12 Points About Tuning Cars
Why are there so many disagreements about how to tune a stereo system in a car??
People in our field are largely self-taught, and we tend to view our hard-won experiences as part of our identity. Humans in general are great at coming up with explanations to explain our experiences, and terrible at letting go of those explanations if they don’t fit the facts. When someone comes along with an explanation which seems to contradict our experience, we often react as if they are attacking us personally. It’s rarely about us, of course, but this kind of reaction is very common when discussing tuning a car-audio system.?
Neither our ears nor our brains can always be trusted. We sometimes hear differences which aren’t really there. We notice some things at certain times, but not at other times. A great read on this topic of perception can be had in the book Blink, by Malcolm Gladwell (although it's not specifically about hearing) - especially the part about the creation of New Coke. It’s a well-understood phenomenon that what listeners see influences what they notice.
Michael Lawrence, a pro audio engineer, has noticed a tendency he has named the Magnitude Fallacy.? It is the best explanation I’ve heard about how some of us tend to latch onto one specific aspect of acoustic sound-system adjustment, and elevate it far beyond its actual relevance. “Put simply, the Magnitude Fallacy states that some things are absolutely real, verifiable occurrences, but are not nearly as important or critical as we might tend to think. Situations for which the proper response is “yes, but don’t worry about it".” As far as tuning audio systems goes, it’s important to emphasize the most important aspects.?
There is often insufficient understanding about how to measure what we hear, and how to interpret what we measure. If we don’t understand how to make these connections, we might react by shunning measurements completely - and this takes away an invaluable analysis tool! If we can’t always trust our hearing from moment to moment, objective measurements are a way to help us collect information so we can evaluate it.?
First, let’s see if we can agree on our goal. Let’s leave out things like “setting input sensitivity” and “setting crossovers”, because even without a DSP processor, we should be doing these things. After all these system design tasks, here’s what’s left:
Left and right need to be largely in phase with each other at the listening position, left and right need to sound reasonably similar, and the sum of the two sides needs to sound pleasant (what that means, we will get to in a moment!) When this is achieved, the system will sound good, the stereo image will be in between the speakers, and the bass will seem to be up front (even if the subwoofer is in the back).?
OK, that said, here are 12 Points About Tuning:
1. The most important part of sound quality is the Frequency Response. This has been demonstrated in many studies. If we don’t manage the frequency response, we aren’t taking responsibility for the results. If we don’t manage the frequency response, we can’t fix it afterwards with better cables, sound damping, or High Resolution Audio. Read on for more discussion of how we can define a good-sounding frequency response.?
2. A real-time analyzer is a basic way to measure sound-pressure at various frequencies, and that makes it the perfect tool for measuring frequency response. Single-microphone RTAs are what most of us have available to us, but it should be noted that with single-mic systems, we may see small variations above 1500 Hz which will go away if we move our microphone a small amount, and which should not be addressed with EQ (more on this matter another time). Multiple-microphone systems often support averaging the frequency response at several points, eliminating these artifacts. 1/3-octave resolution - RTAs with 30 or 31 bands, using the ISO-defined center frequencies - have been very common, but are of limited use with electrical signals, and do not show us detail in the bass. Using higher resolutions, such as 1/48th and 1/24th octave, can reveal important details in electrical signal analysis, but often look confusing and overwhelming when used for acoustic analysis in a vehicle.?
When first measuring sound, we often measure the overall system at one time. This is reasonable, but it will only tell us a partial story about how the system sounds. We often need to measure one speaker at a time, or one channel at a time, or one side at a time - and usually some combination of these. That will help us understand how the speakers are performing, and how they combine into a speaker system. I didn’t understand this for years, and it limited the value I could get from using an RTA.
When some of us have struggled in correlating our measurement results to our listening experience, we may have abandoned measuring sound completely, and make all our tuning adjustments by ear. Since this by-ear process is very inefficient, and cannot be effectively taught, tuning by ear is not a viable commercial endeavor (that is to say, it's very difficult to do profitably). This is not to say that no one should listen to a system after tuning it with an RTA, and make adjustments (if needed) based on their experience! It simply means that proper use of measurement tools is a key part of efficient and repeatable results.?
