Optimal crossover point for my measurements

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
Hi Bernt @juicehifi

Please see attached filtered measurements for left and right (2-way) and sub, for 2.1 setup.

Currently I've chosen XO points as 80 Hz and 1600 Hz.

From the attached can you advise if you would choose different XO points?

Target is minimum phase and crossover is linear phase as per your advice on another thread.

It's already sounding spectacular and I'm really happy with sound quakity.

I'd be happy to try other XO points if you can suggest. Or do the points I've chosen look fine?

Thanks !

4FFDEEE5-F359-4CD4-9F99-DC64263ABF8D.png




5EB9DED7-6658-40E2-84EB-2F795778C981.png
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
It’s looking pretty good, but check if the 65Hz dip in the left midrange causes issues in the simulation.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
It’s looking pretty good, but check if the 65Hz dip in the left midrange causes issues in the simulation.
Thanks Bernt, looks like right channel shows a dip at 65 Hz in the simulation. Left looks fine. Both attached below.

Also group delay filtered measurement and simulated step response attached below.

Why does my step response peak overshoot the target so much? Is that bad?

And the XO at 1600 Hz no need for concern? Nothing more "ideal" to try from the measurements?

I'm so happy with result but happy to try other suggestions.


F56D6AFA-8845-435A-823A-6DEA5AF1ACC7.png


437A0A0A-B9F0-444F-AB6B-8B0B81BC8C6B.png


489B6CD1-14DB-437D-A759-9B0046047B32.png


3ABF7919-168D-4316-9C53-3F2BF0CB4530.png



4F541F7B-C46D-4430-A93E-C4A887B65509.png
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
It is not the target step response. It\s the subwoofer step response. I believe the pre-ringing is inaudible, and the step response is pretty good, but you should test different correction procedure settings and see if you can achieve a cleaner entry.

Please post the frequence responses of the correction filters.

And give us a zoomed view of the impulse response of the correction filters too.

Left speaker will suffice.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
It is not the target step response. It\s the subwoofer step response. I believe the pre-ringing is inaudible, and the step response is pretty good, but you should test different correction procedure settings and see if you can achieve a cleaner entry.

Please post the frequence responses of the correction filters.

And give us a zoomed view of the impulse response of the correction filters too.

Left speaker will suffice.
Thanks Bernt.

I had selected min phase target curve so is the pre-ringing only for the lin phase XO I selected?

In the simulated step response screenshot above, the subwoofer is not properly time aligned? Seems to be misaligned with the left and right step response.

Requested correction filter FR and IR screenshots below.

Appreciate any tips / advice for things to try.


8457C3AC-89D9-4F24-BFA2-7E7DA89E3CAF.png



Left:


182986F0-2C6D-472A-A5C8-9C4807B289DF.png


2D9880F7-C0E7-497B-8CA6-0A3D10A3BF1B.png



Right:


37B20FD8-1D42-4E03-A052-C062DE45CD76.png




2388080B-4EE3-46E6-88ED-97DCF0A9CF4D.png
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
Not bad at all, but I think it is worthwhile to check if things can be improved.

You seem to have a strong cancellation around 65 Hz. It is causing a little bit of pre-ringing you see in the correction filters for the sub and midbass. Corresponds with the peak in the group delay plots further up in the thread.

There are a couple of things to try out to work around this. If you send me the measurement file, or post it here, I'll check it out.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
Not bad at all, but I think it is worthwhile to check if things can be improved.

You seem to have a strong cancellation around 65 Hz. It is causing a little bit of pre-ringing you see in the correction filters for the sub and midbass. Corresponds with the peak in the group delay plots further up in the thread.

There are a couple of things to try out to work around this. If you send me the measurement file, or post it here, I'll check it out.
Thanks heaps Bernt, I emailed to you today and referred to this thread in this email.

I emailed you latest measurement file, measured today as I had to rotate left speaker a few degrees, as it was a few degrees off axis compared to right channel. I like to sit directly on axis.

But strangely now step response seems out of phase, between left and right channels. I don't understand how this is possible but maybe you can see some clues in the measurement file I emailed. No speaker/amp connections were touched.

This is the output of step response "simulation + target"

34ED0097-1828-46E8-8847-1FBF5F5B03FD.png
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
After examining your measurement I think you're already close to as good as it gets here. You've done a good job here.

I could only find two, rather small, potential improvements.

