DiyMobileAudio.com Car Stereo Forum banner

221 - 240 of 283 Posts

·
Registered
Joined
·
4,184 Posts
So one important thing to say about how Dirac handles things as a new old discovery that I think I can speak to a little better.

So in a 2ch Dirac calibration we all know it will fix the impulse to better match the Dirac impulse.

I’ve been measuring some strange delays in my impulses that I know were not there with driver measurements. (I miss the old forum look this is already hard)

Basically the impulse has been stretched or smeared and time has been added to my midbass speakers the most and the phase is wrapping as if there’s delay in the measurement. A group delay that I couldn’t quite figure out till today I finally see what’s going on.

So on a impulse responce if the phase is wrapping out of the 180deg boundary’s and heading upwards of over 1000deg in some frequencies has had me wondering what has been the exact reason.

SO a two channel Dirac tune will add delay between left and right for all speakers on that channel. What if you had a stereo pair of tweetes in the center channel location right next to each other.

What would happen is it would only add like .03ms to the left channel and convolve the impulse and stretch the impulse to add delay as a constant group delay or an all pass and shape it just just right to add delay to the left and right so that the impulses matched for the 1st mic position. That way the left channel gets its 1.xxms delay but without raw delay. In fact , it changes the entirety of left and right so the shape matches. That’s the wierd part. It’s an equivalent to raw delay on the left side but it dosent behave as such.
(I did that exact experiment have two tweets in center and midbass on a 2ch Dirac) and this explains why after Dirac if you add delay to the left channel the stage moves right! The actual placement of phase angle could be whatever. Probably what takes the least processing and nets the most symmetry as far as an impulse.

So what does this mean. It means if it will stretch the impulse and add a group delay where needed or take time out of a measurement from unequal path lengths it can cause some unusual crossover alignment artifacts between left and right only because of unequal path lengths.

If the impulses match so what. It’s not that important. But for stage shaping it can be a bit of a problem if one side crossover region now has a completely different summing scheme. Not saying this is a problem for most speakers but for a mid to midrange or a horn to mid it could make the stage sound like one speaker is playing more than the other compared to the boundaries of the room.

It’s almost worth either using extremely high order crossovers so the spot where the stretching starts is right at the speaker that needs it and doesn’t carry over to the adjacent driver. Or use low order filters so the phase between the drivers phase is much more “fuzzy” and diffuse.

On the contrary because of this effect if you will and in a car cabin to maintain the stage symmetry it seems (at least to me) in the modal range (200-1k) there should be any time delay used on those drivers as the crossovers will reach up or down into the harmonics and if stretched out you would want them to have the same timing coming out.

What I mean by that is , for example the left mid and midbass if they have a pathway difference to the first measurement point to not try to do any “pre delays” so that the two speakers originate at the same time.
The will ensure that the low frequency harmonics stay in phase between the same side components. Again a tweeter crossed up above 2.5k where the wavelength is less than one cycle of the path length differences between the components it won’t make any difference. So a pre delay on a tweeter to the mid to the mic location will actually lessen the amount of all pass impulse stretching. Where a mid to midbass or to sub you do the opposite and don’t worry about doing any pre delays at all. It’s going to convolve it no matter what, might as well keep the LFE intact. It very much could make it a lot better......

In my experiments this has proved to keep better stage symmetry and keep low frequencie harmonics in phase across components.

Dirac will solve any timing differences in its all pass or convolution. So it’s better to know what it’s doing for your situation to help keep stage symmetry.

So awhile back I said something to the effect that a 12db crossover on mids and highs nets a better more open sounding ambient deep stage. I finally know why this seems to be a fact in my situation. It’s definitely worth exploring. Your situation could benefit (maybe not) but definitely worth exploring
 

·
Registered
Joined
·
1,528 Posts
...
...
What I mean by that is , for example the left mid and midbass if they have a pathway difference to the first measurement point to not try to do any “pre delays” so that the two speakers originate at the same time.
The will ensure that the low frequency harmonics stay in phase between the same side components. Again a tweeter crossed up above 2.5k where the wavelength is less than one cycle of the path length differences between the components it won’t make any difference. So a pre delay on a tweeter to the mid to the mic location will actually lessen the amount of all pass impulse stretching. Where a mid to midbass or to sub you do the opposite and don’t worry about doing any pre delays at all. It’s going to convolve it no matter what, might as well keep the LFE intact. It very much could make it a lot better......
Interesting. I will experiment with this. Thanks
 

