Time coherency (MTM Project)

[Twisted85]

Hi,

I started this project about a year ago, and I'm still not sure if I've made everything as good as I can with this setup.

The speaker is a floorstanding MTM with Dayton RS180 woofers and Peerless HDS tweeter.

Picture 1
Picture 2

At the moment both high and lowpass filters are 2nd order.

Now what I've been thinking here, is the time coherency of this design. The phases seem to align well, but I have to admit that this whole time coherency is a bit unclear to me.

I tried to generate impulse responses with SoundEasy some time ago, but didn't manage to do this. Could someone guide me how to do this?

This time coherency thing has also made me thinking of a modification to the tweeter, because one flaw of this design is the power response. The tweeter and woofer has unequal directivity characteristics at the Xo point around 1700Hz, so I've been thinking of making a small waveguide to the tweeter. This would maybe add a little more directivity to the tweeter low end and make a better match to the woofer high end, and also lead to a better power response. This would also move the tweeter backwards, and affect the impulse response.

Any thoughts?

[Tenson]

Hi,

This might help. http://www.geocities.com/kreskovs/TimeAligned1.html

Adding an 8" waveguide to the tweeter sounds like a good idea, but the draw back is that the woofers will have to be spaced further apart from the tweeter to fit the waveguide. This means you will need a lower xover frequency to maintain good vertical dispersion and driver intergeneration.

I think I would opt to keep the design as you have it, and use an asymmetrical xover to align the phase.

Though, maybe a bi-radial shaped waveguide would help

http://www.p-audio.co.uk/products/db..._14_ph-170.htm

[djarchow]

Are you trying to measure the impulse response in SoundEasy? Is your design from your own measured data of your drivers/baffle etc?

Regards,

Dennis

[Twisted85]

I've made all of my own measurements using a calibrated mic.

Here are the simulated phases:

Red is woofer and green is tweeter and Xo is just under 1800Hz

And both high/lowpass filters are electrically 2nd order.

I'm still quite unsure what effects symmetric/assymmetric crossovers has. I've read the page Tenson supplied, but that really didn't make everything that clear to me.

Does high/lowpass filter cause any real time delay to the signal output from the driver on some frequencys?

I'm sorry, I messed up with the right terms. Instead of impulse response, I meant Step response!

Few months ago I tried to generate the step response using the SE manual, but didn't get nowhere. (I've always found the manual to be bad for people who need more precise guides, like myself)

Maybe I should set up my measuring equipment again have a second try.

I cannot use a big waveguide anymore, as the final cabinets are allready built. (I had prototype cabinets, but as I was happy with the design then, I built the real cabinets).
But I was thinking of replacing the peerless faceplate with custom faceplate with equal diameter and with a waveguide as wide as possible. (there are even smaller waveguides in seas tweeters.) But mayde it doesn't have an affect to frequencys low enough.

[Tenson]

Originally posted by Twisted85


Does high/lowpass filter cause any real time delay to the signal output from the driver on some frequencys?

I don't believe it does.

Its quite tricky to understand because there are many variables that all effect each other.

Lets assume a speaker with aligned acoustic centers of drivers. The crossover filters, such as LR 4th order will add a phase slope to each driver. With a symmetrical crossover and aligned drivers, the phase slppe will be the same for high and low pass, and the phase of each driver will still match together.

However, although the phase of each driver still matches, the phase response of the speaker system as a whole will no longer be linear. The result is that the phase relationship of each frequency is differnt to that of the input, and a differnt waveform is created. You must of course remember that nearly all sounds are made from sine waves that interact with each other with certain phase relationships. You can keep the same energy at each frequency and change the phase relationship and you get a differnt waveform. This speaker would be phase aligned, and time aligned, but due to the over-all phase slope added by the xover it is not transient accurate. There are some transient accurate xover designs but they put big demands of the drivers.

Luckily, our ears don't seem to be able to hear this very easily and regardless of the phase relationships, if each frequency has the same amount of energy then it will sound the same to us.

If we now have a speaker where the acoustic centers of the drivers are not aligned, then the phase slopes of the high and low pass filters will not match up any more. So we can put a slightly differnt slope of filter on each driver to make the phase slopes meet at the crossover point. This is phase aligned but not time aligned.

I hope that makes some sense. Too many people talk about 'delay' in xovers because phase shifts can be related to a time delay, but it is not the same thing. At least as far as I understand it!

[john k...]

Now that everyone it totally confused here is a link to a new page I have been working on with regard to crossover and transient distortion.

[djarchow]

John,

Intersting new page. with a lot of good info. Another good reference, especially for someone new to the topic of transient response, time alignment etc is your "Just A Phase I Am Going Through" paper:

http://www.geocities.com/kreskovs/Phase-B.html

Regards,

Dennis

[Tenson]

Hi John,

Thanks for the link and willingness to share your wisdom.

I have read most of the page now and I have a small question that I hope makes sense! Its hard for me to articulate exactly what my question is, but I will try.

You say on the site that the examples all relate to a system with an acoustic response of the stated filter slopes. I therefor assume that the examples hold true for any system with an acoustic slope as described, regardless of the electrical filtering used to achieve it.

You also say that a system can be broken down to its minimum phase and all pass components.

Now I understand that all frequency amplitude variations will have a 'minimum phase' component. I can also see how an electrical filter network could exhibit an additional all pass component.

Where I stumble is understanding where the all pass component comes from in an acoustic slope, regardless of the electrical filter used to get there. I thought that a purely acoustic slope, like a drivers natural roll off, would only exhibit a minimum phase component, and no all pass.

If this were the case though, the all pass component would be defined by the electrical network alone. I guess this is not the case?

[rcw]

If you measure the acoustic offset with a sine wave you find that as frequency increases so does acoustic offset, this is what you would expect as the outer parts of the cone decouple and the acoustic centre moves towards the apex.

The actual curve this effect produces depends upon the details of the drivers cone shape, size and material, but it is definitely all pass having an inverted logistic function shape.
rcw

[john k...]

Originally posted by Tenson
Hi John,


Where I stumble is understanding where the all pass component comes from in an acoustic slope, regardless of the electrical filter used to get there. I thought that a purely acoustic slope, like a drivers natural roll off, would only exhibit a minimum phase component, and no all pass.

If this were the case though, the all pass component would be defined by the electrical network alone. I guess this is not the case?

Don't confuse the drivers and the system. We are talking about the system response here. The system is broken down in to a minimum phase component which accounts for the low and high frequency cut off of the system (woofer box alignment and high frequency cut off of the tweeter). The all pass response includes all aspects of the crossover required to match the acoustic targetes for the crossover. So it would be the combination of the electrical filters combined with the the characteristics of the woofer above the crossover point and the tweeter below the crossover point. If the woofer and tweeter were flat from DC to infinity then the MP component of the system would be just a flat, zero phase response and the AP component would be that of the crossover, and any eq required if the HP and LP sections did not sum flat.

Most of what I presented there is a rehash of things I have discussed on my old web site but I wanted to add the last section of using physical offset to approximate a transient perfect crossover. While it may yield results which are better than standard crossovers It will always have some degree of transient error, even when mixed order/type HP and LP sections are used.