REW and DSP help - Acoustical vs Electrical crossover.

Acoustical vs Electric crossovers.

Do you get electrical vs acoustical crossover differences to the tune of several hundred hz?

I'm using a JL Twk88 DSP with their TUN software. I have an Imm-6 measurement mic and REW. I EQ'd my system about a month ago, it was my first time to REW, measurement mics, and DSPs. It went really well. I ignored the acoustical vs electrical crossover stuff because I'd already used the EQ to hammer it in line and it was too late to go back. I used boost (realized after talking to an installer and fellow owner of a Honda fit who also was using the same mid-bass that we were fighting the same nulls) in a couple areas. I re-adjusted my gains to compensate for the added power.

I'm starting from scratch as I have a few new components. New subwoofer(s), and some horns. This post is only focused on the mid-bass and the subwoofer. The rough goal is no boost, only cuts. I'm trying to start by ignoring the nulls. I might experiment with them later. What I'd like to focus on for the comments to this post is acoustical vs electrical crossovers and what your experience with how far off they can get.

My Midbass previously was ~20-30hz off of my LPF electrical crossover. Same thing is happening now. So far I only applied the EQ suggestions from REW using my house curve. I'm only showing 2 averages on the photos. With EQ and without EQ, no change in crossover points. The subwoofer got no change in EQ because it was pretty flat, and I didn't really know how to address the crossover being THAT far off.

My midbass HPF should be dipping down at ~640, but isn't starting until past 1khz. My subwoofer is ridiculous. Its playing to 140hz when it has an electrical crossover of 80hz. Last time I EQ'd a different subwoofer it played to ~100hz and I just EQ'd it into shape.

But before I get too deep into tuning I wanted to ask this board some questions and figure out what I should do and if this is normal.

Subwoofer is set to HPF 20hz - LPF 80hz.
Mid-bass is set to HPF 80hz - LPF 800hz
Horns then kick in above 800hz - I'm leaving those out of this discussion for now.

Attachment 13443Attachment 13444Attachment 13445

What you're seeing is not uncommon at all.
For example, below is a picture of me using a 510hz high-pass filter to hit a 300hz crossover target.
And that's just the driver's side .....

https://photos.smugmug.com/photos/i-.../i-Qb3NmSQ.gif

Just a quick one, is there any reason you don't want to Upgrade to a newer version of rew? That looks very old? I’ve been using the beta versions for over a year now and they are very very stable

as for your responses, it’s likely a big hump in response and a wide dip lower down, a crossover will look accurate on paper with a perfectly flat response before hand, that’s why you adjust the crossover so the actual response matches the target curve, it’s the dropping of level that gives the phase shift so in theory you could eq in a crossover if you had big enough cutting capability’s

if the actual crossover point isn’t low enough lower the electrical crossover some more

Quote:

---

Originally Posted by dumdum

Just a quick one, is there any reason you don't want to Upgrade to a newer version of rew? That looks very old? I’ve been using the beta versions for over a year now and they are very very stablee

---

screenshot looks old because it was done years ago, so it's old : )

Oh he was talking to you. Lol. I shamefully updated to the latest version of REW.

I got my subwoofer tamed very easily. It fell into line with a change in electrical crossover and 2 bands of EQ.

Like I said previously, I have to take a drive to go EQ and play obnoxious sine waves and pink noise. I was playing in my driveway at VERY low volumes and thought it was interesting that without changing anything, the acoustic is almost the exact same as the electrical crossovers. It must just be with volume that it gets wily. I don't know enough about how active crossovers work on an electrical level, but if its consistent that they are more true at low volumes then one would think that the companies producing DSPs could maybe find a correlation to how volume affects the crossover drift ... It would be interesting to see if this is consistent. I'm curious Jazzi if your midrange is "true" to the acoustic and electrical crossover at very low volumes and drifts as it gets louder.

I drove all the way up the hill to my usual spot, got my mic hooked up and everything set up. I was about to play my first sine sweep when my buddy pulls up out of nowhere on his bicycle .... 45 minutes later and no EQ work was done at all... better luck tomorrow.

Quote:

---

Originally Posted by DirtyBumOAK510

Oh he was talking to you. Lol. I shamefully updated to the latest version of REW.

---

Was that aimed at you instead? What's going on!? lol

Quote:

---

Originally Posted by DirtyBumOAK510

Like I said previously, I have to take a drive to go EQ and play obnoxious sine waves and pink noise. I was playing in my driveway at VERY low volumes and thought it was interesting that without changing anything, the acoustic is almost the exact same as the electrical crossovers. It must just be with volume that it gets wily. I don't know enough about how active crossovers work on an electrical level, but if its consistent that they are more true at low volumes then one would think that the companies producing DSPs could maybe find a correlation to how volume affects the crossover drift ... It would be interesting to see if this is consistent. I'm curious Jazzi if your midrange is "true" to the acoustic and electrical crossover at very low volumes and drifts as it gets louder.

