3 way project with twin woofers + port. Using LspCAD.
Currently doing nearfield measurements of port + woofer1 + woofer2 then merging them to the farfield measurement to achieve a combined low-end response.
In my LspCAD filter project I'm specifying only one driver in the drop-down since logically I'm using a combined low-end response, that is, like one big bass driver.
Correct, or are there better ways to do this? I'm thinking that this approach causes some issues with specifying Dx values as I have to average the position of the low end drivers + port and regard it as a virtual driver for the Dx position.
Currently doing nearfield measurements of port + woofer1 + woofer2 then merging them to the farfield measurement to achieve a combined low-end response.
In my LspCAD filter project I'm specifying only one driver in the drop-down since logically I'm using a combined low-end response, that is, like one big bass driver.
Correct, or are there better ways to do this? I'm thinking that this approach causes some issues with specifying Dx values as I have to average the position of the low end drivers + port and regard it as a virtual driver for the Dx position.
IF the drivers are identical there should be no issue of having them in a combined space however you need to account for port size if ported.
However, if the speakers are going to be out in the room and not right against the wall it would be wise to use one woofer as a baffle step compensation so it might be crossed over at a higher frequency. IF so it might be better to separate the woofers to avoid muddy mid bass region.
IF the woofer are not identical, my feeling is that they should be in thier own enclosures and each aligned accordingly.
However, as far as the crossover is concerned I don't think you can treat the two drivers as a single driver. But then again someone with more crossover design experience will have to guide you in this regard.
However, if the speakers are going to be out in the room and not right against the wall it would be wise to use one woofer as a baffle step compensation so it might be crossed over at a higher frequency. IF so it might be better to separate the woofers to avoid muddy mid bass region.
IF the woofer are not identical, my feeling is that they should be in thier own enclosures and each aligned accordingly.
However, as far as the crossover is concerned I don't think you can treat the two drivers as a single driver. But then again someone with more crossover design experience will have to guide you in this regard.
Yes you can.....
In fact if you are using speakerworkshop, you pretty much have to treat them as a single driver, since the optimizer will not work on multiple drivers. Unless you are doing a 3.5 ways as the poster above suggests, treat them as one with a single FR and their combined impedance for modeling purposes. They get very tough to measure indoors....you will have to turn the whole box sideways to get anything accurate at all.
Dick
In fact if you are using speakerworkshop, you pretty much have to treat them as a single driver, since the optimizer will not work on multiple drivers. Unless you are doing a 3.5 ways as the poster above suggests, treat them as one with a single FR and their combined impedance for modeling purposes. They get very tough to measure indoors....you will have to turn the whole box sideways to get anything accurate at all.
Dick
The best way to measure the drive units in LspCAD I have found anyway, is pretty much what you are doing now, BUT in reverse. The absolute phase of the tweeter/mid and woofer is not important. What is vital however is that the phase/drivers are measured all with respect to one another.
In otherwords in far field place the mic infront of the tweeter, measure it, and set the mic distance (software) to something reasonable so that the phase looks nice. Then lower the mic infront of the mid, measure, then the woofer, measure, but do not alter the mic distance either in software or physically. Dont forget to take a measurement of all the drivers connected together.
Those measurements are your reference and all near field measurements need to be done to match those. Take a near field measurement of the woofer and then when you splice the far field and near field together, alter the 'phase' with JMLS time delay thing on the near field to fit smoothly with the phase of the far field. Unless you are going to be doing bass EQ on the speaker measuring the port probably isnt required for crossover work, as any crossovers will be working well out of the ports effective range.
What you effectively have now is measurements for your three drivers in the box they will be used in. When importing the data into LspCAD you set up the drive units exactly as they are in the speaker, so indeed you set the bass drivers to "two" and select either parallel or series connection. LspCAD will make the required changes to impedance and SPL. After setting the X and Y parameters to correspond with your box, load up that measurement you made of all the drivers together as the target response. Now alter the 'Z' dimension and possibly the individual SPL of the drivers till you get as close a match as you can. IE you are getting LspCADs simulated response with the four drivers, to match up with the actual measured combination of the four drivers.
