Questions regarding electronic crossovers, bi-amping and impedance equalization

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Member
Joined 2004
Paid Member
Hi, I have some questions regarding electronic crossovers, bi-amping and speaker impedance/impedance equalization.
I'm planning on building a pair of speakers over the holidays, a 2 way TQWT design. The low/mid frequencies are handled by a Tang Band W6-623C paper cone woofer, and the remaining mid/high frequencies are handled by the 28-847SE silk dome tweeter. I'd like to ditch the passive crossover and go active.

The speaker design I'll be using is called "Pipe SIX", a 40 liter TQWT enclosure tuned at ~25Hz (see impedance peak in attached impedance/phase plot), originally with a 3rd order (18dB/octave) Butterworth-type crossover at ~2500Hz and an impedance equalization network for the woofer. Since the drivers have a very similar sensitivity, there are no series resistors necessary for attenuating the tweeter, and the crossover is about as simple as it gets.
Thus, an ideal candidate for bi-amping, methinks. I have two chipamps, an LM4780 single chip stereo and an LM4780 bridged dual mono, a cute little mini Aleph, and also a pair of tube amps, so I got the amp section covered.

I would run the LP of an electronic crossover into a chipamp for low/mid, and directly feed the woofer from said amp. The HP section I would like to run into the mini Aleph, and feed the tweeters through a large series connected MP motor run cap (for mandatory DC protection), which I also happen to have on hands. How big a cap do I need? With a crossover frequency of 2.5kHz, one octave down would be at 1.25kHz, one and a half octaves down would be at ~940Hz, and two octaves down would be 625Hz. The 28-847SE tweeter has a nominal impedance of 8R, so with a 23uF cap I would get a high pass with -3dB at f=1/(2*pi*Z*C)=865Hz, which should be fine if I aim for >1.5 octaves. So far, so good.

Now to my questions:

Is this electronic 18db/octave Linkwitz-Riley crossover a good choice? Project 123 (Figure 1)
I would have all the necessary parts on hands. A matched sextet of 13.6nF S&H styroflex caps for the high pass filters, and a bunch of 10nF Wima FKP1 5% caps for the low pass, some 1% MF resistors are easy enough to get, and I also have some LME49720 opamps and LM317/337 pos/neg voltage regulators for the supply.
Do I need input buffers? I guess output buffers with some level setting ability would be nice too, perhaps even with a DC coupling cap? Would some unity gain configured opamps (I don't want excessive gain in my system) and some 20k pots (trimpot even) suffice?

Do I need impedance equalization networks?
The original design incorporates one for the woofer (6.8uF and 10R), and none for the tweeter.
If I calculate the values myself, I get the following values:
Some sources say R_z=R_dc, others say R_z=1.25*R_dc. There seems to be agreement that C_z=L_e/(R_z)^2.
The W6-623C has a nominal impedance of 8R, a DC impedance of 6R (measured by K+T 6.11R), and L_e=0.61mH.
The 28-847SE has a nominal impedance of 8R, a DC resistance of 6R, and L_e=0.016mH.
R_z would thus be 6R or 7.5R for the woofer, and the same fot the tweeter. The resulting C_z/R_z values would thus be 16.9uF/6R or 10.8uF/7.5R for the woofer, and 0.44uF/6R or 0.28uF/7.5R for the tweeter. Do I need impedance equalizing networks for both? How critical are the values?

Well, that'd be it for now. Am I missing something?

Attached you'll find the original crossover schematic, an impedance/phase plot, the drivers' datasheets, and construction drawings (everything in millimeters).
 

Attachments

  • impedance.png
    impedance.png
    67.9 KB · Views: 315
  • xover.png
    xover.png
    15.5 KB · Views: 311
  • W6-623C.pdf
    452.6 KB · Views: 39
  • 28-847se.pdf
    48.7 KB · Views: 32
  • PipeSiXe_Bauzeichnung.pdf
    148.3 KB · Views: 49
Administrator
Joined 2004
Paid Member
OK, I'll take a stab at answering a few of your questions.

Is this electronic 18db/octave Linkwitz-Riley crossover a good choice? Project 123
Yes, it's a good choice. But it's not a Linkwitz-Riley filter, it's 3rd order. I like 3rd order electrical filters.

Do I need input buffers?
Maybe, but I doubt it. You can start without.
Do I need impedance equalization networks?
Probably not. The chip amps should have a low enough impedance that the impedance peaks won't effect the current flow. It won't hurt to leave them in, but I doubt it you'll hear any difference with them out (unless that Zobel cap is really nasty).


Well, that'd be it for now. Am I missing something?
Just that it will not sound the same. The passive crossover's filter function is determined by the impedance of what it's driving. The speaker is not a flat line impedance, even if the Zobel wants to make it that way. So the stock, passive filter isn't a textbook 3rd order, tho it might be close. Your active filter will be close to textbook, and that's gonna sound different.

If I have the time, I might run this thru the Passive Crossover Designer to see what you've really got. But it's probably close enough to nominal that it won't matter. You may like the change that the active filters bring.
 
