I actually have another guy on my team. He is the one who told me, Dude you gotta start using simulation software. He was right. I can't say who it is, because it's top secret shit. Plus I am not sure he really knows what he's talking about. LOL
@PWRRYD said:
That's sure a lot of parts to achieve sort of first order slopes. I don't get it.
I view this as an educational opportunity to improve my understanding of how xovers work. This way, if and when I drop the big bucks on super high end drivers, I will have more tools in my kit to pound the squigglies flat.
That looks pretty good to me. Let me punch in your values as a crossover variant on my end. Then I can quickly flip back and forth between crossover variants to see how they stack up against each other. How do you like using VituixCAD?
@Eggguy said:
I like it but I cant figure out how you got your target tilt lines on there or how much tilt is recommended
Man VituixCAD is great. There have been a few things that I wished it could do, but it turned out that it could already do them and I just want aware of it. The target line for In Room response is tricky to find. I'll post a few screen shots and recommended targets tonight when I'm on my computer (unless Bill beats me to it).
@Eggguy said:
I like it but I cant figure out how you got your target tilt lines on there or how much tilt is recommended
You hold down the ctrl key and then click the left mouse button and they magically show up. Then you can drag the slope up and down on each end of the line with your mouse. The slope target that you set depends on the RT60 reverb time of your listening room. If you have a room that is very dead with thick carpet, heavy drapes, etc., then you use shallow slopes, like maybe .1 to .2dB/octave for the listening window and 0.6 to 0.8dB/octave for the In-Room response curve. If you have a room with alot of hard surfaces (very reflective), then you want maybe .2 to .4dB/octave for the listening window and about 0.8 to 1.2dB/octave for the In-Room response curve. Also, keep in mind that the slopes of our curves are somewhat "fudged" because we only used polars from 0 to 90 degrees instead of 0 to 180 degrees. Our slopes are therefore not a steep as they should be.
In addition to ommiting the 1.5mH shunt leg off the tweeter, I took your model and dropped the tweeter series resistor from 10 to 8.2 ohms. Then I added a 1.5uF/2.4 ohm zobel network across the tweeter. This improves the power response and flattens the tweeter a bit more from 5 to 20kHz. The default preference rating increases from 7.033 to 7.138 with this change.
@Eggguy said:
I like it but I cant figure out how you got your target tilt lines on there or how much tilt is recommended
You hold down the ctrl key and then click the left mouse button and they magically show up. Then you can drag the slope up and down on each end of the line with your mouse.
I just remembered what the 4th limitation of this model is. There are probaby other limitations, but here is a revised list based on my level of understanding:
1) Single channel measurements using raw phase data in XSim or VituixCAD are valid only at the 1 meter distance that is used to make the 3 measurement "mod delay" phase correction. They are not valid at the typical 2.5 meter listening distance. You need a model with accurate X, Y, and Z coordinates in order to triangulate the 1 meter measurements out to a 2.5 meter listening distance.
2) Single channel measurements suffer from a small amount of phase error on the polar plots.
3) OmniMic can only measure polars from about 0-90 degrees. It cannot get a phase lock on the peak of the impulse beyond 90 degrees or so. This causes the slopes of the Power and In-room response curves to be somewhat less than they should be.
4) The model does not contain accurate information below about 400Hz. OmniMic's blended gate mode includes alot of room bounce below 400Hz. Taking additional near field (NF) measurements and then merging them with far field (FF) curves and a diffraction model would be better, but this is a somewhat difficult and time consuming process that is also somewhat limited in accuracy below 400Hz.
Good info, Bill. There is much to consider. It is a very nice tool. Not the only tool needed. It seems to me that we mostly need a thorough understanding of the basics. During my walk today I thought to myself that the zobel on the woofer can be eliminated because it really only had an effect on the phase tracking around 10Khz which is an area in which the woofer is not contributing much. However it did make the squiggles look nicer. Then I wondered if I was just playing with the harmonics of cone break-up, or noisy polar samples. Gonna have to build and listen to more crossovers.
I agree, that woofer zobel is probably inaudible. One thing that I have discovered, when taking measurements with OmniMic, is that the phase curve tends to bounce around somewhat at high frequencies during the "whoop - whoop - whoop" cycling process. At least it bounces around a little bit on my particular setup. I'm using an old HP laptop with Windows 7 on it. So I always pause and save the FRD file at a cycling point where the phase curve is relatively stable at high frequencies (10kHz or so).
Not at all man , got a little happy face going on , little dip going on in the 2 k range. That 2 - 3 k range drop seems to be pleasant on human ears. Nice little bump on the ass end too.
The scale you used is 3dB/div, so it looks more skewed. I see +/-2dB outside of the +3dB bump at 150Hz. That really is not awful from the on axis balance perspective. The off axis seems to rebound a bit, so the room may help or hinder. They certainly won't be forward by any stretch, but maybe a touch heavy or thick in male vocals.
Of course, I'm judging from paper, and could be totally wrong.
Comments
That looks pretty good, Eggman! Bill's isn't awful either.
InDIYana Event Website
Worth a try.
.
Nice, parallel phase tracking going on there!
Thanks, You guys said it was important and that made sense to me, so I pursued it.