Once both the simulated and the actual measured response are as close to one another as they can be (this should be pretty accurate) you can start to deisgn your filters.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=59804
Further down in that post shows the measured response vs simulated response of two drivers, to show exactly what I mean.
WIth regards to bafflestep I am of course assuming the bass drivers are identical. You should have put in a step frequency using Just MLS when making your original measurements.
I am assuming you can cross the mid over ~250hz. If the baffle width is say 250cm then roughly speaking by 250hz the BS transition will be as good as over. Your bass drivers in parallel will at this frequency and below will radiate pretty much omni directionally and thus be 6dB down across their entire passband. Any bafflestep compensation is ignored for the bass driver as its playing out of the BS range. This means the midrange is playing throughout the entire BS transition and ofc something needs to be done here. This is where it becomes nice and simple, you can compensate for the BS loss through the midrange with the highpass placed on it.
So to sum up.
Bass drivers are say 88dB sensitive. wired in parallel give 94dB, after bafflestep they are placed back at 88dB. Or in real world terms more like 90 dB due to some room gain.
Midrange driver is 90dB sensitive to start with, you apply 4dB of lift through its lower range with the highpass to compensate for BS loss and it fits perfectly with the bass drivers.
Thats just one example, you should be able to correct for battle step loss, if you cross in the right places, simply with the highpass/low pass on the midrange and bass units.
What drivers is it you are going to be using?
I know I wrote a lot, and probably overlapped with a lot of what you already know, hope it helps answer your question.
Matt
In otherwords in far field place the mic infront of the tweeter, measure it, and set the mic distance (software) to something reasonable so that the phase looks nice. Then lower the mic infront of the mid, measure, then the woofer, measure, but do not alter the mic distance either in software or physically. Dont forget to take a measurement of all the drivers connected together.
Those measurements are your reference and all near field measurements need to be done to match those. Take a near field measurement of the woofer and then when you splice the far field and near field together, alter the 'phase' with JMLS time delay thing on the near field to fit smoothly with the phase of the far field. Unless you are going to be doing bass EQ on the speaker measuring the port probably isnt required for crossover work, as any crossovers will be working well out of the ports effective range.
What you effectively have now is measurements for your three drivers in the box they will be used in. When importing the data into LspCAD you set up the drive units exactly as they are in the speaker, so indeed you set the bass drivers to "two" and select either parallel or series connection. LspCAD will make the required changes to impedance and SPL. After setting the X and Y parameters to correspond with your box, load up that measurement you made of all the drivers together as the target response. Now alter the 'Z' dimension and possibly the individual SPL of the drivers till you get as close a match as you can. IE you are getting LspCADs simulated response with the four drivers, to match up with the actual measured combination of the four drivers.
Once both the simulated and the actual measured response are as close to one another as they can be (this should be pretty accurate) you can start to deisgn your filters.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=59804
Further down in that post shows the measured response vs simulated response of two drivers, to show exactly what I mean.
WIth regards to bafflestep I am of course assuming the bass drivers are identical. You should have put in a step frequency using Just MLS when making your original measurements.
I am assuming you can cross the mid over ~250hz. If the baffle width is say 250cm then roughly speaking by 250hz the BS transition will be as good as over. Your bass drivers in parallel will at this frequency and below will radiate pretty much omni directionally and thus be 6dB down across their entire passband. Any bafflestep compensation is ignored for the bass driver as its playing out of the BS range. This means the midrange is playing throughout the entire BS transition and ofc something needs to be done here. This is where it becomes nice and simple, you can compensate for the BS loss through the midrange with the highpass placed on it.
So to sum up.
Bass drivers are say 88dB sensitive. wired in parallel give 94dB, after bafflestep they are placed back at 88dB. Or in real world terms more like 90 dB due to some room gain.
