me so far?
If not let’s summarize. If your
speakers are rated at 8 ohms, the amplifier expects all of the drivers in the
speaker to be 8 ohms so it can send equal amounts of power to them all. Since one or more of the drivers in your
speaker are not equally as loud, you must make the louder drivers more quiet to
make the sound from each driver “balanced.”
Attenuators (called L-pads or T-pads) are used to quiet down these loud
drivers. When designing attenuators,
optimum values produce optimum results. The
greater the deviation from these optimum values, the more the sound of the
driver will change.
Poorly designed attenuators result in large changes in the
way a speaker driver will sound. These
changes are indicated in this
calculator when the VSWR is not exactly equal to “1” and the Output
Reflection Coefficient is not exactly equal to “0” for either the Input or
Output sides of the attenuator. When
these values are precisely 1 and 0, the Input and Output Return Loss values are
as high as they can be (what you want to achieve in the highest-quality design
possible).
Enough of the calculator and theoretical explanation, let’s
get down to the real thing and what this all means to what you hear. It seems that all things in life are a series
of compromises and so it is also with tweeter tuning. The Peavey RD1.6 tweeter I use in my Bozak
speakers is much louder than the woofer.
To balance the sound, I must use a well-designed attenuator to retain
the quality of sound this driver is capable of producing. So let’s see what audible subjective
observations can tell us about the measurable effects of using the same
attenuator with resistor values of different tolerances (1% tolerance, 5%
tolerance, etc.). I will use RTA
measurements from my speakers to demonstrate the measurable changes and
subjective comments on what these changes sounded like. In other words, we won’t change a single
thing about the speaker, just the accuracy of the resistors in the attenuator.
First, let’s look at what the speaker did with no
attenuation on the tweeter to demonstrate what an unbalanced level between
drivers is. Know that everything that
was needed to assure I was measuring apples-to-apples was done and that nothing
I did in the measurement process changed the findings.
Tweeter
with No Attenuation
It doesn’t appear that much is needed to balance the sound of the drivers, so the first design used on-hand components. What was possible with the readily-available parts I had in my parts drawer was a -6.69dB attenuator, pretty close to the estimated -5dB. With a bit of solder and some jumper wires, I fashioned this attenuator and attached between the driver and the crossover network. The measured results are shown below.
Tweeter
with -6.69dB Attenuation
This measured nice despite the loudness mismatch and HF roll-off above 8KHz. Subjectively, the sound was pretty dull and the ambience in concert halls was fairly suppressed, although present if you strained your ears and listened through these mismatched levels. The tininess of the overall sound was completely gone and had a very attractive warmth especially in the woodwinds. The timbre of drums was excellent and vocals natural and unstrained. The quality of the sound was acceptable although not what one would consider to be high-end (yup, this means the design is junk; back to the
Next was a properly designed attenuator again using available components and -3.98dB of attenuation. Also added was a 0.01uF Teflon shunt capacitor across the 4.0uF polystyrene crossover network capacitor to help with the high-frequency roll-off issue above 8KHz.
drawing boards).
Tweeter
with -3.98dB Attenuation
The level measured a pretty decent match with a surprising smoothing in the 4-8KHz region (non expected but greatly appreciated). An interesting side effect was the additional smoothing of the midrange in the 500Hz-2KH range. These are the “trickle-down” effects of a well-designed attenuator on the sound of the entire speaker. Now the sound of this attenuator is still a little on the bright side with sibilance noticeable but not objectionable. Hall ambience is very rich and full creating a large sound stage, one of the largest the system has ever produced. But after prolonged listening, it was clear that the minor over-emphasis in the 2-5KHz region needed to be tamed ever so slightly. This attenuation level is pretty “close” to what I audibly perceive as a good quality level of sonic reproduction so I knew I was getting close to the final design worthy of the term high-end.
