Hi,
To the best of my ability I am only able to get a non problematic frequency response if I switch the polarity of either my woofer or my mid. The flattest response seems to be with the mid switched.
Are there any negative effects to switching the polarity on a midrange such negative effects on imaging etc?
The frequency response that I'm getting is:
with blue being just the mid reversed, red being just midbass reversed and green being all in phase.
The crossover being used is:
The drivers being used are:
Tweeter - Dynaudio t330d
Mid - Dynaudio m560d
MidBass- AudioTechnology C-quenze 21i52
It's been a while since I've worked on speakers and any advice would be of great help.
To the best of my ability I am only able to get a non problematic frequency response if I switch the polarity of either my woofer or my mid. The flattest response seems to be with the mid switched.
Are there any negative effects to switching the polarity on a midrange such negative effects on imaging etc?
The frequency response that I'm getting is:
with blue being just the mid reversed, red being just midbass reversed and green being all in phase.
The crossover being used is:
The drivers being used are:
Tweeter - Dynaudio t330d
Mid - Dynaudio m560d
MidBass- AudioTechnology C-quenze 21i52
It's been a while since I've worked on speakers and any advice would be of great help.
Attachments
Speakers arent resistors. So its possible that polarity wouldnt be as expected with such simple arrangements without impedance correction.
From last picture iz seems to me midbass is not attenuated enough, so it overlaps to much with mid.
Have you tried to increase the value of L2, to say 3,3mH.
You may get smoother response, but still have to reverse the polarity of mid.
To be honest, I dont know for sure about negative effects of reversed driver polarity to imaging.
One of my speakers had reversed tweeters and still very nice imaging.
From last picture iz seems to me midbass is not attenuated enough, so it overlaps to much with mid.
Have you tried to increase the value of L2, to say 3,3mH.
You may get smoother response, but still have to reverse the polarity of mid.
To be honest, I dont know for sure about negative effects of reversed driver polarity to imaging.
One of my speakers had reversed tweeters and still very nice imaging.
You wire one driver in reverse polarity to counteract the phase shifting done by the crossover.
So by wiring one driver out of phase the acoustic output is put back in phase (more or less).
So by wiring one driver out of phase the acoustic output is put back in phase (more or less).
I haven't tried L2 with 3.0mh however when I model it in XSim, anything over 1.8mh starts to have an effect on the lower 200-300hz band.
I'll try it and see.
Is it worth while trying a higher order x-over with the woofer and would it be possible to keep the mid/tweeter the same?
I'll try it and see.
Is it worth while trying a higher order x-over with the woofer and would it be possible to keep the mid/tweeter the same?
The flatest system frequency response doesn't always equate into proper phase behavior and best acoustic integration around the crossover point. The system response needs to be checked mid and far field to know for sure if the drivers mesh properly and the driver polarity is set right. I usually verify this by simply listening to how it sounds overall. Your ears won't lie to you. Its not always text book black and white where you can flip the driver phase in relation and obtain 24+ dB of suckout at the xover point.
For a crossover with 2nd order slopes, it's necessary (and normal) to reverse one driver polarity - so, for a 3-way you'd have reverse polarity on the midrange. This is for proper phase-tracking in the crossover region, and avoidance of nulling (which is a feature which can be used to demonstrate an accurate 2nd-order implementation).
Other orders don't feature this need for inversion. However, as others have said, other orders (1st and 3rd) may or may not benefit, owing to the fact that drivers could have significant inductance (for example).
Other orders don't feature this need for inversion. However, as others have said, other orders (1st and 3rd) may or may not benefit, owing to the fact that drivers could have significant inductance (for example).
When deciding on the polarity, measure both, fr response and phase.
Do not complicate decision because its 3-way. First, measure mid and tweeter only, and make decision. Like its 2-way. Measure in both polarities, decide what is better. Once that is done, proceed to measure with woofer, both polarities, compare results, decide.
Easy.
Do not complicate decision because its 3-way. First, measure mid and tweeter only, and make decision. Like its 2-way. Measure in both polarities, decide what is better. Once that is done, proceed to measure with woofer, both polarities, compare results, decide.
