As per thread title, this is my first attempt at a passive crossover so I am looking for comments, constructive criticism etc. I am here to learn!!!!
I guess my key areas of interest are the dip in impedance between 1 and 5 khz and also the hump in the "in room" similarly between 1 and 5 khz (related ?). Maybe the tweeter crossover is a bit high at 6.5khz, is this a reasonable tradeoff for good phase alignment ? (for info the preference rating is 7.1 which I can increase to about 7.3 by increasing R14 to around 4hz reducing the in-room hump).
Measurements were taken in an anechoic chamber so no room reflection (above 100hz anyway) and no gating. All measurements from the same point at 1m from baffle at tweeter height. Woofer nearfield merged at around 350hz (measurement taken at 1cm from woofer dome) - this is a sealed box approx 40L filled with 1.25lbs of stuffing (aiming for Q=0.7). Measurements taken at 0,10,20,30,45,60,75,90degs left and right so a semi-spinorama.
I also measured pairs of drivers and calculated the offset (20uS/7mm) for the tweeter and (270uS/93mm) for the woofer which work out very similar to the geometric/physical offsets. These are inputted into the driver setup on VituixCad.
I have a flat on axis active crossover with good phase alignment that sounds pretty good so trying to replicate that with a passive crossover i.e. flat on axis, 6db drop from 100hz to 10khz in room, good phase alignment / deep nulls.
Woofer xover is 2nd order electrical with tank for 3khz breakup (acoustic LR4 at 830hz)
Mid xovers are 2nd order electrical with a shaping resistor on the low pass (acoustic HP LR4 at 725hz, LP LR4 at 7250hz)
Tweeter xover is 4th order electrical with a notch to reduce level above 8khz (acoustic BUTT2 at 7400hz) - inverted polarity.
Woofer-Mid phase aligned form 200hz to 3khz, Mid-Tweeter phase aligned form 1khz to 13khz.
I guess my key areas of interest are the dip in impedance between 1 and 5 khz and also the hump in the "in room" similarly between 1 and 5 khz (related ?). Maybe the tweeter crossover is a bit high at 6.5khz, is this a reasonable tradeoff for good phase alignment ? (for info the preference rating is 7.1 which I can increase to about 7.3 by increasing R14 to around 4hz reducing the in-room hump).
Measurements were taken in an anechoic chamber so no room reflection (above 100hz anyway) and no gating. All measurements from the same point at 1m from baffle at tweeter height. Woofer nearfield merged at around 350hz (measurement taken at 1cm from woofer dome) - this is a sealed box approx 40L filled with 1.25lbs of stuffing (aiming for Q=0.7). Measurements taken at 0,10,20,30,45,60,75,90degs left and right so a semi-spinorama.
I also measured pairs of drivers and calculated the offset (20uS/7mm) for the tweeter and (270uS/93mm) for the woofer which work out very similar to the geometric/physical offsets. These are inputted into the driver setup on VituixCad.
I have a flat on axis active crossover with good phase alignment that sounds pretty good so trying to replicate that with a passive crossover i.e. flat on axis, 6db drop from 100hz to 10khz in room, good phase alignment / deep nulls.
Woofer xover is 2nd order electrical with tank for 3khz breakup (acoustic LR4 at 830hz)
Mid xovers are 2nd order electrical with a shaping resistor on the low pass (acoustic HP LR4 at 725hz, LP LR4 at 7250hz)
Tweeter xover is 4th order electrical with a notch to reduce level above 8khz (acoustic BUTT2 at 7400hz) - inverted polarity.
Woofer-Mid phase aligned form 200hz to 3khz, Mid-Tweeter phase aligned form 1khz to 13khz.
One suggestion: You need to add a y-dimension to each driver to match their vertical spacing. For instance, if the tweeter is the design axis, the tweeter y=0, the midrange y = - 140 (? approximately ?), and the woofer y = -350 ?? you supply the correct values.
The offsets are already in the measurements. Putting them into Vituixcad rather than leaving it at zero, creates double spacing.
For info I upped the Mid LP to a third order which brings the Mid-Tweeter crossover point down to around 3900hz. Not sure if this is any better or not but the upper frequency impedance is a little better.
