DiyMobileAudio.com Car Stereo Forum banner

421 - 440 of 837 Posts

·
Registered
Joined
·
1,999 Posts
Discussion Starter #421
Oh I HaTE it when that happens ugh!!
Or I save something as horn and over rite it as mid .

So I'm positive Dirac filters clip the signal now if the magnitude isn't within about 6db. A pretty good pre tune is a absolute necessity. I think I've solved my clipping issues now for good. I was thinking as long as I'm setting a target under its 0db line it won't use boost well that's 0dbfs not 0db. It's normalized and even says it on the screen so if I draw my target anywhere above any of the "all before" it in fact does add boost. It brings everything up to 0dbfs which is the absolute max in the digital world. Can't boost past 0dbfs or massive clipping starts. So the only way to get a really really good tune would be to have as flat as possible magnitude responce so that no part of the "all before" is too low.
So that's a test signal and I don't know it's crest I'm assuming it's about 3 to 6db as is most signal generators. But I don't know so I'm going to have to hook up a scope and test. If music has a transient louder than the test signal guess what, yep the closer you get to 0dbfs the more shit will clip. So any erratic changes in magnitude could cause a boost if set wrong and if set right won't leave enough gain to get enough useable output meaning system will loose a lot of signal to cutting. Inverse filters are a bitch. So tomorrow I'm going to spend some quality time with good old fashioned RTA and get shit flat than tune. I've just haven't (still) because Dirac does such a good job at it. But it really is important now. Use eq cuts on output and save precious digital headroom

I got a much better tune but tomorrow I'm going to make the end all (for this week ) tune.
Ok this is a bit over my head, even if it should be the base...
I sometime had some clippings depending of the tune, but I only noticed it on highs. But it can also be my tracks, in fact more often.
Or when I noticed some heavy boosts in Dirac (on mids) I usually lowered the ddrc knob (or default preset) to -3 or 6db. Seemed to do the trick.
Is this a good fix?
But I try to stay safe, either in the Cdsp (input -Xdb, X being the max PEQ I used), and then low target in Dirac.
Of course the final output is much lower, but I never missed juice for now.
But I could imagine that it’s not optimum.

I should try to keep the laptop synced during a drive, to see what output level I usually listen to.
Btw I have a dbC meter that I check sometime when loud, to not go to crazy
But I use it more as relative levels as I’m not sure it’s accurate.


Then I managed to get back my good tune, or at least find out which one it was.
By disconnecting each speaker and playing sweeps, got the XO and levels of the tune.
But the strange thing is, it should not be the one that correspond to the Dirac preset (preset 1 on C-dsp for preset 1 in Dirac, 2 to 2 etc).
I mean, it should not be the best combo but still it’s the one I like :p

I’ll make another Dirac pass once it’s quiet outside...
 

·
Registered
Joined
·
3,278 Posts
When i was playing with Dirac, i found that i never had a clipped signal when doing the chair measurements unless my sub channel was high. It seems to me that it just cant handle a decent spl substage whatsoever.

Im interested in this flat Freq response your talking about. Do you mean eq to a flat response(even 9-12db+ substage?) and then let Dirac do its work?

Whenever you guys figure this out. Id love if someone through together a Hanatsu style tutorial. Reading through this and APL threads get extremely overwhelming.
 

·
Registered
Joined
·
4,304 Posts
Okay I'll try explain better sorry :( :)


So you know I'm make filters tab after you get measurements the blue screen
So you make let's say a measurement and the sub and midbass is above the "0db" line.
And the horns are somewhere above it and some spots below it (because the horns have pretty drastic curves) well , let's say the target by default sets sub about 3db above that line and 16k around -2db part.

So if you just optimize where the default target is it's going to put boosts and cuts to make that target....so the measurement in the "all before" when compared to the reference in "0db" could be high above that line and have parts below it. Well depending on how loud the measurements were taken will show where the "all before" is placed on the graph. So In reference to the 0db line the measurement is just normalized and doesn't reflect any "0db" centering on the graph. They just put the measurement on the screen so you can see it , after all there is no scroll down or scroll up so they have to just stick it in the window with where you want your target.