A good way of getting measured results that correspond to how we hear is spatial averaging. This means taking several measurements of the system, at various points in space, and averaging them. This essentially throws out measurement data which is valid for one point in space but not at any of the others, which means our human hearing system is likely to ignore it. Taking spatially-averaged measurements used to take a great deal of time, but Audison's bit Tune can do it in 50 seconds, at resolutions up to 1/48th octave!
3. A target curve is an example of a pleasing system frequency response measurement which we want to achieve. Studies have shown that humans tend to prefer similar frequency responses, but that hasn’t prevented some industry members from stridently criticizing one curve, and advancing another.
Most differences in these target curves are small, and really not worth arguing over. Most failures to hit a target curve are not small - the deviations from the target are often significant, and many complaints about “this target curves is not satisfactory” actually turn out to have badly missed the target response!?
There is a general consensus that people like the low bass louder than the midrange (this is largely due to the size of the cabin), and that there should be a “ramp” from the midrange to the low bass (rather than a sharp discontinuity from the midrange to the level of the subwoofer). There is a slightly-smaller consensus that a rolloff in the treble is desirable. Beyond these general points, there are some deviations from a linear frequency response which are preferred by some people and not others. However, the vast majority of consumers don’t seem to have significant preferences outside the statistical predictions - most of the individual preferences are from the sound-quality competition realm, and are not related to consumer satisfaction.?
Endorsing a “target curve” is basically a statement that most people prefer a similar frequency response. Disagreeing with this statement requires us to tune every single vehicle for personal preference, and would effectively make tuning impossible to perform in a production environment, where profitability is a requirement.?Fortunately, most people prefer a similar frequency response.
This target is flat between 2500 Hz and 150 Hz.?
Above 2500 Hz, there is a rolloff which reaches -6dB at 20kHz.?
Over the octave between 150 Hz and 75 Hz, there is a 10dB “ramp” to the sub-bass (I recommend this target be achieved with the subwoofer level control at around 50% of maximum, so the listener can adjust to taste).
This sub-bass level is maintained to 30 Hz. Whatever response we get below 30 Hz is fine - we don’t need to EQ down there.?
Now, some people don’t like the treble rolloff. Some people want the bass ramp-up to take place over two octaves, rather than one octave (reaching full output at 40 Hz). Some like to introduce other nonlinearities. There are definitely some adjustments for larger and smaller vehicles, and for tweeterless “wideband”?speaker systems.?
Regardless, all target curves seem to follow this general tilt graphed in a home-audio research project by Bruel and Kjaer in 1974:
This is not to say that every DSP-equipped vehicle needs to sound exactly the same. Think about a home-audio store. They sell several models and possibly several brands of speaker. Every one of these speakers is a design which was improved over and over, until it was ready. They all sound different.
We should think long and hard before we start trying to make every system we install sound exactly the same, regardless of the system design and the products used! We’re trying to let the customer hear the product that they purchased, not trying to re-engineer it after we install it into the car.
4. When we put an individual speaker driver into a car, the speaker’s frequency response is affected. This is largely due to reflections from the interior. Fortunately, the reflected sound which interferes with the direct sound from the speaker attenuates over the greater distance it travels, so the cancellations are partial in nature. We need to use tools - such as an equalizer - to get the best sound from a speaker once we put it into a car. Otherwise, the speaker won’t sound nearly the same in a car as it sounds in a R+D laboratory, or in a display board. DSPs have very good crossovers, level controls, and equalizers - all of which allow us to manage the frequency response.
5. We will need more than one speaker: Stereo creates an illusion of a musical performance between two identical speakers. So we need at least two speakers to get stereo - left and right. Since we sit off-center, the two speakers will not sound identical in the listening position, and part of good system design involves selecting speakers with better performance in the listening positions. Tuning often involves changing the frequency response for each side independently, so that a similar response is achieved in that listening position.
6. We will need more than two speakers: We can’t get a single speaker to play from 20 Hz to 20,000 Hz. A speaker large enough to play 20 Hz audibly would be terrible at playing into the treble, and a speaker capable of playing to 20,000 Hz effectively would rip itself apart trying to play into the low bass. So we will need multiple types and sizes of speaker driver. Most front-stage speakers need at least two tweeters and two mid woofers. Some of these speakers are closer than others, and part of tuning is often to manage their levels to make them match at one listening position.?