I checked the step/impulse graph in REW for the raw measurements and they all seem to be aligned at 20ms (give or take a fraction). I used the "acoustic signal" but mistakenly use the source speaker as the delay signal (rather than using a second speaker) therefore the impulse is aligned with itself. Therefore I think I need to add the offset in from the pairs measurements/alignment. If I had correctly used a second timing speaker and the acoustic signal in REW I would have expected the woofer to have a 250-300uS, Correct ?
Woofer
Mid
Tweeter
Iirc that only applies to measurements taken on axis of every driver. Here everything is measured on tweeter axis and the Δt is already in the measurements.
The delta t is already incorporated into the measurements, yes... but the vertical spacing is not. The vertical spacing is required so that VituixCad can calculate the vertical cancellation effects. Without it, the Power & DI plots are meaningless.
If Ugg10 had made horizontal and vertical polar measurements for all three drivers, using a common mic position, then he could use 0,0,0 as the coordinates for each driver. But with only horizontal polar scans, VituixCad needs the y coordinates of each driver.
The method that was used to take these scans causes some complications when loading into VituixCad. The standard recommended practice is to take scans of each driver along that driver's axis, but always at a consistent distance. When I make scans of a small midrange and tweeter (such as this case), I will use a common mic position on an axis midway between the midrange and tweeter. For the woofer, I would move the mic down to be on-axis with the woofer. The VituixCad Help document says that this is an acceptable deviation from the standard measurement practice.
In this case, I believe that the delta t that is part of the measurements is the hypotenuse of the y vertical distance and the z offset distance. But the algorithms in VituixCad assume that the delta t is purely due to z offset. It may be possible to salvage the measurements. Anechoic measurement conditions are difficult to arrange, so it would be nice if we could find a way for Ugg10 to use them... If @DcibeL could join the conversation, I would like to get his opinion on what could be done with these measurements in order to use them in VituixCad.
@hifijim many thanks for chipping in very useful as always. I had planned on doing "on axis" measurements for each driver but ran out of time in the chamber unfortunately. I guess as the woofer is less affected by the room reflections and will have anything under 400hz merged with the nearfield I could do some measurements at home at 1m on axis and see how that goes if that is going to be easier than untangling the measurements I have got.
Just for completeness here are the results of the pair measurement technique for finding offsets -
Lats image is with the 20/270uS delays and a measurement of all three drivers running wild. As said these tally well with the distances estimated using a bit of Pythagoras.
Just for completeness here are the results of the pair measurement technique for finding offsets -
Lats image is with the 20/270uS delays and a measurement of all three drivers running wild. As said these tally well with the distances estimated using a bit of Pythagoras.
Member
Joined 2003
My recommendation would be to re-measure properly. There's a guide for that somewhere...
If you're splicing data, then the measured phase data is no longer accurate. Did you apply the baffle diffraction response to the nearfield before you spliced? In anycase you'll need to derive new phase data... and that means using minimum phase and X,Y,Z offsets in your VCad sim for all drivers.
Alternatively - use your measured phase (which you said was accurate for all measurements above 100Hz) and forego the bass response in your sims (setting all offsets - X,Y,Z to zero). Given the relatively high XO point between woofer and midrange, you'll have sufficient data points to make the XO meaningful here.
@Dave Bullet many thanks, I loaded up the woofer data without the merge (direct measurement exported from REW) and other than the very low end it made almost no difference to frequency of phase data. The minimum phase button is checked for all drivers in the driver set up pane. I left the old inverted mid nulls in to show low frequency difference. Null at 60Hz is a known chamber issue.
I also did a bit of pythagoras and also took the small offsets from the impulse and step response graphs in REW. The Phythagoras geometric are a closer match to the paired measurements 20/270uS (6mm/93mm) offsets. But none line up that well.
I also did a bit of pythagoras and also took the small offsets from the impulse and step response graphs in REW. The Phythagoras geometric are a closer match to the paired measurements 20/270uS (6mm/93mm) offsets. But none line up that well.
Member
Joined 2003
Probably ought to have said the impedance curves are measured ones using REW, measurements of driver in the box.
P.S. It's late in the UK so if you don't get a response I am not ignoring you, I'll pick up where I left off tomorrow.
Thanks for all of the contributions so far, many thanks indeed.
Just to recap - I have -
Measurements with mic at 1m from centre of baffle at tweeter height for (mic static for all measurements) -
The question is - what do I need to do to use these in Vituix for designing a crossover ? or do I need to start again ? [once I have the import raw data set then I can play to my hearts content designing crossovers].
