Play with Crossover Design and Simulation - Free

[ErikSquires]

Oh, yeah, one other thing. In the distance measurement page I used a prototype baffle which I ended up not using. I wanted to flush mount the woofer but it looked like hell

This is why the Xsim file uses 1.1" instead of 1.32" shown in the tutorial. That 0.22" is the difference between flush and surface mounting.

Also, the FR shown in that page was for an earlier, and much worse, crossover. It won't match the LM-1 now.

[bohdan]

[QUOTE=ErikSquires;
When you measure the tweeter and woofer together you now get the interference pattern caused by the woofer's delay, or the red trace as it where. Now this is a little confusing, but Omnimic will see the entire speaker as 1 driver, it can't tell what's woofer and what's tweeter. It has no way of going "Aha! The woofer is 1.8" behind the tweeter!"

Best,


Erik[/QUOTE]

Hi Erik,

Acoustical delay between woofer and the microphone is exactly the same, regardless if you measure woofer as a single driver, or woofer+tweeter together.

I suspect, that your Omnimic is interfering with the time offset, and referencing to the peak of the impulse response for woofer as a single driver.
When you measure woofer+tweeter together, the Omnimic is now referencing to the peak of the tweeter impulse response - thus leaving the woofer peak behind.
This could cause the discrepancy you are seeing between the blue plot and the red plot.

If this is the case, you need to explain this better in your presentation.

Best Regards,
Bohdan

[ErikSquires]

Bohdan,

Either I'm having a seizure, or we are having a major miscommunication. There is no a discrepancy at all, it's an effect of physics we need to rely on to measure. Yes, the woofer does not move towards or away from the microphone. However, Omnimic references it's FRD based on the leading edge. In a 2 way this is the tweeter. When singly, each driver alone is set to 0 time.

Interfereometry is a method of measuring acoustic differences, and quite useful.Jeff Bagby has a PDF on this subject as well:

http://techtalk.parts-express.com/fi...2&d=1367535655

Measuring the correct accoustic offset of drivers is vital in achieving correct phase alignment. Achieving correct phase alignment is vital to having a flat response as you transition from one driver to another. Note that there is no way to find this offset without a microphone. One way or another, either by interferometry or delay measurement, you must actually play a signal through a driver to find it. There may be some ways of ball-parking it if all you have is a ruler though.

If you want to learn more about the idea. create a new project in XSim with two drivers. You don't need data files at all. Just create two drivers, and wire them in parallel. First, examine what happens when you turn one driver on and off, or disconnect it. The amplitude of the FR will go up and down by 6 dB across the board, from 70 to 76 dB.

To the second driver, add an inch of delay. You'll see that instead of purely constructive (additive) interference you now have a combination of constructive and destructive. This is what happens in real life, and is in fact how the drivers would sound like. Getting this right is important for proper driver phase alignment. Examine what happens now when you turn one of the drivers on or off. The FR goes up and down, but now you also see the destructive interference occuring, where one speaker's output is negating the other.

The process used in my tutorial is a way of discovering the relative distance this when you don't know the actual acoustic centers of your drivers and is very accurate.

If you look at this page:

http://speakermakersjourney.blogspot...surements.html

You will see that the second chart shows the "normal" LM-1 and the case when the tweeter is inverted. This is again, using interfereometry, to guage how well we have integrated the two drivers across the frequency and time domain. That dip occurs when we have perfect phase matching at the crossover. If you heard a speaker like this you could in fact hear that missing note. This it not just a measurement artifact but a very real subjective and objective result.


Best,


Erik

Hi Erik,

Acoustical delay between woofer and the microphone is exactly the same, regardless if you measure woofer as a single driver, or woofer+tweeter together.

I suspect, that your Omnimic is interfering with the time offset, and referencing to the peak of the impulse response for woofer as a single driver.
When you measure woofer+tweeter together, the Omnimic is now referencing to the peak of the tweeter impulse response - thus leaving the woofer peak behind.
This could cause the discrepancy you are seeing between the blue plot and the red plot.

If this is the case, you need to explain this better in your presentation.

Best Regards,
Bohdan

[bohdan]

Hi Eric,

".....Yes, the woofer does not move towards or away from the microphone. However, Omnimic references it's FRD based on the leading edge. In a 2 way this is the tweeter. When singly, each driver alone is set to 0 time....."

