Ask an Acoustic Engineer (me)

[Justin Zazzi]

What is the electrical/phase impact of using dissimilar electrical crossovers?

In the past I've heard all of the following stated as fact:
"The acoustic response is what matters. Choose the filter type/frequency that most closely approximates the target"
"You can use dissimilar frequencies and slopes, but stick with the same filter type"
"Use identical acoustic and electrical filter type, frequency, and slope, and fix the rest with EQ"

They can't all be true.
What does an acoustic engineer say?
Thanks!

Hah, great question.

"The acoustic response is what matters. Choose the filter type/frequency that most closely approximates the target"
Jdunk is right that I prefer the first statement and it is the foundation of the tuning spreadsheet I made (that AudioGal is now evolving).

"You can use dissimilar frequencies and slopes, but stick with the same filter type"
The second phrase might have some usefulness somewhere, but am not familiar with the advantage. This might be a practical constraint if you are using a device that doesn't allow different kinds of slopes or filter types. Many radios will simply have "highpass frequency" or "off". In this case you simply cannot select other options. This is more of a constraint on the equipment you have, not a best practice or something I would prefer to do if I didn't have to. This might also be one less variable to juggle when you're first learning how all this stuff works, so it might make that initial learning curve a little bit easier. I would still avoid this if I could because you'd have to re-learn stuff again later, doesn't really save any confusion imo.

"Use identical acoustic and electrical filter type, frequency, and slope, and fix the rest with EQ"
This is an interesting trick that I like to use sometimes. The advantage is if I want to change the crossover frequencies later on, it is super super easy. For example if I want an acoustic crossover at 100hz, then I set the electronic crossover filters at 100hz. I'll need to use much more agressive EQ and shelf filters to actually get the acoustic response I need, because I'm not using the best fitting electronic crossover filter for the job. However, I can change the crossover frequency to 125hz or 80hz later on very, very easily. In contrast, if I use the "whatever filter that works" method like the first quote you shared, changing the crossover frequency later on is much more involved.

Pros and cons to each. I tend to use the first method of "whatever filter works best" if I am confident I will not be adjusting the crossover frequency later on. If I think I will play with crossover frequency later on, I tend to use the third method of "match xo filters and use more eq instead" (for example I'm doing this right now with my home theater since I'm not sure where I want to cross over my tower speakers and my subwoofers exactly).

[Uberverbosity]

I appreciate the insight.

Can you speak to the debate between "knock down the peaks - no boost" vs 'beat that curve into submission'?
Aside from maybe losing some natural transients and dynamics through overly-aggressive eq-ing, electrically, is there any difference between a 1khz/75db signal that comes direct from the source and a 1khz/75db signal that came into the processor at 90db and was cut down to 75db?
Depending on width I can see how surrounding frequencies might be affected by a filter, but does the relationship change in some way at varying levels, or in the time domain?
In short: Is there an acoustic basis for the belief that 'less is more', or is it really a question of tuning prowess?

[Jdunk54nl]

I appreciate the insight.

Can you speak to the debate between "knock down the peaks - no boost" vs 'beat that curve into submission'?
Aside from maybe losing some natural transients and dynamics through overly-aggressive eq-ing, electrically, is there any difference between a 1khz/75db signal that comes direct from the source and a 1khz/75db signal that came into the processor at 90db and was cut down to 75db?
Depending on width I can see how surrounding frequencies might be affected by a filter, but does the relationship change in some way at varying levels, or in the time domain?
In short: Is there an acoustic basis for the belief that 'less is more', or is it really a question of tuning prowess?

Boost is ok, as long as you have the digital headroom and the boost is a natural frequency response issue with the driver, and not a reflection cancellation issue between the driver and some surface the sound is reflecting off of. The digital headroom part is you can't do anything to a signal that is above 0dbfs. So if at 0 boost, you are already at 0dbfs, then you can't boost any above that. You also can't boost if you are putting your driver in danger. Even REW will give you headroom required in the EQ filter window. I try to keep this at 0db headroom required. You also can't boost a cancellation, so if you apply 6db of boost, and it doesn't go up by 6db, then you are boosting something you shouldn't. Each PEQ change should impact the frequency response by what rew predicts (as long as all of the math matches between REW and the DSP.....looking at you JL AUDIO!!!!)

