Article: The Essentials of Sound Quality: IMHO

Beaming:

• Beaming is kin to the acoustic low-pass (discussed in the following post).
• Beaming is a function of the effective driver size: the dispersion narrows as the wavelength (frequency) becomes smaller than the size of the drive unit.
  Note: Effective driver size is taken from 1/2 surround to 1/2 surround.
• All drivers beam! Though, some drivers are built to extend further on a given axis than others, but often at a cost (sensitivity, cone breakup, etc).
• Frequency response measurements illustrate beaming easily; look for the on and off-axis responses to diverge. This is your beaming point.
• The formula to approximate the beaming point is: 0.5*(Speed of Sound)/(Driver Size)
• Below is a table illustrating approximate beaming points for a round driver with a given cone diameter:
  http://i18.photobucket.com/albums/b1...bleexample.png
  

In the example below of a 4” midrange with an effective diameter of about 3”, beaming occurs approximately at 2khz.

http://i18.photobucket.com/albums/b1...ld/Beaming.png

Acoustic Rolloff:

• Acoustic high-pass determined by:
  
  • Fs – Free air resonance of a driver
  • Qts – Bandwidth of driver resonance

Fs and Qts can be obtained by manufacturer Thiele-Small specs or derived by an impedance chart if available:

http://i18.photobucket.com/albums/b1.../impedance.png

• Acoustic low-pass:
  
  • Determined by things such as motor force, suspension, and (namely) inductance.
  • If the inductance (essentially resistance to current, for our purposes) is high, the output of the speaker is less, regardless of the axis.

Distortion:

There are two types of distortion:

• Linear:
  
  • Any divergence from flat in the frequency response would be considered a form of linear distortion.
  • No driver is completely flat. Though, typically below the beaming point, drivers are fairly flat. Outside of beaming is where breakup occurs and is typically the area where linearity in response is compromised.
  • Does not change with volume.
  • This is typically used to determine low-pass values as cone break-up is the worst offender after the beaming point.
  
  
  
  Below is an example of poor linear performance. Notice after the beaming point of this 4” fullrange driver how it’s on and off-axis measurements differ?
  
  http://i18.photobucket.com/albums/b1...Distortion.png
  
  
  
  Below is an example of good linear performance. Notice after the beaming point of this 4” driver how it’s on and off-axis measurements follow the same trend?
  
  http://i18.photobucket.com/albums/b1...Distortion.png
  
  
  
  
• Non-Linear:
  
  • Distortion that changes with volume.
  • Referred to as harmonic distortion, THD, etc.
  • This is typically used to determine your high-pass values.
  • While argued as to it’s merit of audibility, a good rule of thumb is to avoid the 3% THD range (Note: 3% THD is 30dB down from the fundamental).
  
  
  
  Below is the measured distortion of a 4” fullrange driver. Note the THD has reached the 3% mark at about 150hz. In addition, high frequency breakup exceeds 3% THD at approximately 3khz.
  
  http://i18.photobucket.com/albums/b1...ents96dB1m.png

Tuning Advice:

This section covers some of my “lessons learned” with tips based on my experiences tuning my car and helping tune others. As with anything, I should caveat by saying your mileage may vary. Not everything that works for one will work for another, but maybe some of my failures can lead to your success.




Resources:
Here are a couple recourses I'll reference in the body of text below. Go ahead and check them out. And you may want to go ahead and burn the CD linked below.

• Test CD:
  This disc was given out at my GTG in November. There's a lot of tracks on here but for the purpose of this post, we'll only focus on the pink noise tracks in the last half of the disc.
• Time Alignment Site:
  Robert Mcintosh (Pockets5) helped me make this site. The goal was for someone to easily punch in measurements of each speaker to the listening position and get the numbers needed to delay each speaker so it arrives at the listener at approximately the same time. It's not perfect, but should cut about 80% of your time off the manual way.

 

Setting Gains:


Intro:

• Setting amplifier gains can be as complicated as using an oscilloscope with a dummy load or simply using a $10 digital multimeter (DMM) and Ohm’s Law. A scope will let you know when the output signal is clipping (distorting). The DMM method assumes you know at what voltage for a given load the output signal will clip. I will discuss the latter.
• Since most (brand-name) amplifiers will do their rated, continuous power per spec, all you need is to connect the multimeter to the speaker output of the amplifier and raise the gain until you achieve the required output voltage while playing a tone through your cd player
  
  • Ohm’s law is used to determine amplifier output voltage. An online calculator can be found here: Ohm's Law Calculator
  • Use an attenuated tone, such as -10dB or -5dB as this will allow additional volume for ‘quiet’ tracks. 0dB tones may be used in lieu of attenuated tones IF you listen to heavily compressed music.
  • Tones can be downloaded here: Realm of Excursion
• The tone should be in the bandpass of what you are setting the gain for. For example:
  
  • If you are setting subwoofer amp gains, use a 40hz tone.
  • If you are setting midrange amp gains, use a 1khz tone.

Setting gains with a DMM:

• Here is a quick guide to properly setting your system gains using nothing more than a DMM.
• Reset all your cd player DSP features, balance, fader, etc to ‘default’.
• Disconnect all speakers to keep from damaging them (and your hearing).
• We will assume your cd player puts out clean, undistorted signal at near maximum. Play your tone (-10dB suggested) and increase the headunit volume to a click or two below max.
• Knowing the amps’ rated power at a given load (ohm) obtained from manufacturer’s spec (assuming they are a reputable brand), use Ohm’s law to determine voltage needed.
• For example, with a 4 ohm load and 100wRMS rating, the output voltage should be 20VRMS.
• Set the DMM to “VAC”, which is AC Voltage.
• Connect the leads of the DMM to the amp’s appropriate speaker output and increase the amp gain until you achieve this voltage
• Repeat for each channel.