·
Premium Member
Joined
·
220 Posts
What I mean by that is , for example the left mid and midbass if they have a pathway difference to the first measurement point to not try to do any “pre delays” so that the two speakers originate at the same time.
The will ensure that the low frequency harmonics stay in phase between the same side components. Again a tweeter crossed up above 2.5k where the wavelength is less than one cycle of the path length differences between the components it won’t make any difference. So a pre delay on a tweeter to the mid to the mic location will actually lessen the amount of all pass impulse stretching. Where a mid to midbass or to sub you do the opposite and don’t worry about doing any pre delays at all. It’s going to convolve it no matter what, might as well keep the LFE intact. It very much could make it a lot better......

In my experiments this has proved to keep better stage symmetry and keep low frequencie harmonics in phase across components.

Dirac will solve any timing differences in its all pass or convolution. So it’s better to know what it’s doing for your situation to help keep stage symmetry.

So awhile back I said something to the effect that a 12db crossover on mids and highs nets a better more open sounding ambient deep stage. I finally know why this seems to be a fact in my situation. It’s definitely worth exploring. Your situation could benefit (maybe not) but definitely worth exploring
So if I am understanding this correctly, 2 things.

1) The pre-tune I recently did using Dirac to get TA mostly correct maybe is not the best route when using Dirac for a 2-Ch tune? Perhaps we should leave TA alone and let Dirac deal with it?
2) I usually use 24db slopes to try to help with any phase issues. Are you suggesting 12db slopes if Dirac is going to "clean things up" with the final tune?
 

·
Registered
Joined
·
4,184 Posts
So if I am understanding this correctly, 2 things.

1) The pre-tune I recently did using Dirac to get TA mostly correct maybe is not the best route when using Dirac for a 2-Ch tune? Perhaps we should leave TA alone and let Dirac deal with it?
2) I usually use 24db slopes to try to help with any phase issues. Are you suggesting 12db slopes if Dirac is going to "clean things up" with the final tune?

Yeah sorta

I think I’m mostly implying try it both ways
Try it a few different ways , see what sounds more correct.

It’s looking at the sum no matter what in a 2ch tune, if your pre tune sets things up to where the after causes any cancellations in the crossovers especially.

If your pre tune has the minimum phase regions only with pre eq that should’ve fine
The problem with that is how many comb notches there are in the mid and highs makes that somewhat difficult without smoothing and or the use of high Q eq.

At that point it’s not even worth it and it needs to be made invertible at least in wider sections that left and right are both doing the same thing.

So with that in mind, and the fact that Dirac will do what it wants to do to get a end result no matter how it gets there: so to blindly just eq the amplitude or set delays the way you want it may or may not be counterproductive and cause more processes for Dirac to achieve its end game.
So I am of the mind , let it do it. But definitely try it a few different ways and see if one has better ambiance and clarity or better staging. I definitely wouldn’t spend any considerable amount of time doing any pre tuning. Maybe set some pre delays only if needed because the end result isn’t as good. Let Dirac see the room and the speakers the way they actually are. That is something I would do first and than try other pre setups and try to beat that.

I usually look for 3 big criteria

Stage placement is number 1 (not necessary just center placement that might be remedied usually post Dirac (depends). but actual placement across the stage and depth and symmetry between LandR)

Than ambiance.
Is it transient and have a good sense of space.


Than problems in the crossovers , listen for a crossover to be fighting another channel up/down starting with the midrange
 

·
Registered
Joined
·
2,418 Posts
Maybe set some pre delays only if needed because the end result isn’t as good. Let Dirac see the room and the speakers the way they actually are. That is something I would do first and than try other pre setups and try to beat that.
Pre-delay (having Dirac give me the delays) is about all I do as far as a "pre-tune" goes and so far that seems to have given me the best results. I have tried with doing nothing before Dirac, doing some EQ before Dirac and now just pre-delay. With doing nothing or pre-eq the stage for me would seem just not quite right. Typically it would be that the left side seemed to be, for lack of a better word, compressed or missing something, with pre-delay it seems to be much more open on the left than without.