So when making a target , the center of the measurement will depend on how much boost and cut Dirac can accomplish. So if the measurement has more than about 6db dip in it from its center. To find center you would have to take the tallest peak/null and get the range and divide it by 2 to get the center of where Dirac will make its correction.

Remember Diracs behaviors , Diracs IIrs are not like PEQs on a dsp, dirac uses one single filter to do the whole correction, it would be similar to a single biquad that programs banks of biquads, but it's not limited to banks of biquads. It's a IIR inverse filter literally a mirror image of the averaged frequency response. So it will use boost and cut. Always! You can't make the target line in the blue screen under all the measurements and think it will only use cuts that's not how inverse filters work.

So after you take measurements the "all before " doesn't mean anything compared to what your viewing. So Dirac 1st makes its inverse frequency response filter. Than it uses the reference point and makes a inverse phase filter but before it does that it takes into account what the 1st inverse filters effect on minimum phase so that it doesn't over correct phase.

It determines minimum phase by the reflections in the IR. By avoiding those and feeding back the projected reflection sound with an inverse filter it should effectively change the speakers behavior to make the reflections part of the acoustical stream. So it does some funky shit.... and up there I said "should" because that part wasn't fully explained in the video I saw with the maker of Dirac in it...

So when dealing with inverse filters you gotta know your digital headroom. So 1 you get a stronger signal out of Dirac if the magnitude is already pretty flat. If it's not flat like elgroso said you have to turn down the volume out of Dirac. Well that sucks because than the amp gain has to go up or boost on outs in dsp. Which is why I think the newer minidsps have +12db boost on outputs now. So, if your measurement are more than a 6db swing you may crash into 0dbfs and that's the absolute max in digital. So I figure the test sweep has at least 6db crest noise so a 6db swing (+/-6db or 12db boost and cut together) should be safe I think Dirac come in default with -10db lowered out.

So yes my measurements show a huge cut at 3k (which I did on purpose) and it's about 12db deep on Dirac measurements so it will boost that 12db in its filter and I've been playing Dirac a 0db thinking as long as my target is under 0db it will only cut which is false

:)

And yeah I hear distortion in the high mids too. And that's where I have lots peq on horns. So I gotta rethink things a bit and dial up a really good flat tune 1st. Or have very good level setting I should say and maybe 3 pre. Eq cuts only on big peaks to flatten things first.
 

·
Registered
Joined
·
1,999 Posts
Discussion Starter #427
When i was playing with Dirac, i found that i never had a clipped signal when doing the chair measurements unless my sub channel was high. It seems to me that it just cant handle a decent spl substage whatsoever.

Im interested in this flat Freq response your talking about. Do you mean eq to a flat response(even 9-12db+ substage?) and then let Dirac do its work?

Whenever you guys figure this out. Id love if someone through together a Hanatsu style tutorial. Reading through this and APL threads get extremely overwhelming.
What kind of amplitude did you have in total for these measurements?
I know my target is not crazy but I’ve seen 25db and it worked fine.
Only time I had Dirac mic clipping was this WE with a wrong high Q peq on horn.
 

·
Registered
Joined
·
1,999 Posts
Discussion Starter #428
Okay I'll try explain better sorry :( :)


So you know I'm make filters tab after you get measurements the blue screen
So you make let's say a measurement and the sub and midbass is above the "0db" line.
And the horns are somewhere above it and some spots below it (because the horns have pretty drastic curves) well , let's say the target by default sets sub about 3db above that line and 16k around -2db part.

So if you just optimize where the default target is it's going to put boosts and cuts to make that target....so the measurement in the "all before" when compared to the reference in "0db" could be high above that line and have parts below it. Well depending on how loud the measurements were taken will show where the "all before" is placed on the graph. So In reference to the 0db line the measurement is just normalized and doesn't reflect any "0db" centering on the graph. They just put the measurement on the screen so you can see it , after all there is no scroll down or scroll up so they have to just stick it in the window with where you want your target.