I actually have another guy on my team. He is the one who told me, Dude you gotta start using simulation software. He was right. I can't say who it is, because it's top secret shit. Plus I am not sure he really knows what he's talking about. LOL
I view this as an educational opportunity to improve my understanding of how xovers work. This way, if and when I drop the big bucks on super high end drivers, I will have more tools in my kit to pound the squigglies flat.
Mine has 50% more components, other than that, do you see any problems?
Its even order now
That looks pretty good to me. Let me punch in your values as a crossover variant on my end. Then I can quickly flip back and forth between crossover variants to see how they stack up against each other. How do you like using VituixCAD?
I like it but I cant figure out how you got your target tilt lines on there or how much tilt is recommended
I ommited that 1.5 mH shunt leg off the tweeter. It was causing more harm than good.
Man VituixCAD is great. There have been a few things that I wished it could do, but it turned out that it could already do them and I just want aware of it. The target line for In Room response is tricky to find. I'll post a few screen shots and recommended targets tonight when I'm on my computer (unless Bill beats me to it).
Thanks speaker buddies
You are welcome... oh wait, I really didn't help one bit
You hold down the ctrl key and then click the left mouse button and they magically show up. Then you can drag the slope up and down on each end of the line with your mouse. The slope target that you set depends on the RT60 reverb time of your listening room. If you have a room that is very dead with thick carpet, heavy drapes, etc., then you use shallow slopes, like maybe .1 to .2dB/octave for the listening window and 0.6 to 0.8dB/octave for the In-Room response curve. If you have a room with alot of hard surfaces (very reflective), then you want maybe .2 to .4dB/octave for the listening window and about 0.8 to 1.2dB/octave for the In-Room response curve. Also, keep in mind that the slopes of our curves are somewhat "fudged" because we only used polars from 0 to 90 degrees instead of 0 to 180 degrees. Our slopes are therefore not a steep as they should be.
I agree. The preference score actually increases slightly when you get rid of it because this change flattens the power response curve just a tad.
In addition to ommiting the 1.5mH shunt leg off the tweeter, I took your model and dropped the tweeter series resistor from 10 to 8.2 ohms. Then I added a 1.5uF/2.4 ohm zobel network across the tweeter. This improves the power response and flattens the tweeter a bit more from 5 to 20kHz. The default preference rating increases from 7.033 to 7.138 with this change.
The phase looks excellent (to me anyway) with drivers running parallel to each other
Nice. What do you think about soldering the zobel across the driver terminals to save space on the crossover board, or boards?
Sounds like a plan. Also, since there are so many parts, maybe break it out on 2 separate boards; one for the tweeter and one for the woofer.
We a re starting to think alike.
I had no idea you could do that!
I decided to post some VituixCAD tips in a separate thread. @Eggguy - the tricky way to modify the in-room response target is toward the bottom of the post.
I just remembered what the 4th limitation of this model is. There are probaby other limitations, but here is a revised list based on my level of understanding:
1) Single channel measurements using raw phase data in XSim or VituixCAD are valid only at the 1 meter distance that is used to make the 3 measurement "mod delay" phase correction. They are not valid at the typical 2.5 meter listening distance. You need a model with accurate X, Y, and Z coordinates in order to triangulate the 1 meter measurements out to a 2.5 meter listening distance.
2) Single channel measurements suffer from a small amount of phase error on the polar plots.
3) OmniMic can only measure polars from about 0-90 degrees. It cannot get a phase lock on the peak of the impulse beyond 90 degrees or so. This causes the slopes of the Power and In-room response curves to be somewhat less than they should be.
4) The model does not contain accurate information below about 400Hz. OmniMic's blended gate mode includes alot of room bounce below 400Hz. Taking additional near field (NF) measurements and then merging them with far field (FF) curves and a diffraction model would be better, but this is a somewhat difficult and time consuming process that is also somewhat limited in accuracy below 400Hz.
Good info, Bill. There is much to consider. It is a very nice tool. Not the only tool needed. It seems to me that we mostly need a thorough understanding of the basics. During my walk today I thought to myself that the zobel on the woofer can be eliminated because it really only had an effect on the phase tracking around 10Khz which is an area in which the woofer is not contributing much. However it did make the squiggles look nicer. Then I wondered if I was just playing with the harmonics of cone break-up, or noisy polar samples. Gonna have to build and listen to more crossovers.
I agree, that woofer zobel is probably inaudible. One thing that I have discovered, when taking measurements with OmniMic, is that the phase curve tends to bounce around somewhat at high frequencies during the "whoop - whoop - whoop" cycling process. At least it bounces around a little bit on my particular setup. I'm using an old HP laptop with Windows 7 on it. So I always pause and save the FRD file at a cycling point where the phase curve is relatively stable at high frequencies (10kHz or so).
Take a look at this crap that I thought sounded good
Not at all man , got a little happy face going on , little dip going on in the 2 k range. That 2 - 3 k range drop seems to be pleasant on human ears. Nice little bump on the ass end too.
The scale you used is 3dB/div, so it looks more skewed. I see +/-2dB outside of the +3dB bump at 150Hz. That really is not awful from the on axis balance perspective. The off axis seems to rebound a bit, so the room may help or hinder. They certainly won't be forward by any stretch, but maybe a touch heavy or thick in male vocals.
Of course, I'm judging from paper, and could be totally wrong.
InDIYana Event Website
What Wolf said. +/-3dB on axis need not sound bad at all. There are many speakers worse than that.