7. We need a subwoofer to play the low-frequency content in modern recordings. Since we cannot discern the direction of low bass notes, we can get away with a single subwoofer. Subwoofers are large and usually can’t be up front. We can get away with a subwoofer in the back, as long as we can create the illusion that it’s in front. That means it has to be in phase with the fronts.
领英推荐
8. So now we are up to five speakers. Beyond this, we may add a set of front midranges, which solves several acoustic problems related to driver diameters and dispersion, but gets us up to 7 speakers. Then we may add some rear speakers, which can take us to 9 speakers, 11 speakers, or more!?
9. Whenever we put multiple speakers playing the same notes in a car, those speakers can interfere with each other - which drastically affects the frequency response, as speakers partially cancel each other out at some notes and not others! This problem is made worse by the fact that the various speakers are different distances from the primary listening position, and since sound travels relatively slowly, these sounds arrive at different moments in time. The different path lengths which the sound travels result in multiple arrivals, and multiple arrivals cause periodic phase cancellations.?
Different path lengths to the listening position, and the relatively slow speed of sound, means that left and right speakers partially cancel, front and rear speakers partially cancel, and high and low speakers partially cancel (where they overlap at the crossover point).
These multiple arrivals make it impossible to meet the baseline objectives of a pleasing frequency response, and of left and right channels which match in frequency response and phase. The cancellations ruin our frequency response.?
Is it really that bad, you may wonder??
Here is an example of two perfectly-flat frequency responses interfering with each other due to 27” of path-length difference between the two, and the corresponding phase problems. The path-length difference of 27” was chosen because it roughly corresponds to the path-length difference for door speakers in many modern midsize vehicles.?
The blue and red lines are Channels 1 and 2 respectively, and the green line is both together, Channel 1+2. For the sake of simplicity, the results of room reflections have been eliminated.
As you can see, these cancellations can ruin the frequency response.?
Equalizers are not effective in solving these problems - adding energy to a cancellation doesn’t magically make it disappear - so we have to use other means to manage the cancellations before we start to use equalizers to address the frequency response. DSPs have tools, such as various crossover filter types, delay, and direct phase manipulation, which let us manage cancellations before we equalize.?
A common way to address this problem is by adding delay to various channels of the system, so that the sound from the various speakers arrives at the listening position at the same time. This will eliminate the cancellations caused by path-length differences to the speakers. (It turns out that we cannot hear the delays caused by distance in cars - the distances involved are too short - but these cancellations are what we hear.)
It is important to note that there are phase cancellations in any room, and so after properly applying delay to resolve our path-length differences, there will still be some phase cancellations remaining. They are caused by reflections in the vehicle (they may also be caused by errors in our system design or implementation, but hopefully we have prevented these by using best practices in our system setup). These remaining cancellations may be significant enough to be objectionable. This does not mean that delay should not be calculated by distance.
Some people want to set the delays by ear, listening to a specific test tone or song, and making adjustments until the center is in the center.?If we experiment by listening to different delay settings - delay values not derived by the distances to the speakers - we will hear the frequency response change, as cancellations occur at different frequencies when we use different amounts of delay. Some of these frequency responses may be more pleasing than others, but that does not make them more accurate a solution for cancellations caused by path-length-differences.?
The problem with the “setting delays by ear” approach is that it presumes that left and right already match in level and response - but they don’t. Some notes are louder on one side, some on another, and that muddles our ability to discern the proper delay setting. Until we address the cancellations caused by various path-length differences, we can’t get the frequency response matching on the left and right - and if we can’t, setting delays by ear to place the center in the center is a compromise at best.?
So, distance-based delay doesn’t eliminate all sources of cancellations, but that doesn’t mean it doesn’t work or that we shouldn’t use it.
10. Measuring phase: Phase is a complex topic, and we won’t go into it in detail - but it is important to note that we do not hear absolute phase, and we do not need the phase to be 0 degrees at every frequency to have good sound. Studies have shown that when music is played over a speaker, phase manipulations are not audible. Lipshitz, et al, 1992:?
“On normal musical material heard via loudspeakers in an average listening room, we have not thus far detected the effect of midrange phase distortions of up to two cascaded all-pass networks… We do not have evidence to conclusively demonstrate whether phase distortions of this amount can be heard in normal reverberant loudspeaker listening to normal musical or acoustic transients, but it is clear that the effect, if audible, is extremely subtle.” ?