Midrange driver is 90dB sensitive to start with, you apply 4dB of lift through its lower range with the highpass to compensate for BS loss and it fits perfectly with the bass drivers.
Thats just one example, you should be able to correct for battle step loss, if you cross in the right places, simply with the highpass/low pass on the midrange and bass units.
What drivers is it you are going to be using?
I know I wrote a lot, and probably overlapped with a lot of what you already know, hope it helps answer your question.
Matt
Thanks, Matt, for the detailed reply.
Three things:
1. Are you saying I should measure only one woofer and allow lspcad to calculate the gain for the second parallel woofer?
Yes, both woofers are the same, Vifa M22, and they are in parallel.
Re baffle step, I'm crossing to a Peerless HDS134 5.25 mid in a sealed enclosure that's rolling it off at around 300hz, so yes, that's a good spot for BSC.
2. I can see your point about the port not needing measuring, but while I'm there I may as well, and add it to the woofer response matching it up with the woofer output below resonance as suggested by the JustMLS manual.
3. Re measuring all drivers: your method calls for measuring each driver on it's axis. So, where are you measuring all drivers together? On the tweeter axis? If that's the case shouldn't all drivers be measured on the tweeter axis, individually and together?
Thanks for your reply.
Mos
NB: FYI, measuring outside. Tweeter is XT25.
Three things:
1. Are you saying I should measure only one woofer and allow lspcad to calculate the gain for the second parallel woofer?
Yes, both woofers are the same, Vifa M22, and they are in parallel.
Re baffle step, I'm crossing to a Peerless HDS134 5.25 mid in a sealed enclosure that's rolling it off at around 300hz, so yes, that's a good spot for BSC.
2. I can see your point about the port not needing measuring, but while I'm there I may as well, and add it to the woofer response matching it up with the woofer output below resonance as suggested by the JustMLS manual.
3. Re measuring all drivers: your method calls for measuring each driver on it's axis. So, where are you measuring all drivers together? On the tweeter axis? If that's the case shouldn't all drivers be measured on the tweeter axis, individually and together?
Thanks for your reply.
Mos
NB: FYI, measuring outside. Tweeter is XT25.
Exactly.
I measure all the drivers together, on the axis I intend to listening to the speakers.
The drivers should be measured on axis, LspCAD takes into consideration the off axis response when you set up the X and Y parameters. As far as I remember you also need to state the measurement distance somewhere when trying to match the simulation to the measured combination.
If for example you will listen on the midrange axis, the midrange would take the position 0,0,0 (x.y.z) and everything else should be altered relative to it.
If you were to measure the speakers at the distance you were going to listen to them at, then by all means you could measure all the drivers on the tweeter axis, or the axis you intend to listen to them on. The reason why you dont do this; if the mic is at 60cm the angle created with the bass drivers (relative to the mic at tweeter level) is likely to be perhaps 30-45 degrees, whereas when listening its more like 15 (relative to your head), thus altering the off axis performance and what your filter will be designed around.
Of course this is implying the off axis response is going to be an issue. At 300hz the off axis and on axis response of the 8" will probably be identical, so measuring at the listening axis (not listening distance) wouldnt be a problem. The same goes for the tweeter/mid, the angle there is probably going to be very low , 5 degrees or so, if that, and will have little impact on the off axis response.
If you choose to measure all the drivers in one place, then make sure the driver piston radius is set to a very small value, as you dont want LspCAD trying to compensate for offaxis performance too, you already did enough of that when measuring. The X and Y values must be added in though as its important for the crossover design.
Where measuring like this might become an issue is if the crossover was going to be in a position where the drivers off axis and on axis performance was significantly different. This I'snt desirable as it indicates the final off axis and on axis response will be different, which isn't a good thing.
I have done both. Measured all drivers on one axis and measured all drivers on their own axis, the result was pretty much identical.
I am sure someone around here with infinitely more knowledge then me, can come up with a reason for not measuring all the drivers on the same axis (apart from the LspCAD room simulation), but from measurements I've made, its had very little impact on the final result, if any at all.