1) disable TTD per driver
2) increase the low TTD window to 7 cycles.

40987


With the settings I mentioned, the pre-sound drops straight down to -40 dB and drops rapidly from there (looking from right to left), which is pretty good ... and slightly better than with your current settings. I'll leave to you to decide if the difference to your current setting is audible - and if so which sounds the best.

The pre-oscillation in the sub and midbass correction will sometimes be audible, although the simulation is very good (like here), but not necessarily. The only way to get rid of this is probably to losten closer and/or acoustic treatment.
40988



In the taste department, you could try my kind of target. Basically a straight line that falls 4-6-ish dB towards the top. It sounds different from what you are using now, less colored and more rhythmic to my ears. A lot of users lean towards what you're doing, though. So trust your ears and stick to what you like.

40990
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
Thanks heaps Bernt, I emailed to you today and referred to this thread in this email.

I emailed you latest measurement file, measured today as I had to rotate left speaker a few degrees, as it was a few degrees off axis compared to right channel. I like to sit directly on axis.

But strangely now step response seems out of phase, between left and right channels. I don't understand how this is possible but maybe you can see some clues in the measurement file I emailed. No speaker/amp connections were touched.

This is the output of step response "simulation + target"

View attachment 40989
This happens on occations, but not very often. Most likely, the polarity has been corrected wrong by Audiolense.

In the measurement you sent me, the sub and tweeter both had inverted polarity, while the midbass had positive. The safest solution is to ensure that everything is connected with positive polarity and disable the automatic polarity correction
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
This happens on occations, but not very often. Most likely, the polarity has been corrected wrong by Audiolense.

In the measurement you sent me, the sub and tweeter both had inverted polarity, while the midbass had positive. The safest solution is to ensure that everything is connected with positive polarity and disable the automatic polarity correction
Noted, thanks Bernt.

Using same measuremen file I emailed you, I disabled auto polarity correction and generated new step response and new correction filters, without doing a new measurement.

Step response Looks better now. All drivers time aligned it seems
5CCBD0A9-7523-4169-BE4F-79AB47C3998A.png
 
Last edited:

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
And impulse response looks good. But zooming into step response simulation I wonder if improvements can be made


46ADF898-9016-49D4-9A38-FD5AF3C231C5.png




6F9CA690-1D22-45A4-9991-9F0B8738A004.png
 

BenToronto

Member
Joined
May 22, 2017
Messages
52
In post #1, are those the individual drivers with or without the crossover filters?

Lots of considerations in choosing XO points. I'd start by seeing what bands each driver is capable of (FR, distortion, etc), so you know what you have to work with; which is what's wonderful about this approach. But has to be done first without XO or other filters (and not too loud for tweeters).

B.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
After examining your measurement I think you're already close to as good as it gets here. You've done a good job here.

I could only find two, rather small, potential improvements.

1) disable TTD per driver
2) increase the low TTD window to 7 cycles.

View attachment 40987

With the settings I mentioned, the pre-sound drops straight down to -40 dB and drops rapidly from there (looking from right to left), which is pretty good ... and slightly better than with your current settings. I'll leave to you to decide if the difference to your current setting is audible - and if so which sounds the best.

The pre-oscillation in the sub and midbass correction will sometimes be audible, although the simulation is very good (like here), but not necessarily. The only way to get rid of this is probably to losten closer and/or acoustic treatment.
View attachment 40988


In the taste department, you could try my kind of target. Basically a straight line that falls 4-6-ish dB towards the top. It sounds different from what you are using now, less colored and more rhythmic to my ears. A lot of users lean towards what you're doing, though. So trust your ears and stick to what you like.

View attachment 40990
Thanks Bernt, I seemed to not get a notification to this post of yours, so completely missed it.

Thanks heaps for the feedback.

I tried to copy your settings but I don't get the same "log step simulation".

You get -40dB pre-oscillation but mine is slightly higher with same settings (version 6.16 XO)

41141
 
Last edited:

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
The pre-oscillation in the sub and midbass correction will sometimes be audible, although the simulation is very good (like here), but not necessarily. The only way to get rid of this is probably to losten closer and/or acoustic treatment.
Also just to clarify Bernt, is this pre-oscillation because I'm using linear phase XO at 80 Hz? Or caused by something else?

Overall target curve is minimum phase.
 