Easy.
When I listen to tones, I can hear noticeable nulls around the 300-500hz points when all the drivers are in phase.
When the mid or midbass is inverted everything sounds flat except for around 3000-4000hz where it sounds a bit bright but seems to measure flat enough.
When I listen to music through them everything sounds fine, there is no sibilance or any noticeable problems. But I have only been listening to one speaker.
I only have small rooms to measure these in so all my measurements are gated at -1 to 4.5us so I'm not sure how much I can trust them.
When listening for phase, what should I be listening for?
Do I need both speakers or can I do my listening test using only one?
Does it matter if they're only going to be used nearfield in a small room?
Thank you for all the great advice.
Below is the frequency response with phase measurements:
I'm not sure what to look for here, when listening to tones they sound a bit bright around 3-4k.
Should I still do 2-way measurements? I'm not sure how useful these will be as I think I get overlap between the midbass and the tweeter due to the 6db crossover and the low tweeter crossover point.
sometime I do 'unwrap the phase' selection in upper right corner, to see it as continues line without those jumps up and down
otherwise it seems ok to me...any sudden change indicate wrong polarity selected
if you have little bit too much energy in the 3-4kHz region, it may be too much overlap between tweeter and midrange. Try to mod tweeter crossover a little higher, lower the cap value, that should decrease the 3-4kHz bump and perhaps sound less bright
Switching the polarity of a driver is pretty common and often unavoidable. 6dB/oct isn't the best choice for a 3-way because the huge overlap. Your simulation might look good but that's not gonna cut it because of several things.
1. Your nice simulation does not show the distortion. A midrange dome speaker will distort - even the huge Dynaudio m560d - if it gets down to 300Hz with just 7dB attenuation, that's even below the fs. But much more important: These drivers are ~30 years old and the ferrofluid likely dried up, all the old Dynaudio dome drivers got that problem. The biggest changes are happening close to the fs and since you plan to use all of them at the fs with that 6dB slope, that will be a tough challenge.
2. The Dynaudio drivers are assembled with glue, it will be difficult to open them and replace the ferrofluid. The FF changes its behaviour/damping with rising temperature, so using them near the fs isn't a good idea anyway.
The impedance peak will rise with the temperature (higher Q), the driver will behave differently cold/warm, the crossover will fit one state but never both.
3. Your simulation doesn't show the decay. You will get several serious ridges, the midrange fs ie for sure and tweeter too, most likely on the mid bass also in the upper range.
4. Your nice simulation does not show any dispersion. Each driver is simulated at 0° angle. The mid woofer starts to beam slowly from ~1500 on upwards and over 2k it gets much narrower than the dome drivers. On axis the plot looks good but off-axis they will add up a lot differently and since the bass still works up to 3-4k with very slowly dropping slope, the upper mid and hights will drop off axis much sooner.
To get the simulation to a point it reflects the actual behaviour, you need to do measurements of the drivers also off-axis (at least 15° and 30°), decay and distortion. You probably need to change the the xo to 2nd or 3rd order.
1. Your nice simulation does not show the distortion. A midrange dome speaker will distort - even the huge Dynaudio m560d - if it gets down to 300Hz with just 7dB attenuation, that's even below the fs. But much more important: These drivers are ~30 years old and the ferrofluid likely dried up, all the old Dynaudio dome drivers got that problem. The biggest changes are happening close to the fs and since you plan to use all of them at the fs with that 6dB slope, that will be a tough challenge.
2. The Dynaudio drivers are assembled with glue, it will be difficult to open them and replace the ferrofluid. The FF changes its behaviour/damping with rising temperature, so using them near the fs isn't a good idea anyway.
The impedance peak will rise with the temperature (higher Q), the driver will behave differently cold/warm, the crossover will fit one state but never both.
3. Your simulation doesn't show the decay. You will get several serious ridges, the midrange fs ie for sure and tweeter too, most likely on the mid bass also in the upper range.