I may or may not have seen someone boost like 21db before.....on the low frequencies....it was a driver natural response issue (I believe anyway) and the driver was fine with the boost and so was the amp, so it was fine.



The over eq thing comes from what you can actually hear (smoothing settings help a lot). Some people try to beat the curve into submission putting PEQ with q's of like 10, 15, and 20. You can't perceive a dip/boost that narrow (higher q, narrower frequencies impacted). I like to stay under a q of 5. I started doing this after reading Floyd Toole's book on his research and after doing the harmon how to listen stuff and where I can no longer consitently perceive what frequencies are changing. Depending on the frequencies, we are better at perceiving changes than others, so sometimes we don't even need a q of 5, as we can't even hear that narrow of a change.

So, why do we still do it? A lot of times we are looking at a screen and trying to match perfectly the shape (I too am guilty of this) of the house curve, and we are boosting and cutting a lot, and most of those things you can't even hear. It is a fun exercise to sit in a vehicle and make a change, and see if you can actually hear a difference. If you can't, then that eq is probably not even needed.

I try to limit myself to 10 bands of PEQ for fixing the curve. Then I can use the other 21 bands for fixing the imaging by ear and all pass filters. Usually I need about 15 or so eq changes to get everything dialed in.

[Jdunk54nl]

Also good questions so far! You are clearly doing some research and are asking the right things!

[Uberverbosity]

What I'm interpreting is that there's no issue with a boost/cut in and of itself, but let's say I stack 10 filters that affect the same range, some boosting some cutting... while the resulting level of my target may be where I want it, I've likely wreaked havoc on the surrounding frequencies in an unnatural manner.
This relates directly to your comments about minimal eq and the interplay between what looks right and what sounds right.

[Jdunk54nl]

Right a boost or dip in and of itself is fine. Stacking filters can increase the overall amount of boost at certain frequencies. And yes, if you need to stack 10 filters to get it "flat" in a relatively short region (say like 500hz to 700hz), you are too worried about the prettyness of the graph. Usually on my door midbass drivers I have like 4-6 eq for 80hz-350hz. My wifes door midbass goes from 80hz to 2400hz. That needs all 10 filters in the dayton dsp 408 in her car to cover that range.

Like the example below. I cut at 100hz -6.3db but boosted at 100hz by 10db, overall the headroom needed is 5.9db at 100hz as seen by what I circled in red. So I would either need to not do this, or cut the overall level of that driver by 5.9db, so my dsp "gain" would have to come to -5.9db for that driver. That way at 100hz, it will only hit 0db max. Normally I try to get 0db headroom required in that circled part, especially on lower frequency driver. Higher frequency driver I am ok with a little bit of boost due to how music is actually made (very similar to pink noise with a -3db/octave slope).

[Justin Zazzi]

I appreciate the insight.

Can you speak to the debate between "knock down the peaks - no boost" vs 'beat that curve into submission'?
Aside from maybe losing some natural transients and dynamics through overly-aggressive eq-ing, electrically, is there any difference between a 1khz/75db signal that comes direct from the source and a 1khz/75db signal that came into the processor at 90db and was cut down to 75db?
Depending on width I can see how surrounding frequencies might be affected by a filter, but does the relationship change in some way at varying levels, or in the time domain?
In short: Is there an acoustic basis for the belief that 'less is more', or is it really a question of tuning prowess?

Jdunk's reply is great, as usual!
Here are some more things to think about.

In your example, the signal coming into the processor at "90dB" is +15dB hotter than the other signal, and so I would expect it to have a better signal-to-noise ratio. I don't think this is what you're asking about though.

If we take the example to the extreme like boosting +50dB and then also cutting -50dB, the end result should (in theory) be +0dB of gain or neutral. In the real world, however, the processor can only make something so loud before it runs out of numbers. This is the same problem of using a simple handheld calculator that has 7 digits on the front. You can do math with up to seven digits all day long without any problems
like 1 + 5,000 - 5,000 = 1
But if you try to do math with numbers larger than seven digits then the calculator can't display it and gives you an error:
like 9,999,999 + 1 = error

This is the same problem as taking a picture and shrinking it down really small then trying to blow it back up to full size. You will lose some information and the image will be blurry when you try to make it bigger again. In this case, you can run out of really really small numbers just like you can run out of really big numbers.

