Site Links
Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Categories
In this Discussion
Who's Online (9)
- ani_101 10:18PM
- DanP 10:18PM
- Ed_Perkins 10:18PM
- Inertiaman 10:18PM
- jholtz 10:18PM
- mb1973 10:18PM
- Mitchell_Cote 10:18PM
- rjj45 10:18PM
- Steve_Lee 10:18PM
First TMWW - VituixCAD / Crossover advice and critique
Working on my first TMWW. Trying to follow the VituixCAD instructions very closely but going back over things I think I made a mistake during measurement. Once I get this figured out, I'll likely have other xo related questions.
For the two woofers:
- Impedance: I measured impedance in the box with both drivers wired in parallel, just as they will be in the final speaker.
- Far field: I measured SPL at 1m, on axis of the top woofer, with a pillow covering the bottom woofer. Both drivers connected in parallel.
- Near field: I measured SPL at 0.25" from cone of top woofer, with a pillow covering the bottom woofer. Both drivers connected in parallel.
Doing some more research on forums (not the manual), while the nearfield should be measured this way, Kimmo recommends the far field be measured by disconnecting the bottom woofer rather than using the pillow. He suggest the effect of the disconnected woofer acting like a passive radiator will be irrelevant once merged since its effect is only around tuning like a port.
I think...IF I would have measured with the 2nd woofer disconnected I would have an accurate SPL measurement for the single driver and would need to scale the impedance by 2, since my measured impedance was with both drivers connected in parallel. I would add both drivers to my crossover and would get +3db for doubling drivers and +3db for the lower impedance.
In my case...I have SPL of only one driver, but because it was still wired in parallel I still have +3db due to the lower impedance. So when I add my driver, I need to scale impedance by 2, and scale SPL by -3db.
Does that sound right??
Is the image below the right way to enter two drivers in parallel, setting each's y-offset correctly?
Comments
In both cases you have measured the response of a single driver, the SPL does not need to be adjusted. The driver is provided the same voltage whether 1 or two are connected in parallel, SPL out of a single driver will remain the same whether one is covered by a pillow or physically disconnected.
Scale impedance accordingly for the driver physical wiring, that is , scale the impedance by 2 for a pair of drivers measured in parallel, and scale by 0.5 for a pair of drivers measured in series.
Yes, those are 2 drivers in parallel. Enter physical x and y axis coordinates accordingly.
Is the upper woofer really 1 mm to the right (or left) of the other drivers? That typo probably makes no difference in the real world.
I have no doubt that 1mm is crucial to the design of this speaker.
Ha, nice catch. I'll claim that it was a typo and not the fact that my measuring and routing of driver recesses is off by 1 mm.
Just messing with you David
Okay, next question. I understand how baffle step loss works and that I would pad down my tweeter and midrange to the level of my woofer if I wanted full baffle step compensation. But I often see people only partially compensate for baffle step loss, say 4.5db. In xo software, would the SPL curve just look about 1.5db low and gradually rising to the baffle step frequency, relative to a flat line? Below is a simplified illustration I've worked out for my own thinking. (By the way, other than perfectly flat SPL curves, these are my actual drivers - 93db tweeter, 90db midrange and dual 87db woofers.)
How does it sound currently with the actual measurements in room ? Im looking at the mid in your measurement and wonder if you couldnt recess the mids a bit?
excuse me if I missed something
All of my speakers ended up with BSC in the 5-6dB range, none I feel are boomy or unbalanced sounding. Most of that I believe will come from poor room placement, or poor measurement and response processing that leads to inconsistent results. For room placement issues such as jamming a speaker in a corner in too small of a room, I would still prefer to design the speaker for "every room" and fix the room placement issue with EQ after the fact. This will allow the speaker to be used anywhere with predictable results. The only exception would be a woofer right at floor level, where completely reflection free measurements may be a bit skewed as the floor can be considered a constant here and provides bass reinforcement that probably should be included in the design process.
To answer your question, yes the full BSC implementation will be "flat on-axis", 4.5dB implementation would simply mean that >1kHz response will be 1.5dB lifted vs the 100Hz response, so you can expect a slight uphill slope from 100Hz to about 1kHz.
Since you are using VituixCAD, I suggest turning your focus to the in-room response in the power & DI chart. This is a spacial average that approximates that "every room" as a typical in-room response. I find that setting up the optimizer using listening window at -0.2dB/oct from 100Hz to 10kHz and in-room response at -0.8 to -0.9dB/oct from 100Hz to 10kHz provides excellent results. You can dive into the preference rating tool for optimization as well, which balances many factors and not at a specific constant slope like I've described.