---

The behavior of an audio system should be linear, meaning it behaves the same at low volume and high volume. This is certainly true for DSPs and amplifiers and other electronics unless you're pushing them *way* too hard. This is also true for speakers at medium and low and "very low" volumes. So I doubt your system would behave different between medium, low, and very low volumes.

It is more likely the very low volume measurements were corrupted by ambient noise or some other non-speaker-thing that is usually quiet enough to not matter when you're measuring a medium volume levels.

Unless you're using part of your factory radio system? Some factory radios have a crossover frequency shifting as volume increases to protect speakers from the lowest frequencies as you start to hammer on it. This is a really neat feature when you have lots of design constraints and cannot use the speaker you want, so you use the speaker you have in the best way you can at each volume level.

I think what I'm saying is if you're really seeing this behavior, it's not on accident because normal DSPs don't work that way. It takes a lot of effort to setup this kind of feature on purpose. You can very easily check for this behavior by taking multiple measurements as you slowly increase the volume on your radio a couple clicks at a time then look at all the measurements on the same screen at the same time.

Quote:

---

Originally Posted by Justin Zazzi

Was that aimed at you instead? What's going on!? lol



The behavior of an audio system should be linear, meaning it behaves the same at low volume and high volume. This is certainly true for DSPs and amplifiers and other electronics unless you're pushing them *way* too hard. This is also true for speakers at medium and low and "very low" volumes. So I doubt your system would behave different between medium, low, and very low volumes.

It is more likely the very low volume measurements were corrupted by ambient noise or some other non-speaker-thing that is usually quiet enough to not matter when you're measuring a medium volume levels.

Unless you're using part of your factory radio system? Some factory radios have a crossover frequency shifting as volume increases to protect speakers from the lowest frequencies as you start to hammer on it. This is a really neat feature when you have lots of design constraints and cannot use the speaker you want, so you use the speaker you have in the best way you can at each volume level.

I think what I'm saying is if you're really seeing this behavior, it's not on accident because normal DSPs don't work that way. It takes a lot of effort to setup this kind of feature on purpose. You can very easily check for this behavior by taking multiple measurements as you slowly increase the volume on your radio a couple clicks at a time then look at all the measurements on the same screen at the same time.

---

I was going to say just that... floor noise ���� And the rew upgrade was for the OP :D

I've learned to EQ before applying xover filters. Makes it easier to identify areas outside of the passband that needs correction. Using LR4 xover filters; I'll EQ up to an 1.5 octaves outside of the passband and maybe more if I have filters left. I found this usually results in acoustical xovers matching electrical or very close at least.

Quote:

---

Originally Posted by Justin Zazzi

Unless you're using part of your factory radio system?

---

Not a factory radio. I'm gonna chalk it up to background noise and noise floor issues because I was doing this with VERY low volume. I'll take your word for it.

So I just realized I've been living in bonehead land. Somewhere I could have sworn I read that you should EQ with Sine sweeps and not pink noise. So thats how I eq'd my last build. Now I'm halfway through EQing this one and realizing the error of my ways after chatting with someone. So I'm supposed to have been using periodic pink noise. After searching the forums, this has become clear. Doh!

So my questions of the day, which I ask after scouring threads all afternoon and still not feeling confident.

1. When do we use sine sweeps? What purpose do they serve?
2. I'm setting my EQ by using 8 measurements, (Left speaker = 3 left ear, 5 right, Right speaker = 2 left ear, 6 right ear), then taking an average. Then plotting that against my custom house curve (Andy's bass, Jazzis dip (verified by Eric Stevens)) So if I'm using periodic pink noise... How long do I need to take each measurement (per ear)? It seems link pink noise repeats itself every 3.5 seconds, so is that as long as I need to take a measurement?
3. When downloading Periodic pink noise ... I know that I match the FFT on my RTA graph. But when downloading the periodic pink noise, I'm downloading it at 0dbfs? Left only? (for left speakers) and vice versa R only for right speakers? Because L+R pink noise is no longer mono ...

Thanks in advance.

Quote:

---

Originally Posted by DirtyBumOAK510

So I just realized I've been living in bonehead land. Somewhere I could have sworn I read that you should EQ with Sine sweeps and not pink noise. So thats how I eq'd my last build. Now I'm halfway through EQing this one and realizing the error of my ways after chatting with someone. So I'm supposed to have been using periodic pink noise. After searching the forums, this has become clear. Doh!