Thanks, Matt.
Since I've been doing some measuring on the same axis and getting less than good results when compared to the combined response, I'll try a fresh approach and measure on axis for each driver.
So, plan X:
JustMLS:
Measure nearfield response of one woofer & port. Add.
Merge with woofer on-axis farfield response.
Measure mid & tweeter on axis.
Measure combined response (this time with both woofers in parallel) on mid axis.
In LspCAD:
Set bass driver dialog to 2 drivers, parallel
Set mid as 0,0,0 and other drivers relative to that until matched to combined response.
Does it matter if the driver radius is set before or after d,x,y positions? I assume before as we are trying to match with combined response and driver size will affect that response.
Mos
Since I've been doing some measuring on the same axis and getting less than good results when compared to the combined response, I'll try a fresh approach and measure on axis for each driver.
So, plan X:
JustMLS:
Measure nearfield response of one woofer & port. Add.
Merge with woofer on-axis farfield response.
Measure mid & tweeter on axis.
Measure combined response (this time with both woofers in parallel) on mid axis.
In LspCAD:
Set bass driver dialog to 2 drivers, parallel
Set mid as 0,0,0 and other drivers relative to that until matched to combined response.
Does it matter if the driver radius is set before or after d,x,y positions? I assume before as we are trying to match with combined response and driver size will affect that response.
Mos
I would start with the tweeter.
Get the mic positioned at say 60-80cm (whatever distance you are using for the farfield.
Measure the tweeter, set the mic distance in MLS to a value around the measurement distance. The phase should look simple and uncomplicated. No more then two-three major jumps from -180 to +180 degrees (vertical line) within the tweeters range, lets say from 1khz up to 20khz.
Save the response as S1 in MLS.
After doing that lower the mic down to the midrange axis. Hit measure. Hit save to S2. Thats it nothing more nothing less.
Then connect all four drivers in parallel, hit measure. Save it to S3.
Then lower the mic down to the woofer axis (bare in mind the distance to the loudspeaker must be kept identical for each measurement), hit measure. Hit save to S4.
Move the mic nearfield (2-3cm from woofer).
Recall S4 and look at it.
Have a look at the phase of the woofer, note a point of interest, say there is a -180 to 180 degree jump at 800hz, keep that in mind.
With the mic now 2-3cm away hit measure. Then alter the mic distance in MLS so that you try and get a -180 to 180 jump at around 800hz. Save this to S5.
Recall S2 and inspect the midrange phase for a similar point.
Measure the near field of the midrange and again, alter the mic distance in MLS trying to get a reasonable match to the point you remembered. Save it to S6.
Recall S1, export it as the tweeter response.
Recall S3 export it as the combined response.
Recall S2 and select the merge option. Recall S6 as the lower half. Set the frequency you wish to merge the two at, this can be anything if they match up well. Then alter the lower halfs parameters (this is vital DO NOT TOUCH the upper half) so that the phase transition from the upper and lower part is as smooth as can be. Perform the merge. Now put in the centre frequency for bafflestep. Export this as the midrange response.
Now recall S4 and do the same as detailed for the midrange, but using S5 as the lower response. As the tweeter data has been exported you can save this combined response under S1.
If you wish to add in the port data, take a measurement of the port and save that under S3. Recall S1 and now you want to 'add' the response of the port S3 to that of S1. THere is no real need to measure the port and its just going to add in additional complications, but you can do it if you want. Export S1 as the woofer measurement.
Now you should have four sets of saved FR data. The tweeter, the midrange, the woofer and the four drivers combined.
The midrange driver had the nearfield added in as the lower half, with the lower half altered accordingly so that the frequency, phase and dB level all made a smooth transition. You then added in bafflestep.
The woofer was done in an identical way.
Now, open up LspCAD.
Import the driver data. Select two drivers for the bass and connect them in parallel.
The midrange paramaters should read 0,0,0.
Alter each drivers X and Y position relative to the midrange point.
Set the cone radius accurately, this is not 10cm for the bass driver. Measure the actual diameter of each cone and devide by 2.
I use LspCAD pro 5.25 so do not know version 6. I dont know what version you have. BUT..
Open up the general options, there is a part that pertains to the setting of the Measurement distance. Deslect infinate and alter this to the value you took your far field measurements at.
Now import S3/ the combined response as the target.
Play around with the Z values until the Simulated and mesaured response match up. You *may* need to alter the level of the tweeter or the bass drivers to get the match perfect, but this shouldnt be by much.
Hopefully after having done all that you will have ended up with a reasonable match. The mid/tweeter match is the most important to get correct, the mid to bass crossover is at a much lower frequency where the wavelengths are far longer, meaning there is more room for "error" in the crossover.
Once thats done, hide the target. Change the measurement distance to the listening distance and get designing your crossover
Get the mic positioned at say 60-80cm (whatever distance you are using for the farfield.
Measure the tweeter, set the mic distance in MLS to a value around the measurement distance. The phase should look simple and uncomplicated. No more then two-three major jumps from -180 to +180 degrees (vertical line) within the tweeters range, lets say from 1khz up to 20khz.
Save the response as S1 in MLS.
After doing that lower the mic down to the midrange axis. Hit measure. Hit save to S2. Thats it nothing more nothing less.
Then connect all four drivers in parallel, hit measure. Save it to S3.
Then lower the mic down to the woofer axis (bare in mind the distance to the loudspeaker must be kept identical for each measurement), hit measure. Hit save to S4.
Move the mic nearfield (2-3cm from woofer).
Recall S4 and look at it.
Have a look at the phase of the woofer, note a point of interest, say there is a -180 to 180 degree jump at 800hz, keep that in mind.
With the mic now 2-3cm away hit measure. Then alter the mic distance in MLS so that you try and get a -180 to 180 jump at around 800hz. Save this to S5.
Recall S2 and inspect the midrange phase for a similar point.
Measure the near field of the midrange and again, alter the mic distance in MLS trying to get a reasonable match to the point you remembered. Save it to S6.
Recall S1, export it as the tweeter response.
Recall S3 export it as the combined response.
Recall S2 and select the merge option. Recall S6 as the lower half. Set the frequency you wish to merge the two at, this can be anything if they match up well. Then alter the lower halfs parameters (this is vital DO NOT TOUCH the upper half) so that the phase transition from the upper and lower part is as smooth as can be. Perform the merge. Now put in the centre frequency for bafflestep. Export this as the midrange response.
Now recall S4 and do the same as detailed for the midrange, but using S5 as the lower response. As the tweeter data has been exported you can save this combined response under S1.
If you wish to add in the port data, take a measurement of the port and save that under S3. Recall S1 and now you want to 'add' the response of the port S3 to that of S1. THere is no real need to measure the port and its just going to add in additional complications, but you can do it if you want. Export S1 as the woofer measurement.
Now you should have four sets of saved FR data. The tweeter, the midrange, the woofer and the four drivers combined.
The midrange driver had the nearfield added in as the lower half, with the lower half altered accordingly so that the frequency, phase and dB level all made a smooth transition. You then added in bafflestep.
The woofer was done in an identical way.
Now, open up LspCAD.
Import the driver data. Select two drivers for the bass and connect them in parallel.
The midrange paramaters should read 0,0,0.
Alter each drivers X and Y position relative to the midrange point.
Set the cone radius accurately, this is not 10cm for the bass driver. Measure the actual diameter of each cone and devide by 2.
I use LspCAD pro 5.25 so do not know version 6. I dont know what version you have. BUT..
Open up the general options, there is a part that pertains to the setting of the Measurement distance. Deslect infinate and alter this to the value you took your far field measurements at.
Now import S3/ the combined response as the target.
Play around with the Z values until the Simulated and mesaured response match up. You *may* need to alter the level of the tweeter or the bass drivers to get the match perfect, but this shouldnt be by much.