So when making a target , the center of the measurement will depend on how much boost and cut Dirac can accomplish. So if the measurement has more than about 6db dip in it from its center. To find center you would have to take the tallest peak/null and get the range and divide it by 2 to get the center of where Dirac will make its correction.

Remember Diracs behaviors , Diracs IIrs are not like PEQs on a dsp, dirac uses one single filter to do the whole correction, it would be similar to a single biquad that programs banks of biquads, but it's not limited to banks of biquads. It's a IIR inverse filter literally a mirror image of the averaged frequency response. So it will use boost and cut. Always! You can't make the target line in the blue screen under all the measurements and think it will only use cuts that's not how inverse filters work.

So after you take measurements the "all before " doesn't mean anything compared to what your viewing. So Dirac 1st makes its inverse frequency response filter. Than it uses the reference point and makes a inverse phase filter but before it does that it takes into account what the 1st inverse filters effect on minimum phase so that it doesn't over correct phase.

It determines minimum phase by the reflections in the IR. By avoiding those and feeding back the projected reflection sound with an inverse filter it should effectively change the speakers behavior to make the reflections part of the acoustical stream. So it does some funky shit.... and up there I said "should" because that part wasn't fully explained in the video I saw with the maker of Dirac in it...

So when dealing with inverse filters you gotta know your digital headroom. So 1 you get a stronger signal out of Dirac if the magnitude is already pretty flat. If it's not flat like elgroso said you have to turn down the volume out of Dirac. Well that sucks because than the amp gain has to go up or boost on outs in dsp. Which is why I think the newer minidsps have +12db boost on outputs now. So, if your measurement are more than a 6db swing you may crash into 0dbfs and that's the absolute max in digital. So I figure the test sweep has at least 6db crest noise so a 6db swing (+/-6db or 12db boost and cut together) should be safe I think Dirac come in default with -10db lowered out.

So yes my measurements show a huge cut at 3k (which I did on purpose) and it's about 12db deep on Dirac measurements so it will boost that 12db in its filter and I've been playing Dirac a 0db thinking as long as my target is under 0db it will only cut which is false

:)

And yeah I hear distortion in the high mids too. And that's where I have lots peq on horns. So I gotta rethink things a bit and dial up a really good flat tune 1st. Or have very good level setting I should say and maybe 3 pre. Eq cuts only on big peaks to flatten things first.


I think I get what you're saying.
So you mean that whatever shape and wherever we put the target, with the orange dots, dirac will kind of normalize it and boost/cut as if you had drawn the same target, but placed right in the middle of the total amplitude?
Then the final output level, the knob or the drop down menu in ddrc plugin, will adjust the final max level (0dbfs)
A bit like with APL and "target level".
Makes sense.

But,
I think I saw few times dirac having troubles to match perfectly some higher targets.
I mean, if I draw it lower, or mostly lower than the "all before", I usually get a perfect prediction match.
(Sometime with some squiggles down low, LF & taps limitations I guess. But by playing with the total fr selection window, the handles, deleting some points or smoothing here and there I can get back some taps to fine-tune)
But if I draw the same target, and placed higher, I can clearly see it struggling.
I would then get few holes in few spots, like it's getting out of juice to match the target.
I'll put some screenshots to illustrate.
 

·
Registered
Joined
·
1,999 Posts
Discussion Starter #429
Here, see how lower the better? Of course it eats some juice.
It’s not my target just the default one moved up and down.







It’s today’s tune, 100-1700Hz.
No PEQS on mids, because around the XO they looked pretty nice by themselves, so I wanted to let Dirac fix the bandpass.
Then just few PEQS on sub and horns to get a smooth slope.
But it’s a miss, note how the mids play higher than expected, at least than what I saw in REW.
Also note the left midbass, weak around 100-150Hz.
I get that when I use a large «*cube*» of measurements.

