I suppose track is wrong. I see the SPL of both contributing equally, meaning as one driver is dropping off, the other is taking over without causing cancellation where they both are providing equivalent SPL
I confess that tracking / divergence only makes sense here re Phase. Re the cancellation, this asymmetric crossover with 50mm acoustic offset to show cancellation is trying to exhibit what I mean. I can see with a RN the constructive / destructive effect on the summed response. Yes I can see phase diverge and yes you can read the amount of cancellation from phase. I wasn't saying you couldn't.
The orange focusses on the more or less audible passband I usually focus on (~24dB below reference)
The orange focusses on the more or less audible passband I usually focus on (~24dB below reference)
If I decided I wanted 35 degrees positive phase difference between the two drivers held constant across the crossover region, I could see that on a phase plot but I couldn't easily see that on a null plot without deeper analysis.
That's right, but phase plays a role in what that response will be.
We will hear the phase if it's problematic. For example, if the drivers sum 6db at the crossover point then they are in phase there, but if we only see them summing 3 or 4db higher together then they could be 90 degrees out of phase and I promise you will hear that and it can be weird. This can often be seen better with a variety of off-axis measurements. Also, I have seen situations where drivers are in phase at the crossover point but out of phase at a different point below or above that. It can get weird when you're combining acoustic and electric slopes or manipulating the Q of the slope too far.
I will also add to this discussion that I think keeping this tweeter crossed higher is a better option and I think the 2nd order crossover above is the best looking one. I also think that these are simulations for a crossover and it's important to take actual measurements because they will almost always be different once the drivers are in a box.
Update.
I built one speaker. No measurements yet. But it's obious that they are too tilted down. I measured the resistors and they measure as 6.2 not the 4.7ohm thats written on them. I'm going to go out and get some 3.3, 3.9, and new 4.7 and see if that makes a positive difference before I try to measure. I think something around 4 ohm will be the ticket. Also - there's a LOT of bass from that 830656, which probably doesn't help with the notion that they are "too polite".
I built one speaker. No measurements yet. But it's obious that they are too tilted down. I measured the resistors and they measure as 6.2 not the 4.7ohm thats written on them. I'm going to go out and get some 3.3, 3.9, and new 4.7 and see if that makes a positive difference before I try to measure. I think something around 4 ohm will be the ticket. Also - there's a LOT of bass from that 830656, which probably doesn't help with the notion that they are "too polite".
Seems like your learning the software quick.
Curiosity is what the driver positions are.
and what center to center distance the tweet
and woofer are
and how that is established in the model.
few ways to do it.
But if the model assumes the wrong position.
Your just working with phase responses that dont
resemble real world.
and if people are just playing with z or playing
with y to make the sim " work"
you have to make sure those positions are
actually physically possible in real life.
Curiosity is what the driver positions are.
and what center to center distance the tweet
and woofer are
and how that is established in the model.
few ways to do it.
But if the model assumes the wrong position.
Your just working with phase responses that dont
resemble real world.
and if people are just playing with z or playing
with y to make the sim " work"
you have to make sure those positions are
actually physically possible in real life.
Is anybody the wiser after reading this thread. Get on and build it.listen, modify,listen again.there is no techno theory that is going to tell you how it sounds and if your creation is to your liking.
X and Z are pretty close to reality (+-5mm maybe less) as per my sim screen caps.
Also, since I'm looking at the Xover again, I tried just a 1st order on the midwoofer and it might even be better than the current iteration. Also - fewer parts - so stays with the cheap and cheerful theme. My worry is that distortion and breakup will be more likely as there only 4db diff between the midwoofer and tweeter at ~5400hz.
Also, since I'm looking at the Xover again, I tried just a 1st order on the midwoofer and it might even be better than the current iteration. Also - fewer parts - so stays with the cheap and cheerful theme. My worry is that distortion and breakup will be more likely as there only 4db diff between the midwoofer and tweeter at ~5400hz.
Measurements tomorrow if providence allows. But comparing to a Revel F35 - it isn't terrible by any stretch of imagination.