Hopefully after having done all that you will have ended up with a reasonable match. The mid/tweeter match is the most important to get correct, the mid to bass crossover is at a much lower frequency where the wavelengths are far longer, meaning there is more room for "error" in the crossover.
Once thats done, hide the target. Change the measurement distance to the listening distance and get designing your crossover
Thanks very much, Matt, you've given me everything I was looking for.
I did consider measuring the nearfield of the mid, even though it rolls off pretty well 2nd order below 300hz. Never a bad thing to be too accurate.
I'm sure many others will benefit from your feedback. Thanks again. I'll let you know how I go this weekend.
NB: I use LspCAD 5.25, with the Behringer ECM8000 & UB802 mic amp/mixer. Seems like a good combo.
I did consider measuring the nearfield of the mid, even though it rolls off pretty well 2nd order below 300hz. Never a bad thing to be too accurate.
I'm sure many others will benefit from your feedback. Thanks again. I'll let you know how I go this weekend.
NB: I use LspCAD 5.25, with the Behringer ECM8000 & UB802 mic amp/mixer. Seems like a good combo.
An Alternative.......
the measuring method described is pretty conventional, and nicely offered to beginners. A method of extracting what I assume is acoustic center information is offered:
"Play around with the Z values until the Simulated and mesaured response match up. You *may* need to alter the level of the tweeter or the bass drivers to get the match perfect, but this shouldnt be by much.
Hopefully after having done all that you will have ended up with a reasonable match. The mid/tweeter match is the most important to get correct, the mid to bass crossover is at a much lower frequency where the wavelengths are far longer, meaning there is more room for "error" in the crossover."
There is another way to do this, which is also in common use. Place your microphone a little farther away from the speaker (1-2 metres) and position it exactly on the intended listening axis. Take an FR of all the drivers in sequence without moving the speaker or microphone, or adjusting anything else in the setup.
These measurements will accurately show you the phase relationship between the drivers without resorting to the "fitting" exercise described above. Since you are designing the speaker on some specific axis, the microphone positioning dictates the phase between the drivers....on that axis of interest.
Now, it is possible that the measuring system used in the method exhibits too much "launch jitter", that is, sample to sample variuations in the latency of the start of the pulse, for this to work. Mine does allow this method though.....convince yourself by taking 10 samples in a row and seeing of the phase varies by more than a small amount between samples.
Dick
the measuring method described is pretty conventional, and nicely offered to beginners. A method of extracting what I assume is acoustic center information is offered:
"Play around with the Z values until the Simulated and mesaured response match up. You *may* need to alter the level of the tweeter or the bass drivers to get the match perfect, but this shouldnt be by much.
Hopefully after having done all that you will have ended up with a reasonable match. The mid/tweeter match is the most important to get correct, the mid to bass crossover is at a much lower frequency where the wavelengths are far longer, meaning there is more room for "error" in the crossover."
There is another way to do this, which is also in common use. Place your microphone a little farther away from the speaker (1-2 metres) and position it exactly on the intended listening axis. Take an FR of all the drivers in sequence without moving the speaker or microphone, or adjusting anything else in the setup.
These measurements will accurately show you the phase relationship between the drivers without resorting to the "fitting" exercise described above. Since you are designing the speaker on some specific axis, the microphone positioning dictates the phase between the drivers....on that axis of interest.
Now, it is possible that the measuring system used in the method exhibits too much "launch jitter", that is, sample to sample variuations in the latency of the start of the pulse, for this to work. Mine does allow this method though.....convince yourself by taking 10 samples in a row and seeing of the phase varies by more than a small amount between samples.