P.S. It's late in the UK so if you don't get a response I am not ignoring you, I'll pick up where I left off tomorrow.
Thanks for all of the contributions so far, many thanks indeed.
Just to recap - I have -
Measurements with mic at 1m from centre of baffle at tweeter height for (mic static for all measurements) -
- 0degs for individual drivers, Mid/Tweeter together, Mid/Woofer together, All three together
- Measurements at +/-0, 10, 20, 30 , 45, 60, 75, 90degs horizontal (sorry no vertical) for each driver individually
- Nearfield of woofer at 1cm from the dust cap (no port as a sealed box)
- REW impedance measurements of each driver in the box.
The question is - what do I need to do to use these in Vituix for designing a crossover ? or do I need to start again ? [once I have the import raw data set then I can play to my hearts content designing crossovers].
Last edited:
Except that there is no reliable position information in this case due to the USB mic. Ugg wrote..
The delay will be different at every angle (even if you could reliably work it out).
Ugg10 - I have some more thoughts on this.
Now that I see your driver geometry includes both vertical and horizontal offsets, using your original data set becomes even a bit more complicated.
I also noticed in your original post you mentioned polar scans from 0 to 90, left and right. Did you do any scans from 90 to 180? If not, VituixCad will not be accurately calculating sound power.
As AllenB noted, you are using a USB mic and your timing data may or may not be useful.
Perhaps it might be best to collect new data. I do this all the time. I can not recall a single project where I did not have to repeat at least some measurements.
Now that I see your driver geometry includes both vertical and horizontal offsets, using your original data set becomes even a bit more complicated.
I also noticed in your original post you mentioned polar scans from 0 to 90, left and right. Did you do any scans from 90 to 180? If not, VituixCad will not be accurately calculating sound power.
As AllenB noted, you are using a USB mic and your timing data may or may not be useful.
Perhaps it might be best to collect new data. I do this all the time. I can not recall a single project where I did not have to repeat at least some measurements.
@hifijim thanks again and no I did not do 90-180 unfortunately as my time was very limited in the chamber. The Mic is a Behringer ECM8000 run through a Behringer UMC204HD interface. I do have two mics and have tested using the acoustic signal system with a second speaker in VituixCad so know I can use this method.
One of the down side of living in the UK is that we do not have nice open yards like in the US so measuring outside is not really an option as there is always a wall or a neighbour close by and also our rooms are much smaller gating needs to be down nearer to 3.5mS range to get past first reflections (typical living room is 12ft x 15ft with 8ft ceilings).
Does anyone have a link to the step by step measurement instructions that @DcibeL mentioned in post #10 or have an idiots guide I can crib?
One of the down side of living in the UK is that we do not have nice open yards like in the US so measuring outside is not really an option as there is always a wall or a neighbour close by and also our rooms are much smaller gating needs to be down nearer to 3.5mS range to get past first reflections (typical living room is 12ft x 15ft with 8ft ceilings).
Does anyone have a link to the step by step measurement instructions that @DcibeL mentioned in post #10 or have an idiots guide I can crib?
Sorry for the late entry. From the first design, I think your impedance is too low between the mid and tweeter, possibly because they are overlapping too much. Sharpen the LP filter on the mid and extend the tweeter down and make it roll off more sharply. Also related is that your mid should be cut off ~ 3-4 kHz based on dispersion. And yes to everything about measuring the acoustic offset.
It seems you are attempting a 4th order HP on the tweeter, but your slope is kind of shallow.
It seems you are attempting a 4th order HP on the tweeter, but your slope is kind of shallow.
Last edited:
Member
Joined 2003
I think that's mostly related to the size/directivity of your drivers and your cross points/roll-off shapes. Since the midrange has very wide dispersion still at 2 kHz, it's going to be a balancing act to get smooth off axis behavior above and below that range.
Shifting the woofer cross point up or adjusting its roll-off to be a little more gradual may help (assuming that doesn't screw up other things).
Just bringing the midrange level down slightly may be useful.
You may want to put a target curve on the midrange low-pass and the tweeter high pass and try to match those a little better also. The tweeter roll-off looks a little shallow, so that is probably adding a little to the midrange off-axis flare as well.
If none of that gets you where you want to go, just accepting that aspect as part of using the dome midrange can make sense also. Many highly regarded speakers have flares of that magnitude in their off-axis response. It's not egregious.
Last edited: