Hi wesayso
Thank you for taking your valuable time for a comprehensive explanation. It is too much for me to digest all at once but I will surely seek your golden advice when I am ready. I am a novice when it comes to interpreting impulse plots.
My main aim in this thread is to establish whether my woofer and tweeter are time aligned passively. My delay measurements with REW show they reach the mic at the same time but members have rightly pointed out the negative spike suggests otherwise. Furthermore, unlike many here, my spike is not very pretty. It seems to slew rate limit on the way up, followed by a short ring before decaying. I really don't understand what is happening. Is it because my spl was too high in the sweep. Looking at it again, it averages 95dB.
In this digital age, it may seem odd as to why I'm investing my time with out-dated technology. That's because I want to know how my time-aligned speakers compare to something like the JBL 4206 Studio Monitors. Not high end stuff like Dunlavy. How time-aligned are the JBL 4206 anyway? They can't surely be as accurate as those using DSP.
By the way, my mic is 37 ins (0.94m) to the speaker, on tweeter axis.
Hi Michael,
Just consider me another enthusiast like you, that's all I am.
Could you post the IR that belongs to this STEP? On the same scale please.
There are a couple of questions I do have, you showed the speaker responsible for this result. How exactly was the tweeter aligned with the woofer to get the right timing?
What I'm curious about is this post:
What exactly moved here, did the tweeter physically move 20 mm? We see two graphs, is one with filter and one without? And you talk about 0.06 ms corresponding to 20mm.
What I don't get is: what did you do to go from graph 1 to graph 2.
Were both graphs made with the same filters on? Or was one with HP on tweeter and LP on woofer and the other without.
That part was not clear to me. Maybe I'm just slow
. It may also help to know what kind of filter (what order) we are talking about.
I could look at the graph posted before these to figure it out, it's just the talk of that 20mm that confuses me.
Just consider me another enthusiast like you, that's all I am.
Could you post the IR that belongs to this STEP? On the same scale please.
There are a couple of questions I do have, you showed the speaker responsible for this result. How exactly was the tweeter aligned with the woofer to get the right timing?
What I'm curious about is this post:
What exactly moved here, did the tweeter physically move 20 mm? We see two graphs, is one with filter and one without? And you talk about 0.06 ms corresponding to 20mm.
What I don't get is: what did you do to go from graph 1 to graph 2.
Were both graphs made with the same filters on? Or was one with HP on tweeter and LP on woofer and the other without.
That part was not clear to me. Maybe I'm just slow
. It may also help to know what kind of filter (what order) we are talking about.I could look at the graph posted before these to figure it out, it's just the talk of that 20mm that confuses me.
Last edited:
Maybe your "too close" monitoring position gave the weird spikiness?
Does it change as you move the mic further out?
Does it change as you move the mic further up? or further down? and maybe below lining up with the bass/mid driver.
Can you estimate the acoustic centres of the two drivers and then try to get the mic equidistant between the two?
I could be wrong here, but to me it looks like the slopes are approaching 12 dB/oct acoustically.
Based on the STEP you posted I expect the IR to dip down and up again, meaning we have an inverted tweeter that hands over nicely to the woofer.
If that's what we are looking at, we cannot try and compare it to that STEP posted from a speaker using first order slopes. It should not look like that.
Just as Mitch said just below your post, it looks like a second order graph with the inverted tweeter leading the woofer. Nothing wrong with that though. We are after all looking at a passive crossover without DSP.
I'll wait until we see the actual IR belonging to that STEP.
A frequency graph showing phase might help too.
Based on the STEP you posted I expect the IR to dip down and up again, meaning we have an inverted tweeter that hands over nicely to the woofer.
If that's what we are looking at, we cannot try and compare it to that STEP posted from a speaker using first order slopes. It should not look like that.
Just as Mitch said just below your post, it looks like a second order graph with the inverted tweeter leading the woofer. Nothing wrong with that though. We are after all looking at a passive crossover without DSP.
I'll wait until we see the actual IR belonging to that STEP.
A frequency graph showing phase might help too.
Don't upload from a remote server.
Attach your pics/plots using the Forum's preferred method.
How to attach images to your posts.
Hi wesayso
Have fun. Best wishes to every one too for the coming New Year.
Thanks for the tip Andrew. I'm good now.
This is the Time-Aligned Delay plot for Tweeter with HP Filter and Woofer with LP filter.
Red (tweeter) is the reference. Green is woofer.
These are the readings in the side panel.
Both drivers registered 0.56 +- 0.02ms Delay from Acoustic Reference, hence the perfect overlap of the two bottom peaks.
This is the Time Delay between the tweeter and the woofer. For this, I disconnected their respective filters. Power amp is now directly connected to the drivers.
And here are the readings at the side.
It registered a Delay of 0.06ms which works out to 20mm. We can conclude that the acoustic centers of the tweeter and the woofer is 20mm apart.
Without physically offsetting the woofer with a step or tilting the front panel to compensate for the 20mm, with the mic not moved, the signal out of the tweeter with the HP and the signal out of the woofer with the LP are perfectly aligned as shown in the first plot.
I did this strictly with passive crossovers, without any lattice network to delay the tweeter. For this particular design, it is 2nd order for LP, 3rd order for HP. But this is not fixed. I can end up with both 2nd order. But the thing is, I'm able to align the time arrival of the woofer and the tweeter using passive components only.
I will follow up later with the Step Responses. May be a couple of hours or more.
Have fun. Best wishes to every one too for the coming New Year.
Thanks for the tip Andrew. I'm good now.
This is the Time-Aligned Delay plot for Tweeter with HP Filter and Woofer with LP filter.
Red (tweeter) is the reference. Green is woofer.
These are the readings in the side panel.
Both drivers registered 0.56 +- 0.02ms Delay from Acoustic Reference, hence the perfect overlap of the two bottom peaks.
This is the Time Delay between the tweeter and the woofer. For this, I disconnected their respective filters. Power amp is now directly connected to the drivers.
And here are the readings at the side.
It registered a Delay of 0.06ms which works out to 20mm. We can conclude that the acoustic centers of the tweeter and the woofer is 20mm apart.
Without physically offsetting the woofer with a step or tilting the front panel to compensate for the 20mm, with the mic not moved, the signal out of the tweeter with the HP and the signal out of the woofer with the LP are perfectly aligned as shown in the first plot.
I did this strictly with passive crossovers, without any lattice network to delay the tweeter. For this particular design, it is 2nd order for LP, 3rd order for HP. But this is not fixed. I can end up with both 2nd order. But the thing is, I'm able to align the time arrival of the woofer and the tweeter using passive components only.
I will follow up later with the Step Responses. May be a couple of hours or more.
Your 2nd order and 3rd order filter will have different phase lead/lag at the crossover frequency.
That combined with the acoustic roll off that will be present will be quite complicated. Then add on the 20mm difference in acoustic centre distance and you have a complex alignment to analyse.
That combined with the acoustic roll off that will be present will be quite complicated. Then add on the 20mm difference in acoustic centre distance and you have a complex alignment to analyse.
Hi Michael,
Thanks! Very interesting! What program you use for this measurements? But what about sourse phase + amplifier phase + speakers mesuared with microphone(elaktrical + own drivers phase) = ?
Its can be heard? Its worth it? fi at x-over phase -+10 for ahample or +-45 its can be heart? Maybe more important for example responce between 100 -350 hz and 1-4khz?
MIcrophone own phase?
Thanks! Very interesting! What program you use for this measurements? But what about sourse phase + amplifier phase + speakers mesuared with microphone(elaktrical + own drivers phase) = ?
Its can be heard? Its worth it? fi at x-over phase -+10 for ahample or +-45 its can be heart? Maybe more important for example responce between 100 -350 hz and 1-4khz?
MIcrophone own phase?
Last edited:
So you aligned the phase by picking the crossover slopes that would get you alignment.
I'm still wondering about your plots though. Are we looking at a close-up of the IR to view the timing? Or is this a limited signal to perform alignment.
The IR of a filtered tweeter is going to present a different picture than one of a filtered woofer.
To be save one would like to look at what happens at the crossover frequency.
One could use the Filtered IR tab in REW, set a filter around the frequency of interest (for example 1/3 @ 2500 Hz) and jump back to the IR tab.
That would show a (filtered) result at or near the crossover frequency.
With the filters applied, on both woofer and tweeter, you can check the alignment at that frequency by looking at the overlays and comparing those two plots at 2500 Hz. Do they line up? If they do you can be sure the alignment is at that frequency.
I would still recommend checking the alignment at the listening position or at listening distance. It's at that distance you want it to be good, right?
Last edited:
Hi wesayso
Here are the IR and STEP of the tweeter with the HP filter. I'm sorry I didn't post them earlier. I was having issues with my pc. Still trying to sort it out. I'm using my laptop for the time being. So, I may not be able to respond as quickly as I would like to.
I believe that to time-align the woofer and the tweeter, it must be done with their respective networks. If it's simply based on aligning their acoustic centers, they will not be time aligned once their filters are installed because of the delay caused by phase lead and lag. That made me question the common perception of aligning acoustic centers for time alignment. That only holds true with a complete speaker. For example, to time align left and right speakers because both speakers have the same delay.
Regarding the Impulse and Step responses, they are not the FR of the tweeter with the HP filter. It is a high freq beep that is used for time alignment. This is part of REW. When you select REW for Acoustic Timing, it starts the sweep with a beep first. This beep lies somewhere between 2kHz~3kHz. I don't really need to sweep from 20Hz ~ 20kHz. In this instance, it is from 2kHz~3kHz. Upon activation, REW will still do a full sweep but what is needed is this pulse. The FR is not important.
The same procedure is repeated with the woofer+LP filter. The difference between the 2 pulses is the Delay. I may be wrong but I think that's the way REW measures Delay.
IMPULSE RESPONSE OF BEEP
STEP RESPONSE OF BEEP
Here are the IR and STEP of the tweeter with the HP filter. I'm sorry I didn't post them earlier. I was having issues with my pc. Still trying to sort it out. I'm using my laptop for the time being. So, I may not be able to respond as quickly as I would like to.
I believe that to time-align the woofer and the tweeter, it must be done with their respective networks. If it's simply based on aligning their acoustic centers, they will not be time aligned once their filters are installed because of the delay caused by phase lead and lag. That made me question the common perception of aligning acoustic centers for time alignment. That only holds true with a complete speaker. For example, to time align left and right speakers because both speakers have the same delay.
Regarding the Impulse and Step responses, they are not the FR of the tweeter with the HP filter. It is a high freq beep that is used for time alignment. This is part of REW. When you select REW for Acoustic Timing, it starts the sweep with a beep first. This beep lies somewhere between 2kHz~3kHz. I don't really need to sweep from 20Hz ~ 20kHz. In this instance, it is from 2kHz~3kHz. Upon activation, REW will still do a full sweep but what is needed is this pulse. The FR is not important.
The same procedure is repeated with the woofer+LP filter. The difference between the 2 pulses is the Delay. I may be wrong but I think that's the way REW measures Delay.
IMPULSE RESPONSE OF BEEP
STEP RESPONSE OF BEEP
Hi Michael,
I'm glad you made that point clear, as it is crucial to get the alignment right at the crossover point itself. Does REW use the other channel as a reference for this timing?
(one for the actual sweep, the other as a loopback signal as reference)
As even repeated measurements using soundcards, USB or otherwise, can have small deviations in their timing. Using REW with an Asio output this feature is available as a loopback.
It would make sure the timing is the same for every repeat measurement.
I know a couple of features have been added to REW, maybe this loopback isn't needed anymore. I did not look into those features yet as I did not need it myself.