You’ve now set up your system to maximize output and performance with a wide variety of music.

Keep in mind, when level matching, you can use your decrease amp gains as opposed to decreasing output level on your DSP. Though, I choose to maximize the gain structure up the chain as much as possible and make adjustments at one place (such as a DSP) just to keep things simple.

DSP Basics:


Intro:

• Most DSPs have these same features:
  
  • Crossover
  • Polarity (0/180 deg phase)
  • Time Alignment – ability to delay signal so all channels reach the listener at the same time
  • Levels – essentially gain for each output channel
  • Equalizer
  
  
  
• The difference in DSPs is usually small, in regards to features. The difference typically comes in the resolution of adjustment of these features or (most notably) the interface. Here are some examples*
  *Note: This is NOT intended to be a summary of all features of every DSP. I’ve just chosen specific examples.:
  
  • Resolution of adjustment:
    
    • Alpine PXA-H800 uses 0.5dB increments for levels; Helix DSP uses 1dB
    • Alpine PXA-H800 uses 1/3 octave frequencies for crossover points; Helix DSP has 1/12 octave (in other words, Helix has a whole lot more flexibility); ARC Audio PS8 allows you to type in any number you wish
    • There are some exceptions where a DSP has something different (Helix offers adjustable phase in 15 degree increments on the subwoofer channel)
  • Interface:
    
    • Most DSPs are now controlled via a computer interface, though some are still controlled directly through the DSP (ie; Pioneer P99 or Alpine H800)
    • This is user preference. Some prefer to have a laptop and some prefer to be able to tune ‘on the fly’.

Crossovers:

Crossover Point/Slope should be evaluated as a set. Using just a number and a willy-nilly slope isn't exactly a good method to use. There needs to be some reason for setting these values. The frequency dictates at which point you want to start rolling the speaker off. The slope will dictate not only absolute and relative phase but also attenuation.

Crossovers are made of both the frequency and the slope you use. So, let's look at that...

Crossover Frequency:

Namely, there are four aspects I am considering for low/high pass crossover values. Each of these are discussed in the driver basics section above but I will recap.

1. Beaming. Where does the driver's high frequency response begin to separate from on/off-axis? Using the Polar Response and Beaming reference above, you can see some of the math behind it and reasons as well.
2. High frequency breakup. All speakers begin to break up at some point. It's just an effect of the cone material and shape. Typically drivers don't break up until about an octave or so above their beaming point. Some drivers control break-up better than others through cone design. What you want to avoid is the area where the break up is severe enough to be heard outside of the crossover point. BUT, since this breakup usually occurs above beaming, you should be crossing before breakup occurs.
3. Low frequency distortion. You should know what I mean here... take a tweeter for example. If you cross a tweeter at 500hz, odds are, you're going to get all sorts of distortion and ultimately fry it. The general rule of thumb seems to be to cross the tweeter at 2*Fs (Fs=resonant frequency of the voice coil). BUT, this isn't a one-size-fits-all solution. Some drivers may have a low Fs but may not be well suited for a low crossover (ie; some drivers have an Fs of 700hz but I wouldn't run them at 1400hz full tilt with any slope). The key here, really, though is matching the dispersion pattern as well as you can to the driver before it. In the case of a 3" mid which beams at 2.5khz, you'd want to cross your tweeter somewhere in this area to keep from having a null at the crossover point, not fixed by any phase/polarity changes. Again, see the post I mentioned in #1, above.
4. Natural Rolloff. On the low frequency end, the driver rolls off naturally. This is dictated by the Qts and Fs (the Qts dictates how much the rolloff is and Fs tells you at what frequency it occurs). These 2 pieces of information can be found in an impedance plot. Most people will try to set their crossover point and slope to essentially follow that same natural rolloff so you combine both the acoustic rolloff of the driver with the electrical crossover and don't alter the phase severely.

Crossover Slope:

The slope you use depends on the following:

1. Level of attenuation needed.
   This should be self-explanatory. Basically, the steeper the slope, the faster the rolloff.
2. Phase:
   
   • I can't say enough how non-trivial this is. There's the "set it and forget it" method which can be made to work or there's the "spend a lot of time on it until you get it as good as you can" which I propose. The latter option will save you a lot of headaches in the future.
   • Think of two sine waves. If both are in phase, they play together and the amplitude is increased. If they are out of phase by 180 degrees, they cancel each other out. Your goal at the crossover is to essentially allow one speaker to 'carry' in to the other, without evidence or calling attention to anything in the crossover region. You want in phase sine waves... for a lack of better analogy.
   
   
   What you will HAVE to do is EXPERIMENT. Using the above crossover frequency info, pick a point that makes sense to start with. From there, change the slope of one driver with respect to another or even both drivers. Take notes. Which settings sounds better? What happens when you change the polarity of the tweeter but leave the midrange polarity the same? What happens if you change the slope of the midrange from 24dB/Octave to 12dB/Octave? Then try flipping it's polarity.
   
   FWIW, I typically shy away from anything less than Linkwitz-Riley 12dB/octave (LR2 - "2" for second order). The reason why is pretty simple: power handling. If you know me, you know I like some volume. Then there's the case of out of band EQ'ing that may be necessary and if you don't have this ability with your DSP, it's a problem. A steeper slope keeps less information from overlapping. Especially the high frequency content you don't want playing due to beaming and breakup. And for those convinced that low order is the best way to go in a time domain case, keep this in mind: although a 6dB/octave butterworth has no group delay, it doesn't mean your summed response won't result in such. What you care about is the total response. Not just a resistive load single crossover. ;)

There's no guaranteed slope or crossover point that will work for you. The reason why is because the crossover network that works best is dependent on the characteristics of a single driver (breakup, natural rolloff, distortion) AND the crossover point/slope of the driver you are mating it to!

