My first try at designing something

[Dan B]

I'm using the PE data files in Passive Crossover designer. The woofer is the 4inch Goldwood 290-373 and the tweet is Dayton 275-070. The tweeter has a 10uf cap, and is padded down with a 3.3 resistor in series and a 2.7 parallel. The woofer has a 1.5 inductor in series and a 15.o cap parallel. For the box I put the PE Data into WinISD and thought a closed box of 7L would work out ok.

Any thoughts?

Heres what it looks like> <A HREF="http://speakerbuilder.blogspot.com/2005/09/my-first-try.html">http://speakerbuilder.blogspot.com/2005/09/my-first-try.html</A>

[cjd]

1. Do you have any baffle step compensation?

2. Did you extract minimum phase (Hilbert transform)?

3. Did you set the driver offset parameters?

Since the drivers are measured on-axis every time, you really need to do the Hilbert transform and set driver offset close to accurately or you're going to discover that the drivers do not integrate as you think.

If you have no baffle-step, you're going to have a *very* forward and brash speaker - it already is a bit peaky through the upper midrange. Or, another way of thinking of no baffle-step compensation is that you're going to have no bass.

Already commented, but the area through driver integration is a bit hot. A small inductor on the tweeter may help things a LOT by softening the low edge of the tweeter response.

C

[curt_c]

Dan, I would suggest you avail yourself to the BDS and FRC programs over at the FRC consortium. You can ‘massage’ PE’s files to add the diffraction effects of the baffle, sum them to PE’s anechoic responses, and derive minimum phase of the result. Then you can import these files to the PCD and have a pretty accurate scenario of what you can expect. The only fly in the ointment is you have to guess the difference in acoustic centers, which has been suggested to be close to the difference in distance between the flanges and front plates of the two drivers in question.

I also see you are crossing the tweeter at about 1500 Hz. The silkie will be much happier if you cross it higher, say above 1800Hz. This will likely require additional components in the tweeter network, but this will preclude distortion due to the over excursion of the tweeter dome.

C

[cjd]

At least, excursion based. :)

RS28A showed itself to maybe have some issues where I was trying to cross (theoretically, a 1400Hz target with a very steep slope... 3-way at HTGuide). Sounds like there's a fan between me and the speaker. :) I'm not convinced everythign is what it's pretending to be though and have some new hardware on the way to hopefully provide me with better measurements.

You're right though, that's definitely a low point for that setup. Of course, I was more subtle in my suggestion ('cause adding that inductor would also help out excursion issues here...)

I know, however, that you *can* do pretty darned well just using the FRC/BDS/PDC3 tools. I'm pretty sure you used those (or similar) for the Triune, and I did that RS150 MTM that way as well. :)

C

[pin2hot]

Chris & Curt,
The phrases "extract minimum phase (Hilbert Transform)" and "derive minimum phase of the result" exceed my current vocabulary. Any thoughts on a source to get up to speed on this concept?

Heath

[curt_c]

Yup! If you want to hear Diana Krall sound just like Janis Joplin, just cross the tweeter low…

And yes, just how low can one cross the RS28A, and just how consistent are they? Mine tolerate a 3rd order @ 1650 Hz with aplomb, yet Dennis finds my design for Brian, 5th order @ 1500 Hz to be unacceptable. It would be interesting to find out the cause of this apparent issue. Are we just pushing the envelope too far, or is it something simple like variations in height of the felt pad behind the dome? Granted, excursion increases to the square of the frequency, but this seems like more of a driver variation issue.

C

[cjd]

> And yes, just how low can one cross the
> RS28A, and just how consistent are they?
> Mine tolerate a 3rd order @ 1650 Hz with
> aplomb, yet Dennis finds my design for
> Brian, 5th order @ 1500 Hz to be
> unacceptable.

I wish he would try Jon's version (also hold out hope he might try my version, but much less so).

Odd beasts to be sure.

In my case, I'm wondering if things aren't crossing lower than they seem due to some odd measurements. So, that road first.