So if you have a very heavy hand with EQ filters, it is possible you can start losing information with a combination of lots of boosting and lots of cutting. Modern processors have the ability to use really big and really small numbers. You'd have to do something pretty odd before this becomes a problem. Probably.

This is where Jdunk says "if you have the digital headroom". You can, actually, manipulate the signal to larger than +0dBFS (full-scale) and have music at +10dBFS ... inside the processor anyways. But when you try to output it from the DAC, it cannot do anything greater than +0dBFS and then you run into problems and things sound really awful.

I may or may not have seen someone boost like 21db before.....on the low frequencies....it was a driver natural response issue (I believe anyway) and the driver was fine with the boost and so was the amp, so it was fine.

That particular boost wound up in a motorcycle sound system. It's hard to get bass out of a motorcycle so ... I boosted it? I didn't do that blindly though, I had the good fortune of knowing everything about the whole system and being able to use protection schemes like bass management and compressor/limiter to prevent over-driving the system too much. Fun stuff when you can work with a closed system like that.

A very similar example is the open-baffle loudspeaker towers at my home. These are the Linkwitz LX521 design and have four 10" long-through woofers running open-back without an enclosure, so they are very inefficient, so I need lots and lots of bass boost to make them sound correct. This project I also have access to advanced protection mechanisms like compressor/limiter and I can set the limits to be just on the edge of what my amplifier and woofers are capable of. When I set the protections just right, it becomes indestructible, and then I can do something silly like this:

[Justin Zazzi]

One more thing: I've been reading a lot lately about speaker correction vs room correction. It seems like the direct sound, the sound that reaches us first, has a large impact on the tonality that we perceive. If your speakers are pretty good to begin with, and you start EQ'ing them a whole bunch, then you'll be disrupting the tonality of the direct sound. You would be trading better tonality due to the room's influence vs better tonality coming directly from the speaker itself.

I haven't experimented with this concept much because there is almost no difference between the direct and reflected sound in a car. I will be playing with this concept a lot more in my home stereo system and other rooms that are not-cars. This also means using an RTA and constantly playing pink noise would not work because you cannot separate the direct from the reflected sound. You would need to use chirps or other impulsive measurement techniques, and the moving mic method doesn't work either. Seems like a lot more work, potentially. Maybe.

In these cases, I believe EQ'ing a system aggressively could show up as a problem more commonly than doing the same in a car.

[Justin Zazzi]

Assuming you could get an airtight seal. What would the effects of filling an enclosure with a different gas be? Say helium vs air.

https://www.diyaudio.com/forums/subw...afluoride.html

Sent from my XT1710-02 using Tapatalk

I finally found a good answer for this one. Thank you for being patient! My great friend (and coworker at Rockford) Garrett is exceptional at math and did a wonderful derivation of this. See images below.

In the math below, the final equation states:

k_box = .. [(..)^ɣ -1]

You want the stiffness of the box (k_box) to be as low as possible so it behaves like a large box. The only variable that changes when you use a different gas is ɣ (gamma) which is the ratio of specific heats. Garrett calls it the adiabatic index which is the same thing but from physics textbooks.

Looking at the table at the very bottom, find a gas with a small value for gamma ɣ and you'll make the box appear larger. It looks like some of the best choices are the highly flammable hydrocarbons like propane lol.

[Jdunk54nl]

One more thing: I've been reading a lot lately about speaker correction vs room correction. It seems like the direct sound, the sound that reaches us first, has a large impact on the tonality that we perceive. If your speakers are pretty good to begin with, and you start EQ'ing them a whole bunch, then you'll be disrupting the tonality of the direct sound. You would be trading better tonality due to the room's influence vs better tonality coming directly from the speaker itself.

I haven't experimented with this concept much because there is almost no difference between the direct and reflected sound in a car. I will be playing with this concept a lot more in my home stereo system and other rooms that are not-cars. This also means using an RTA and constantly playing pink noise would not work because you cannot separate the direct from the reflected sound. You would need to use chirps or other impulsive measurement techniques, and the moving mic method doesn't work either. Seems like a lot more work, potentially. Maybe.

In these cases, I believe EQ'ing a system aggressively could show up as a problem more commonly than doing the same in a car.

I've been playing around more with the omnimic software, which gates the response and goes "anechoic" at certain frequencies. I know you can do this with REW (and possibly smaart????) but omnimic is easier to use for this stuff.