Last edited:

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
It is caused by a strong cancellation in the room. Strictly to do with acoustics. I believe yours is borderline audible, and this is also highly dependant on what kind of music you listen to. I believe it will be inaudible with all kinds of acoustic performances, for instance.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
It is caused by a strong cancellation in the room. Strictly to do with acoustics. I believe yours is borderline audible, and this is also highly dependant on what kind of music you listen to. I believe it will be inaudible with all kinds of acoustic performances, for instance.
Thanks Bernt. I have to ability to demo (for free) some big absorbers.

From the measurements, what frequencies need to be absorbed? Which graph shows this

So I can demo the correct room treatments
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
If you zoom in on the entry to the step response you should be able to estimate the duration of one cycle pretty well. From there, you can calculate the frequency as 1000/ cycle time. And then you can probably find one or two corresponding depressions in the measured frequency responses. And that's it.

Another approach is to look at the measuremd group delay and frequency response. Look for group delay peaks that corresponds with magnitude depressions.

Sometimes the problems can be hidden behind later reflections that fills in the depressions, but usually not.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
If you zoom in on the entry to the step response you should be able to estimate the duration of one cycle pretty well. From there, you can calculate the frequency as 1000/ cycle time. And then you can probably find one or two corresponding depressions in the measured frequency responses. And that's it.

Another approach is to look at the measuremd group delay and frequency response. Look for group delay peaks that corresponds with magnitude depressions.

Sometimes the problems can be hidden behind later reflections that fills in the depressions, but usually not.
Thanks Bernt, I'll do some investigating.

I am super super happy with SQ already, the improvement Audiolense XO has made is 'night and day'.

So this is more just about having some fun to see if any can be audibly improved, nothing too serious.

But in Post #13 above I don't seem to have the same log step response that you have, even with same settings as you can see in my screenshot.

Is it because you have a different target curve to mine?

Hence simulated will be different?
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
If you zoom in on the entry to the step response you should be able to estimate the duration of one cycle pretty well. From there, you can calculate the frequency as 1000/ cycle time. And then you can probably find one or two corresponding depressions in the measured frequency responses. And that's it.
Zoomed in pics attached below.

I've chosen one cycle from 787.25 ms to 787.292 ms = 0.042 ms

Frequency = 1000 / (787.292 ms - 787.25 ms) = 1000 / 0.042 ms = approx. 23 kHz

Is this the right approach?

23kHz seems high? I thought it's more a bass issue?

1 of 4.png



2.png


3.png


4.png
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
I looked at the large cycles (dips) before the step response and calculated

One cycle from 767.2 ms to 778.9 ms = 11.7 ms

frequency = 1000 / 11.7ms = 85 Hz.

Is it these larger/slower cycles (deep bass) that you were referring to as being the main problem?

I don't see corresponding frequency response anomolies at around 85Hz.




41178
 
Last edited:

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
This is the log step view. This is also relevant. The step is 38 dB down before it rises, which is pretty good.

If you look at the regular step response (step response of simulation plus target) it is easier to asess the most problematic frequencies.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
If you look at the regular step response (step response of simulation plus target) it is easier to asess the most problematic frequencies.
Hi Bernt, which area should I zoom to see the most problematic frequencies, to do the f = 1000/cycles estimate ?

The red circled or orange circled area?

Thanks again

41330
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
Yes a bit of both. Low frequency oscillation (the ellipse area) is where audible problems tend to occur most often. This is looking pretty good, though.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
Yes a bit of both. Low frequency oscillation (the ellipse area) is where audible problems tend to occur most often. This is looking pretty good, though.
Thanks Bernt. For the bass (ellipse area) problematic frequency works out to be 60-80 Hz

There's nothing in the frequency response in that area that is 'strange' apart from the fact this is crossover region (linear phase XO) with the subwoofer.

I guess I could try some huge bass traps to see if that changes anything after re-measuring. I can ask my dealer if I can borrow some for free, for home trial and re-measure.

One final question about my step response simulation though.

Regarding the below: the pink circled area is obviously the subwoofer.

Is this hump an acoustic 'post ringing' / 'standing wave' issue?

Or is this something I can try to improve in Audiolense? Or not important?

The measurement zip file is attached



41332
 

Attachments

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
That "hump" is a function of the low frequency roundoff. Some version of it will always be present. A more "low q" target will probably take it down somewhat.

This is imo a very fine result. I would have easily lived with this kind of performance.
 
Top Bottom