Here’s mine:
 

·
Registered
Joined
·
3,278 Posts
What kind of amplitude did you have in total for these measurements?
I know my target is not crazy but I’ve seen 25db and it worked fine.
Only time I had Dirac mic clipping was this WE with a wrong high Q peq on horn.

This was my original post and try with it awhile back

Ok so i did some more testing. Keep in mind that the FR shown in the below pics are using different mic positions, meaning that for REW i use Hanatsus and for DIRAC i use the positions on the DIRAC software.

This is my base tune using using the Hanatsu method and The C-dsp for everything. This is without Dirac. VVVVV


Here is my Dirac measurement with time alignment done beforehand. The same T/A settings as in the first pic but different mic positions. VVVVV


As you can see. i get drastically different FR at certain points.

Now here is my Dirac measurement with all T/A settings on my dsp set to 0. VVVV



I used the same curve on both and then uploaded it to the Dirac software part 2 which just lets you control gain, T/a, and up to 4presets . So i uploaded with and without T/a as preset 1 and preset 2.

In order to listen with an A/B method. I had my settings(that gave me the curve on the first picture) on my C-dsp saved to preset one and the same exact settings on preset 2 but without any T/A applied, then i had t go into the DIRAC program and swap accordingly as well. So Preset 1 on c-dsp with preset 1 on Dirac, preset 2 on c-dsp with preset 2 on Dirac. . Sounds more complicated than it is. Initially both preset 1(with ta) and preset 2(without Ta) were too far left but with a little fiddling in the Dirac program i had perfect center stage within 2ms of change and 2db of change.

This means that this thing does time alignment for you. Dope. There was definitely a different sound to a song even with the same exact curve when comparing A/B to the curve done with and without Initial time alignment. I enjoyed no inital T/A more and i attribute that to less boosting done to fill in weird gaps.

As for negatives. i heard a lot of distortion and there is a clipping meter on the Dirac software that was going off plenty. Im not sure the reason behind this. The computer was playing music to the headunit through an Aux cord so im hoping it was either the cheap computer feeding distortion into the system because of just a simple built sound card or something along those lines. I dont know what the Dirac clipping indicator actually indicates. At what point is what clipping in the software.... Idk.

Also in order to really get a curve, you have to cut which to me is taking away even more headroom and making my loudest possible volume that much lower. Im curious if it would be better to tune most of your system to a flat curve without any PEQ or GEQ involved at all. What i mean is, level match, do your best as crossover matching. Dont do any time alignment, dont do any PEQ or GEQ and let DIRAC do all of the FR shaping after that.

anyway, this thing is dope. im definitely getting a DDRC-24 in the future.
 

·
Registered
Joined
·
1,999 Posts
Discussion Starter #431
As you can see. i get drastically different FR at certain points...
Yeah I remember your post, with or without T/A.
But so in the laptop version you have T/A per channel?
Not on mine, it's only 2ch.
Anyway, I think to compare both FR you should use the same sweep points.
But they don't look that different right now. Seems just accentuated in Dirac. Probably the SPFR that accentuate the lower output around XO.

For the mic clipping yeah 43db is a lot :D
It’s also strange that it doesn't distribute your measurement equally around 0db. Up limit is 10db and goes to -30db here.
(well I just looked at mine and it’s the same)

For the clipping output, do you have a master volume control in this soft version? In the plugin there's a -3/-6/-10db option.

About pre-tuning flat VS target, I’m more for target.
Just because you can keep control of output/final gain to not lose too much headroom. And I use always the same kind of target so I know my system like that now. Can also listen to it without Dirac when I want to change something else.

But I agree on the noPEQ pre-tuning. But only if the XOs are clean enough, using some PEQS here to optimize the slope will really help.
In the bandpass Dirac is better than me.
 

·
Registered
Joined
·
4,304 Posts
Yeah, the lower the better indeed. That would leave more digital headroom
There's a theoretical 96db of headroom in digital audio. I think we use about 20db and the music we listen to rides so close to 0db it's like 5db and a good cd is like 12db headroom or something I can't remember exactly what it is but I read it somewhere reputable...

