Can anyone help with Alignment?

[levimax]

Thanks for taking a look.

Can the delays not be estimated by viewing the overlayed Impulse response?

View attachment 479333

I read elsewhere that the delays can be set by aligning the rise of the tweeter and mid with the woofer?

Being a DIY speaker designer is hard and has a learning curve. Trying to do "timing" measurements in room is one of the hardest things. I recently went through this and used 6 different time alignment methods and got 6 different results. I definitely do not have all the answers but a few things to consider.

1. What Mic are you using? If a UMIK-1 it probably is not accurate enough for this type of work. A UMIK-2 or a "regular MIC" using a 2 channel interface so you can do loopback measurements will work. See link for REW designers comments https://www.avnirvana.com/threads/acoustic-phase-measurements.13300/post-99941

2. To align you need to do identical full range sweeps of the drivers without any filters. The REW alignment tool works well for this for the "mid" to "Tweeter" but woofer alignment is extremely difficult with "in-room" measurements no matter what anyone says. If possible timing measurements are best done "outside" in a large open space as far above the ground as possible but often this is not possible. If not I would try several different methods on the woofer (REW alignment tool, REW Wavelet tool, overlayed impulse method (the problem is what do you align? front? first crossing, peak, ?) , nearfield woofer and mid impulse response and then a tape measure and calculator. See if you can get any consistency between the different methods and if not try to figure out why and correct.

Once you get the timing sorted you can start with the crossover filters. As mentioned something is not right with the filters you have now. Good luck and have fun.

 

[Flaesh]

I’m trying to setup a DIY 3 way speaker.

Will the crossovers be active?
What horn will be used for the HF144? It can handle lower than 1k.
The Sovereign's whizzer is, in my opinion, unnecessary and can even be harmful for such use.
I'm not sure the TL16H is truly necessary; but that's a matter of personal preference and setup specifics.
And I personally use a subwoofer system for home hi-fi))

 

[Keith_W]

This is what you need to do, in this specific order. The reason it needs to be done in this order is because each step affects the next. For example, you can not do time alignment first and then apply the XO later, because XO's rotate phase and mess up your time/phase alignment!

1. First, decide if you will be working in linear-phase or minimum-phase. The procedure is slightly different for each. If it's MiniDSP, passive XO, etc it's minimum-phase. I will proceed assuming that you will be working with minimum-phase. In general, linear-phase is much simpler because there are fewer moving parts. You don't have to worry if it will sum properly - it will always sum properly.

2. Measure each individual driver with no crossover under free-field conditions (i.e. no room reflections) to confirm that the drivers in your home perform the same as published specifications, and that the left and right drivers are the same. Depending on QC some manufacturers are known to ship duds and there is always inter-sample variability. These are electromechanical devices, after all. Mount the drivers in your speaker cabinet, place the speaker in the middle of your room (if you have a large room) or take them outside. Place the mic 1m on-axis to each driver, then sweep each driver across its usable range.

3. Using this information, generate appropriate crossovers for each driver. If you want to linearise the driver, do it now. You might need to linearise the driver if (a) left and right driver pairs do not measure the same, (b) you want to account for the baffle step, (c) the frequency response is not flat across the bandpass. There are additional considerations when selecting XO points and slopes, e.g. some XO configs don't sum properly, where the Fs is, how much power handling, etc. In general, distortion is highest around the driver's Fs, so we usually apply the HPF an octave above the Fs.

4. Convolve the crossover with the measured driver response (REW Trace Arithmetic A*B), then sum the woofer, mid, and tweeter (REW Trace Arithmetic A+B) to see if there are any summation errors. If the XO's do not sum to flat, then additional measures are necessary. You could (a) consider choosing a different XO slope (b) design an all-pass filter to improve summation (c) consider inverting the polarity of one of the drivers. Dealing with summation errors is quite complex if you are working with minimum-phase, there is an entire chapter dedicated to it in Vance Dickason's book. I can not reproduce all that information here, sorry. For a start, I would recommend LR4, but the best config depends on your specific situation.

4. Once your drivers are individually linearised AND with the XO in place, put the mic 2-3m away and adjust the gains of each driver so that the result sums to approximately flat.

5. NOW you time and phase align them. There are many methods, which I'm not going to go through because I can see from your measurements that you haven't done all the previous steps properly.

 

[kemmler3D]

@Keith_W has delivered some good advice. I was just going to point out that you actually need to do time alignment last, because your xover filters will tend to affect phase considerably unless you go all FIR. Don't worry about the time alignment until you're completely happy with the overall frequency response.