It's certainly been good enough that I have not touched it in months now.


My Helix is for sale. I will soon be joining the DL crowd... :)
Another Helix user making the change.... coming from a Helix, I personally prefer the layout and some of the features of the Helix software, but those I mostly noticed when manually tuning using the 8x12, since upgrading the firmware to Dirac those features are pretty much negated.
 

·
Premium Member
Joined
·
220 Posts
I still have my Helix P Six and love it! However, I think it would be "end-game" if someone came out with an all-in-one DSP/AMP/Dirac unit! Something like the P Six WITH Dirac, or MiniDSP 8x12 with an amp! I would then be ready to upgrade my setup to this, as I am sure many others would also. I am sure in time these will come out. I will be ready to buy once they are!
 

·
Registered
Joined
·
4,184 Posts
If it can do the delays between same side drivers as part of the convolution, it makes sense to let it do it as such in the low frequency.

To better articulate

Let’s say your using a LR4 and you have a PLD of .2ms between the midrange and midbass , and a 1ms between the midbass and sub for the left channel

For the right channel let’s say the mid to midrange is .1ms and .8ms to the sub

So .2ms can be easily made up in convolution, so can .8

If the lr4 in the stopband can affect the phase of the lower driver into its inband by adding a delay would disrupt the two drivers
From coupling at the lower frequency.

It might be a PLD to one spot in space but the two drivers would be out of phase with each other as far as wavelengths go. The added delay in the stop band could be viewed as an excess phase, which in turn might try to be corrected for as a sum thus throwing the crossovers out of phase

Here’s a good analogy

If you have two subs in a box and there 20” apart from each other and you add delay to one of them , it’s almost certain they will fight each other. You wouldn’t try to delay subs to a single location, the wavelength is too long and adding delay would move one too far out of phase and cause a group delay that is destructive.

So the same thing going on between mids midbass and subs in the stop bands and the sum of the phase between them.

If Dirac corrects the sum , it can move the timing as well on same side drivers , if you leave them all at the same timing (except a tweeter perhaps) it’s more likely to have better overall phase


If any cone from any speaker is vibrating even the tiniest amount in the lowest area of its crossover stop band , it will greatly affect the ajacent band. By Haas depending on its timing/phase it may reinforce or cancel or take over the imaging cues if it’s ahead of the other driver of its first. So ideally , having the same timing avoids any strange behaviors.
 

·
Registered
Joined
·
4,184 Posts
Pre-delay (having Dirac give me the delays) is about all I do as far as a "pre-tune" goes and so far that seems to have given me the best results. I have tried with doing nothing before Dirac, doing some EQ before Dirac and now just pre-delay. With doing nothing or pre-eq the stage for me would seem just not quite right. Typically it would be that the left side seemed to be, for lack of a better word, compressed or missing something, with pre-delay it seems to be much more open on the left than without.

It's certainly been good enough that I have not touched it in months now.




Another Helix user making the change.... coming from a Helix, I personally prefer the layout and some of the features of the Helix software, but those I mostly noticed when manually tuning using the 8x12, since upgrading the firmware to Dirac those features are pretty much negated.

What’s up buddy!

Yeah if the delays Dirac suggests is a excellent proven good way....although not always necessary. Again. Do what works! Yes definitely


For anyone that wants to wrap there brain around what I’m saying. Think of this

Go take a close mic measurement and look at phase , now add 13.5” (1ms) and remeasure. The phase will have a constant delay and wrap.

So Dirac can simply do “wraps” at targeted frequencies with a shape that matches the rolloff of each speaker and there actual distance to compliment each other perfectly. Adding the delay appropriately and not changing crossover interactions.

Phase is angular and so is path lengths.
A path length difference of 1ms is 180deg at 1k or 90deg at 500 or 22.5deg at 125hz or 11.25deg at 63hz or 5.125deg at 31.

All it has to do is shape the phase at a very minor amount and shape it into the next driver in a multi-way.

With a 23ms window I don’t think a .5ms PLd is going to use up much fir , and it’s going to be dealing with other groups of delay simultaneously.
 