EDIT - also disagree - I am comfortable with predictions that Vituix provides. If the resistor was a 4,7 - it would have been even harder to tell that the tilt is too much.
over time you will learn how to not have 30 ohm peaks
at crossover points.
Using impedance compensation and adjusting component values
to the driver impedance points at those frequencies.
Crossover points are likely not 8 ohms.
then of course most the issue is " pulling out"
all the baffle step from using too small of baffle.
at crossover points.
Using impedance compensation and adjusting component values
to the driver impedance points at those frequencies.
Crossover points are likely not 8 ohms.
then of course most the issue is " pulling out"
all the baffle step from using too small of baffle.
I think you're right @WhiteDragon - impedance and bafflestep - are where I feel I need more understanding amongst many other things.
I would have stopped with the simulation from your post #3, build and measured it.
At some point you have to decide whether your hobby is playing with a simulator or building speakers.
As long as you don't have the measurements of YOUR physical existing chassis, build into your baffle, you are chasing ghosts.
In your case, Peerless is a Danish brand from yesterday, only tolerated by Indonesian Tymphany as a cash cow. Do not expect too much from the data sheet.
I always have a look at the date of a data sheet first, in many cases they are a decade old. As hardly any chassis with a reasonable price comes from some country you will ever visit today, even if the manufacturer is a honest Joe, the basic materials will change, changing the final product. Sometimes more, sometimes less. It is worthless to complain about that. No one in trade or production takes any responsability for any totally f***ed up data sheet ever. This doesn't mean the drivers are bad, often they even improve, you just can't simulate to the last dB.
This advice can save your from much frustration in your further DIYS speaker life.
At some point you have to decide whether your hobby is playing with a simulator or building speakers.
As long as you don't have the measurements of YOUR physical existing chassis, build into your baffle, you are chasing ghosts.
In your case, Peerless is a Danish brand from yesterday, only tolerated by Indonesian Tymphany as a cash cow. Do not expect too much from the data sheet.
I always have a look at the date of a data sheet first, in many cases they are a decade old. As hardly any chassis with a reasonable price comes from some country you will ever visit today, even if the manufacturer is a honest Joe, the basic materials will change, changing the final product. Sometimes more, sometimes less. It is worthless to complain about that. No one in trade or production takes any responsability for any totally f***ed up data sheet ever. This doesn't mean the drivers are bad, often they even improve, you just can't simulate to the last dB.
This advice can save your from much frustration in your further DIYS speaker life.
Your doing good with the software.
to maybe fast track that.
Having to not remove it, or less helps.
So use a wider baffle.
Always run a tweeter, tight tight as physically
possible close to the woofer based on its mounting
areas.
To remove baffle step without using high impedance.
And if a very small baffle is wanted.
The price would be payed with more passives.
Since various notch filters can be used, or a series
notch.
Also use the software scrolling impedance to fine tune
the filter at the drivers impedance. which isnt always 8 ohms.
The woofer crossover point is climbing to 12 or 20 ohms likely
So the inductor values especially often need to jump or down
to compensate.
Also understand zobel or impedance compensation can be
used before or after the filter to flatten out the impedance curve.
The natural rising impedance of a driver can be lowered.
So yes more friendly value inductors can be used.
Can often help improve the cutoff slope sharper.
Since the " theoretical" filter vales are correct.
But the impedance rise ruins the actual real world cutoff
For the record as well what is the baffle height and width
being used
to maybe fast track that.
Having to not remove it, or less helps.
So use a wider baffle.
Always run a tweeter, tight tight as physically
possible close to the woofer based on its mounting
areas.
To remove baffle step without using high impedance.
And if a very small baffle is wanted.
The price would be payed with more passives.
Since various notch filters can be used, or a series
notch.
Also use the software scrolling impedance to fine tune
the filter at the drivers impedance. which isnt always 8 ohms.
The woofer crossover point is climbing to 12 or 20 ohms likely
So the inductor values especially often need to jump or down
to compensate.
Also understand zobel or impedance compensation can be
used before or after the filter to flatten out the impedance curve.