So if our music is so close to full blast that doesn't mean we have a lot of room above the 0db line in our target , unless there was cuts made in the filter than what we see would be elusive...

So yeah your getting it I think. If you look at your screen it shows the measurements as normalized , when we make the measurements we can put out output at whatever I set my test sweeps to -35db and 0db would be 0dbfs , obviously that would probably blow all the speakers in the car if we did that. However......that's the test sweep. As soon as I make my filters I been jacking up the outpolut to 0db. And on my deck toslink out I turn it up all the way when I listen. So there for leaving whatever digital headroom in the music as means of dynamic headroom and it's safe because it's media and wasn't recorded with clipping. However if my filter has boost in it than I eat up whatever left over headroom I have. So what we do it lower out filter target below the 0db line and all seems ok right, yes. And it is, however we need to first look at the "all before" and take the biggest peak and biggest null and count it out in db on the graph divide it by two and that should be how low we set the output gain to.....which means to truly get no clipping the "all before" really needs to be within just a few db of flat otherwise we have to gain it down so far that we have hardly any output. We would have to gain our amps up a lot or boost the output in dsp to compensate.

What I'm saying is the measurement is normalized, if you made that measurement at -35db. Why is it show way above the 0db line on the blue screen graph? Because it's fake and it's showing up the curve and has nothing to do with target....the target is the sum of the measurement and inverse filter than averaged over all 27 measurements or whatever. The inverse would make a perfectly flat line electricity but it's an average so that's why we see it as a flat flat but squiggly yellow line....

So a good pre tune to be made as flat as possible and use up as much bits in the output peq. Would be the way to go first. That way the input is as strong as possible and will lead to a lot higher noise floor and a lot more gain....

It's okay how Dirac dose it for idiot proofing but gosh dam I would so much prefer making a filter and tweaking it on my own...

My problem is I keep changing shit and need to stop and actually set my gains and get it all right. So I was going to do that today, I forgot about Valentine's Day so shit the wife wanted to do something and my B&C 10s came today so crap. I'll have to do it tomorrow or next day :)
 

·
Registered
Joined
·
4,304 Posts
Yeah I remember your post, with or without T/A.
But so in the laptop version you have T/A per channel?
Not on mine, it's only 2ch.
Anyway, I think to compare both FR you should use the same sweep points.
But they don't look that different right now. Seems just accentuated in Dirac. Probably the SPFR that accentuate the lower output around XO.

For the mic clipping yeah 43db is a lot :D
It’s also strange that it doesn't distribute your measurement equally around 0db. Up limit is 10db and goes to -30db here.
(well I just looked at mine and it’s the same)

For the clipping output, do you have a master volume control in this soft version? In the plugin there's a -3/-6/-10db option.

About pre-tuning flat VS target, I’m more for target.
Just because you can keep control of output/final gain to not lose too much headroom. And I use always the same kind of target so I know my system like that now. Can also listen to it without Dirac when I want to change something else.

But I agree on the noPEQ pre-tuning. But only if the XOs are clean enough, using some PEQS here to optimize the slope will really help.
In the bandpass Dirac is better than me.

Yep I felt that way before I notice the clipping issue...

So maybe your doing something different or have REALLY quiet amps and can gain them up to the 500mV setting. I can't I get noise if I gain my amps up. But I use class AB which is so much more sensitive to gain noise than a D. So maybe you got me beat on this...

If I could just gain my amps up with no noise I would just do that.... I can get to about half way or 2v setting and it starts to show hiss :( the three Class D I have have zero noise. It's a shame maybe I'll get Another 600/4 and pull the focals for good.
 

·
Registered
Joined
·
4,304 Posts
When I say use peq first I am strictly talking about a peq that affects both left and right together. A averaged of the entire system on with all speakers playing no eq no Dirac but yes let Dirac do the L/R filters on it's own. It would be too much if not.

And find the big two or three peaks and cut them down. That's is just try to make Dirac correction not use up so much dynamic range.

I'm convinced this will work. So I will try tomorrow. If I have time .
And let you know..... you don't hear any clipping in the horns on yours after you make a target filter?
 

·
Registered
Joined
·
4,304 Posts
And I don't think this would be a issue for non horn users . I hear zero clipping in my mids
Just horns ....and subsequently that's the only driver that has up to 18db differences in peaks and nulls ....


I will get answers sorry but it's buggin me ...

I'm write a letter to Drac himself .
 

·
Registered
Joined
·
1,999 Posts
Discussion Starter #436 (Edited)
Yeah, the lower the better indeed. That would leave more digital headroom
There's a theoretical 96db of headroom in digital audio. I think we use about 20db and the music we listen to rides so close to 0db it's like 5db and a good cd is like 12db headroom or something I can't remember exactly what it is but I read it somewhere reputable...

So if our music is so close to full blast that doesn't mean we have a lot of room above the 0db line in our target , unless there was cuts made in the filter than what we see would be elusive...

So yeah your getting it I think. If you look at your screen it shows the measurements as normalized , when we make the measurements we can put out output at whatever I set my test sweeps to -35db and 0db would be 0dbfs , obviously that would probably blow all the speakers in the Car if we did that. However......that's the test sweep. As soon as I make my filters I been jacking up the outpolut to 0db. And on my deck toslink out I turn it up all the way when I listen. So there for leaving whatever digital headroom in the music as means of dynamic headroom and it's safe because it's media and wasn't recorded with clipping. However if my filter has boost in it than I eat up whatever left over headroom I have. So what we do it lower out filter target below the 0db line and all seems ok right, yes. And it is, however we need to first look at the "all before" and take the biggest peak and biggest null and count it out in db on the graph divide it by two and that should be how low we set the output gain to.....which means to truly get no clipping the "all before" really needs to be within just a few db of flat otherwise we have to gain it down so far that we have hardly any output. We would have to gain our amps up a lot or boost the output in dsp to compensate.

What I'm saying is the measurement is normalized, if you made that measurement at -35db. Why is it show way above the 0db line on the blue screen graph? Because it's fake and it's showing up the curve and has nothing to do with target....the target is the sum of the measurement and inverse filter than averaged over all 27 measurements or whatever. The inverse would make a perfectly flat line electricity but it's an average so that's why we see it as a flat flat but squiggly yellow line....

So a good pre tune to be made as flat as possible and use up as much bits in the output peq. Would be the way to go first. That way the input is as strong as possible and will lead to a lot higher noise floor and a lot more gain....

It's okay how Dirac dose it for idiot proofing but gosh dam I would so much prefer making a filter and tweaking it on my own...

My problem is I keep changing shit and need to stop and actually set my gains and get it all right. So I was going to do that today, I forgot about Valentine's Day so shit the wife wanted to do something and my B&C 10s came today so crap. I'll have to do it tomorrow or next day
Hum… ok
It really sound like one of our old discussion, what I understand is that we have 3 options:
A - tune to flat first
B - tune to target first
C - tune to target but non optimized (only around XO, not bandpass)

A - means higher gains on horns, or lower gains on sub/mids, or big dsp output discrepancies
> loss of power and potential clipping (well depending of the target shape)

B - means good levels, good gain chain, less or no clipping
but potentially more errors created by the multiple PEQS per driver for Dirac to fix

C -well a mix of both, but due to the high amplitude variation on horns, could drive to clipping

Makes sense?



Yep I felt that way before I notice the clipping issue...

So maybe your doing something different or have REALLY quiet amps and can gain them up to the 500mV setting. I can't I get noise if I gain my amps up. But I use class AB which is so much more sensitive to gain noise than a D. So maybe you got me beat on this...

If I could just gain my amps up with no noise I would just do that.... I can get to about half way or 2v setting and it starts to show hiss the three Class D I have have zero noise. It's a shame maybe I'll get Another 600/4 and pull the focals for good.
I have the same thing with my horns amp. It's a prs-a900, AB too, and it's the one driving the final global gain in a way.
I can't reach 500mV, have to lower it till I have acceptable noise, but if I set it to no noise, then I miss too much final power. No idea of the final Voltage selection.
(btw 500mV means average and 2V max?)



