DiyMobileAudio.com Car Stereo Forum banner

81 - 100 of 210 Posts

·
Registered
Joined
·
530 Posts
Much better this time. I will remake the test for both fronts - to see how big TA values should be for their alignment (both on one graph). In any case, both fronts have to be delayed appr. 16-18 ms for better connection with the sub.
 

·
Registered
Joined
·
1,999 Posts
No. Place it in the absolute center, no different positions for left/right ;)
OK!

You do not have that ability.
yep,
but I have a 180 switch on sub maybe?

Much better this time. I will remake the test for both fronts - to see how big TA values should be for their alignment (both on one graph). In any case, both fronts have to be delayed appr. 16-18 ms for better connection with the sub.
Thx Alextaastrup,
this soft is easy in fact, the easiest to read maybe.
Anyway I think I'll try the Cdsp soon to have more control (after the rain :p)
 

·
Premium Member
Joined
·
1,148 Posts
This is the best I have been able to achieve. But as you can see my sub is 20+ ms out and I am out of delay on my tweeters. The Helix only goes to 15.6ms. So any suggestions on what I should try?
Right now I using LR 24's for all crossovers. Should I try using a lower order crossover and maybe Bessel or Butterworth instead of LR?

[/URL][/IMG]
 

·
Registered
Joined
·
4,973 Posts
Discussion Starter #87
Acoustic crossovers are what matters, see the crossovers in your DSP as response shaping. There seem to be sub/mid integration issues as well as T/A and/or L/R FR inconsistancies. I would do individual measurements on all driver by RTA/noise so you get as much averaging as possible. Then compare all the individual measurements to a full left side+sub and a full right side +sub, check if all drivers sum at crossovers. Check actual slopes. Do a sweep of left side and one of the right side in RoomEQ for example, check excess group delay for large peaks which indicates destructive interference.

If the L/R EQ is set correctly you should be able to set T/A by ear very easy with correlated pink noise. I can measure my car tomorrow and show you why some of these issues occur.


Sent from my iPhone 6 using Tapatalk.
 

·
Premium Member
Joined
·
1,148 Posts
Ok, sounds like I just need to start back at step one. I want to change my house curve anyways. I will go back and make sure I have everything dialed with with EQ before trying to utilize TDA.
 

·
Registered
Joined
·
4,702 Posts
I have to stress this:

DO NOT SIT IN SEAT WHILE MEASURING. DON'T TOUCH THE MICROPHONE.

Clamp it in headrest, in the middle where your head is normally. If you move the mic even an inch the measurement will be messed up. If you are aiming for repeatable results while correcting time coherency the mic must be absolutely stationary.

Elgrosso, The results you seeing on right side are probably due to destructive interference. I assume there's dips in the frequency response (AFR) at the same spots. Is this a 2 or 3-way? If it's a door mid it's perfectly normal. A small dedicated midrange should do better if placed optimally. Where is the crossovers? ;)
Great thread, and an important venture in an important topic. :)

I've worked with CSDs in the past but the lack of software to sum up time domain responses from multiple mic positions is a huge drawback. The only way to receive an input signal unbiased is to average many many times from correct positions, with your body in the seat. That way your measurement is centered on the "true" signal your brain receives and measurement error is mitigated through averaging.

If the "measurement is messed up" by moving one inch, you have big problem. The solution is not to keep the mic fixed but to implement some form of averaging and to place the mic correctly around your years when your head is next to it.

I have no doubt that you measurement is correct for that location and an empty car, but whether that is relevant to what you hear when in your seat is ultimately a very important topic. Even if you repeat the measurement exactly by mimicking your previous test position it does not mean your results are accurate, just that the software has high accuracy. You still have to feed it the right signal.
 

·
Registered
Joined
·
4,702 Posts
Ms8 and APL, so it's really not me :D
But check that, left and right:





Sub could be better on both, i have somehting strange on the right woofer, and some ringing on the right tweeter?
I'll try some more with ms8 on/off and apl on/off.
Check your vertical axis. It appears you changed the axis settings between left and right. The software likely gives you the option to focus on certain frequencies and you may have done so not knowingly.
 

·
Registered
Joined
·
4,973 Posts
Discussion Starter #91
Great thread, and an important venture in an important topic. :)

I've worked with CSDs in the past but the lack of software to sum up time domain responses from multiple mic positions is a huge drawback. The only way to receive an input signal unbiased is to average many many times from correct positions, with your body in the seat. That way your measurement is centered on the "true" signal your brain receives and measurement error is mitigated through averaging.

If the "measurement is messed up" by moving one inch, you have big problem. The solution is not to keep the mic fixed but to implement some form of averaging and to place the mic correctly around your years when your head is next to it.

I have no doubt that you measurement is correct for that location and an empty car, but whether that is relevant to what you hear when in your seat is ultimately a very important topic. Even if you repeat the measurement exactly by mimicking your previous test position it does not mean your results are accurate, just that the software has high accuracy. You still have to feed it the right signal.
TDA is some kind of mashup between a CSD and Group Delay plot.

You are correct that you need to do a lot of averaging when doing CORRECTIONS in the frequency response.

This is different. We are measuring ONE point in space, time domain measurements are always and only available at one specific point in space. If you are sitting in the car while doing time domain measurements or impulse aligning, it will most likely be messed up because the head cause "shadowing". It can cause significant changes in the frequency response and it will impact the phase response / time domain as well. You are also correct that we need to correlate this with what we are hearing, the head shadowing is a distortion in the measurement we don't want, we don't perceive the sound that way, there is always acoustic crosstalk. I've done extensive testing into this and found that an empty car with the mic is usable for two types of testing.

1. Time domain measurements.

2. Subwoofer measurements (matters very very little if you sit in the car or not, no averaging is required either).

Don't understand what you mean by "feeding it the right signal" though. TDA creates the sweep and measures at the same time.
 

·
Registered
Joined
·
4,702 Posts
TDA is some kind of mashup between a CSD and Group Delay plot.

You are correct that you need to do a lot of averaging when doing CORRECTIONS in the frequency response.

This is different. We are measuring ONE point in space, time domain measurements are always and only available at one specific point in space. If you are sitting in the car while doing time domain measurements or impulse aligning, it will most likely be messed up because the head cause "shadowing". It can cause significant changes in the frequency response and it will impact the phase response / time domain as well. You are also correct that we need to correlate this with what we are hearing, the head shadowing is a distortion in the measurement we don't want, we don't perceive the sound that way, there is always acoustic crosstalk. I've done extensive testing into this and found that an empty car with the mic is usable for two types of testing.

1. Time domain measurements.

2. Subwoofer measurements (matters very very little if you sit in the car or not, no averaging is required either).

Don't understand what you mean by "feeding it the right signal" though. TDA creates the sweep and measures at the same time.
It GD and CSD are basically the same thing no? I can see this platform has various ways to show you a plot but the information contained should be the same. The 2D plot is very nice.

As long as you can move the mic and collect another series of time domain measurements (and they should be different) then averaging is a matter of "how to" not "is it possible?" In this case if you could export your time measurements in a raw format from multiple mic positions you could average them.

Ok, if you restrict your use to TA and subwoofer measurements then I agree, you are taking in the right signal. Go on. :D
 

·
Registered
Joined
·
4,973 Posts
Discussion Starter #93
Group delay is the negative slope of the phase response. A CSD displays ringing (decay) in the magnitude response (derived from the IR). Both depict linear distortion. The 2D/3D view is kinda special, I haven't seen a program that combines delay and decay/ringing in the same plot. It's certainly useful since you get data that would require way more work to attain otherwise.

Yes you can average them that way but the thing is, what are you gonna use it for? :p

We need some kind of "zero point" to use as reference, it can be useful to move the mic and see if some issues goes away but to attain repeatability you would need some kind of jig so you can set the mic at fixed positions so you can eliminate an altered mic position while troubleshooting an A vs B measurement. We must know if we are applying corrections to fix the mic position or the speaking tuning, that's the reasoning behind my statement ;)

APL Workshop is the exact opposite of TDA, there you average hundreds of points in the sound power domain.
 

·
Registered
Joined
·
5,674 Posts
Finally loaded this bad boy up to mess with it. Here goes nothing. :)
 

·
Registered
Joined
·
4,973 Posts
Discussion Starter #95 (Edited)
I have a few plots to upload tomorrow, measured my new build. Right side measured almost flat from 50Hz to 20kHz +/- 2ms, with only level matching and T/A. No EQ at all. Was actually surprised how good it looked, only two minor hiccups around 1-2kHz for some reason.

Edit: I had them on this computer lol.

No tuning at all, no T/A, No level matching. Right side only.



Level matching and T/A applied. No EQ. APL bypassed.

That ringing around 400Hz is due the high Q small pods of the midrange drivers. I might have to vent them better... You can also see the high Q peak for midbass enclosures around 130Hz.



Sub got pretty high GD even though it's in a large sealed box.

 

·
Registered
Joined
·
1,999 Posts
Check your vertical axis. It appears you changed the axis settings between left and right. The software likely gives you the option to focus on certain frequencies and you may have done so not knowingly.
Here they're really almost the same.
I have trouble managing the view on this, it jumps from front 2d to 3d with the slightest trackpad move. Maybe it's // desktop, or just windows...
 

·
Registered
Joined
·
4,973 Posts
Discussion Starter #98
Nice, where do you cross your mids here, 200hz??
And how did you change the time window on the graph? Or I guess you have the full version now

80/200/3700Hz.

There's a box where you can enter the time window. Yep I'm running the "full" version


Sent from my iPhone 6 using Tapatalk.
 

·
Registered
Joined
·
5,674 Posts
K so am I doing something wrong? Or will it simply not allow seeing the TDA plot or 3D plot in demo mode?




Sent from iPhone using Tapatalk
 
81 - 100 of 210 Posts
Top