The natural rising impedance of a driver can be lowered.
So yes more friendly value inductors can be used.
Can often help improve the cutoff slope sharper.
Since the " theoretical" filter vales are correct.
But the impedance rise ruins the actual real world cutoff
For the record as well what is the baffle height and width
being used
Impedance is better here.
2nd peak is smaller. Thanks for your suggestions @WhiteDragon. Lets see how this translates to the real world.
2nd peak is smaller. Thanks for your suggestions @WhiteDragon. Lets see how this translates to the real world.
Speaker is built. Measurements forthcoming... Assuming I can make good measurements, we will see.
I would caution against using a 1st-order crossover on your particular midwoofer. Your simulations show that it has some significant cone breakup resonances.
An extra capacitor shouldn't add a lot to the parts cost. Just buy one of the less expensive ones.
The resonance at 5.4 kHz is highly likely to be audible as it will continue to ring after the tweeter's response has subsided.
A first order crossover will not work, be low quality or be more expensive than a 2nd order one.
You can write that into your little red book of speaker building rules.
If you have very linear, mostly expensive drivers and use parts to smooth out the drivers impedance, you can get away with it.
As you have to invest into the impedance correction, there is no money to be saved compared to 2nd order.
Maybe keep the idea for a later project, this one will be fine with second order.
If you start to measure and you are new to that, if possible put the speaker on a stand about 1.5m high and measure outside, microphone 50cm-1m away.
Then take some measurements inside, so you get an idea how things change because of room and distance.
Near field measurements inside are nice and convenient for crosssover tuning, but don't give a realistic on baffle response.
I use both, always make a final outside measurement, at least from 3 different angles.
Do not overestimate the low end measurements from 200 Hz down. Better measure TSP and attach a simulation of your box to the free field measured curve. The low end depends too much on room and placement, so the actual result will never match. Because of this the simulation gives a better idea of the speakers capabilities.
You can write that into your little red book of speaker building rules.
If you have very linear, mostly expensive drivers and use parts to smooth out the drivers impedance, you can get away with it.
As you have to invest into the impedance correction, there is no money to be saved compared to 2nd order.
Maybe keep the idea for a later project, this one will be fine with second order.
If you start to measure and you are new to that, if possible put the speaker on a stand about 1.5m high and measure outside, microphone 50cm-1m away.
Then take some measurements inside, so you get an idea how things change because of room and distance.
Near field measurements inside are nice and convenient for crosssover tuning, but don't give a realistic on baffle response.
I use both, always make a final outside measurement, at least from 3 different angles.
Do not overestimate the low end measurements from 200 Hz down. Better measure TSP and attach a simulation of your box to the free field measured curve. The low end depends too much on room and placement, so the actual result will never match. Because of this the simulation gives a better idea of the speakers capabilities.
Hi all - just wanted to update. As mentioned I've built the speakers and today I got a chance to measure them (windowed to ~300Hz).
Executive summary: I built a short floorstander that doesn't sound good from above (D'oh!).
However, raising and tilting to get to woofer axis and below - it's really not too bad, considering the negligible parts cost (~$200CAD for the entire project). I am pleased with my first ever effort. (Lessons learned at the end).
Pics:
Impedance sweep:
There's a ~200Hz resonance in both speakers. Also, the "port dip" is different in one speaker - same freequency but less deep - anyone know why that may be so? Yes - that 33Ohm mountain will be in the "lessons learned".
Measurements:
Measurement validation - I measured my M22 first, to see if what I get is reasonable compared to published data and it seems OK-ish (emphasis on the "ish").
And mine:
Not sure how to scale it to match ... but it seems OK to me. Dip between ~1K and 5K seems to match and the rising response towards 20K. I dunno - if this is too far off, then what follows is not good either.
My speaker - on axis: Eeeewww what's happening at 6K!???
Horizontal (15, 30, 45): Hmmm - 6K dip smooths out a bit off axis?
Vertical Below (Woofer, 10 below, 20 below): Also smooths out at 6K.