The HDs are great, absolutely no noise, I could try the horns on them to check.
And if good I could get another HD, all stacked in the rear wing they would fit, stealth! :p
It’s really cool that, bridged, they push the same power for 4 or 8 ohms.
I wonder which one I should use if I want to try 16 ohms drivers (1x1200 per mibass? :D)




When I say use peq first I am strictly talking about a peq that affects both left and right together. A averaged of the entire system on with all speakers playing no eq no Dirac but yes let Dirac do the L/R filters on it's own. It would be too much if not.
And find the big two or three peaks and cut them down. That's is just try to make Dirac correction not use up so much dynamic range.
Ok I see, so you use PEQS, pre-Dirac, for target correction and then a minimal PEQ for driver correction (on horns usually).
I do the same, + sometime the sub. Because the sub can drive the total amplitude of the "all before" up. As I cut it 10/15db around 40Hz.



And let you know..... you don't hear any clipping in the horns on yours after you make a target filter?
It's not that clear in fact. Most of the time if I have clipping it's track dependent.
I mean maybe I'm already at 0dbfs but don't notice. Not loud enough to reach it except with very bad recordings?
Or it's just the bad recording that is already clipped (still have some very old mp3).

But yes when I had clipping, lowering volume either on the phone (source) or the ddrc fixed it. Not at the final dsp (6x8).
Lowering the target for a new filter works of course.


And I don't think this would be a issue for non horn users . I hear zero clipping in my mids
Just horns ....and subsequently that's the only driver that has up to 18db differences in peaks and nulls ....
Ok so for both of us. Could it be that clipping on mids is harder to notice?
About the big difference between peaks & nulls, that would push for the good pre-Dirac EQ too.
 

·
Registered
Joined
·
4,304 Posts
I'll answer that post ^ next

But first ide like to say I got an answer...

So apparently Dirac is only capable of 10db correction. Period . If there is more than 10db it won't go past it and push all of the correction to -10db and stop. So whatever is left would be uncorrected and would be additional signal that would have to be attenuated even farther than the -10db so literally eating up all the dynamic headroom.... so that solves it.

Our measurements need to be within 10db of flat and all should be okay. :)
 

·
Registered
Joined
·
4,304 Posts
Hum… ok
It really sound like one of our old discussion, what I understand is that we have 3 options:
A - tune to flat first
B - tune to target first
C - tune to target but non optimized (only around XO, not bandpass)

A - means higher gains on horns, or lower gains on sub/mids, or big dsp output discrepancies
> loss of power and potential clipping (well depending of the target shape)

B - means good levels, good gain chain, less or no clipping
but potentially more errors created by the multiple PEQS per driver for Dirac to fix

C -well a mix of both, but due to the high amplitude variation on horns, could drive to clipping

Makes sense?





I have the same thing with my horns amp. It's a prs-a900, AB too, and it's the one driving the final global gain in a way.
I can't reach 500mV, have to lower it till I have acceptable noise, but if I set it to no noise, then I miss too much final power. No idea of the final Voltage selection.
(btw 500mV means average and 2V max?)



The HDs are great, absolutely no noise, I could try the horns on them to check.
And if good I could get another HD, all stacked in the rear wing they would fit, stealth! :p
It’s really cool that, bridged, they push the same power for 4 or 8 ohms.
I wonder which one I should use if I want to try 16 ohms drivers (1x1200 per mibass? :D)






Ok I see, so you use PEQS, pre-Dirac, for target correction and then a minimal PEQ for driver correction (on horns usually).
I do the same, + sometime the sub. Because the sub can drive the total amplitude of the "all before" up. As I cut it 10/15db around 40Hz.





It's not that clear in fact. Most of the time if I have clipping it's track dependent.
I mean maybe I'm already at 0dbfs but don't notice. Not loud enough to reach it except with very bad recordings?
Or it's just the bad recording that is already clipped (still have some very old mp3).