Yep, This is what is missing from your original description in your blog, as it is this is essential in your method. If you do not provide this information, people wanting to duplicate your results will go and measure separate woofer, separate tweeter and then woofer+tweeter together, and will get different results if you do not specify exactly how the time is referenced in each measurement.

That's pretty much all there was to it.

Best Regards,
Bohdan

[Brian Steele]

Pretty much any change you want to do to a part, if you can make the change, is through the "Tune" menu. Right click on a part and select "Tune."

When you select a driver you'll see the distance entry which is called "mod delay" While you can add delay to any driver, I set the tweeter to 0" and the woofer to 1.1"

Ooh, great stuff. I didn't know that before. I've remodeled my Blastorama V2 update using XSim now because XSim does offer a few advantages that the other x-over modelling tools don't. For example, my high-frequency impedance compensation experiment (my attempt to flatten impedance at high frequencies so the speaker's FR isn't impacted when connected to a cheap class D amp) can be modelled in XSim. And XSim can also model power dissipation across each component in the speaker - very handy when the user is trying to put together a design for pro audio use. The graph below for example shows the power dissipation across the midbass driver, the tweeter and the resistor in the impedance compensation circuit for my Blastoramas, given an input signal equivalent to 75W into 8 ohms. I think that the tweeter is going to be very safe at this level. I'm not so sure about that resistor...!

[ErikSquires]

Hi Brian,

Yeah, examining the power across resistors is one of the features I use a lot. Also good to look at the Voltage transfer graph for tweeters to ensure your crossover is rolling off the bottom end fast enough. I've gotten into trouble when trying add a little low-end boost to a tweeter. I ended up negating most of the effect of the crossover.

One thing it does not do automatically however is calculate music power dissipated. For this I usually set the watts to 10 when looking at the tweeter, then 100 for the woofer if doing a 100W 2-way.

FYI, I was curious about why the resistor Tune window has Watts. According to the author, Bill Waslo, the "Watts" value doesn't do anything. It was put there when he was thinking of selling XSim to part vendors, so you could have it spit out a bill of materials with specific part numbers. It never went anywhere.


Best,


Erik

[ani_101]

Erik and bodhan, I don't use omnimic but Holm impulse. And I take tweeter response and woofer response and then tweeter + woofer. Can I use these three measurements in xsim to get the z offset?

[ErikSquires]

ani,

For a measurement tool, you need to be able to:

- Measure detailed frequency response (around 1/6th to 1/12th octave is what I like to use)
- Measure phase

If you can only take 1 octave measurements then it becomes difficult.

To infer driver distance you need a tool with the following 3 features:

- Ability to sum 2 charts together (tweeter + woofer)
- Ability to set relative distances of the 2 charts you are summing
- Ability to view the summed plut AND The measured combination plot at the same time, so you can keep changing distance until they overlap

In addition to XSim, I believe OmniMic has this feature (but it's more difficult to use) and I'm sure other tools can do this as well.


Best,

Erik

[ErikSquires]

Hi Eric,

".....Yes, the woofer does not move towards or away from the microphone. However, Omnimic references it's FRD based on the leading edge. In a 2 way this is the tweeter. When singly, each driver alone is set to 0 time....."

Yep, This is what is missing from your original description in your blog, as it is this is essential in your method. If you do not provide this information, people wanting to duplicate your results will go and measure separate woofer, separate tweeter and then woofer+tweeter together, and will get different results if you do not specify exactly how the time is referenced in each measurement.

That's pretty much all there was to it.

Best Regards,
Bohdan

Thanks for reading that post and for your suggestion but I'm OK with that. That blog post is meant to be a simple and short step by step tutorial I deliberately omit those discussions which seem to have caused you some confusion.

So long as the steps are right, and the results are correct, I like the post as is.

Best,


Erik

[bohdan]

Hi Erik,


".....I deliberately omit those discussions which seem to have caused you some confusion......"

It's the fact that you "deliberately" omitted the essential information (which is the crux of your method), that is causing the problem - not the other way around.
I am not sure how you can get it so wrong.