Vertical Above (Tweeter, 10 above, 20 above): OMG! Not good. Not good at all.
Bass: I attempted to do nearfield measurements and splice earlier with indoor measurements, but I'm not confident of it. So lets just leave this here as an FYI
And here is outdoor on-axis without IR windowing, also as an FYI
Subjectively there is enough bass / sub-bass in there to be satisfying.
VituixCad Model and Crossover again:
Vertical directivity:
Lessons learned:
Executive summary: I built a short floorstander that doesn't sound good from above (D'oh!).
However, raising and tilting to get to woofer axis and below - it's really not too bad, considering the negligible parts cost (~$200CAD for the entire project). I am pleased with my first ever effort. (Lessons learned at the end).
Pics:
Impedance sweep:
There's a ~200Hz resonance in both speakers. Also, the "port dip" is different in one speaker - same freequency but less deep - anyone know why that may be so? Yes - that 33Ohm mountain will be in the "lessons learned".
Measurements:
Measurement validation - I measured my M22 first, to see if what I get is reasonable compared to published data and it seems OK-ish (emphasis on the "ish").
And mine:
Not sure how to scale it to match ... but it seems OK to me. Dip between ~1K and 5K seems to match and the rising response towards 20K. I dunno - if this is too far off, then what follows is not good either.
My speaker - on axis: Eeeewww what's happening at 6K!???
Horizontal (15, 30, 45): Hmmm - 6K dip smooths out a bit off axis?
Vertical Below (Woofer, 10 below, 20 below): Also smooths out at 6K.
Vertical Above (Tweeter, 10 above, 20 above): OMG! Not good. Not good at all.
Bass: I attempted to do nearfield measurements and splice earlier with indoor measurements, but I'm not confident of it. So lets just leave this here as an FYI
And here is outdoor on-axis without IR windowing, also as an FYI
Subjectively there is enough bass / sub-bass in there to be satisfying.
VituixCad Model and Crossover again:
Vertical directivity:
Lessons learned:
- I need to pay attention to vertical dispersion/directivity also! Maybe this would have been a more successful project if the woofer was above the tweeter? I made the layout decision without really thinking about it. I can definitely hear that 3-6Khz chasm when I stand up. Oh well... next time.
- Watch the impedance - I knew not to let it get too low (under 4Ohm) but I need to also not make 33Ohm mountains.
- Measurements are hard - not at all confident that the above represents anything aproaching the truth.
- THE BIG ONE: I have very forgiving ears - raise and tilt these things and they make lovely music that doesn't really bother me with anything offensive. Making speakers is fun and I am looking forward to honing my skills in objective measurement so that I am more confident that my next speaker is better and mabe even reaches for "good".
Nice one.
Re your points:
1. This is inevitable with relatively higher XO points and larger midwoofer increasing the CtC spacing between drivers. The solution here is to either cross lower or reduce the CtC spacing (thereby using a smaller midwoofer). Of course there are compromises (smaller woofer, less bass and possibly dynamics). One possible solution to lower the tweeter XO point is to increase the order of your XO slope to provide more protection. You'd need to model excursion and measure distortion to see if this is a problem.
2. Impedance mountains aren't a problem for a decent solid state amp. Phase angles are the next most important thing to look for (after minimum impedance). A good rule of thumb is to keep within 45 degrees of phase swing and a decent amp will be happy.
3. I agree. Full range / non-gated measurements are tricky. We only have 3 solutions.
a) Jack the speaker way up to about 3.9 meters off the ground to get a reflection free measurement down to say 50Hz at a 1000mm measurement distance
b) Do a Ground plane measurement. I've found these hard to setup and doing off-axis measurements difficult. You also get up to +6dB "mirror" image boundary gain that you need to subtract if you want to get an accurate SPL idea (2.83v @ 1m)
c) Quasi-anechoic. We splice the nearfield and farfield (gated) as you have done, but you must remember to add in the baffle diffraction and step losses to your nearfield BEFORE you splice, otherwise you will artificially boost the bass response by up to 6dB thinking you are getting more bass than you are in reality
4. At the end of the day, this is what really matters. However, once the euphoria wears off, we tend to become a bit more picky and critical of our designs. I'd suggest you go through your catalogue and find the worst recorded piece of music that you love and listen to most often. The key here is to make this enjoyable, but not "voice" your speakers so much that it makes good recordings "dull". This is the art of it to me.
