Electrical Crossover and Acoustical Response

[SkizeR]

I was talking about the low end having a tom of bossy and playing very low

Sent from my SM-G975U using Tapatalk

[Brego]

I was talking about the low end having a tom of bossy and playing very low

Sent from my SM-G975U using Tapatalk

I totally understand what he is saying, just 3 more beers and when I say it out loud it will make even more sense... lol

[Justin Zazzi]

Could you accomplish the same result if you used 600/12dB instead of 750/36dB? 12dB Xo's gives a more natural sound, and you could reduce the gap in your Xo's to the Speaker Curve...

I'd like to reinforce jtrosky's response to you.

Every rule of thumb and every comment you've ever heard about crossovers are supposed to be about the acoustic response, not the electrical response.

The electronic crossovers and filters and magical stuff you can do with a DSP doesn't have a "sound" or a performance of any kind by itself. You can only evaluate the sound and performance and tightness and whatever else after you combine the filters with the speaker, the enclosure, the room they are playing in, and from your listening position.

For example:
-remove the speaker from the system and you get no sound, you can't evaluate anything
-remove the enclosure from the system and all your bass will evaporate, so it will "sound" thin
-remove the room from the system or play the speaker outside the car and it will sound very different, likely less bass because there is a lack of cabin gain
-remove your listening position from the system and the sound will change entirely based on where you are standing. put your head in the passenger footwell or listen to the radio with your head in the engine compartment and the "sound" will be terrible.
-remove the filters from the system and you'll have a raw un-tuned response of a system that is not optimized, and very likely it will not sound great.

[Justin Zazzi]

This topic is so outrageously important and so commonly confused that I want to share more.

The following two posts are from my build log over on DIYMA from December 2013 (wow it's been a while!). These posts are quoted word for word with the graphs I made at the time. This was a huge ah-ha! moment for me that changed everything.

I posted lots of useful stuff in that build log if you want to learn more things. Wow that was a fun thread.

[Justin Zazzi]

Set the speaker to run wide open without any crossovers, eq, loudness controls, or other processing of any kind. We want to measure the natural response of the speaker as it is installed from the listening position. Here is what my front left fullrange speaker looks like:

The pink line is the speaker's frequency response at the listening postition, inside the car, installed in the door. The blue line is the target curve I built in excel using the equations from the Texas Instruments page linked in my previous post. The equation is shown if you want to try this method too. Using that equation and Excel, create a "house curve" text file that room EQ wizard can use in the format shown on REW's help pages here.

Fun fact: this is where my inspiration came to build the "Jazzi's tuning compantion for Room EQ Wizard" !

The Mosconi 6to8 has a maximum boost of +12dB, so I placed my target curve 12dB above the lowest point of my speaker's response that I anticipate to be boosting. The crossover point will be 300hz, and you can see the target curve replicates the LR4 filter perfectly because it is down by 6dB at the crossover point, and maintains a steady 24dB/octave slope thereafter.



So here's the cool new stuff.



Notice the speaker's response is very similar in slope to the target curve without any tuning. This is interesting because minimal processing is needed to reach the target. I'll start with a 2nd order variable-Q high-pass filter in REW and watch what the "predicted" response will be in real time without having to re-measure the system again and again (very, very cool). By varying the crossover frequency and the Q, I arrived at the following filter and after measuring the speaker again, this is my new response:

If you look at those numbers, you might be thinking ... what the heck? The one filter I'm using is contrary to just about every "rule of thumb" out there. It's not the steepest one I can use, nor is it branded by some cool name like "Butterworth" because I'm using a Q factor that doesn't match anything in particular. Also, the frequency of the filter is about 2/3 of an octave above the crossover frequency I was aiming for.


But wait a minute. With just a single 2nd-order filter and no EQ whatsoever, my speaker now resembles a textbook Linkwitz-Riley 4th-order. The response is even +-5dB up to about 1.5khz without any other tuning. How cool is that? One filter!


So I took it a little further and added very aggressive parametric EQ to get the response to match the target curve as close as possible (except for 8khz and above where I'll tune by ear later).

Next step, do the same with the left front midbass.
Then, compare the phase of each speaker and see what happens.

[Justin Zazzi]

continuing from my build log...