But yes when I had clipping, lowering volume either on the phone (source) or the ddrc fixed it. Not at the final dsp (6x8).
Lowering the target for a new filter works of course.




Ok so for both of us. Could it be that clipping on mids is harder to notice?
About the big difference between peaks & nulls, that would push for the good pre-Dirac EQ too.
Yep you got it buddy. We on same page now.

I truly think it's a issue only us (or horn users) will experience. Direct radiators frequency response is much more manageable . So it forces us to have some sort of pre-equalization before we run the Durack SoftwRe .

Man I'm not kidding I was getting some pretty bad clipping and I had everything all the way down it wasn't making any sense but now it makes perfect sense we were both doing it the right way the first time and I was thinking too hard into it all it is is there's no more than 10 DB correction and we're good .

If we had direct radiators or we would have to do is sit or gains so that the frequency response is flat between drivers because direct the radiators don't have such a large variances

This makes sence why joeysycon2 (sp?) is not experiencing these problems although he is running the trial version but it shouldn't matter should still work the same and less the computer sound card is Jank , but overall that's what it comes down to.

So 10 DB is our number, as long as our gains and pre eq has a magnitude with no more than 10 DB variance and We set out output level in Dirac to -10db it really won't matter where we put the target in Dirac. Which would also make sense that if you had let say one or two or three DB variance total you would be able to boost the target above 0db about 6,7,8 db. Or whatever the variance is.

I was confusing the over zero DB Line as 0dbfs. Albeit that 0dbfs is still part of it but we would have to set the target like 15db above the 0db line before we worry too much about that. So

Overall. If your measurement has more than 5db variance it's a good idea to have the target no more than about 3-5db above the 0db line and if it has more than 10 the target line HAS to be below the 0db line. Doesn't matter where could be 1db below 0db just needs to be below 0db. And than we set the output to -10db and wallah clip free music alas
 

·
Registered
Joined
·
4,304 Posts
This also makes sence why any bending of the target makes a undesired sound and could make left or right stand out more. Because there's a separate correction for left and right . If one channel has 13db variance in one frequency band while the other channel doesn't have over 10 DB variance in that frequency band but may have a variance that's greater than 10 DB in another frequency band could cause any bending of the target so not sound right or have one side sound louder in one frequency.....

But being that it is an average I would still do all my pre-equalization on both channels at the same time however I would look up left and right channel separately with an RTA and just look at the behavior of the frequency response and make a decision based on what you see in the RTA weather will be good to do a Pre tune on separate left and right channels but I will personally think it would be better to let dirac that single correction. Otherwise he cues are probably counteracting each other and that just seems to be not the good way to go .

So yes I would pre-tune a 100average or more Pink PN large fft RTA average to both left and right together. And just focus on the big peaks and knock them down with wide Q. = .7-to 1.7... and get less than 10db variance.

For that matter what I would do is set for my PEqs flat use some output eq to tame stop band interference, but I wouldn't go too crazy with it , ya know knock down the 5k interference hump frommidbasses that are crossed at 800-1k . And some minor stop band spikes but I wouldn't focus too much on it just get the big stuff that will be audible , especially if you using Iir crossovers , there's no pre-ringing to worry about from stop-band interference on a crossover, Dirac will sum the interacting drivers and correct whatever combined wave that happens to hit the measurement microphone for phase and FR
 

·
Registered
Joined
·
1,999 Posts
Discussion Starter #440
Yeah nice find man. I should have seen it, I thought the squiggles were related to taps limit but in fact it's super obvious on the third pic I posted, everything that passes the +10db limit follows the exact same curve than "all before".
Cool so we were a bit overthinking all this and the first approach was the right one.
So just stay within 10db of the measurements, less if possible, check max predicted boost and adjust either target level or output attenuation.
Much simpler :)

Ok next step, focus on pre-EQ with good old fashioned tools!
 
421 - 440 of 837 Posts
Top