Re your points:
1. This is inevitable with relatively higher XO points and larger midwoofer increasing the CtC spacing between drivers. The solution here is to either cross lower or reduce the CtC spacing (thereby using a smaller midwoofer). Of course there are compromises (smaller woofer, less bass and possibly dynamics). One possible solution to lower the tweeter XO point is to increase the order of your XO slope to provide more protection. You'd need to model excursion and measure distortion to see if this is a problem.
2. Impedance mountains aren't a problem for a decent solid state amp. Phase angles are the next most important thing to look for (after minimum impedance). A good rule of thumb is to keep within 45 degrees of phase swing and a decent amp will be happy.
3. I agree. Full range / non-gated measurements are tricky. We only have 3 solutions.
a) Jack the speaker way up to about 3.9 meters off the ground to get a reflection free measurement down to say 50Hz at a 1000mm measurement distance
b) Do a Ground plane measurement. I've found these hard to setup and doing off-axis measurements difficult. You also get up to +6dB "mirror" image boundary gain that you need to subtract if you want to get an accurate SPL idea (2.83v @ 1m)
c) Quasi-anechoic. We splice the nearfield and farfield (gated) as you have done, but you must remember to add in the baffle diffraction and step losses to your nearfield BEFORE you splice, otherwise you will artificially boost the bass response by up to 6dB thinking you are getting more bass than you are in reality
4. At the end of the day, this is what really matters. However, once the euphoria wears off, we tend to become a bit more picky and critical of our designs. I'd suggest you go through your catalogue and find the worst recorded piece of music that you love and listen to most often. The key here is to make this enjoyable, but not "voice" your speakers so much that it makes good recordings "dull". This is the art of it to me.
The woofer and tweeter have a nice and close centre-to-centre distance, so they have the potential to work well off axis. 👍
I note that the tweeter is close to equi-spaced from the top and the sides, which could lead to some frequency response effects caused by strong diffraction effects. ❓
I also noted that your VituixCAD simulations indicate that the woofer and tweeter are in-phase through the crossover region. If this carried through to the actual build, then the system should have been tolerant of listening above and below the design axis, as the main radiation lobe should be pointing directly forwards, perpendicular to the front baffle. 👍
That might be due to an enclosure resonance.
Not sure what's going on there? Is one of the ports partially covered over somehow?
The measured impedance in the 20Hz to 20kHz range shows a minimum value of about 7.8 ohms. That high impedance peak of about 33 ohms at 1.3kHz shouldn't be a problem, as the associated swings in phase response around it are quite moderate.👍
That's a good question. It does seem a bit odd. Looking at the VituixCAD simulations, you will note that at 5.8kHz there is a large phase difference between the woofer and the tweeter. If the simulations underestimated this, and the woofer had a somewhat higher output, then some phase cancellation might be coming into play. Or it could merely be a cancellation due to tweeter diffraction effects. Check by measuring and averaging the distance from the center of the tweeter to each of the three nearby edges. Even better, doing a simulation in VituixCAD using your baffle dimensions and the location of the tweeter near the top of the baffle might show something.
I'm a bit intrigued as to why you may have chosen the design axis to coincide with the woofer. Can you shed some light on why you may have made that choice? Many systems tend to place the design axis on the tweeter.
It's also quite evident that there is a lot of overlap in the acoustic responses of the woofer and tweeter below the crossover frequency at 3.5kHz or so. It could have been better to ensure that the tweeter filter network did a greater job of filtering, as the acoustic response is only 13dB down at 1kHz, which isn't a lot of attenuation. Of course, that would result in a need for tweaking the woofer's filter as well, as its acoustic response is characterized by a rather slowish initial roll-off.