Hi Nick. The graphs were just lines and curves in Excel to illustrate what I "think" the intuition is. I have what I hope are good measurements and now the speaker is disassembled while I paint and veneer it this week. I'm trying to work on the xo and this is where I start questioning my understanding and ability to take accurate measurements.
Yes, I am working on that. I had worked up what I thought might be a good xo to then realize that I had only put in ONE of the woofers. I realize that if my measurements are wrong OR I process and get them into VituixCAD wrong then a great looking xo in the program isn't going to reflect anything in reality. I do think my measurements are good, so hopefully will make progress soon. I will then post here for more feedback on the crossover as well as in-room response, directivity, etc.
Here's something that might help provide some context to the charts and graphs and make sure the first crossover you build is a good one. I wrote up this document a while back on how to use VituixCAD along with APO EQ to simulate any crossover designed in VituixCAD digitally. It can be active blocks or passive components, the result can be exported to APO EQ and simulated as a digital crossover using your PC as the DSP. You just need enough audio channels and amplifiers available for the speaker you have in mind. Many desktop PCs these days have 5.1 or 7.1 audio built in which is enough for a pair of 3-way speakers.
https://diy.midwestaudio.club/discussion/1832/simulating-crossover-filters-with-vituixcad-and-eq-apo#latest
Looking at other screenshots on the internet, the SPL curve of dual woofers is represented by a single line showing the combined effect. The way I have them, VituixCAD shows two separate SPL curves. (I have exaggerated the y-offset just to make it obvious.) Do I have them hooked up wrong or do I need to change some option to have it shown a single combined line??
You will only see a single line for drivers where the x,y dimensions are the same (incorrect configuration) or in an MTM configuration where distance from speaker to listener and angle are the same.
You have a pair of drivers with significant difference in the y dimension, so the response from each should be different. Each driver has a different distance and angle to the listener. so what you are seeing in VituixCAD is correct.
But the two drivers combined are 6db higher than the individual drivers. This does show in my reference angle and listening window curves...but when I am trying to match for example an LR2 on the midrange and an LR2 on the combined woofers, the crossover "appears" to be about 12db down on the individual drivers but really is only 6db down (as it is supposed to be) on the dual woofers. Is that just normal and you just mentally take it into account when trying to fit the target curves?
(I appreciate everyone's patience, as I get that a lot if this is obvious to those who already know what they are doing.)
With such a distance between drivers, you can expect some phase interaction as you approach 1kHz. Design your speaker for the resulting sum of the two drivers together.
The large distance between drivers was added to specifically make the lines diverge. The actual offsets are -286mm and -508mm. What was confusing me was having two lines (even if right on top of each other) vs a single line 6db higher.
I'm pretty sure I made a mistake measuring the woofer and can't re-measure until Saturday so I'm stuck on hold for a few days.
2 drivers, 2 traces, sorry but I’m not understanding the confusion. Turn on reference trace and you will see that it’s roughly 6dB higher as the two individual drivers add together.
There was something "off" on my measurement of the woofer so I redid my measurements. Measured outdoors, tweeter 77" from the ground, nothing other than the ground within about 12 feet of the speaker and mic setup. I should be able to put components together and actually listen tomorrow. Speaker is 40" H x 9.5" W x 15.5" D
Below are two starting points - does anyone see any obvious issues/problems or have general recommendations? I get the basics - different orders of xo, L-pad, zobel, and notch filters and that is about it. 2nd crossover notches out the peak 1.5khz which I'm quite sure is baffle step induced.
The images got squished down when you posted so it's a bit hard to read. Considering overall response, the second would be much preferred by myself. First one needs the mid padded down a bit more. Second crossover has a impedance dip around 1.5kHz, not a big deal but something to be wary of.
I would consider a series notch around 7kHz for the midrange breakup peak. The tweeter as well has a fairly wonky response, I would consider lifting its response a bit if you find it a bit dull as-is.
I would also consider turning on in-room response (orange trace) on Power & DI and using that as the optimizer target.
Coming back to this project...I am definitely going to need some help with the xo, but first I want to run my measurements by anyone interested. Other than squiggles, dips and peaks that I wish weren't there, I think this is just the reality of what I have to work with. Let me know if you see anything that just looks wrong or doesn't make sense.
Measured outdoors with the tweeter more than 6 feet from the ground at 1m on axis of each driver being measured. I can see how small differences in merging my farfield and nearfield responses will make a pretty big difference considering I'm likely crossing the woofer around 350 to 400Hz. (Farfield and nearfield frd files attached if anyone wants check my merging decisions.)
Tweeter: The original tweeter GRS PT2522 planar in a diy waveguide with a diy deep back cup replaced with GRS PT2522C with new GRS faceplate and flush mounted. Actually measures pretty smooth compared to before except the big spike at 15K which is NOT in the mfg's curves. My baffle is 9.5" wide and has about a 1/4" drop about 5/8" from the edge and then the edge has no roundover (see pic below (not with the new tweeter)). I'm assuming that might be the cause?