The post from "ani-101" and your ( finally more detailed, but still murky ) response to him proves my point - lack of information still confuses some of your readers.
I was able to work out the missing information from your blog, but as you can see, others are still asking questions - if I was the writer, I would respect this.

BTW, impulse response and exact FFT windowing presentation would be very helpful to see.

Anyway, good luck with your blog.

Best Regards,
Bohdan

[dlr]

Erik and bodhan, I don't use omnimic but Holm impulse. And I take tweeter response and woofer response and then tweeter + woofer. Can I use these three measurements in xsim to get the z offset?

If Xsim is capable of 3-dimensional driver positional placement, then yes. I haven't done any testing with it, however.

One thing that I've pointed out a number of times over the years is that if you are using measurements in a software modeling program, you can use those three measurements because the measurement system does not have to be capable of measuring phase. An example of such a system is LMS, part of the LMS/LEAP software package that has been around since the 90's, maybe the 80's. That was the first design software that I used. It uses stepped sine waves for SPL, you then extend the slopes and generate minimum-phase, but there is no measured phase. Yet you can still use that to determine offset because offset is a relative value between driver models, not an absolute offset of any sort. If you were to determine offset, then change the slope of a model for some reason, the offset may no longer be valid and may need to be adjusted for the change in the model.

Ultimately what you are doing is adjusting the relative offset (z-axis only since you know the x- and y-axes) for the driver models that you have created so that the summed raw model result matches the measured summed result.

You can see an example in this article:

Finding Relative Acoustic Offsets Empirically

dlr

p.s. Looking at some older threads (google found this PE thread) and if I understand it, Xsim uses a pure delay that you can enter for drivers. This works as well because all you are essentially doing in a 3-D software system is determining the total delay between two driver models to the point of measurement if you aren't using measured phase. The one difference with pure total delay this is that you can only model the summed response for a single point, but it works for that. So if I understand it correctly, you don't find what is typically called relative acoustic offset, as in the z-axis component, but you do find the total time offset that still works for a single point. It's essentially the same as using two drivers with correct measured phase and leaving the delay at 0, since all relative delay is accounted for in the individual measurements. This only works for a single point in space, however, the actual measurement point. You can't model any off-axis points in this scenario.

[ErikSquires]

Hi Erik,


".....I deliberately omit those discussions which seem to have caused you some confusion......"

It's the fact that you "deliberately" omitted the essential information (which is the crux of your method), that is causing the problem - not the other way around.
I am not sure how you can get it so wrong.

The post from "ani-101" and your ( finally more detailed, but still murky ) response to him proves my point - lack of information still confuses some of your readers.
I was able to work out the missing information from your blog, but as you can see, others are still asking questions - if I was the writer, I would respect this.

BTW, impulse response and exact FFT windowing presentation would be very helpful to see.

Anyway, good luck with your blog.

Best Regards,
Bohdan

Bohdan,

My point in writing the blog was to focus on the steps, as opposed to being an exhaustive manual on the theory. A DIY speaker designer has so much to learn that I make choices to keep things short instead of attempting to put them through a rigorous class.

If you feel a better blog post would be one with theory I encourage you to contribute your own.

Best,


Erik

[ErikSquires]

Sorry, XSim is not capable of modelling speakers in 3D. Only a Z axis is part of the modelling equation.

Compared to more elaborate modelling systems, this does mean you will be unable to model FR at any given point. You can only model it for the point of the microphone.

However, for hobbyists this is plenty to start with.


Best,


Erik

[viyohepav]

hi eric, tq verymuch
if i may suggest, can you do a video tutorial with example for a dumb like me will be perfect.
i tried a view free software including excell spreadsheet out there but i always got stuck.
i hope this software will be something better.
maybe a video example for crossover design for high pass or low pass, and
a video example for crossover design for with 2 way TM.

[ErikSquires]

I"m surprised if there are not already very good video tutorials here at PE.

I can however offer you a little help with my blog on crossover basics.

http://speakermakersjourney.blogspot...er-basics.html

I suggest you try those examples with XSim first. Play with the parts, and play with the blocks. Get a visceral feel for how the components fit together. Once you get past that, read about speaker distance. After that, try the A26 crossover simulation, or the LM-1.


Best,


Erik