Dick
I believe doing it like this was mentioned, just not with the mic that far away, but it was said that a further distance (listening distance) would be prefered. Thats the way I tend to measure as it doesn't involve moving anything
lazy! Although with the mic around 80-100cm as the room the PC is in for measuring doesnt allow for more space.I think you misunderstand the fitting I described. This is just done once on the very first measurement, so that you get a nice phase response on the tweeter. IE sometimes when I measure a tweeter and put in the mic distance in, I will get a couple more jumps from -180 to 180 then I would like. So i alter MLS software distance (usually not by much) to get a nice smooth phase from +180 at say 1khz drifting diagonally down to -180 at around 20khz or beyond.
The phase response of the other drivers isnt altered a single bit, its all done with respect to this first measurement of the tweeter.
Matt
Re: An Alternative.......
This is how I've been trying up until now, with less than great results. The phase appears to move quite a lot, and even when I get good readings, the individual driver measurements never satisfactorily line up with the combined measurement.
This method is simpler than the one Matt outlined for me as you're not moving the mic.
dickmorgan22 said:
This is how I've been trying up until now, with less than great results. The phase appears to move quite a lot, and even when I get good readings, the individual driver measurements never satisfactorily line up with the combined measurement.
This method is simpler than the one Matt outlined for me as you're not moving the mic.
Re: Re: An Alternative.......
Its interesting because the method I described still relies on the fact you get consistency between the measurements. After all everything is measured with respect to the first driver, all phase being relative to it.
If you do as Dick mentioned
and end up with less then acceptable consistency then I'd be at a loss on how to get very accurate results.
You will of course still have the summed response of all the drivers, so you *know* how they should combine. The phase of the drivers should be accurate just not precise, as in the reference point is moving, if I understand right, this is the launch jitter. Thinking about it, altering the 'Z' values should get you pretty close to the target response even if there is a small amount of launch jitter.
The only issue here would be your reference point. In the program 0 0 0 is the midrange, you are of course expecting, in the measurement of all the drivers, for the midrange to sit on 0 0 0 too. If there is launch jitter, your measurement of the midrange could be "out" by a few ms in comparison to the summed measurement. So in actual fact the midrange should be sitting on 0.0.5 say. Or more precisely the combined measurement should be sitting at 0.0.-5. If you get what I am saying. This shouldn't really make a huge difference either, I wouldn't have thought, everything else being good.
My initial problem was understanding how to take measurements so the phase of each driver was relative to one fixed point (the tweeter in my case). After I solved that, everything else fell into place and I've never had a problem since. With either measurement technique.
It may just require some trial and error.
Mos Fetish said:
Its interesting because the method I described still relies on the fact you get consistency between the measurements. After all everything is measured with respect to the first driver, all phase being relative to it.
If you do as Dick mentioned
and end up with less then acceptable consistency then I'd be at a loss on how to get very accurate results.
You will of course still have the summed response of all the drivers, so you *know* how they should combine. The phase of the drivers should be accurate just not precise, as in the reference point is moving, if I understand right, this is the launch jitter. Thinking about it, altering the 'Z' values should get you pretty close to the target response even if there is a small amount of launch jitter.
The only issue here would be your reference point. In the program 0 0 0 is the midrange, you are of course expecting, in the measurement of all the drivers, for the midrange to sit on 0 0 0 too. If there is launch jitter, your measurement of the midrange could be "out" by a few ms in comparison to the summed measurement. So in actual fact the midrange should be sitting on 0.0.5 say. Or more precisely the combined measurement should be sitting at 0.0.-5. If you get what I am saying. This shouldn't really make a huge difference either, I wouldn't have thought, everything else being good.
My initial problem was understanding how to take measurements so the phase of each driver was relative to one fixed point (the tweeter in my case). After I solved that, everything else fell into place and I've never had a problem since. With either measurement technique.
It may just require some trial and error.
I never noticed that the thread I linked to was a discussion of you trying to solve the very same problem heh. You seemed to arrive at a successful measuring technique then, so know roughly what you are aiming for. And I imagine, having solved it before, are now wondering why its not working.
I hope you manage to work it out.
I hope you manage to work it out.
Re: Re: Re: An Alternative.......