Midrange (Dynavox LY401F): This is a full-range driver with quite flat mfg's curve. I have to assume all of the squiggles are just baffle related. Peak at 1.4Khz is consistent with what BSC diffraction modelling says will happen.
Woofer (Dayton Audio DC200-8x2)(Raw SPL of top woofer only in image below): Only odd thing here is the dip and peak between 200Hz and 300Hz. This is from the nearfield measurement and shows up in every version I have. I measured right in front of the dustcap as well as slightly to the side where the dustcap meets the cone and it didn't make any difference.
A4eaudio, I had a look at the response data you've provided and can give you some commentary on the data:
DC200:
Wiggle at 250Hz could be a cabinet mode from the height, or potentially a port resonance. Was the cabinet fully stuffed for the measurement? If it's port resonance, it might be worthwhile to plug the port for another nearfield measurement, also a method to determine if the resonance is caused by the port.
The big dipper at 1kHz is normal DC200 response.
LY401F:
Manufacturer response is 1/3 octave smoothed with squished virtical scale to make it look good, so if you want a direct comparison, it's easy to smooth to 1/3 octave in VituixCAD when you load the driver data. The midrange wiggle is unfortunate, but I don't think is diffraction related as it follows through with all the off-axis responses.
If it were me I'd be tempted to spend $30 on a sacrificial driver and cut out the whizzer with a scalpel and see how that changes things. For midrange duty it might be an improvement, hard to say without trying.
Minor detail but the diffraction file here doesn't have the mic located in front of the driver. Not a big deal since you've entered 10m for the distance so it doesn't really affect anything significantly enough to care about.
PT2522:
Looks like a fine tweeter. I wouldn't worry at all about the response above 14kHz.
Thanks dcibel! I'll post xo attempt tomorrow or late tonight. I took this speaker to SDC and overall it sounded
"okay", middle of the pack probably, but with the new tweeter I'd like to get better than okay. As long as the measurements are correct I think I can get there with help. See some comments below.
I just finished a tiny little speaker that had a port resonance - peak right where WinISD said it would be. Port resonance here should be 850Hz so most likely a cabinet mode. It was lined pretty well but I can work on that. It will also get tamed a little bit with the xo in place. Just wanted some confirmation that it made sense.
Hmmm...I might try that. I actually had one that measured poorly and bought a replacement a few months ago, but pretty sure I tossed that bad one.
Well crap, thanks for pointing that out.
Good to hear that. Even though I can't hear above 13KHz I would have tried to address it if needed, as everyone else in my house can hear to 17Khz+.
It's not a problem really, in fact 10m distance is what's in the screenshots in the VituixCAD measurement guides. I think distance for this should be typical 2-3m listening distance, but there's not much difference from 2m to 10m. 10m avoids interaction from multiple drivers such as MTM.
I also don't know what's going on <800Hz in the midrange response you show above. I assume that's some sort of blended in-room response?
Here's what I've got with your merged responses. Spliced the near field around 200Hz to remove the woofer wiggle from the nearfield measurement.
If it sounds anything like it looks, it's going to kick butt.
Thanks! They definitely look better than they sound right now, but hopefully that will be rectified soon.
I have no idea what happened there, but it was obvious once you pointed it out. I think I must have loaded something related to another measurement in the LF section.
I was able to replicate you merged responses for both the midrange and woofer.
That made a big difference, whereas I think I was splicing closer to 450Hz before. How did you chose 200Hz, just a rule of thumb you always follow or was there something in the nearfield and farfield data that you saw that indicated that was the appropriate point?
Usually the choice is just the location where you get a good overlap between the near field and far field response. For myself this is usually in 300-500Hz range because of the gating limitation from measuring indoors with 8ft ceiling. You had mentioned that you measured outdoors with the tweeter elevated above 6ft, so the far field response and window length should be a fair bit longer that usual and provide higher resolution far field data allowing the merge frequency to be lower, so response down to 200Hz is usable.
I think I mentioned previously, that when is comes to the low end of the windowed response, the type of windowing used can make a difference of whether you get overshoot or undershoot in the low end of the response where there's little resolution. Based on Kimmo's recommendation, using Cosine windowing here provides the best result, with Tukey 0.75 a close second.
Ok, here is a stab at a xo. I understand different orders of electrical orders and acoustic slopes, zobels, L-pads and notch filters and that is about it - I don't know any of the nice little tips and tricks that the more experienced people know.
A couple of points and a question:
Please let me know if you have suggestions, see any problems, or see anything that can be done better. (FRD and ZMA files attached in case you are interested in experimenting.)