The next attenuator was about -1dB quieter. Although consisting of more than one resistor per leg (there are three legs in a T-pad), all of the resistors used in the attenuator were of the non-inductive type and one was matched to only 11.5% of the required value (not as precise as I wanted but the only values I had on hand). Now I suspected that this minute level of change would be inaudible so I anticipated that measurements were really where any differences would be noted and my ears would be able to subjectively detect nothing. Let’s first see first what the meter tells us.
Tweeter
with -4.97dB Attenuation
The meter shows a pretty significant change in the overall character of the system. Starting with the bass, there is a new peak at about 250Hz and a more pronounced dip at 500Hz. The midrange between 500Hz and 2KHz is not as uniform as in the well-matched -3.98dB attenuator, and the tweeter is not really attenuated as anticipated. The sound is very different where the emphasis of the system is shifted to the regions around the crossover point at 2.8KHz. There is still clarity in stringed instruments and sibilance is a little worse and less controlled, but nothing to write home about. What is most depressing is that there is a definite non-uniformity in the overall sound again with an overbearing dullness not indicative of a true high-end system and something that is a definite mistake. Cat Steven’s “Father and Son” is very revealing of this attenuator’s shortcomings so much so that I need not ask my wife if “this is a good change” for indeed it is not.
What did the calculator predict with the mismatch? Let’s see and use what it tells us to correlate what I heard against what was predicted. For this attenuator, input impedance=8, output impedance=7, and attenuation=4.97. Optimal values are therefore 3.084, 1.149, and 12.391. I used 3.0, 1.0, and 12.0 resistors causing the VSWR to change from 1 to 1.026 and 1.039, apparently minor shifts by the numbers (less that 4%). But the significant change in the sound told more than the VSWR would be a good indicator of how it would sound (good or bad). The VSWR for the better-matched -3.98dB attenuator was 1.013 and 1.003 (less that 1.5%). Putting it another way, a 4% VSWR is not good enough and less than 1.5% is audibly much, much better. Back to the drawing boards.
I have returned the attenuator to the -3.96dB version and I am far more pleased with the overall sound. While still not pleased with the entire performance of this attenuator, the sound is far superior to the -4.97dB more than 10% mismatched version. One day, when I order matched non-inductive resistors for this attenuator version, I will re-visit it to confirm the suspicion that it was the mismatch that caused the huge change in tonal balance. Hopefully the shortcomings of the present -4.97dB attenuator will also be addressed and this particular link in the chain will be made as strong as it can possibly be.
This exercise hopefully demonstrated to you the need to invest serious time in precisely matching values in if nothing else the modest audio speaker attenuator. If you correlate these findings with other potential sources of similar sonic anomalies in amplifier and preamplifier design (transistor matching, tube matching, resistor matching, capacitor matching), you can begin to appreciate what it takes to produce a thoroughbred piece of audio gear. What reared its ugly head in the audio attenuator is a symptom of what care must be taken when designing and even more important in manufacturing a fine piece of audio gear regardless of type (speaker, amplifier, CD player, active equalizer, etc.). High tolerance capacitors and resistors are essential as are matching active devices. This will assure all of the sound each piece of gear is capable of producing will in fact make it unaltered to the output.
Like any chain, your audio system is only as strong as its weakest link. Extreme attention to detail AT ALL LEVELS must be made to assure the entire chain is as strong as it can possibly be. Hand matching components in the circuitry of your existing equipment to within 0.1% (or better) of optimum values should be something DIY hackers could easily accomplish and improve the sound coming from an average piece of gear (not from your high-end gear since hopefully this has already been done).
Yours for higher fidelity,
Philip Rastocny
I do not use ads in this blog to help support my efforts. If you like what you are reading, please remember to reciprocate, My newest title is called Where, oh Where did the Star of Bethlehem Go? It’s an astronomer’s look at what this celestial object may have been, who the "Wise Men" were, and where they came from. Written in an investigative journalism style, it targets one star that has never been considered before and builds a solid case for its candidacy.