C

[pin2hot]

Curt, could you elaborate on estimating the difference in acoustic centers of two speakers as being "suggested to be close to the difference in distance between the flanges and front plates"? I think that I understand the "flange" as being the speaker mounting flange, but don't get the "front plate".

Thank you, Heath.

[dlr]

>It would be interesting to
> find out the cause of this apparent issue.
> Are we just pushing the envelope too far, or
> is it something simple like variations in
> height of the felt pad behind the dome?
> Granted, excursion increases to the square
> of the frequency, but this seems like more
> of a driver variation issue.

I suspect that it's the latter. I had one RS28 in a pair that sounded fairly normal when tested with an MLS signal. It's surprising how much one can tell audibily from a test tone at times, hollowness, low end extenstion, etc.

However, it wasn't until I ran a swept-sine impedance test that one of them made a problem obvious. Above about 800Hz all was fine. But I always sweep from 20Hz up, even on tweeters. There was distinct distortion starting around 400Hz (IIRC) that lasted up to about 800-900Hz. Fairly strong.

I swapped coils. The problem stayed with the motor. The tolerances have to be fairly good given the usage of the dual gap copper plates. Apparently the problem was rubbing of the former and/or coil. I wasn't able to determine the exact problem (concentricity, oval shape, etc.) of the gap.

I know that Darren has been addressing these things and is likely getting better product now, but only sampling could determine that.

Pushing the crossover down to 1400Hz is a bit problematic, but that's true of most tweeters. Only the best never have an issue when doing that.

dlr

[dlr]

that PE replaced the problem unit with no complaint. Excellent customer support!

dlr

[paul_n_oneal]

Hmmm, what about all those without measuring equipment that might not be aware of this problem? I just ordered 2 RS28's and now you have me worried.... Paul.

[curt_c]

http://www.partsexpress.com/resources/speaker.gif


For each individual driver frequency response (SPL) there exists a unique phase response. The Hilbert transform derives phase from the SPL plot that is minimum phase with respect to the acoustic center of the driver. This acoustic center is some magic place in the driver where the wave front is initiated. More about acoustic centers later…

When a driver is measured with a mic, the phase response is the same, but includes the additional phase wrap due to the microphone position not being at the acoustic center of the driver. For example, measuring the phase of a driver at 1 kHz, and one meter away would include almost another 3 complete 360 degree cycles of phase (Called phase wrap due to the way its charted on the plots.) as 1 wl at 1 kHz is roughly one third of a meter.

Now this ‘excess’ phase can be left as is or electronically removed in the measurement software, but it doesn’t really matter as when we measure the drivers on the baffle, it is all at the same position with respect to the baffle. Along this ‘design axis’ the relative phase between each driver is measured with respect to each other, so their phase relationships are accurate. If you measured two different drivers at two different times, you lose this relationship. Even if you put the mic ‘exactly’ the same difference from the baffle each time, the difference in acoustic centers in each case will be different, so you have no common reference between the two measurements. Phase measurements made in this instance mean precisely nothing.

Using the Hilbert transform gives us a common reference; the acoustic center of the driver. Unfortunately, since no one has been able to determine the absolute acoustic center of any driver, we have to ‘guess’ its position on each driver. If we guess correctly, the phase response of the drivers are then referenced to each other (with respect to the baffle) similar as measurement with a mic.

How to guess? Vance Dickason suggests the acoustic center of a driver is somewhere around the front plate. The ‘front plate’ is the plate between the basket and the magnet.

Measure the distance from the basket flange of the woofer to the front plate. (In the diagram below this would be distance 'c' subtracted by disctance 'e'.) Do the same to the tweeter, then subtracts the tweeter distance from the woofer. This is an approximation of the difference in acoustic centers of the drivers in question. This difference is usually entered into the crossover software as the ‘z’ location of the woofer,and simulates the woofer acoustic center being behind the tweeter with respect to the baffle.

Well, I guess it wasn’t very short, was it?