However, the stepped baffle or slanted baffle can have their own advantage for time alignment, for instance the way Troels uses it. With his (preferred) second order symmetrical filters, normally one of the drivers should be inverted to get the needed alignment.
However a physical offset of the tweeter can accomplish the same time-alignment without the need to invert a driver. He has done many speakers this way.
What would be interesting for the documentation of this thread here is the IR and STEP of the complete speaker, without any time alignment signal in advance of the sweep. In other words a more normal sweep from 20-20k, presenting the IR and STEP to see how this asymmetric filter acts both in phase as well as in summing. We could even do a virtual theoretical ideal of that filter to see how that would look.
I call it asymmetric because we have a different order on the tweeter than on the woofer. This in itself does not need to be a problem but it will influence the hand-over from one driver to the next. It would need to be an informed choice to deviate from the more commonly accepted symmetrical filters. It will behave different than a symmetrical crossover.
For instance the Harsch filter has gained in popularity on this forum: S. Harsch XO
This Harsch filter topology has different slopes for woofer and tweeter, a quick note: it does not sum flat at that crossover point. It does have less phase turn than for instance a LR4 and it probably isn't a bad idea to still 'have a steeper filter' on most woofers anyway to suppress any breakup region. It's another alternative which gives a STEP which looks closer to the idealised first order STEP which was presented as an example earlier in this thread. Much more info in that thread.
I'm glad you made that point clear, as it is crucial to get the alignment right at the crossover point itself. Does REW use the other channel as a reference for this timing?
(one for the actual sweep, the other as a loopback signal as reference)
As even repeated measurements using soundcards, USB or otherwise, can have small deviations in their timing. Using REW with an Asio output this feature is available as a loopback.
It would make sure the timing is the same for every repeat measurement.
I know a couple of features have been added to REW, maybe this loopback isn't needed anymore. I did not look into those features yet as I did not need it myself.
However, the stepped baffle or slanted baffle can have their own advantage for time alignment, for instance the way Troels uses it. With his (preferred) second order symmetrical filters, normally one of the drivers should be inverted to get the needed alignment.
However a physical offset of the tweeter can accomplish the same time-alignment without the need to invert a driver. He has done many speakers this way.
What would be interesting for the documentation of this thread here is the IR and STEP of the complete speaker, without any time alignment signal in advance of the sweep. In other words a more normal sweep from 20-20k, presenting the IR and STEP to see how this asymmetric filter acts both in phase as well as in summing. We could even do a virtual theoretical ideal of that filter to see how that would look.
I call it asymmetric because we have a different order on the tweeter than on the woofer. This in itself does not need to be a problem but it will influence the hand-over from one driver to the next. It would need to be an informed choice to deviate from the more commonly accepted symmetrical filters. It will behave different than a symmetrical crossover.
For instance the Harsch filter has gained in popularity on this forum: S. Harsch XO
This Harsch filter topology has different slopes for woofer and tweeter, a quick note: it does not sum flat at that crossover point. It does have less phase turn than for instance a LR4 and it probably isn't a bad idea to still 'have a steeper filter' on most woofers anyway to suppress any breakup region. It's another alternative which gives a STEP which looks closer to the idealised first order STEP which was presented as an example earlier in this thread. Much more info in that thread.
I believe the loopback was previous method. The latest version of REW has a provision for measuring Delay. It simplifies the procedure. You can even use a USB mic. In my case, it was the mic that came with my OmniMic.
MiniDSP has a good writeup regarding this feature.
Measuring Time Delay
Thank you for the link. I am not aware of his time-align designs. I do not visit other designers site mainly due to lack of time. My thread is not meant to disparage his or other designer's work. I am on a personal journey in understanding the
finer aspects of audio engineering. Mainly for self enrichment. 2018 marks my 19th year working full time on audio. The first 10 was in designing power amplifiers. When I was able to command 20K for one amplifier, I knew I had reached
my destination. I then embarked on the next journey, which is the speakers. I am presently at the tail end of this second leg. I will be ready to launch my speakers in 5 yrs time. Whether I want to do that is something I have yet to decide on.