·
Premium Member
Joined
·
220 Posts
What’s up buddy!

Yeah if the delays Dirac suggests is a excellent proven good way....although not always necessary. Again. Do what works! Yes definitely


For anyone that wants to wrap there brain around what I’m saying. Think of this

Go take a close mic measurement and look at phase , now add 13.5” (1ms) and remeasure. The phase will have a constant delay and wrap.

So Dirac can simply do “wraps” at targeted frequencies with a shape that matches the rolloff of each speaker and there actual distance to compliment each other perfectly. Adding the delay appropriately and not changing crossover interactions.

Phase is angular and so is path lengths.
A path length difference of 1ms is 180deg at 1k or 90deg at 500 or 22.5deg at 125hz or 11.25deg at 63hz or 5.125deg at 31.

All it has to do is shape the phase at a very minor amount and shape it into the next driver in a multi-way.

With a 23ms window I don’t think a .5ms PLd is going to use up much fir , and it’s going to be dealing with other groups of delay simultaneously.
When I get time, here is what I would like to A/B test compared to my current pre-tune of "Dirac TA" configuration:

24db XO tweeter TA only
24db XO no TA
12db XO tweeter TA only
12db XO no TA

If anyone thinks this is a good idea and can test as well, please reply back and let us know how it worked. I will do the same once I can test.
 

·
Registered
Joined
·
4,184 Posts
When I get time, here is what I would like to A/B test compared to my current pre-tune of "Dirac TA" configuration:

24db XO tweeter TA only
24db XO no TA
12db XO tweeter TA only
12db XO no TA

If anyone thinks this is a good idea and can test as well, please reply back and let us know how it worked. I will do the same once I can test.

Love it.....

And the listen. As soon as you turn it on you can’t immediately say it’s better or worce based only how strong the center is.
(I did that for a long time before ever giving it a chance to find the best parts of Dirac)

Some tracks may have a perfect center and some may be a little diffu’se , listen for overall ambiance, spectral balance, center , stage depth, placement, and system linearity.

I would throw a 48db test in there also at least between mids and tweets and mids and midbass, especially if tweets close to the mid)
I love it ! I’m very curious what anyone down for the test thinks. And the test should have 1,2days listening time and a first impression critique.

Again, this might not be the best test for the guy/gal that really only wants a defined center that’s boundry is inside the car. The soundfield may or may not be like that, I think that depends a lot on your locations and speaker size and dispersion characteristics etc..... may be surprised none the less :)
 

·
Registered
Joined
·
66 Posts
I hope too many people don’t mind if I ask a quick noob question: Does anyone what the most elegant way to integrate this unit into a BMW M240i system would be?

Under ordinary circumstances I’d get a Helix UP7 amplifier that is plug and play and integrates 5 channels for center and RLFR + 2 sub channels along with an 8 channel DSP.

That would be the most elegant. But then I wouldn’t have Dirac, and I need Dirac after experiencing it at home. Any advice on what the most elegant way to get Dirac on 5 channels + 2 subs in a BMW system would be greatly appreciated.
 

·
Registered
Joined
·
3,277 Posts
Im curious if turning off XOs for the first measurement would be a good idea. Less phase anomalies from crossovers when it's looking for timing ques.

Sent from my XT1710-02 using Tapatalk
 

·
Registered
Joined
·
377 Posts
Picked up one of these 8x12 DL units for Christmas. It is replacing an aging Zapco Z8. Looking forward to seeing what it can do. I did to the individual speaker measurements to get the TA numbers and I did it a couple of times to see how consistent the number are. I also fiddled a bit with the mic input sensitivity and it seems that that calibration is very important. Following the recommendation in the Dirac app to calibrate the input to be slightly below -24dB seems to yield best results, meaning, the numbers looks close to what I had arrived at using HOLMImpulse when doing it by hand. Just wanted to mention that.

Also, the Dirac app on OSX Mojave does not work. First, it does not request microphone access which prevents you from calibrating the input gain on the mic. MiniDSP needs to update the app to make the API call to request the access from the user (OSX security feature). I got around that issue by starting the app from CLI which triggers an access request from Terminal app which wraps the Dirac app. But, once you start running the sweeps, the app crashes. So, Dirac app is a no-go on OSX Mojave right now. I reported this on MiniDSP forum. I will probably just get a USB adapter for my surface tablet and use that for calibration.