So my questions of the day, which I ask after scouring threads all afternoon and still not feeling confident.

1. When do we use sine sweeps? What purpose do they serve?

---

You can set your frequency response using sine wave sweeps (chirps) or pink noise. Either method works good. When using REW, I find it's a little easier to wave the mic around for a while to get a spatial average while using pink noise, compared to making multiple sweeps in multiple locations and then averaging a bunch of traces in the program. However, chirps can give you polarity and phase information which pink noise cannot, so I find chirps useful for sanity fhecking the polarity and setting the time delays.

Quote:

---

Originally Posted by DirtyBumOAK510

2. I'm setting my EQ by using 8 measurements, (Left speaker = 3 left ear, 5 right, Right speaker = 2 left ear, 6 right ear), then taking an average. Then plotting that against my custom house curve (Andy's bass, Jazzis dip (verified by Eric Stevens)) So if I'm using periodic pink noise... How long do I need to take each measurement (per ear)? It seems link pink noise repeats itself every 3.5 seconds, so is that as long as I need to take a measurement?

---

What do you mean by "verified by Eric Stevens"?
Also, you're right that periodic pink noise is a repeating pattern. That is the "periodic" part. If you have the FFT size of the RTA set the same as the sequence length of the pink noise, and if you have the window on the RTA set to "none" or "rectangular" then to get the RTA result fully updated you only need to measure for a cycle or two of the pink noise which is around a second or two. If you are doing a spatial average to capture near your left ear, right ear, and so on ... then you'll want to ensure the "averaging" feature is set to infinite and capture a cycle or two of noise at each location. So if you're waving the mic around near your left ear, we could consider that 3-5 locations and you'll want something like 5 seconds of capturing in that area. Then another 5'ish seconds near your right ear.

The exact amount of time is not critical. The important thing is can you make the measurements reapeatable? Like if you make a "measurement" using whatever method you want, can you make a second (separate) measurement and then compare the traces and have them agree really nicely? That is when you know you're moving slow enough, measuring for long enough, and doing it "right".

Quote:

---

Originally Posted by DirtyBumOAK510

3. When downloading Periodic pink noise ... I know that I match the FFT on my RTA graph. But when downloading the periodic pink noise, I'm downloading it at 0dbfs? Left only? (for left speakers) and vice versa R only for right speakers? Because L+R pink noise is no longer mono ...

Thanks in advance.

---

Pink noise should be 0dBFS meaning the loudest momentary peaks in the noise file should be at maximum level (in the recording). You'll want to also pay attention to the crest factor which is the difference between the loudest peaks and the average energy. Aim for a crest factor of 12dB if you can, and REW should let you choose this number. If you set the crest factor to a really small number in REW then it will start sounding very distorted and likely not give you a good measurement.

Quote:

---

Originally Posted by Justin Zazzi

What do you mean by "verified by Eric Stevens"?

If you are doing a spatial average to capture near your left ear, right ear, and so on ... then you'll want to ensure the "averaging" feature is set to infinite and capture a cycle or two of noise at each location.

You'll want to also pay attention to the crest factor which is the difference between the loudest peaks and the average energy. .

---

Re: "Eric Stevens Verified". Verified wasn't the appropriate word. I asked him what his preferred house curve was and he told me the same dip in the ~1.5khz-5khz region. I believe he even used the same exact numbers that your text file uses.

Averaging to "Infinite" is Averaging "Forever"? That is how REW words it.

Where are you finding the "Crest factor"? Is it that "2.15dbfs peak"? I googled it and it popped up that you can change it to "CTA-2034" to increase that number to ~9. But I'm not exactly sure what that will do. I'm not sure what CTA-2034 is or if it will affect the "Full range".

You don’t want your generator digitally clipping (the red on the wave form and your peak is +2.15 dbfs. That is not going to be good. The peaks should never exceed 0dbfs

Also, I just found this out thanks to Justin, if you double click on the image after uploading, you can adjust size of them. That way people can see them without having to click on them

So how do I change it?
I'm a bit confused now. I want that big number to be 0, and the small number to be +12? or -12?
I'll post some shots of some of the options I have.

Sorry I'm not on the Beta branch of REW so my thoughts above will be misleading.

I read the beta documentation and it looks like you're doing everything right but the big number should read "-6.00dBFS" or smaller, as in more negative like "-12.00dBFS". That number sets the RMS or average number and works different than I thought it did. Oops!