Administrator
Joined 2004
Paid Member
On second glance, I'm not sure about your passive crossover values being 2.5 KHz, 3rd order. Need to do the math.

EDIT: The passive values you show for the woofer don't make sense. Is there a mistake there?
For the tweeter it looks like a perfect 3rd order Butterworth at 3500Hz.
 
Administrator
Joined 2004
Paid Member
Thanks for the update, Dave. Those woofer values are still kinda odd, but they may be there to correct FR bumps in the woofer and baffle.

To me it looks like a spread frequency crossover. Woofer ~2K, Tweeter @ 3K. You can see that in the combined impedance plot you posted.
 
Hi,

Its all wrong, except for the bass loading perhaps, and going
active won't help at all with the basic flaws of the design.
Which is wrong, for equal sensitivity drivers the tweeter
always needs attenuating for any half decent design.

An active copy of a crap passive is still crap.

rgds, sreten.
 
Member
Joined 2004
Paid Member
Hm, I never really dealt with crossovers until two days ago. Never heard of a spread frequency crossover before, but it makes sense. Setting frequencies is easy enough with the crossover design in question.

Now having found the formulae for 2-way 3rd order Butterworth filters, the suggested crossover values do look odd. The high pass seems to be set for -3dB at around 3400 Hz, and the low pass at around 2500Hz, maybe?

And to address the critique by sreten...is the design really that bad? Can you elaborate what leads you to say that?

Also, here are the TSP for both drivers:

28-847SE
Fs = 850 Hz
Qms = 2,37
P-Dia = 28 mm
Qes = 1,87
Re = 6 ohms
Le = 0,016 mH
Z = 8 ohms
Pe = 8 watts
Qts = 1,05
1-W SPL = 90 dB



W6-623C
Fs = 46 Hz
Qms = 3,18
Vas = 23,88 liters
Cms = 0,858 mm/N
Mms = 13,8 g
Xmax = 4 mm
Sd = 140 sq.cm
Qes = 0,44
Re = 6,11 ohms
Le = 0,68 mH
Z = 8 ohms
BL = 7,4 Tm
Pe = 30 watts
Qts = 0,39
1-W SPL = 89 dB
 
Administrator
Joined 2004
Paid Member
The crossover isn't "all wrong", tho the low pass is certainly odd. I do agree that copying a bad passive is no way to go. :no: I just don't know how bad this is. I still suspect that either the low pass values are wrong, or the designer was doing something strange electrically because of the acoustical response.

Do you have Microsoft Excel? If you do, I would suggest a copy of Jeff Bagby's most excellent Passive Crossover Designer. It also allows active crossovers or a combination of the two.
 
Administrator
Joined 2004
Paid Member
No. The design is clueless and until you really know
your stuff don't even think about an active version.
This tells us almost nothing. What is "clueless" about it? What do you object to? Can you give details about your objections?

The high pass filter is pretty much textbook 3rd order Butterworth @ 3500Hz. (given the published tweeter impedance curve) and that's not a bad place to be for these drivers. There may be better crossover point, but 3500Hz isn't a terrible choice. The woofer values are strange, and probably a mistake. It's worth looking at that further. We do NOT know the acoustic response of the woofer in this box, and that is important to the crossover design.
 
Account Closed
Joined 2001
Maybe I missed it but, is this passive crossover you want to "ditch" an existing crossover (that was designed by someone else) and is in use on the speakers currently? And it sounds good?

If so, that might help your progress (and establish a good base line) considerably when trying to go the active direction.

Cheers,

Dave.
 
Hi Rodeodave,

The impedance correction (zobel) of the woofer is only used by the passive low-pass filter, you do not need it in an active setup. To convert a passive design to active, you have to find filters giving exactly the same transfer function as the passive crossover does at the driver terminals. I know the review of this speaker in K+T 4-2008 and also have a simulation of it. Therefore I can retrieve the driver voltages from my speaker simulation program:

DriverVoltagesPipeSix.jpg

Unfortunately the passive filter interacts with the driver impedances and yields transfer functions far from text book. You will have to design active filters having approximately the same response. Good luck!
 
Account Closed
Joined 2001
Hi Rodeodave,

The impedance correction (zobel) of the woofer is only used by the passive low-pass filter, you do not need it in an active setup. To convert a passive design to active, you have to find filters giving exactly the same transfer function as the passive crossover does at the driver terminals. I know the review of this speaker in K+T 4-2008 and also have a simulation of it. Therefore I can retrieve the driver voltages from my speaker simulation program:

View attachment 319305

Unfortunately the passive filter interacts with the driver impedances and yields transfer functions far from text book. You will have to design active filters having approximately the same response. Good luck!

If that transfer function plot uses measured driver impedances (and it looks like they do), then the results contain the interaction with driver impedances and those would be equivalent to actual electrical measurements taken at the driver terminals. This would be the "baseline" I was referring to.

Cheers,

Dave.
 
Last edited:
Member
Joined 2004
Paid Member
Maybe I missed it but, is this passive crossover you want to "ditch" an existing crossover (that was designed by someone else) and is in use on the speakers currently? And it sounds good?