You will have to experiment. You can use your ears or a mic. I think it's useful to try using your ears as it makes you a more attentive tuner. The mic can do a lot of things great, but it can also mislead you as well. So, you have to really listen in between adjustments made via mic. One thing I suggest is to narrow the best settings down to a few. Then from those, listen. Use pre-sets if you can and switch back and forth between them.

I also recommend using correlated pink noise to help you set your crossovers. The CD link I provided in the Resources section has this.
Mute everything but the mid and tweeter (or whatever it is you're setting the crossovers on) and listen to them as a pair while you alter the crossover settings. Listen for the sound to be fuller and as one. If you hear more of one driver than another keep in mind this may be simple levels, which we will address later. For now, focus on trying to listen for cohesion at the crossover point, though. If you're crossing your mid and tweeter at about 3khz, don't worry about what's going on at 500hz or 10khz. Pay attention to the 2-4khz range and see how that area changes as you adjust your crossovers' settings. Something else you can try, which is what I do, is to use pink noise 'tones' at the frequency around the crossover point. This is really helpful with time alignment which I'll get in to later as well.

Something else I've noticed is when the two drivers are out of phase, I'll hear a phantom image in front of me; away from the drivers I'm tuning. It's a total mind trip.

Finally, your goal is to get it as close as you can. It likely won't be perfect, because most car installs require a large delta between drivers and thus there will most likely be lobing (this is again why #1 in the Crossover Frequency section is important) but what you should find is after spending some QUALITY time doing this, you'll have a much fuller soundstage and more cohesion between drivers. Time alignment can be used to further tweak this. That's for later, though. Right now, focus on getting the crossover point and slope as dialed as possible between each set of drivers.



Additional Crossover Notes::

While the previous section focused on a single driver characteristics and understanding them, the import factor is the implementation of a driver in a system… with other drivers.

Good power response should be your goal but it’s not easy.

• For example, mating a 1” tweeter to an 8” woofer isn’t as easy as mating a 1” tweeter to a 6” woofer. Why? Because the tweeter doesn’t have to cross as low to match the 6” woofer’s dispersion. If you tried to cross the tweeter to match the 8” woofer, you’d increase the tweeter non-linear distortion considerably. Conversely, if you increased the 8” driver’s low pass filter you’d likely exceed it’s beaming point and the dispersion wouldn’t match between the mid/tweeter.

Let's look at another example using a typical 2-way setup:

• A typical 2-way setup consists of a 6.5” woofer and ¾” tweeter. The following is a generic analysis based on typical drivers:
• ¾” Tweeter Crossover:
  
  • THD reaches 3% at 2khz (raw driver)
• 6.5” Woofer Crossovers:
  
  • Most 6.5” woofers with moderate linear throw (3-5mm one-way) have 3% THD by 80hz.
  • Fs/Qts will drive the enclosure which drives Qtc which drives the crossover as well. A lower value Qtc means less cone control and more attention should be placed here to not damage the suspension or fry the voice coil.
  • Beaming will occur by about 1.2khz with an effective diameter of 6”. Based on this, you can assume a 2khz would suffice to keep the driver’s on and off-axis response fairly well matched. Much higher above this and the separation is more severe.

Therefore, a nominal crossover point between mid and tweeter in this generic example would be in the 2khz – 3khz range, in order to mitigate woofer breakup and beaming but also to lessen tweeter distortion. Crossing the woofer above 80hz on the low end will help mitigate it’s distortion at higher volumes.

How We Hear: Interaural Time Delay and Interaural Intensity Difference:

Our ability to determine location of sounds is rooted in what can be boiled down to a few key points (and for the sake of this presentation will be kept very basic):

• Interaural Time Difference (ITD) – Localization determined based on the time a sound takes to arrive at each ear
• Interaural Intensity Difference (IID) - Localization determined based on the intensity of a sound arriving at each ear
  
  • Also known as Interaural Level Difference (ILD)

Each of these contributions can be roughly boiled down to the following*:

• ITD cues contribute mostly up to approximately 800hz and then share with IID through approximately 1400hz where IID then takes over
• * This is a summary of various research. There is additional useful information here: ITD and IID Cues

Below is a graphical representation of ITD/IID and the area where they have some overlap:

http://i18.photobucket.com/albums/b1...ld/ILD_ITD.png




It should be noted this section does not consider any sound source other than laterally in front of the listener. For further reading on the topic, also consider the impact of the Head Related Transfer Function (HRTF) and Head Related Impulse Response (HRIR) as the “Cone of Confusion”.

Time Alignment and Levels

Basic Audio Terms:

Knowing the basic definitions of soundstage can really help you understand why certain DSP aspects are needed and how they can be implemented. So, let’s look at one of the most basic and important aspects of soundstage:

• Balance: Center and Acoustic Boundaries
  
  • To properly balance your system is it important to know your stage boundaries.
  • In acoustics, these boundaries are defined as the left-most stage and the right-most stage.
  • You know a center is the point in space between two lines. Therefore, the acoustic center should be placed mid-way between your acoustic boundaries. The center shouldn’t be an artificial point on your dash: it is the acoustic center of the acoustic left and right boundaries.
  • It’s important to note that not all recordings have the same characteristics; some may have a vocalist to the right of center. Others may place a kick drum to the left side of the stage. Judging your center by a sole vocalist or instrument can be misleading if you don’t know for sure where that location on the stage really is supposed to be. So with that said, it’s a good idea to use correlated pink noise or a centered narrator (both provided in my Test CD) to determine center.
• Depth, Width, Height
  
  • Each of these have importance as well, but I feel focusing on the center and putting that in the appropriate location with respect to your left and right boundaries will allow the other aspects to “fall in place”
  
  

Time Alignment and Levels:

Typically, in a car, the listener is positioned so that none of the speakers are the same distance from him.