 

[ppataki]

This is what you need to do, in this specific order. The reason it needs to be done in this order is because each step affects the next. For example, you can not do time alignment first and then apply the XO later, because XO's rotate phase and mess up your time/phase alignment!

1. First, decide if you will be working in linear-phase or minimum-phase. The procedure is slightly different for each. If it's MiniDSP, passive XO, etc it's minimum-phase. I will proceed assuming that you will be working with minimum-phase. In general, linear-phase is much simpler because there are fewer moving parts. You don't have to worry if it will sum properly - it will always sum properly.

2. Measure each individual driver with no crossover under free-field conditions (i.e. no room reflections) to confirm that the drivers in your home perform the same as published specifications, and that the left and right drivers are the same. Depending on QC some manufacturers are known to ship duds and there is always inter-sample variability. These are electromechanical devices, after all. Mount the drivers in your speaker cabinet, place the speaker in the middle of your room (if you have a large room) or take them outside. Place the mic 1m on-axis to each driver, then sweep each driver across its usable range.

3. Using this information, generate appropriate crossovers for each driver. If you want to linearise the driver, do it now. You might need to linearise the driver if (a) left and right driver pairs do not measure the same, (b) you want to account for the baffle step, (c) the frequency response is not flat across the bandpass. There are additional considerations when selecting XO points and slopes, e.g. some XO configs don't sum properly, where the Fs is, how much power handling, etc. In general, distortion is highest around the driver's Fs, so we usually apply the HPF an octave above the Fs.

4. Convolve the crossover with the measured driver response (REW Trace Arithmetic A*B), then sum the woofer, mid, and tweeter (REW Trace Arithmetic A+B) to see if there are any summation errors. If the XO's do not sum to flat, then additional measures are necessary. You could (a) consider choosing a different XO slope (b) design an all-pass filter to improve summation (c) consider inverting the polarity of one of the drivers. Dealing with summation errors is quite complex if you are working with minimum-phase, there is an entire chapter dedicated to it in Vance Dickason's book. I can not reproduce all that information here, sorry. For a start, I would recommend LR4, but the best config depends on your specific situation.

4. Once your drivers are individually linearised AND with the XO in place, put the mic 2-3m away and adjust the gains of each driver so that the result sums to approximately flat.

5. NOW you time and phase align them. There are many methods, which I'm not going to go through because I can see from your measurements that you haven't done all the previous steps properly.

This is a great list (as always ), fully agree!

Just a dummy question regarding point #5:
Let's say we apply min phase filters - if we set the delay values with all the LPFs and HPFs on, wouldn't that cause misalignment in the summation that was properly configured in the preceding steps?
It is kind of like a chicken and egg question....

 

[Keith_W]

This is a great list (as always ), fully agree!

Just a dummy question regarding point #5:
Let's say we apply min phase filters - if we set the delay values with all the LPFs and HPFs on, wouldn't that cause misalignment in the summation that was properly configured in the preceding steps?
It is kind of like a chicken and egg question....

You will notice that in the previous steps, the drivers were measured independently and not time or phase aligned to each other. They are effectively speakers all on their own. The summation between drivers is assumed to be correct at some arbitrary point, independent of the physical location of the driver on the baffle. Once all the corrections are done, then we time/phase align them to find that arbitrary point.

 

[OCA]

@Keith_W is your go to person with driver alignment and crossovers, you are lucky because it's tough to even know where to start with these things. All I can contribute to you for some time saving is that your tweeter seems like it needs to be inverted relative to woofer and mid.

 

[Robbie010]

Thanks to you all.

I have taken some native measurements this morning, with out any crossover filters applied. For the time being I think I need to ignore the woofer response up to around 500hz, it isn't correct as I am waiting on a new front baffle and some acoustic filler for the cabinet but you get the idea:

Looking at these responses I'm thinking either 2.7khz or 4khz crossover between mid and tweeter and either 750hz or 1.2khz between woofer and mid.

 

[Flaesh]

What horn, LTH142 or another?..

 

[Robbie010]

What horn, LTH142 or another?..

18sound XT1464

 

[Keith_W]

Please post the MDAT with all your drivers. I can see that your L/R tweeters don't match, and I would like to take a closer look. Are you sure you took the measurements correctly? You don't want to correct a measurement artefact, this is absolutely critical.