I will be nearly 70 by then. I don't think I want to spend the rest of my remaining years nuturing a business.
Thank you for taking your precious time to review the sweeps. I have attached them at he end. They were all done without the time reference feature (normal 20-20k)
Thank you for the subtle tip. I am aware my crossover is asymmetrical but that's what it took to time-align the woofer and the tweeter. It can be symmetric, it all depends on the transducers. I am very open minded when it comes to problem solving. The science helps in understanding what is happening but if prevailing methods don't solve the problem, I will have to look elsewhere. Admittedly, my method of time alignment is not very scientific. There are no maths involved. It's more like "flying by the seat of your pants" approach. But based on what I did, my phase is aligned and so is the time-arrival at the crossover frequency. I would be grateful for your analysis. Am I right or is there a fundamental flaw in my thinking.
FR of Speaker
Impulse of Speaker
Step of Speaker
Nope. Loopback is active, accurate and easy method. It is standard system in most of the measurement programs I know for measuring IR with acoustical or electrical timing. For example SoundEasy, ARTA, justMLS and REW. Acoustic reference is first aid to support cheap usb mics (and single channel sound cards). Reference sound source needs space, work and time. Extra objects nearby are not good thing while measuring acoustical signals.
Manufacturers of Umik and OmniMic should make products having signal output and fixed internal loop in channel 2. Single channel gear could be useful with HT systems if amplifier does not have analog input.
Last edited:
Thank you for your advice. I will make more measurements using the loop technique. My soundcard is a M-audio Delta 66 I think. Balanced out and balanced in. That should be fine. I will need to get a suitable balanced mic though. Maybe I can adapt my mic from LMS. It has balanced input. I'm not sure whether it requires phantom power. If so, I may just buy a mic instead or maybe run it through a mixer with phantom power. The loop should cancel out whatever is in the chain. Is this loop technique similar to that in RightMark?
I understand your point that this acoustic timing reference method is not as accurate as using the loop. Right now, that level of accuracy is not so important. Proof of concept is my objective.
For example, instead of physically displacing the woofer with a step or tilting the baffle to align their acoustic centers, my work suggests that there are other ways to achieve the same results. This is why I'm presenting my findings here. I would like to invite members to review and even criticize robustly my method. I have no problem if proven wrong. In fact, I'll be grateful because I won't need to pour my time and resources in the wrong direction.
I understand your point that this acoustic timing reference method is not as accurate as using the loop. Right now, that level of accuracy is not so important. Proof of concept is my objective.
For example, instead of physically displacing the woofer with a step or tilting the baffle to align their acoustic centers, my work suggests that there are other ways to achieve the same results. This is why I'm presenting my findings here. I would like to invite members to review and even criticize robustly my method. I have no problem if proven wrong. In fact, I'll be grateful because I won't need to pour my time and resources in the wrong direction.
Very sorry, but different approaches for minimizing excess group delay, linearizing phase response, shaping step response close to ideal etc. has been available for few decades. Someone already mentioned Harsch and I can add Le Cleach to the list of gurus, but everyone can be his own guru: test and optimize the most suitable timing concept for available drivers with a simulator. All choices are not always available due to limits in acoustic slopes, mechanics and filtering. Dual-channel frequency response measurement (with phase) is needed for each driver.
Time alignment changes every single frequency in the band by the same amount.
Phase alignment changes every single frequency by different amounts, depending on the angle and frequency.
Both can be used to align a woofer and tweeter at the crossover point.
In your original image, according to Tom Danley this is not a problem. As long as you are within 1/4 wavelength, it will not be noticeably destructive.
Phase alignment changes every single frequency by different amounts, depending on the angle and frequency.
Both can be used to align a woofer and tweeter at the crossover point.
In your original image, according to Tom Danley this is not a problem. As long as you are within 1/4 wavelength, it will not be noticeably destructive.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.