What we find objectionable is when the sound from two different speakers reach our ears significantly out of phase, and cancellation results. This damages our frequency response, which we have already mentioned as the single largest factor in perceived sound quality. So we are not working at keeping phase at 0 degrees at all frequencies - a practical impossibility - but we are engaged in avoiding phase misalignments between speakers, which result in cancellations which damage the resulting summed frequency response, and thus the perceived sound quality.?
Because of the above problem of speakers interfering with each other, does it help to know the value of phase at each frequency from each speaker? Let’s explore this.?
Some in the industry are now advocating that tuners measure the phase for each individual speaker driver (Since an RTA does not measure phase directly, they are using an impulse-response measurement system), and then use a combination of delay and phase manipulation to try to put every speaker in phase with its counterpart on the other side of the crossover.?
Often, these people also advocate detailed equalization of the speaker in the stop band of the crossover filter, so that this phase alignment is easier to achieve (if a speaker doesn’t reach the ideal frequency response of the intended crossover, the phase of the speaker’s output will not be at the value intended by the designer of the crossover filter topology).?This is reasonable, up to a point. An analogy is building an overall frequency response by making each individual puzzle piece to fit a predetermined shape.?
But if there is a mismatch in phase in the transition region, at what point is it audible? If the two speakers in question are 180 degrees out of phase at the crossover point, that’s definitely audible. If the two speakers are 45 degrees out of phase at the crossover point, is that audible? The resulting damage to the sum would be 0.35dB. What about 60 degrees of misalignment? The damage to the sum would be 0.65 dB. 90 degrees of misalignment inflicts 3dB of damage at the sum, and so we can definitely say that we want our phase mismatches to be less than 90 degrees whenever possible.?
So is measuring the value of phase essential for good sound? The ability to capture phase using impulse response measurement has been around for decades - but it’s not simple, and it’s not quick, so few people have continued to use it in daily work. The question we should be asking ourselves is, is the juice worth the squeeze? If this requires more equipment and a longer process, is the result commensurate with the work? Can we sufficiently resolve these issues in any other ways?
Some, including this writer, feel that many adherents of this approach are placing undue emphasis on the process, and when they do, they are victims of the Magnitude Fallacy. As a good friend of mine says, “It is not that we need to be perfectly in phase; It is very important to not be perfectly out of phase.”
These measurement techniques don’t include the OEM signal path, and this can be a problem. Often, the system would sound great, if we properly corrected the signal before tuning. If we are going to invest time and money into part of the process, measuring the OEM signal is more important than measuring the value of the phase at the crossover points.?
An alternative to measuring phase: If we have the ability to play single speakers and capture the resulting frequency responses with an RTA, and then play combinations of speakers and measure that response with an RTA, then we can discern phase-cancellation problems by comparing the predicted change in frequency response to the actual change. Using this approach, it is not required to measure the phase value to be able to discern and manage phase cancellations between speakers.
A table of phase alignments and summed amplitudes
?
11. OEM systems today address these phase cancellations, in order to get more out of their speakers and in order to let their equalization be effective (EQ cannot overcome bad phase cancellations). They do this with delay, and with direct phase manipulation (using all-pass filters).?
12. OEM integration in modern cars means “detecting and undoing what the OEM tuner did to manage these cancellations”. This is more important than simply reversing their equalization, or summing partial-range channels back together. We have to address their phase-and-time processing before we use our own cancellation-management techniques, because our techniques can’t be used on top of what the OEM system is using. If left and right signals are not in phase with each other at all frequencies, we can’t use delay to get a good stereo presentation - and this is a bigger problem than phase at the crossover!
In conclusion, there are a few different ways to reach a pleasing overall frequency response, with left and right matching in frequency response and phase. Some are faster than others, and some are more consistently replicable than others. If you are involved in this industry as a professional, faster and more replicable processes are probably of interest to you, and you will benefit from using RTAs, target curves, distance-driven delays, and response comparisons in your process. It’s not so much that there is only one path, and all others are wrong, as it is that some paths are more efficient, and we need an efficient process to make DSP tuning a commercially-viable activity.?
American actor and film producer. Cast as Matthew Perry look alike. Lover of scary, comedic roles.
1 年?? Brilliant article!!!??