Make sure you remember what the sequence length is. Your screenshot shows 64k. I like to put it in the filename of the wav file so I never forget it. You'll need that number when you set the FFT length in the RTA module. If you use a alittle shorter length like 32k or 16k then it will repeat itself faster which means the RTA will update faster.

If you'd like to create a wav file and share it, I can analyze it to make sure it works the way we think it does.

Quote:

---

Originally Posted by Justin Zazzi

I read the beta documentation and it looks like you're doing everything right but the big number should read "-6.00dBFS" or smaller, as in more negative like "-12.00dBFS". That number sets the RMS or average number and works different than I thought it did. Oops!

If you'd like to create a wav file and share it, I can analyze it to make sure it works the way we think it does.

---

Hmmmm .... CAJ won't let me upload wav files. Email? What is the preferred method for sharing wavs here?
Thank you for the help. I can share 2 files. One has the big number as -12, the other has the small number as -12. I'm assuming its the small number is the "crest". Both files avoid clipping.

Am I right in assuming that I use a "Left Channel" to tune my Left speakers, and a "Right Channel" to tune my Right speakers? What about if I want to match levels between left and right? For instance if I'm trying to match my midbass to my horns?

Cheers.

Progress!

This is the file I use of Pink PN 64k
This was generated with REW and is about 11 minutes long. This was generated with -12dbfs for the file and peaks are like -9dbfs.

https://drive.google.com/file/d/1c5w...ew?usp=sharing

Quote:

---

Originally Posted by Jdunk54nl

is about 11 minutes long. This was generated with -12dbfs for the file and peaks are like -9dbfs.

---

Looks similar to what REW wants to produce for me. -12dbfs exactly and a "-9.85 dbfs pk"

My favorite tool to use is ACX Check plugin for Audacity since it's free and calculates a bunch of stuff for you. I recommend you give it a try and see if it works for you. It doesn't work for me for some reason. Instead, below is a Matlab script for doing something similar. The results are at the bottom. There is an open-source version called Octave that everyone can download and use if you'd like to give it a try. You can analyze your own files using the code below.


Attachment 13535


Looks like the peak value is -9.75dBFS which matches what you saw. The crest factor is close to 6dB which matches the documentation from the REW Beta help files. But the average value is -15dBFS which doesn't match what you saw on the screen. I'm not sure why that is.

In a couple days I'd be curious to get your input about the differences between sine sweeps vs pink noise. I'll share some pictures once I have re-tuned everything to pink noise.

Question(s) of the day:
So sine sweeps can bring out phase issues, as well as resonance issues, right? So at what point should I use sine sweeps to double-check my measurements and what exactly should I be looking for? Where is the happy medium?

Quote:

---

Originally Posted by DirtyBumOAK510

In a couple days I'd be curious to get your input about the differences between sine sweeps vs pink noise. I'll share some pictures once I have re-tuned everything to pink noise.

Question(s) of the day:
So sine sweeps can bring out phase issues, as well as resonance issues, right? So at what point should I use sine sweeps to double-check my measurements and what exactly should I be looking for? Where is the happy medium?

---

Use pink noise, rta, and moving mic averaging to measure and tune, if something isn't acting right (xovers not summing correctly or other weirdness) then use sweeps to diagnose problems. In short, take both types of measurements and use rta for tuning.

Quote:

---

Originally Posted by Justin Zazzi

The important thing is can you make the measurements reapeatable? Like if you make a "measurement" using whatever method you want, can you make a second (separate) measurement and then compare the traces and have them agree really nicely? That is when you know you're moving slow enough, measuring for long enough, and doing it "right".

---

Thanks again for all the help. I had an issue I was having a hard time wrapping my head around. I am on round 5 of tuning. In Round 1 I checked for polarity of the speakers, and everything was as it should have been. I did this using the IASCA test CD. At the end of rounds 1-4 I verified polarity by reversing polarity on my mids and looking for the major gulch around the crossover. Everything was normal. I just completed round 5, and I got the crossovers to almost perfectly match the house curves. But when I ran the entire left side (subwoofer, mid-bass, and horns), all of a sudden I'm getting a major dramatic dip exactly at the crossover, and if I flip the phase of my horns, the gulch becomes almost completely flat.

So my question of the night...
1. If my crossover between my mids and horns (tweeters/midrange) is out of phase ... and its fixed by reversing the polarity of the horns ... does that mean that the entire range of my horns are now out of phase? Or does it just mean that I'm out of phase near my crossover? (800hz).
2. Do I just flip the polarity of my horns to fix this?
3. Or what do I do to fix the phase issue at my crossover point?