4. Your nice simulation does not show any dispersion. Each driver is simulated at 0° angle. The mid woofer starts to beam slowly from ~1500 on upwards and over 2k it gets much narrower than the dome drivers. On axis the plot looks good but off-axis they will add up a lot differently and since the bass still works up to 3-4k with very slowly dropping slope, the upper mid and hights will drop off axis much sooner.
To get the simulation to a point it reflects the actual behaviour, you need to do measurements of the drivers also off-axis (at least 15° and 30°), decay and distortion. You probably need to change the the xo to 2nd or 3rd order.
First-order crossovers are very demanding - it's difficult to avoid significant IMD. Below the pass-band, excursion increases at 6dB/octave - tweeters especially aren't designed for that sort of abuse.
To be avoided!
Edit: Though below the driver's resonance, 12dB/octave is subtracted from that figure - so a net reduction in excursion of 6dB/octave. Maybe an advantage for a driver with a (well-damped) high resonant frequency.
To be avoided!
Edit: Though below the driver's resonance, 12dB/octave is subtracted from that figure - so a net reduction in excursion of 6dB/octave. Maybe an advantage for a driver with a (well-damped) high resonant frequency.
Last edited:
Thank you for taking the time to type out your reply, it's raised issues that I haven't thought about yet.
1. The responses that I posted (other than the xsim screenshots) were measured, are they ok? I'm not sure what a suitable distortion graph would look like, my understanding is there's a lot of second and third order harmonic distortion with first order cross overs? Below is the distortion tab from REW:
2. I've taken one of the t330d's apart before to fix a dented diaphragm, I should hopefully be able to redo the ferro fluid if necessary. Though it was about 10 years ago, from memory there was no glue needed to take them apart, they are screwed down only. I'm assuming the m560d are the same.
What's the best way to check whether the ferrofluid needs to be replaced?
3. The decay tab in REW is blank, is it possible to do decay measurements in a small room (eg gated measurements to avoid reflections)?
4. The speakers will only be used nearfield and on axis in a fairly small room. Does the off axis response still matter? Will it affect imaging/soundstage?
My apologies if the questions are stupid, It's been a while since I've worked on these speakers, I've forgotten a lot and I feel overwhelmed.
The filler driver xover might be of interest (2nd order on woofer and tweeter, mid/filler driver gets a 1st order bandpass). This has been used in early Band and Olufsen speakers (Erik Bækgård filter).
Errata, SRPP, and More Crossovers
Errata, SRPP, and More Crossovers
@cph2000: The Dynaudio t330d as filler driver is a complete waste of the driver and I doubt he wanted to build a semi-2-way just because of the phase. Don't misunderstand me here, you've got an idea and that can be very helpful, especially because almost noone ever got a filler driver on the radar, but that's clearly not what he has in mind as a goal.
@discl: It is very hard to tell if something is right or not if you don't know the circumstances. If that measurement is actually taken at 107dB, then that's excellent! Though, it looks more like it was taken at maybe 70-75dB and the mic isn't calibrated in spl. It's useful to take the measurement at about 1W (2,83V @8 Ohm), it doesn't have to be exact and the spl calibration isn't that important but without any of these informations it's not saying anything at all.
I suggest you start measuring with setting a recording window (length) with FFT (Fast Fourier Transformation) to cancel the room influence or take measurments outside without reflecting surfaces/walls etc. I am sorry, I don't want to explain it all here, I'm sure you'll find it explained here in the forum. I don't use REW but I'm pretty sure though REW needs a windowed measurement to analyze the decay.
You need to post how you are measuring (distance, mic, height in relation to the drivers, driver placements, recording window, voltage/power, etc) and start to measure different angles and don't forget a loopback check.
How far will you be away from the speakers? Sitting on a desk? The dispersion pattern is in nearfield usually much more important because you are very close to the speakers and leaning forward or sideways (mouse, reading etc) result in a huge change of the angle (compared to a normal listening position). That does not only got huge influence on the frequency response but changes the phase a lot too.
