Early Stages of RS-100 Design

[HareBrained]

Hi Y'all,

I had decided to build a speaker system around the Dayton Audio RS100. I like the openess of small drivers and their enclosures will present a very small cross-section while having a high WAF. It also means it needs to be a 3-way, MTMW. I spent some time reading about different designs and conversing with some people on the board. I was able to narrow my design down to an ND tweeter and the DC160 woofer.

I used some of the tools from Jeff B. to model the drivers, using data from ZaphAudio and datasheets, and here's what I got:

The first set is not my final design because of the ~2.25 ohm impedance but I was impressed at how sensitive these modeled. The configuration is 2 RS100S-8 and 2 DC160S-8 with the ND20FA-6 tweeter. Although I did model the RS100S-8 with other tweeters, this one definitely integrated the easiest. The others were complex to implement or wouldn't mesh with the RS xo.




Another reason to show these is that the DC160 FR is not the modeled box but the straight FR. I was pleased that the simple 2nd-order xo resulted in a smooth slope despite the crazy IB midrange FR.

The design I think I'm going to pursue uses a single DC160S-8, 2 RS100-4 and the ND28FA-6. The sensitivity of this combo modeled out to a reasonable 85db with an easy load for the amp. The system response uses the vented enclosure FR and Z models from Passive Crossive Designer, but I went back and forth with the driver data to ensure the models were coherent in the representations as I tweaked to the xo design. Again, I modeled the RS100-4 with all of the tweeters I was considering but the ND28 fell into this union very naturally.




The crossover for the DC160 consists of a simple 2nd-order cap-coil with l-pad, the pair of RS100's also use 2nd-order cap-coil on both ends and a series RC, and the ND28 is a 3rd-order with a series resistor. The mids and tweeter are wired with reverse polarity. All values are within what would be consider "normal" based on other designs. I include a BSC because the baffle design isn't set and I will be using these in close proximity to walls and furniture.

So, for the data I have, I believe the design is sound. My next step is to buy components (with a few extras for tweaking), measure the actual drivers and see how accurate this data was.

If you have any comments or think I've messed up somewhere, please let me know. I'm just learning this craft.

-John

[Wolf]

Hi Y'all,

I had decided to build a speaker system around the Dayton Audio RS100. I like the openess of small drivers and their enclosures will present a very small cross-section while having a high WAF. It also means it needs to be a 3-way, MTMW. I spent some time reading about different designs and conversing with some people on the board. I was able to narrow my design down to an ND tweeter and the DC160 woofer.

I used some of the tools from Jeff B. to model the drivers, using data from ZaphAudio and datasheets, and here's what I got:

The first set is not my final design because of the ~2.25 ohm impedance but I was impressed at how sensitive these modeled. The configuration is 2 RS100S-8 and 2 DC160S-8 with the ND20FA-6 tweeter. Although I did model the RS100S-8 with other tweeters, this one definitely integrated the easiest. The others were complex to implement or wouldn't mesh with the RS xo.




Another reason to show these is that the DC160 FR is not the modeled box but the straight FR. I was pleased that the simple 2nd-order xo resulted in a smooth slope despite the crazy IB midrange FR.

The design I think I'm going to pursue uses a single DC160S-8, 2 RS100-4 and the ND28FA-6. The sensitivity of this combo modeled out to a reasonable 85db with an easy load for the amp. The system response uses the vented enclosure FR and Z models from Passive Crossive Designer, but I went back and forth with the driver data to ensure the models were coherent in the representations as I tweaked to the xo design. Again, I modeled the RS100-4 with all of the tweeters I was considering but the ND28 fell into this union very naturally.




The crossover for the DC160 consists of a simple 2nd-order cap-coil with l-pad, the pair of RS100's also use 2nd-order cap-coil on both ends and a series RC, and the ND28 is a 3rd-order with a series resistor. The mids and tweeter are wired with reverse polarity. All values are within what would be consider "normal" based on other designs. I include a BSC because the baffle design isn't set and I will be using these in close proximity to walls and furniture.

So, for the data I have, I believe the design is sound. My next step is to buy components (with a few extras for tweaking), measure the actual drivers and see how accurate this data was.

If you have any comments or think I've messed up somewhere, please let me know. I'm just learning this craft.

-John

The peak in the rolloff of the midrange needs to be addressed. It is cancelling with the tweeter, to make the null in the FR directly above it. Also- what's your acoustic-phase and reverse-null look like? That would help.

I would guess that you didn't factor in offsets or BSC, or extract minimum phase for the files.
WRT offsets:




The Response Modeler 2.0 extracts minimum phase, and can help with the bass alignment and BSC.
Later,
Wolf

[nick29498141]

Is that Fostex driver in your "CTC" pic that great one I somehow missed out on for Theta?