Quote:

---

Originally Posted by DirtyBumOAK510

So my question of the night...
1. If my crossover between my mids and horns (tweeters/midrange) is out of phase ... and its fixed by reversing the polarity of the horns ... does that mean that the entire range of my horns are now out of phase? Or does it just mean that I'm out of phase near my crossover? (800hz).
2. Do I just flip the polarity of my horns to fix this?
3. Or what do I do to fix the phase issue at my crossover point?

---

1. If your crossover region is better one way, go with it.
2. Sure.
3. If you want to make the crossover sum together well then use the combination of settings that result in the best summing.

This is about the best you can do with REW that I know of, however John is constantly adding new features so maybe there's cool stuff I haven't seen yet : )

Quote:

---

Originally Posted by Justin Zazzi

1. If your crossover region is better one way, go with it.
2. Sure.
3. If you want to make the crossover sum together well then use the combination of settings that result in the best summing.

This is about the best you can do with REW that I know of, however John is constantly adding new features so maybe there's cool stuff I haven't seen yet : )

---

I need to learn more about phase...
I started off my tune and double checked (even triple checked) that all my speakers were in phase. I played phase tests and it was night and day. I was 100% confident that everything was in phase. After massive EQ work, adjusting levels, matching my measurements to house curves ... sure enough, I get pronounced -10db dips ~where my crossovers are. I reverse phase on my horns and the dips turn ~perfectly flat with my house curve. I played the same "phase test" sample tracks and again, it sounds night and day 100% correct ... but now my horns are reversed from where they started from.

As a small update:
My measurements with pink noise (I dare not go back to sweeps lest I ruin my sanity) are measuring within ~1.5 db of my house curve from 30-18khz. I am fixing a few small issues that came up during my 31 band pink noise tests. The main issue was that at certain points the right would be 1.5db the opposite direction as the left, and vice versa, thus throwing off the imaging @ 160hz, 1200hz and 2000hz. I'm going to try and dial those closer to center this week as I have ~5-6 hours of drivetime this weekend to add my "subjective" notes.
I'm having a small issue where my subwoofer is playing flat from 80hz-160hz which is adding a baby bump in my curve @ ~ 125 hz. It's only ~5db below where my mid-bass come in, so it is showing up on my RTA measurements. I don't know that I can really hear it. It may also be less noticeable because due to the -14db generated pink noise I'm using from REW, the subwoofer flat spot is @ 75db, and my whole system is tuned to 80db (as this is where the pink noise gets me). My normal listening levels are higher than that.
I had a post on DIYMA about trying to eliminate the flat spot on the subwoofer. I tried making cuts using my JL Twk88 DSP, however even with a variety of -12db (max) cuts, it only made a very small ~ -3db difference. The LPF on the amp seemed to make a bigger difference, but the problem is that the -12db/oct slope from the LPF would start to cut into my 60-80hz range. I did set my gains on the subwoofer amp to -10db attenuated tones, so likely I could afford to dial back the gains a bit. The only problem with that is that I'm using a pair of low sensitivity (~83db) 6.5" subwoofers, and right now my measured RTA readings (which match my house curve) goes to +10db from 30hz-60hz (and at f3 of +7db @ 20hz) in a dual sealed box. I like the extra room on my attenuation knob because some music that I listen to frankly lacks sub-bass. I'd rather attenuate than be looking for more.
I'm still confused as to where that flat spot from 85hz-160hz is coming from on my subwoofers...

Quote:

---

Originally Posted by DirtyBumOAK510

I need to learn more about phase...

---

If you want to learn more about phase, a good place to start is to learn the difference between phase and polarity. Anytime you make an adjustment that is labeled as a positive or negative, or a 0 vs 180 degrees, or normal vs. reversed, these are all polarity changes. These are the same as if you took the positive and negative wires on the speaker and reverse them. Polarity means any time the cone should move forward it moves backwards instead and vice versa. Polarity has the same effect at every frequency.

Everything else that you think is phase, probably is, unless it's polarity.

For your other questions, it would be easier with screenshots since we don't know what you're seeing.

Heres a few pictures of subwoofer RTA measurements.

Unfortunately these aren't labeled well, but they represent a mix of electronic crossovers from 60hz-90hz, some with EQ where I tried to eliminate the flat spot at 85-140hz with -12db spikes ... the main reason I show these photos is to demonstrate the flat spot is present no matter how I change my crossovers or EQ cuts.

That doesn't seem too strange. You have a dip around 110hz and a peak around 140hz so if you want to straighten then out you might need one filer for each: a little boost at 110hz and a little cut at 140hz (or whatever the frequencies are).

If you try to fix it with only a crossover filter, then you'll still have that squiggle but it'll be moved up or down as a whole since the crossover filter won't separate that dip and peak that as so close together.