But, that's adjusted by the dx, dz values, or do you mean something else?
My main problem has been getting a good match between combined and individual reponses. That must go back to the measuring of the individual drivers and subsequent manipulation of the data.
5th element said:
But, that's adjusted by the dx, dz values, or do you mean something else?
My main problem has been getting a good match between combined and individual reponses. That must go back to the measuring of the individual drivers and subsequent manipulation of the data.
Phase Jitter is a drag.....
Some measuring systems seem to offer consistent phase measurements, and others do not. In order to get mine to give me consistency that I can trust, I discovered that a stripped down PC worked best, and NO network connection during measurements. Whether this will help you or not I cannot say. Some PC's are beyond redemption, it depends how the drivers for the audio are working.
If you cannot get consistent measurements, then you are forced into the "dithering" method, which was described. Shoot all three drivers in sequence, as well as a composite of all at once. Your phase data is crap, but that won't stop you.
There are an infinite number of ways in which three drivers will add up to a composite FR, but really only only one way where the phases are within pi radians of each other, that will add up to the composite FR for all of them at once. The method described where you remove the excess phase from all three drivers the same amount, then commence modifying TWO of them from the Z perspective (ie, add or subtract distance) until the three add up to your composite - all drivers chirped - FR, will work. Its a pain in the butt, but it works. The absolute result you have is irrelevent, but use that relative result when modeling your XO and you are good to go.
Good luck,
Dick
Some measuring systems seem to offer consistent phase measurements, and others do not. In order to get mine to give me consistency that I can trust, I discovered that a stripped down PC worked best, and NO network connection during measurements. Whether this will help you or not I cannot say. Some PC's are beyond redemption, it depends how the drivers for the audio are working.
If you cannot get consistent measurements, then you are forced into the "dithering" method, which was described. Shoot all three drivers in sequence, as well as a composite of all at once. Your phase data is crap, but that won't stop you.
There are an infinite number of ways in which three drivers will add up to a composite FR, but really only only one way where the phases are within pi radians of each other, that will add up to the composite FR for all of them at once. The method described where you remove the excess phase from all three drivers the same amount, then commence modifying TWO of them from the Z perspective (ie, add or subtract distance) until the three add up to your composite - all drivers chirped - FR, will work. Its a pain in the butt, but it works. The absolute result you have is irrelevent, but use that relative result when modeling your XO and you are good to go.
Good luck,
Dick
Excess Phase........
Whether this construct makes sense in your measuring system, I do not know. In SpeakerWorkshop, it does. Excess phase is effectively related to the time of flight from the speaker to the microphone. During this "delay", the signal will experience a number of complete rotations or phase wraps. This appears in your display as the phase rotating N times from 20-20000 Hz.
By setting a start gate, I can control the time of flight the software assumes, and reduce the nuber of wraps. I can also remove N mSec of "excess phase" from any arbitrary data set with a command. You probably can too.
You never HAVE to do any of this, it makes it easier to read your graphs if there are two phase wraps rather than 71. Just make sure that your remove exactly the same amount of excess phase from all signals....no more or no less. If you do not process them identically, the relative phase between them will be destroyed, and this is the information you are actually interested in.
Clear as Mud?
Dick
Whether this construct makes sense in your measuring system, I do not know. In SpeakerWorkshop, it does. Excess phase is effectively related to the time of flight from the speaker to the microphone. During this "delay", the signal will experience a number of complete rotations or phase wraps. This appears in your display as the phase rotating N times from 20-20000 Hz.
By setting a start gate, I can control the time of flight the software assumes, and reduce the nuber of wraps. I can also remove N mSec of "excess phase" from any arbitrary data set with a command. You probably can too.
You never HAVE to do any of this, it makes it easier to read your graphs if there are two phase wraps rather than 71. Just make sure that your remove exactly the same amount of excess phase from all signals....no more or no less. If you do not process them identically, the relative phase between them will be destroyed, and this is the information you are actually interested in.
Clear as Mud?
Dick
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.