The below illustrates a standard car setup with only a single speaker on each side creating not only a near-side biased stage, but also incoherency in the sound due to poor time arrival differences. There is a clear stage boundary, but no focus and no way to really pick out a center.

http://i18.photobucket.com/albums/b1...sicspeaker.png


In the example above, the Left/Right speaker delta is 11 inches. In time, 11 inches is approximately 0.808 ms, which means that the left speaker’s sound arrives 0.808ms before the right speaker’s sound. This would create a left side bias due to ITD.

The SPL at your seat is approximately 2dB higher from the left side. This would also create a left side bias but due to IID.

The combination of ITD & IID would drive an overall stage that sounds squished to the left; there would be very little focus if any and no well-defined acoustic boundaries or center.


Given what we learned about our hearing based on ITD & IID, we can use our DSP’s time alignment and level features so that each speakers’ sound arrives to you at the same time as opposed to the nearside sound arriving first.
Of course, it should be noted that this really assumes properly acoustic polarity has already been accounted for. A simple 0/180 degree “phase” swap. In some cases wiring your speakers up in electrical polarity the correct way doesn’t necessarily mean you’ll get the correct acoustic polarity. Phase will be discussed more in the following section.

Time Alignment:

Time Alignment and Level Matching’s use is to give the impression that all sounds arrive to the listener at the same time based on ITD & IID. In short, if you want to simulate a sound arriving later from one speaker than another you add time delay and decrease the output which you want to push away.

Time Alignment can be derived in simple (tape measure) or complex (measurement system with loop-back) manners.

The simple method is:

1. Use a tape measure to determine the distance of each driver from your listening position.
2. Measure approximately from the speaker voice coil (if you can’t see it, add the appropriate amount to the cone) to the center of your ears.
3. Use the following site to determine the approximate time delay needed to ensure each speakers’ sound arrives at the same time:
4. http://tracerite.com/calc.html

The complex method would be one using measurement gear to measure the impulse response of your speaker. But, this will not be discussed due to time constraints. Maybe later, though. ;)




Levels:

Given that levels are more dominant in regards to staging in higher frequencies, tweeters are typically the most impacted drivers when it comes to level setting between left and right. Additionally, in acoustics, each doubling of distance would result in the halving of SPL; in this case the left side’s output would *likely be stronger.
*Note: If the left speaker is aimed off axis, it would permit the higher frequency content to roll off a bit sooner thus helping to mitigate the IID bias to the left speaker. This is why some home speaker setups are aimed inward. Something to chew on.... ;)

There is no real easy way to adjust for levels via a calculation. I mean, there is… but real world things cause issues here:

• Did you set the gains the exact same?
• The environment will take a flat speaker response and make it not so flat.
• Are the speakers the exact same sensitivity?
  
  • Bigger concern between driver types such as a midrange and tweeter, as opposed to two of the same midranges.

The best way to start off adjusting levels is to do so using pink noise and listening to high frequency content (>2khz) and adjusting levels until you achieve a good balance between left and right stage. Another option is to use an RTA or Phone app that will approximate the SPL at your seat from each speaker.



Using Time Alignment and Levels to create a Balanced Stage:

• The ITD aspect:
  Using the simple T/A method, you determine that the left speaker delay should be set to 0.80 ms.
• The IID aspect:
  You also have determined via SPL measurement that you need to attenuate the left speaker 2dB to match with the right speaker.

What you wind up with a more equal representation of left and right stage boundaries, resulting in more width and a more realistic center location for your stage.


So, what you should now have is something *CLOSE* to this:

http://i18.photobucket.com/albums/b1...edspeakers.png



Another way to consider it is going from this ...

http://i18.photobucket.com/albums/b1...ishedstage.png


... to this ...

http://i18.photobucket.com/albums/b1.../spreadout.png





Recap:

Remember, when adjusting levels and time delay what you’re adjusting with one speaker is RELATIVE to the other speakers. You’re adjusting the nearest speaker(s) so that it sounds as if it has the same intensity and time of arrival as the furthest speaker. This would then sound as if the other speaker had moved.

Or think of it like this: you’re moving your ‘center’ with respect to the left and right boundaries. You continue to do this until your boundaries sound equidistant from the center vocalist/pink noise.

Equalization: A Primer

So, we've gone over the basics of DSP. So far we've covered time alignment and level matching to help you tune. Another major aspect of tuning is using the EQ to help resolve problems. Now, some may say EQ isn't necessary. I say that depends: what is your goal? What do you want to achieve in a car? What speakers do you have, how is your install set up, did you pay attention to the driver basics section and how it impacts your crossover choice?

Let's put it this way:
If you have a speaker that has few problems outside of it's beaming point then you're ahead of the game. If you cross the speaker before modal issues take over then that's a help. If you aren't doing either of these then you need to know that the speaker is going to cause you all sorts of problems and you're going to be going hard to work on the EQ to correct these problems. Compounding it more is the simple fact that some problems can't be resolved via EQ. You can't fix a cancellation null in a speaker due to surround or basket resonance. Not going to happen. You can't fix a modal issue in one axis and it not affect other axes. What about your installation... do you have resonance in an enclosure? If so, good luck correcting for that with standard 1/3 octave EQ.