FYI, using the TOSLINK input with Clarion HU using 96kHz sampling rate. The clarity from high-res FLAC files on this setup is kind of blowing my mind. Can't wait to hear it once I have a tune working.
 

·
Registered
Joined
·
4,613 Posts
Picked up one of these 8x12 DL units for Christmas. It is replacing an aging Zapco Z8. Looking forward to seeing what it can do. I did to the individual speaker measurements to get the TA numbers and I did it a couple of times to see how consistent the number are. I also fiddled a bit with the mic input sensitivity and it seems that that calibration is very important. Following the recommendation in the Dirac app to calibrate the input to be slightly below -24dB seems to yield best results, meaning, the numbers looks close to what I had arrived at using HOLMImpulse when doing it by hand. Just wanted to mention that.

Also, the Dirac app on OSX Mojave does not work. First, it does not request microphone access which prevents you from calibrating the input gain on the mic. MiniDSP needs to update the app to make the API call to request the access from the user (OSX security feature). I got around that issue by starting the app from CLI which triggers an access request from Terminal app which wraps the Dirac app. But, once you start running the sweeps, the app crashes. So, Dirac app is a no-go on OSX Mojave right now. I reported this on MiniDSP forum. I will probably just get a USB adapter for my surface tablet and use that for calibration.

FYI, using the TOSLINK input with Clarion HU using 96kHz sampling rate. The clarity from high-res FLAC files on this setup is kind of blowing my mind. Can't wait to hear it once I have a tune working.
Glad to have you here and on the MiniDSP forum Ziggy! Thanks for your input!
 

·
Registered
Joined
·
377 Posts
Im curious if turning off XOs for the first measurement would be a good idea. Less phase anomalies from crossovers when it's looking for timing ques.

Sent from my XT1710-02 using Tapatalk
You can try it but curious how you're going to protect your tweeters from playing super low frequencies? Also, I imagine you'd get all sorts of gain from drivers that would be getting frequencies that they usually don't play. Your sub would still be playing above 200hz and your mids would be bleeding into tweeter territory. I suppose you could EQ each driver to cut them but my understanding is that you actually want the xovers in play to have Dirac correct phase in those areas.
 

·
Registered
Joined
·
377 Posts
Got a first pass at the tune done and it is already sounding better than what I did before by hand with my Zapco. The imaging is so precise, it's exceeding my expectations. This is what it looks like right now. I don't care much for that 26Hz dip. Not much is going on that low in the genres that I listen to. Any idea on that 60Hz 4dB dip?
259781
 

·
Registered
Joined
·
3,277 Posts
You can try it but curious how you're going to protect your tweeters from playing super low frequencies? Also, I imagine you'd get all sorts of gain from drivers that would be getting frequencies that they usually don't play. Your sub would still be playing above 200hz and your mids would be bleeding into tweeter territory. I suppose you could EQ each driver to cut them but my understanding is that you actually want the xovers in play to have Dirac correct phase in those areas.
Well yes you would put a hpf on the tweeter. I feel like i remember that hogh frequencies are easier to get time alignment info from. The MInidsp does one driver sweep at a time so you wouldn't have conflicting sonic information between drivers.

And im only talking about the very first Center measurement that it asks for. Id have the Plugin up and re-enable everything for the rest of the measurements.

Sent from my XT1710-02 using Tapatalk
 

·
Registered
Joined
·
3,277 Posts
Got a first pass at the tune done and it is already sounding better than what I did before by hand with my Zapco. The imaging is so precise, it's exceeding my expectations. This is what it looks like right now. I don't care much for that 26Hz dip. Not much is going on that low in the genres that I listen to. Any idea on that 60Hz 4dB dip? View attachment 259781
Whats the before look like? Looks like comb filtering between midbass sub maybe. It won't fix more than 10 db.

Do it with just the midbass mid tweet on. See if the dips are there.

Sent from my XT1710-02 using Tapatalk
 
221 - 240 of 283 Posts
Top