Nick

[Wolf]

Is that Fostex driver in your "CTC" pic that great one I somehow missed out on for Theta?

Nick

Yes it is! It's the "Sarlac" driver you almost had. I just inverted the pic in Paint. It was the only driver with a decent angle to match the picture angle on the tweeter I wanted for the pic.

Now- What's the tweeter, wiseguy?
Later,
Wolf

[HareBrained]

Thanks for the feedback.

My simulation did account for the driver mounting architecture. I don't have exact driver measurements but they're closed than nothing at all. I said I wouldn't account for the BSC until I decided on the cabinet dimensions. I don't know what "exact minimum phase" is nor do I know what to do with it. Could you please explain?

The small rise in the midrange around 14kHz contributes at most -.5db to the tweeter. The tweeter contributes the rest. I had considered modeling a notch filter but decided I'd wait until the actual drivers were measured and then see what happens. I may wait until I hear them.

Regarding phase, the acoustic phase goes a full 720. The reverse null is -40db on the woofer-mid xo point but only about -15db on the mid-tweeter xo point. I'll post the graph when I get the chance.

I'll search for Response Modeler and see what it can do.

Thanks,
-John

[dthomas]

If you traced your frequency and impedance data and did not do a Hilbert transform to extract minimum phase then PCD is not giving you an accurate representation. Because it is assuming that phase is 0 at all points for all drivers.

Jeff's response modeler program has the ability to extract minimum phase from traced data. You simply load the traced file into the program and tell it to extract minimum phase then resave the file. This is a Hilbert transform and is a very accurate model of the phase of the driver assuming all your files are level matched. Magnitude, phase and the relative differences in acoustic center of the drivers are all equally important in determining how the drivers interact. So leaving the phase data out is not giving you an accurate model of what to expect.

As far as the interference the mid is causing at 14KHz it is hard to say if it is audible but it would be simple enough to correct. I doubt you need a notch filter from the look of the slope on the mid it is a second order filter going third order would likely be more than enough to eliminate this issue and only cost you one additional component.

Download Jeff's Response Modeler program it will do this for you in about 60 seconds and you will have a more accurate model of your system.

Dave

[Chris Roemer]

Two things to try:

One very simple (and inexpensive) notch to try is to put a tiny cap (0.10, 0.22, 0.33uF) across the inductor on the low pass filter portion of your midrange XO.

Also, for your impedance minima, try rearranging your band pass components (try HP then LP, LP then HP, both series components THEN both parallel components). Sometines this will bring that droop up.

I'd try to use a pair of the DC160s for handling excursion issues, unless they'll be used with a sub. A vented pair should only need 3/4 to 1 cf.

chris

[someuser]

What's the tweeter

That would be the VISATON KE25SC. http://www.visaton.com/en/high_end/h.../ke25sc_8.html

[Wolf]

That would be the VISATON KE25SC. http://www.visaton.com/en/high_end/h.../ke25sc_8.html

Nice job! By chance have you heard one? I would like to use it- eventually....
Later,
Wolf

[HareBrained]

Dave, thanks for the feedback. I did trace the data. I did not perform the transform but will retrieve the tool and include the data. As my initial post said, the design is in the early stages and I'm still learning.

I did try a 3rd-order LP but was experiencing an issue with a hump from ~2500Hz and up. I think this was related to the phase & tweeter interactions. Maybe when the transform data is added, this won't be an issue. But you're right about the 14kHz bump going away.

Chris, thanks for the xo suggestions. I'll give them a try. I know PCD has some ability to pick and choose where in the topology the components can be placed.

Regarding a 2nd DC160, I'm trying to keep the cost and the overall cabinet size down. And more than likely, I won't be using a separate subwoofer. So, the DC160 will be handling the lower octaves all by it's lonesome. What sort of excursion issues would exist? Does the xo point of ~350Hz change anything? Do I want to go higher or lower for some reason?

I had considered using the DC160S-4 but didn't think the there was a need for the additional SPL after modeling with the S-8. Am I wrong?

The room they're going into is small at 14x11. The sub of my current 2.1 set-up is turned way down and is either overpowering or non-existent, and I don't really see it integrating into a new system.

I am open to the idea of using the MTM down to 100-120Hz and using a pair of powered 6.5" or 8" subs. Any arguments for or against relative to my initial design? Any suggestions better then the DC160?

-John

[Wolf]

Dave, thanks for the feedback. I did trace the data. I did not perform the transform but will retrieve the tool and include the data. As my initial post said, the design is in the early stages and I'm still learning.

I did try a 3rd-order LP but was experiencing an issue with a hump from ~2500Hz and up. I think this was related to the phase & tweeter interactions. Maybe when the transform data is added, this won't be an issue. But you're right about the 14kHz bump going away.

Chris, thanks for the xo suggestions. I'll give them a try. I know PCD has some ability to pick and choose where in the topology the components can be placed.

Regarding a 2nd DC160, I'm trying to keep the cost and the overall cabinet size down. And more than likely, I won't be using a separate subwoofer. So, the DC160 will be handling the lower octaves all by it's lonesome. What sort of excursion issues would exist? Does the xo point of ~350Hz change anything? Do I want to go higher or lower for some reason?

I had considered using the DC160S-4 but didn't think the there was a need for the additional SPL after modeling with the S-8. Am I wrong?

The room they're going into is small at 14x11. The sub of my current 2.1 set-up is turned way down and is either overpowering or non-existent, and I don't really see it integrating into a new system.

I am open to the idea of using the MTM down to 100-120Hz and using a pair of powered 6.5" or 8" subs. Any arguments for or against relative to my initial design? Any suggestions better then the DC160?

-John

Put the 160S in a 0.4ft^3, and further load the cabinet with a SD215-8" PR. Use about 23g added mass to the PR, and line the walls. This will get you an F3 of about 45-50, but the rolloff is very gradual. It will work great on a room-boundary to smooth it out, and you'll be pleased to about 25 Hz. I'm sure you can lookup "Bitty Boom V2" in my blog if you have more questions.
Later,
Wolf

[dthomas]

Get the phase data in your files this will change how the drivers sum with each other. Without you are guessing at what is really going on. A third order electrical network should push the hump down by another 6db I would think that is enough. I would play with the values paying attention to the hump in the RS100s response at 12K (IIRC) getting it as low as possible then adjusting the response at the crossover point to get good summation.

A notch filter will fix it. I find including a target response in PCD is very helpful you want the RS100 down about -25db at where the hump is to get rid of any potential audible issues it might create with the tweeter. But at this frequency it is less likely to be discernable.

Dave

[HareBrained]

Dave, I've been using the built-in target responses for each driver (mostly 4th-order butterworth) but I only posted the summed responses. I saw the option to load a FR target but didn't have anything on hand that seemed better than the auto-generated ones.

I'll work thru the design in your suggested manner of kill the bump then slide the xo point. And I'll play with Chris' suggestion of ordering the components. I'm not too worried about the bump but I won't know how it really looks until I get the transform data into the sim.

Wolf, I had seen your Bitty when I was researching the DC160. Have you done a comparison of your alignment with the driver in an "optimum" vented enclosure? FR, etc? I don't understand the advantage of a sealed enclosure with mass-loaded PR compared to a vented enclosure of essentially the same Vb, especially considering the extra work, extra driver and still have a higher F3. Two DC160 would be more appealing if it's an SPL issue.

-John

[dthomas]

In its simplest form a sealed box with a passive radiator is no different from a ported box of the same size except you have no port tube and the passive radiator acts like the port. One advantage is that you can use a smaller enclosure and tune the PR lower without the need of an excessively long port. Port length can become an issue in smaller boxes where as the PR could be much easier to fit in. Also PRs don't suffer from port noise or resonance.

Unibox will model a PR and help you determine how much mass to add to the PR to get the tuning you want. I have two 12" subs using PRs because the port lengths were just too long to fit into the box. They work great and no more difficult to implement than a standard vented enclosure.

Dave

[Wolf]

In its simplest form a sealed box with a passive radiator is no different from a ported box of the same size except you have no port tube and the passive radiator acts like the port. One advantage is that you can use a smaller enclosure and tune the PR lower without the need of an excessively long port. Port length can become an issue in smaller boxes where as the PR could be much easier to fit in. Also PRs don't suffer from port noise or resonance.

Unibox will model a PR and help you determine how much mass to add to the PR to get the tuning you want. I have two 12" subs using PRs because the port lengths were just too long to fit into the box. They work great and no more difficult to implement than a standard vented enclosure.

Dave

Well said, Dave! Thanks for answering him for me.
Later,
Wolf

[someuser]

Nice job! By chance have you heard one? I would like to use it- eventually....
Later,
Wolf

Not yet. I believe there are a couple people using them over on the DIYAudio forum. I'd like to see some distortion numbers; maybe they are on Zaph's todo list.

[HareBrained]

So, I was able to extract the phase from the FR of the drivers and re-initiate the simulation. I'm glad I did. The xo actually simplified.

First the results, phase and FR...


And then with the midrange polarity inverted...


Well, a couple of things happened on the way to the forum:
1. I found a newer version of PCD. It worked quicker and the auto-entered data seems more accurate.
2. The points where the phase "comes back around" are no longer at the xo points.
3. The FR is as flat as the driver responses.

When I restarted, I loaded the last solution and things were a bit wonky. So, I tried to modify the components and throw in a 3rd-LP on the midrange and a bunch of other "fixes". Although, things didn't look bad, I know they could be better.

Staring at the curves, two things struck me. One, the midrange was supposed to be the limiter for SPL but the components were suppressing it. And two, the hump in the tweeter from Fs to 3000 was creating a flatter response than the rest of the FR. So, I cleared the additional circuits, raised the target db and moved the target upper xo point to below 3000k. Voila!

The woofer xo slope and midrange xo slopes are still 2nd order but they have no attenuation, zobel or notch filters. Straight-up, 2 components per slope. The midrange bump at 13k is barely visible with a max level of a hair over 60db. That's >25db down from the tweeter level. The tweeter is a 3rd order slope with a 1 ohm series resistor to attenuate the highest octave. Otherwise, it has a lift at the end. It's probably something I wouldn't notice and I could probably get rid of it as well.

The inverted polarity has a perfect response at the lower xo point. The upper point isn't perfect but that probably due to the mixing of even and odd order slopes but it's still swinging in the right direction.

These do not include the box responses but from the other modeling, the SPL of the vented DC160 should be right around 87db (again, assuming the data I'm using is accurate.) And because of the thin MTM baffle, I don't think BSC will be necessary, or maybe 0.5-2db max. It just something more to play with. Over all, I'm feeling pretty good about the design so far. I realize it doesn't mean anything until I have the actually drivers but it's still lookin' good.

Again, comments are welcome.
-John

[Wolf]

So, I was able to extract the phase from the FR of the drivers and re-initiate the simulation. I'm glad I did. The xo actually simplified.

First the results, phase and FR...


And then with the midrange polarity inverted...


Well, a couple of things happened on the way to the forum:
1. I found a newer version of PCD. It worked quicker and the auto-entered data seems more accurate.
2. The points where the phase "comes back around" are no longer at the xo points.
3. The FR is as flat as the driver responses.

When I restarted, I loaded the last solution and things were a bit wonky. So, I tried to modify the components and throw in a 3rd-LP on the midrange and a bunch of other "fixes". Although, things didn't look bad, I know they could be better.

Staring at the curves, two things struck me. One, the midrange was supposed to be the limiter for SPL but the components were suppressing it. And two, the hump in the tweeter from Fs to 3000 was creating a flatter response than the rest of the FR. So, I cleared the additional circuits, raised the target db and moved the target upper xo point to below 3000k. Voila!

The woofer xo slope and midrange xo slopes are still 2nd order but they have no attenuation, zobel or notch filters. Straight-up, 2 components per slope. The midrange bump at 13k is barely visible with a max level of a hair over 60db. That's >25db down from the tweeter level. The tweeter is a 3rd order slope with a 1 ohm series resistor to attenuate the highest octave. Otherwise, it has a lift at the end. It's probably something I wouldn't notice and I could probably get rid of it as well.

The inverted polarity has a perfect response at the lower xo point. The upper point isn't perfect but that probably due to the mixing of even and odd order slopes but it's still swinging in the right direction.

These do not include the box responses but from the other modeling, the SPL of the vented DC160 should be right around 87db (again, assuming the data I'm using is accurate.) And because of the thin MTM baffle, I don't think BSC will be necessary, or maybe 0.5-2db max. It just something more to play with. Over all, I'm feeling pretty good about the design so far. I realize it doesn't mean anything until I have the actually drivers but it's still lookin' good.

Again, comments are welcome.
-John

Uhm- you have it backwards on BSC. The thinner the baffle, the higher it needs to be engaged. Wider baffles actually require less BSC. Shoot for 3dB, and go from there. If it's not enough bass, bump it to 4. Rarely do I feel BSC is necessary to be full 6dB.

I'm now betting that your "offsets" are not present, correct? It seemed too easy, didn't it!? Without the offsets present, it is pretty easy to calculate and align.

Your offsets include: Acoustic Centers (Z axis), and Center-to-Center in meters in the program. You can use these as a guideline:




Let the fun/hair-pulling begin!
Later,
Wolf

[tombond]

Hey guys, since we're talking about measurements and I'm starting some today, I have a few questions.

If your using PCD and you have actual measurements ( I'm using ARTA ) in the box you intend to use, you don't have to enter the acoustic offset. Correct?

Again, using PCD, if you have actual FR and impedance files do you still extract minimum phase?

[dthomas]

As long as you place the measurement mic at the desired listening axis and measure each driver from that point. You can use response modeler to calculate the in box response below 300Hz and paste this to your Arta measurement if you are like me and don't like to drag all your measurement equipment outside in the 104 degree heat. Arta will give you pretty accurate measurements in room from about 200 to 300Hz up.

Yes Arta will extract minimum phase from the measurement. I think Dave Ralph wrote a nice paper on averaging on and off axis measurements which is worth a read. It is on his website.