C

[unknownuser]

> Curt, could you elaborate on estimating the
> difference in acoustic centers of two
> speakers as being "suggested to be
> close to the difference in distance between
> the flanges and front plates"? I think
> that I understand the "flange" as
> being the speaker mounting flange, but don't
> get the "front plate".

> Thank you, Heath.

The front plate is the steel plate on the front of the magnet or the side close to the cone/flange. It usually lines up with the voice coil


(Originally posted by: Isaac)

[Jeff B.]

[dlr]

> Hmmm, what about all those without measuring
> equipment that might not be aware of this
> problem? I just ordered 2 RS28's and now you
> have me worried.... Paul.

As I said, I know that he's addressed the issue and pressed to have the quality control improved. Their track record is good. My case could easily be one of very few. I think that the few reports of any problems should show that the numbers are small.

I wouldn't be too concerned at this point. If you think that there is a problem, I'm sure that PE will respond in a positive way.

dlr

[dlr]

> Now this ‘excess’ phase can be left as is or
> electronically removed in the measurement
> software, but it doesn’t really matter as
> when we measure the drivers on the baffle,
> it is all at the same position with respect
> to the baffle. Along this ‘design axis’ the
> relative phase between each driver is
> measured with respect to each other, so
> their phase relationships are accurate. If
> you measured two different drivers at two
> different times, you lose this relationship.

It is possible to reproduce nearly identical measurements at different times.

> Even if you put the mic ‘exactly’ the same
> difference from the baffle each time, the
> difference in acoustic centers in each case
> will be different, so you have no common
> reference between the two measurements.
> Phase measurements made in this instance
> mean precisely nothing.

This does not have to be the case. The measurements, if taken using the same measurement software and hardware (that is, the hardware latency is identical) and the start time marker is set precisely to the same point, then the phase relationships are maintained perfectly.

Of course other hardware, such as a preamp and mic, would also need to be calibrated, but this will correct for any variances in them.

This is how I do it for the measurements that I have posted at my site. It is possible to take any woofer, midrange and tweeter combination and use them directly and have accurate phase relationships by modelling them and realigning the modelled phase to match the measurments.

However, the deviations from baffle influence makes that difficult. One would have to run the files through the FRD tools, then readjust the delays to re-align the phase near the top end for a tweeter. The mid and woofer should be easier since measurements encompass much of the lowpass rolloff whereas we have to guess more with the tweeter. But it is doable.

> Using the Hilbert transform gives us a
> common reference; the acoustic center of the
> driver. Unfortunately, since no one has been
> able to determine the absolute acoustic
> center of any driver, we have to ‘guess’ its
> position on each driver. If we guess
> correctly, the phase response of the drivers
> are then referenced to each other (with
> respect to the baffle) similar as
> measurement with a mic.

> How to guess? Vance Dickason suggests the
> acoustic center of a driver is somewhere
> around the front plate. The ‘front plate’ is
> the plate between the basket and the magnet.

My suggestion is to use the point where the former attaches to the diaphragm. I spent considerable time trying to determine absolute AC, but it wasn't repeatable enough and was extremely time consuming. But all results indicated it to be slightly in front of this attachment point.

Without measurements, though, any of these methods can be difficult. The problem is that so much depends on the SPL model used to generate the Bilbert phase. I would add that someone without measuring equipment should ensure that the lowpass (high frequency) rolloff should be extended, especially for the midrange and woofer units, since they rolloff close enough to the typical crossover areas that they'll influence the relative offset more than the tweeter lowpass will.

A 12db slope is probably the best assumption to make for the lowpass.

dlr

[pin2hot]

Thank you Curt. An elegant explanation of a complex problem. Maybe you can use this for the intro to your next book.

Heath

[pin2hot]

> The front plate is the steel plate on the
> front of the magnet or the side close to the
> cone/flange. It usually lines up with the
> voice coil

Thenk you Isaac. Without this forum and the folks that contribute to it, diy speakers would be a near impossible struggle.

[Dan B]

I haven’t done any of these things. It looks like have some reading to do. I’ll try the BDS and FRC software as Curt suggests as well.

Thanks Dan