Also, are your speakers ported or not?

 

[voodooless]

Realistically, you don't need the Visaton TL16H at all; the HF144 will do well enough. At the very least, you should cross if much higher than the 3.2 kHz you have now. 5 or 6 kHz at a bare minimum. And why use a full-range driver together with a large horn? Why cross the large XT1464 horn at only 1.2 kHz? The horn and driver can easily do 800 Hz or so. But directivity-wise, that's not the best either. A 12" horn would have been a better choice.

 

[Flaesh]

18sound XT1464

Strange FR. At what distance was it measured?
The usual approach is to use the horn as low as is appropriate for the particular horn and driver.

 

[Robbie010]

1. MDAT attached.

2. I went with the full range 12" woofer to give me more options on potential crossover points with the horn.

3. The 12" Fane is (will be) in a sealed 55L cab, the horn and tweeter are freestanding.

4. The measurements were all taken at 1m distance with the driver raised off the floor as much as is practical in my domestic living space.

5. No crossover settings were applied to these measurements

 

[Sokel]

These are the fairly unsmoothed responses with MTW applied (so gated up high)

It seems you have your hands full with it, I don't see in-room measurements that can help you, you'll have to take them out if you want reliable data.

 

[voodooless]

Why does the Visaton tweeter drop down past 15 kHz? According to its datasheet, it should not. The Faital seems to match well with the datasheet. The woofer, who knows, I think there is a bit too much room there, maybe add some windowing to remove some of the reflections.

 

[levimax]

While you are getting good advice especially from @Keith_W, the problem is you are trying to use in room measurements for speaker design which doesn't work.

While there is a big learning curve you may want to take a look at VituixCAD. https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_help_20.html The guide walks you through all the steps of speaker design with taking accurate measurements (both on and off access) being the first step and hardest part. You may just want to read the measurement part if you don't want to go the whole CAD route. The really nice thing is if you can get good measurements into the program you will be able to relatively easily optimise the design (which is an unusual design so many normal rules of thumb won't apply).

Good luck and have fun.

 

[Keith_W]

All right, let's take a look.

The first thing to realize is that there is a wavelength limit where you are able to take a free-field measurement (i.e. no reflections). This limit is determined by proximity of the speaker and microphone to reflective surfaces, including floor and ceiling. Any wavelength longer than this limit will be irretrievably contaminated by reflections. You can find out what this limit is by looking at the Energy-Time Curve. I have overlaid ALL your measurements, and we can see the first reflection arriving at 3.52ms (arrowed). If one period of the frequency in question equals the time-arrival of the reflection, then the limit of your measurement is 284Hz (f = 1000/3.52). This means that if you apply a frequency-dependent window of 1 cycle, you have a clean measurement suitable for DSP from 284Hz up.

I have also gone through all your waterfall plots to make sure you have taken good measurements with adequate SNR. They are all excellent. Well done.

I looked at all the distortion plots to see if you would benefit from crossing over above a certain frequency. Ignore the high "distortion" in the woofer below 50-60Hz, that's an artefact and it's not real. You can see that these are very good drivers (at least when measured at 81dB). Distortion is commendably low, below the 1% threshold. Well, the tweeter has high-ish distortion at the lower end of its operating range and it might benefit from being high-passed at 3kHz or so.

I agree with @voodooless' comment that your tweeter does not match the datasheet. There is a whopping 10dB drop between 10kHz and 20kHz in both tweeters! Very disappointing. I would go so far as to say that if the measurement is correct, these tweeters are unusable. What good is a tweeter that does not tweet?

If one tweeter was OK and the other was not, I would tell you to return the bad tweeter. But for both tweeters to be bad suggests either you made a mistake with your measurement (off-axis measurement, wrong calibration file, wrong microphone orientation, faulty microphone, etc) or the manufacturer is lying. Before you kick up a stink with the manufacturer, let's make sure it's not you first. Make sure your mic is pointed on-axis to the tweeter. Examine your calibration file and make sure it is non-inverted. Try to borrow a mic and repeat the measurement.

 

[Sokel]

Here's a comparison gating the measurements with MTW:

(also shown the settings, it's a fairly new addition to REW)


They short of show how useless they can be in room but one can get some useful data if measuring procedure is sound.
But still, without a spin fairly anechoic it will only show 30% of the speaker if not less.

 

[Robbie010]

Wow, thanks guys, thats great stuff.

I’ll try re-measuring the tweeters again and post the plot and MDAT.