In a small room the surroundings reflect the sound much earlier and often worsen the location and soundstage. But even if they are not obstructed or no reflecting surfaces a constant dispersion will always be feneficial as speaker placement becomes much easier. The dispersion doesn't have to be perfect in every regard but you should know what could be a problem or can be neglected. The more you know the better can the speaker be developed to your own needs (not mine or someone elses). I don't know if you love a wide, deep stage or rather prefer the exact placement of the musicians on the stage. You are the one which is okay or in which way you want to go.
If someone wants to help, it's always easier if you know what you want to achieve and tell them. 😉
@discl: It is very hard to tell if something is right or not if you don't know the circumstances. If that measurement is actually taken at 107dB, then that's excellent! Though, it looks more like it was taken at maybe 70-75dB and the mic isn't calibrated in spl. It's useful to take the measurement at about 1W (2,83V @8 Ohm), it doesn't have to be exact and the spl calibration isn't that important but without any of these informations it's not saying anything at all.
I suggest you start measuring with setting a recording window (length) with FFT (Fast Fourier Transformation) to cancel the room influence or take measurments outside without reflecting surfaces/walls etc. I am sorry, I don't want to explain it all here, I'm sure you'll find it explained here in the forum. I don't use REW but I'm pretty sure though REW needs a windowed measurement to analyze the decay.
You need to post how you are measuring (distance, mic, height in relation to the drivers, driver placements, recording window, voltage/power, etc) and start to measure different angles and don't forget a loopback check.
How far will you be away from the speakers? Sitting on a desk? The dispersion pattern is in nearfield usually much more important because you are very close to the speakers and leaning forward or sideways (mouse, reading etc) result in a huge change of the angle (compared to a normal listening position). That does not only got huge influence on the frequency response but changes the phase a lot too.
In a small room the surroundings reflect the sound much earlier and often worsen the location and soundstage. But even if they are not obstructed or no reflecting surfaces a constant dispersion will always be feneficial as speaker placement becomes much easier. The dispersion doesn't have to be perfect in every regard but you should know what could be a problem or can be neglected. The more you know the better can the speaker be developed to your own needs (not mine or someone elses). I don't know if you love a wide, deep stage or rather prefer the exact placement of the musicians on the stage. You are the one which is okay or in which way you want to go.
If someone wants to help, it's always easier if you know what you want to achieve and tell them. 😉
@cph2000 Thanks for the tip, I've had a brief look at the tubecad link you posted and the filler driver thread at "Filler" driver ala B&O and it looks very promising. As @ICG mentioned, its not really aligned with my design goals of having the two t330d/m560d upfront and doing most of the work however the concept looks promising for integrating the AT 23I52 with the t330d which appears to be my main problem at the moment and I'll do some more reading and will run some tests with the Yamanaka crossover from the tubecad link.
@ICG I ran about 12 measurements and all but the last two measured at about 80db, the last two measured at 107. I must have accidentally changed something somewhere however the curves look the same regardless of the gain. My measurement setup is a calibrated umik-1, measurements are taken in a fairly small bedroom at 1m with the speaker on stands 1.2m between the ground and ceiling. I am setting IR windows between -1us and 4.5us in REW. I live close to a busy road and have about 50db of ambient road noise at all times so I'm not sure how much I can trust the measurements however I have run them several times and they are consistent between passes.
I haven't done a loopback check, I'll take a look into it and how to do a FFT.
Thank you again for the wonderful help. I have enough to go on and hopefully I'll be able to sort the crossover out to do the drivers justice.
@ICG I ran about 12 measurements and all but the last two measured at about 80db, the last two measured at 107. I must have accidentally changed something somewhere however the curves look the same regardless of the gain. My measurement setup is a calibrated umik-1, measurements are taken in a fairly small bedroom at 1m with the speaker on stands 1.2m between the ground and ceiling. I am setting IR windows between -1us and 4.5us in REW. I live close to a busy road and have about 50db of ambient road noise at all times so I'm not sure how much I can trust the measurements however I have run them several times and they are consistent between passes.
I haven't done a loopback check, I'll take a look into it and how to do a FFT.
Thank you again for the wonderful help. I have enough to go on and hopefully I'll be able to sort the crossover out to do the drivers justice.