But, let's say you've got the system set up to the best of your ability (install and basic tuning methods already discussed). When it comes to using your DSP’s equalizer, there are essentially two different methods you can use:

1. Ear
   
   • Correlated Pink Noise or 1/3 Octave Pink Noise: these tracks can be used for tonality adjustments and centering of frequencies that may jump out of band
2. Measurement
   
   • SPL Meter used with 1/3 Octave Pink Noise or tones
   • RTA
   • Impulse

Each of these methods has some advantage over another, but there is no single ‘best’. What I find is a good mix of ear and measurement will net you the best response.


"Why can't I use just one of these methods?"
"Why not just use method only? "
Because of the human factor: We don’t always understand what the data is showing us and misinterpret it, therefore creating more problems than we set out to fix.

• "Measurements don’t lie, right?"
  
  • Well… sure, they tell you what they tell you. But what if you are measuring the wrong way? Doesn’t help you, does it? There are numerous tutorials on how to measure yet there’s still a lot of questions and potential for better ways to measure. Additionally, you may want to believe you fixed a problem because it’s visually not there, but what if you fixed the wrong thing? What if the bump at 1khz you knocked down was merely next to a null that can’t be fixed? So, you knocked down 1khz 9dB and you thought you fixed it only to sometime later realize you didn’t fix the real problem and the truth is that you can’t really do anything about the null anyway. Bummer. :(

• "Why can’t I trust my ear?"
  
  
  • No one says you can’t. You should, however, question your brain. ;)
  • The problem with using your ear is simply that it’s easy to miss things or even to focus too much on something while altering what was good to begin with.
  • For example, let’s say you hear a problem at 1khz; it’s pulling to the left. So, naturally, you start cutting the left side at 1khz and boost the right side at 1khz. Problem solved, right? Until you get in your car the next morning and it’s STILL there. :rolleyes: After some playing around you found that the problem really is due to a reflection from your right speaker. You wanted to believe you were fixing the problem and so you did… but you really didn’t. Another potential problem could be the issue at 1khz is due to a reflection; a harmonic of 500hz is causing the issue and if you wanted to fix the root of the problem, you needed to adjust 500hz. Not 1khz.

Measuring "Right"

Now you're obviously asking how you measure the "right" way, or "listen right". Well, truth be told, it's not very easy to relay over the internet. It really takes time and practice. Luckily there are a few sources already that I think do a good job of illustrating the methods of measuring and listening. Here's some links:

• Dr. Sean Olive's "How To Listen" Software
• http://www.diymobileaudio.com/forum/how-articles-provided-our-members/163234-first-timers-guide-measuring-your-system.html
• My RTA Walk-Through Thread

Analyzing the Results:

One thing that really needs to be done is spatial averaging. The two measurement tutorials above discuss this but I want to reiterate it because it's just that important. If you'd like additional information, please read the attached PDF in this post . The document is by Dr. Earl Geddes and is a study he did for Ford Automotive. It's not too technically heavy which should keep the "I don't like science" excuses to a minimum. ;)
Note: The following is plucked straight from my build log.


The cliffs version is simply this: your car (and home) are wrought with reflection inducing panels/walls. When measuring response in the environment, you have two options:

1. "Gate" the response so you obtain only the response of the speaker you're trying to measure and you essentially ignore everything else.
2. Measure everything: speaker response and reflections.

Doing the first in the car?... good luck. How about... don't bother. At least not until you've gotten really good at measuring and understanding what you're measuring. Let's just say for all intents and purposes you won't be doing the first... like... ever (thank you, Taylor Swift, for making that phrase weird now). Really, it's just pretty much trivial unless you have a very specific goal and understanding of how to achieve it this way.

So, we do the second option. The issue, then, is the fact every measurement you take is a measurement of EVERYTHING occurring at the mic. This is good and it's bad. It's good in the way that there's not a whole lot you can do to the speaker itself so it kind of keeps you from worrying about it. - Although, this is why I really encourage people to study independent tests or do their own to understand the issue(s) with the speakers they've chosen before they use them in the car. - It's bad because, thanks to the nature of the reflective environment, you can't really trust a single point measurement (a measurement taken with the mic in one location). If you move the mic as little as one-half inch you'll get a different result. Most notably in the higher frequencies. This means RTA'ing your car for any desired curve by using one mic measurement is a TOTAL WASTE OF TIME. It's ideal to take multiple measurements in the "head area" and average them together. TrueRTA, OmniMic, and REW allow you to do this pretty easily. Then you have what is known as a good spatial average. It's not an exact method but it's the most realistic and approximates a very realistic response in the seated position.

For this spatial average there are a couple methods: one is a 'live average', discussed in my tutorial, the second is using various single measurements and averaging them together.

For this purpose, I did (6) individual measurements.
Now I've got 6 measurements. What next? Simple: average them all together to get one measurement.

Here's an example....

All six measurements taken by the method described above (no smoothing applied):

http://medleysmusings.com/wp-content...asurements.png




Same as above, but with 1/3 Octave smoothing:

http://medleysmusings.com/wp-content...-one-third.png





All of the above averaged in to one response:

http://medleysmusings.com/wp-content...-one-third.png






Let's talk about the above... at least my personal take on the above. I'm sure others may key on to some other aspects I might otherwise ignore or just overlook.

Notice how the response varies more the higher you get in frequency? This is exactly why I said using a single point measurement to tune to a curve is a very bad idea.

I'm going to ignore the shape of the curve, however, for this post... what I really want to focus on is midbass/subbass response, so let's look below 300hz. Anyone notice the one glaringly different thing about the response below this frequency versus the response above it? No matter where the measurement was taken, the response is pretty much the same. This is the critical frequency area (schroeder frequency (Fs)). Linkwitz gives the most simple definition I can think of here:

Quote:

---

The frequency fs is also called the Schroeder frequency and denotes approximately the boundary between reverberant room behavior above and discrete room modes below.

---

Which makes sense, right? Look again at the graphs I provided. Reverberation is occurring above about 300hz as evidenced by the diverging responses from the 6 head area measurements. Below this, the response is pretty constant in this area so it is modally (sp?) dominated. What does this mean to us? You can ignore spatial averaging (multiple measurements) when focusing on low(er) frequency response! This saves you time! Of course, every car is different so I suggest you always do a spatial average to determine where this Fs occurs in your car, but you can expect it to occur around the 200-400hz area, depending on car size. The larger the 'room' the lower the Schroeder frequency. This means once you do a spatial average you'll know where this frequency is. From then on, when you only care about working on the low frequency response, you can ignore spatial averaging and just put the mic at the seated position and measure, tune, measure, tune, rinse, wash, repeat until you're satisfied. I will caveat this by saying that tuning low frequency response with graphic EQ's isn't easy because modal peaks and dips are often too narrow and too specific of a frequency to effectively be targeted by graphic EQs. This is where my subsequent posts will sort of pick up.



Cliffs:

• Tuning based on one mic measurement is a waste of time. This has a caveat...
• Take a few measurements in the head area, where you sit. Look at them all overlaid. Where do they really start to diverge? This is your car's Schroeder frequency.
• Above the Schroeder frequency you must take multiple measurements and average them if you want to tune via RTA.
• Below the Schroeder frequency, one mic measurement will suffice since the response doesn't change enough to matter.
• Graphic EQs aren't the best tool for fixing response issues low in frequency. Parametric EQs are MUCH better. But, if all you have is a graphic use it to the best of your potential.

Analyzing the Data

As mentioned earlier, one problem with measurements is the incorrect use of the results. Once you achieve your measurements through the aforementioned methods, here’s some things you should ask yourself before heading straight to the EQ:

• Is what I’m seeing audible? Is it a dip or a peak? Here’s a good quote from Floyd Toole regarding this:
• Quote:

  ---

  Attenuation of excessive levels appears to be very safe, but avoid trying to fill deep holes. A narrow dip is probably caused by a null in a standing wave or interference pattern. As such it is the acoustical equivalent of a bottomless pit - it cannot be filled. Narrow dips are difficult to hear in any event, and all that will happen if you dial in a lot of gain is that the amplifiers will have reduced headroom, and the loudspeakers will be working harder to no avail. The result will be increased distortion.

  ---

  • In other words, don’t worry about dips. But do worry about large peaks.
• If it’s a peak, use the EQ to drop it down some and re-measure. Did it go away? If not, are you really sure it’s a peak, or is it just a spot next to a null that looks like a peak? A-ha!!!!
• If you are looking at the overall system response, try measuring each side individually and see if you notice something that may be causing the issue. IOW, take it down a level and see if the issue is caused by just the left midrange (for example). Then evaluate and tweak (or don’t) as necessary.

Let's look at some real examples...




Low Frequency EQ'ing (tuning below the Schroeder Frequency):

Taking off from the above, I'm only going to focus on the response below 300hz (graphs are out to 400hz for the sake of resolution).

The following is with no EQ. Time alignment and levels have been set, however.

First off, let's take a look at the difference measured from the driver's seat vs the passenger's seat.

http://medleysmusings.com/wp-content...-passenger.png


The results show the same response show pretty much the same curves above 70hz. I've seen this numerous times; almost as if the car has varying Schroeder frequencies. One is for the entire cabin; the other is for one location at a time. Of course, I'm not talking about moving the mic to the rear of the car... that's an entirely different can of worms. The point in this measurement, however, is to show that there is actually a sub-band that really needs attention below the seated Schroeder frequency: the midbass band is entirely subject to this. As shown, 70hz is the starting point for different results between seats but 300hz is about the starting point for different results within the same seat. So, 70hz to 300hz is gonna be a total PITA in my car. Through about 5 years of dealing with this same car, my measurements show me what I already know, so it's definitely been vetted. ;)



Next...






After doing that, it's time to get back to the driver's seat and start measuring response from there.

One might choose to measure the system response as a whole and use the RTA that way, but it's a bit more conclusive to study each individual side's response (left and right side response). This is easy to do: just pan the balance to one extreme or the other and measure. When you do, you'll have the left side stereo contribution vs the right side stereo contribution.

So, here we have just that. Panned left is Green. Panned right is Purple. No EQ. 1/12 octave (to show the crappy little modal stuff that 1/3 doesn't get).

http://medleysmusings.com/wp-content...ight-no-eq.png


What this really shows me is that both the left and right side stereo contributions have their own problems. Notice that slight dip around 85hz at the driver's seat? Everyone has that problem to some degree because of their proximity to the speaker. Bottom line, that dip is a cancellation mode. There's nothing I can do to fix it, either. I can EQ it up but what will happen is I'll just keep applying more power to the driver's side midbass, causing distortion to ramp up and likely audible issues due to it. And while it may raise the response there, it'll also make resonant modes more problematic. The potential to damage the driver certainly exists. There's just not a whole lot you can do here. Some EQ will help but if you try to flatten it out by adding 4-5dB of EQ you'll alter the response curve in a negative way and create other issues. The only way to really fix a problem like this is to move from the boundary causing the null or move your driver(s). So, I just ignore this. Truth be told, it's not a real big issue when listening. And this is just one more example of why you should not rely entirely on the RTA. You should always use your own ears to accompany what you've measured. If you have a narrow dip it's not as audible as a broad dip; the same goes for a bump in response.

So, yea... I'm not going to sweat that dip at 85hz measured at the driver's seat. It's a lost cause and serious waste of time to try to flatten it. I just want to smooth it so a bit of EQ here and there will help that.


Now, look at the rest of the curves. That dip around 85hz on the left side is exacerbated by the rise in response around 125hz. After looking at the decay plot, measured by REW, I see why...

http://medleysmusings.com/wp-content...dal-issues.png


This is a plot of response over time, laid out in 2-D. The highest levels are closer to the initial response time. As the graphs change color below one another, you're seeing 'slices' of the response in time. Look at the legend. It shows time in milliseconds (ms). Each color corresponds to a time slice/section. Ideally want to see is each slice dying out quickly and contributing less and less to the results. However, what you actually get is modal issues showing up... these are the ones that linger around and don't taper off smoothly. Looking at these plots is pretty subjective and really should be used with some subjective listening as well. But, I'll give some thoughts on how I look at it...

The 125hz issue showing up in the left side FR plot... now look at the decay plot around that frequency. See how the darker blue looks pretty mountainous here with a dominant spike at about 125hz? Notice how the shade of blue just before this has the same spike? This is an indication of a modal issue. Luckily, I have an EQ band right here... I can cut it some. The problem, however, is cutting here also affects the tonality in other ways. With a parametric EQ, I can set a narrow Q and cut accordingly. But, I don't have that, so I have to cut here with the 31 band EQ. Here is the result when I use the EQ to cut 125hz by 3dB:

http://medleysmusings.com/wp-content...-125hz-3db.png


Not surprisingly, there was no miraculous alteration of the issue. It cut the problem by 3dB as it should but it didn't make the ringing issue go away. It did lessen the effect some. This is where subjective listening will tell you if it helped. The drawback here is you also changed the tonality of the system because the Q (bandwidth) of the 1/3 octave equalizer is so wide; it doesn't just change a single frequency.

This site is a great reference for what frequencies influence what you hear and can help you understand the tradeoffs you deal with when changing EQ bands to fix problem areas:
Interactive Frequency Chart - Independent Recording Network



There are other frequencies that do the same thing. 100hz definitely lingers. 83hz lingers as well. Remember earlier my bit about bumping up 80hz to fill in that hole caused by the left side response? What do you think happens when you do that regarding the modal issues? It's a nasty problem. What you really need is a way to target specific modes without negatively affecting the other areas you want to fix with standard EQ methods. This would be a really good intro in to why parametric EQs are so good. So, I'll stop here and pick up there when I have the chance.

Keep in mind I've only really discussed one component of the system response here. The right side response has it's own problems as well.


Cliffs:

• Room modes suck. They muddy up system response as a whole.
• When the midbass is muddy it overshadows everything good about the rest of the system.
• All cars have modal issues smack in the midbass area. :mad:
• Standard EQ can only go so far. But when properly used, EQ can help tame some of the modes which results in a much more tonally pleasing car stereo and much better blending with sub on the low end and midrange on the high end.

placeholder 1 if needed

placeholder 2 if needed

placeholder 3 if needed

placeholder 4 if needed

Awesomesauce! Thanks for the effort that you've put into this.

Holy crap, now that's in-depth learnings. Finally some real detail on things needed to know.

This is awesome, Erin.

This is great, Erin, thanks.

It gives me more reason to delay my build since I'm a slow reader.

Awesome info, thanks.

new info added in post #13 (Time Alignment and Levels)

Holy SHIT!


HUGE THANKS!!!!!!!!

well, hopefully it's beneficial. :)

Oh it will be!

Now this going into some serious detail. This really helps understand alot of things and definitely helps with the tuning process. Things become much clearer and easier when we all understand the reasoning for the process and how to.

Thanks Erin, great info to be read...and re-read time and time again.

TL;DR



In all seriousness though, this is an awesome resource.

updated posts 14-16.

This is some great info!

On beaming: I have forever been using a different formula, and I'm wondering which one is accurate.

13740 / (driver size).

This is the one that MANY home audio folks use, including folks setting up D'Appolito arrays. I can link sites that use this exact math time and time again. Are you SURE your math is correct?


This is critical to my build because I designed and installed the size speakers I did in the off-axis positions I did, because of the math.

As comparison:

3/4" tweeter begins to beam just at the edge of the 20kHz envelope, right at 18.3Khz.
2" driver begins to beam at 6870 Hz
8" driver begins to beam at 1717 Hz


1145 feet per second, multiply by 12 for inches.




So which is right?

I'm using half wavelength. The reason why is based on the data I've seen. I understand different sources use different fractions (namely, whole, 1/2, or 1/4 wavelength). I think using 1/2 is a good middle grounds and represents what the data typically shows.

For example, look at the graph I provided. It's a 4" midrange with approximately 3.0" effective diameter.

Full wavelength:
Take the speed of sound (13600 in/s is 60*F, 50% humidity if I remember my calculations correctly) and divide it by the effective driver diameter. In this case, you'd have a beaming point at 4khz. Now look at the data; at 4khz your 0 degrees axis vs 60 deg axis measurements are about 8dB apart. That's quite a stretch if your goal is trying to achieve as close to Omni as you can at the crossover point (understanding that different slopes/types have different results).