First thing you need to do is start this playlist of Lindsey Stirling and The Piano Guys on youtube. Their work is truly awesome, and I'm sure a bunch of you will enjoy the mashup of classical string instruments with dubsteb and other modern fusion music.

Ok, now that you have that playing, check this out. It totally worked!

----

After tuning the left channel fullrange driver in my above post, I went back outside and tuned the left midbass speaker using all of the same ideas and techniques. Below is (first) the acoustic response of the midbass without any processing and (second) the response after two 2nd order variable-Q high-pass and low-pass filters were applied along with aggressive parametric EQ to match the target curve as close as possible. Note, the target curve is created in excel using the same techniques as the post above, though this one is a little more trick to make.

Just a quick recap. I built my target response curves based on the 4th-order Linkwitz-Riley crossover alignment at frequencies of my choosing (50hz and 300hz). Then using Room EQ Wizard, I applied minimal amounts of variable-Q high-pass and low-pass filters to shape my acoustic response to match the target curve. Then I aggressively used the parametric EQ to force the acoustic response to very closely match the target curve. Finally, I used the loopback feature of REW to estimate the impulse response delay of each channel, and delayed the rullrange speaker by 0.42 milliseconds.

Here are the results. The following two graphs say it all.

So there it is. It worked. I'm really and truly happy with this result. I did not massage any of the data with the intention of creating this result, and in fact I had the time delays backwards at first and had a completely different (and disappointing) result. To have a hypothesis completely and definitively confirmed on the first try is beyond satisfying.

I hope this has been helpful. I've sure learned a lot.
Please ask questions because I know this isn't the easiest thing to describe!

Conclusions
1) Time delay should be used to align the first arriving waves from each speaker (as a compensation for time only) and not as I previously thought as a tool to find the optimum interference between speakers (as a compensation for phase).
2) Crossovers are not one-size-fits all. The electronic (or passive) crossover needs to be a compliment to the pre-tuned response of the system to result in an acoustic response that matches the crossover alignments we are all so familiar with. The 4th order Linkwitz-Riley is my preferred alignment in this case.

[Brego]

WOW, Nice write-up Jazzi !!... I have to read it about 6 more times,... Thank You for this information !!

Somehow I missed this build log over on the other forum.....

[jtrosky]

This topic is so outrageously important and so commonly confused that I want to share more.

The following two posts are from my build log over on DIYMA from December 2013 (wow it's been a while!). These posts are quoted word for word with the graphs I made at the time. This was a huge ah-ha! moment for me that changed everything.

I posted lots of useful stuff in that build log if you want to learn more things. Wow that was a fun thread.

Whoa - school is *in session*!!! I haven't read any of it yet, but I most certainly will - probably multiple times! Thanks for posting all of the great info!

@Brego - This is probably all mentioned in that info from Justin, but you can actually use REW to "simulate" different crossover settings. For a midbass, for example, take a pink-noise measurement without any crossovers, EQ, etc. Then, once you have that measurement in REW, you can go into the EQ page and "define" different crossover slopes to see what they would do to the response. I use the "LP" and "HP" entries. Just keep in mind that each LP/HP entry is a 12dB slope, so if you wanted to simulate a 500hz, 24db slope, you would create two exact same entries in the EQ sections (Type: HP, Frequency: 500) - and REW will show you how that HP/LP filter will effect your response. If you want it to simulate a 36dB slope, you would do 3 exact same entries in the EQ page. That is how I determined that 725hz/36dB would work well for me on that one midbass (instead of 500hz/24db, which is the acoustical xover I wanted) - I just tried different combinations in REW until I found one that worked the best - then tested it in the actual car.

Just make sure you are careful with mids/tweeters - you don't want to take any loud measurements without *any* xovers defined or you could damage the speakers. I just used this method for my midbass and rear-deck coaxial speakers.

I'm sure Justin has better, more detailed info on this in his links above though.

[Brego]

Very Helpful, very very Helpful information, Thank You! I remember reading that somewhere in the past, but never tried it... I'm going to give it shot.

Did you keep that 725hz/36dB setting for your mid-base. Does it sound really good?

[jtrosky]

Did you keep that 725hz/36dB setting for your mid-base. Does it sound really good?

Yes, I've been using the 725/36dB setting for a few months now and it works out really well for my setup.