Half-Wavelength:
Now use my formula: 13600/Effective Diameter/2. You get ~2.2khz. Now look at the data. That's practically right on the point where on/off axis diverge. I'm more concerned with omnidirectionality; if the driver is nearly 10dB down at 60 degrees at one full wavelength then it doesn't really do me any good trying to match it to a tweeter that's playing omni at it's HPF. Additionally, you're closer to the breakup point of the driver. If the driver breakup is pretty considerable (say, >15dB modal peak) then it's going to be hard to attenuate that via the crossover if the crossover point is practically right there. Furthermore, that breakup really screws up your power response because the driver's behavior off-axis (at least with poorer drivers) is quite different than it's on-axis behavior). If you're an octave lower (by using the 1/2 wavelength), then you're in that much more "clear". Especially if you're using steep slopes.

You're gonna have to link me to these other people's comments. Measuring from both seats is a waste of time in this environment unless your goal is to attain a balance (and believe me, there WILL be compromises made at your primary listening position) between the seats.

Also, NEVER, EVER... EVER EVER EVER use only a single point measurement of the system as your tuning guide. This is EXACTLY the stuff I talk about when I say "you're doing it wrong". SPATIAL AVERAGE. The guy who's measuring your car with only one position and not doing a head area spatial measurement is doing it wrong. Plain and simple. You can see why in my examples. The only time a single space measurement is worthy of doing is when you're measuring the response below the Schroeder Frequency; all of which was covered in my earlier posts and examples provided as to why.

That's averaging multiple measurements. Over a head area. One of those threads is actually mine. And that particular one would have the mic placed in 6 spots over about a 4x6" rectangle. And then those results would be averaged together.

I thought you were talking about putting the mic in one spot and using it to obtain only a single measurement. That would be bad.

Quote:

---

Originally Posted by BlackHHR

And this is what we did , from the drivers seat ONLY ....

If you have TrueRTA (like most of us do) then it's a total breeze.

Here's how you get a spatial average:
Measure Center. Save results by going to the 'view' menu and saving or simply hit 'Alt+1'.
Move the mic about 45 degrees to the right or left of center. Mesure. Save results in #2 (Alt+2).
Move the mic to the opposite side of center (left or right). Measure. Alt+3 to save.
If you can, move the mic up about 2-3". Measure in the center. Save as #4. Alt+4.
With the mic in the same Y position (height position), move the mic left/right of center. Measure. Save. Alt+5.
Move the mic to the other side of center. Measure. Save. Alt+6.
Finally, average all the results together by doing this: Go to the 'Utilities' menu. Then 'Average'. 'Select Input': this will be #1-#6. Now 'Select Memory to use for Result': this will be #7 or whichever number you want to use. Just don't override one of your measurements.
That's it. Now, look at the difference between each singular measurement and the 'averaged' measurement. Notice the difference?

You have now measured and completed a spatial average of your system. Use this averaged measurement to tune for best results.

---

While this is certainly doable and Erin even shows examples I believe, some (including several who have taught me) just move the mic as you measure. Sweep it side to side in an arc in other words. Set your averaging on your trueRta to 40-50 or so. Much, much faster! You don't have to sweep fast either, just keep it moving.

As some know, Mark E built a stand that is motorized to do just this. It sweeps the mic in a roughly 270* arc. That way he can sit outside the car with the computer and run the processor and just let the mic sweep while he does it. When I do it I don't particularly like smacking my face with it, just I sweep about 180*, or side glass to side glass! that way you are accounting for the reflections you will get off the glass. An alternative would be to sit in the passenger seat, holding the mic in the area where the listeners head would be sweeping. Or, as I once did while Todd was tuning on my car before the gtg, I sat in the back seat and swept it.

Regardless, that way you aren't concerned with whether or not you had the mic in the same 6 points in space as you did the previous time when you go to do it again.

Just a thought :)

What happened to the Test CD? I was going to start messing around with some of this this weekend, and was planning on picking up some blank CDs on the way home. Is it somewhere else now?

Jay

just re-read this for the third time. Going to try a few things when I get the new amps in before I let someone fine tune. You mentioned phone apps for level matching L/R. I have the JL app with the DB meter. Is that accurate enough to match left and right? just thinking of wedging it between the head rest and seat back or taping to head rest and monitoring

Quote:

---

Originally Posted by erinh

Setting gains with a DMM:

• Here is a quick guide to properly setting your system gains using nothing more than a DMM.
• Reset all your cd player DSP features, balance, fader, etc to ‘default’.
• Disconnect all speakers to keep from damaging them (and your hearing).
• We will assume your cd player puts out clean, undistorted signal at near maximum. Play your tone (-10dB suggested) and increase the headunit volume to a click or two below max.
• Knowing the amps’ rated power at a given load (ohm) obtained from manufacturer’s spec (assuming they are a reputable brand), use Ohm’s law to determine voltage needed.
• For example, with a 4 ohm load and 100wRMS rating, the output voltage should be 20VRMS.
• Set the DMM to “VAC”, which is AC Voltage.
• Connect the leads of the DMM to the amp’s appropriate speaker output and increase the amp gain until you achieve this voltage
• Repeat for each channel.

---

If I'm mixing 4 ohm and 8 ohm speakers, and I *don't* know my amp's rated power at 8 ohms (who does? lol) do I assume half power, or do I still try to match to total power?

In my case, my tweeters are 4ohm, but my mids are 8. and my amp is rated at 50w/ch@4ohm. They are right next to each other. At 4ohms, 50w is 14.14V, and at 8ohms 50w is 20V. So, would I set tweeter channels to 14.14v, and the mid channels to 20v? or would I assume 25w/ch @ 8ohm which would still end up being 14.14v?

Thanks.