If so, that might help your progress (and establish a good base line) considerably when trying to go the active direction.

Cheers,

Dave.

The speaker design and passive crossover were published in K+T in 2008, and it has been built by numerous people (though not by me yet), and is received quite well from what I read.
I can get the drivers for very little money, and at a first uninitiated glance the crossover network seemed to be straight forward and easy enough to replace by an active design. Little did I know...


Hi Rodeodave,

The impedance correction (zobel) of the woofer is only used by the passive low-pass filter, you do not need it in an active setup. To convert a passive design to active, you have to find filters giving exactly the same transfer function as the passive crossover does at the driver terminals. I know the review of this speaker in K+T 4-2008 and also have a simulation of it. Therefore I can retrieve the driver voltages from my speaker simulation program:

View attachment 319305

Unfortunately the passive filter interacts with the driver impedances and yields transfer functions far from text book. You will have to design active filters having approximately the same response. Good luck!

Thanks, I think I get it now. It's there to keep the woofer's rising impedance from messing with the low pass filter response, right? It's not there for the amp, it's for the LCL filter's sake.
Your whole response has been very helpful, thanks!

Thanks Dissi, that's pretty much what my sims show. This may have been done to correct for baffle peak.
Dave, if you have any way to measure the frequency response, it would be a great help.

Well, I do have RTA/spectrum analyzer apps on my phone (Galaxy S), and have calibrated/zeroed them as the apps demand. They are quite sensitive, but I doubt that the microphone's response is sufficiently linear in terms of SPL. I can definitely measure something though, for instance a relative change in SPL at a certain frequncy.


Dave, here is what I get from the published impedance plots, which looks a lot like the graph above.

Are you saying you plugged the driver and passive crossover data into a simulator and got that plot? And it's similar to Dissi's simulation? Well, at least that would be consistent then.


And thanks for the links, sreten, they should keep me busy for some time.


I guess I'll have to look further into filter design to emulate the odd LP behaviour. I'll get there eventually. Thanks for your input, guys!

:hohoho:
 
Administrator
Joined 2004
Paid Member
Well, I do have RTA/spectrum analyzer apps on my phone (Galaxy S), and have calibrated/zeroed them as the apps demand.
How do you use it? Play pink noise and look at it on the app? That would be interesting. Of course, you have to build the speaker first. ;)

Are you saying you plugged the driver and passive crossover data into a simulator and got that plot? And it's similar to Dissi's simulation?
Yes, that's right. That's the electrical response those passive components would give with those drivers. Since the drivers are pretty flat thru the crossover range, acoustic response should be close. With that box, there should be a baffle peak from about 600-1200Hz, tho not a very big bump. Don't know if that's why the weird shelf in the low pass.

You don't have to this crossover, you can do your own. We can help.
 
Member
Joined 2004
Paid Member
How do you use it? Play pink noise and look at it on the app? That would be interesting. Of course, you have to build the speaker first. ;)
...

Exactly, somehow suspend the phone in mid air on some sort of stand in front of the speaker, play pink noise, observe graphs. I have these: https://play.google.com/store/apps/details?id=radonsoft.net.rtapro and https://play.google.com/store/apps/details?id=com.electronchaos.SpeedySpectrumAnalyzer. For calibration one can either play pink noise through a known-good system (yeah, not that lucky...) or have it auto zero in a very low noise environment (my parents live in a very rural area, zero measurable noise there). I did the latter and it seems fairly usable, but like I said, the mic's linearity regarding SPL response is questionable when calibrated below target SPL. Still better than nothing.

...
Yes, that's right. That's the electrical response those passive components would give with those drivers. Since the drivers are pretty flat thru the crossover range, acoustic response should be close. With that box, there should be a baffle peak from about 600-1200Hz, tho not a very big bump. Don't know if that's why the weird shelf in the low pass.

You don't have to this crossover, you can do your own. We can help.

I like that attitude. Would you still recommend a 3rd order design?
 
Member
Joined 2004
Paid Member
I tried to simulate the enclosure/woofer combination in hornresp, but of course I'm not sure if I'm doing it right since I first opened the program only like an hour ago, and I'm under the influence of eggnog. I'm using hornresp version 30.20

The driver parameters are seemingly straight forward and can be copied from the datasheet or published specs.

Ang is set to 2pi halfspace
Eg is set to 2,83V
Rg is set to 0,2Ohms

S1, area at the beginning of the TQWT, is 5,6*18,7=104,72cm^2
S2, area at the position of the driver (center), is 11,05*18,7=206,64cm^2
S3, area at the end of the TQWT, is 17,3*18,7=323,51cm^2
S4, area of the port, is 3,5^2*pi=38,48cm^2
S5, since the port is not expanding, is S4

L12, distance from TQQT's beginning to driver (center), is 86,7-18,7=68cm
L23, distance from chassis to TQWT's end (tricky because of the fold), is 11,4+86,7=98,1cm
L34, ?, is set to 0,01
L45, the port length, is 10cm

For contour I chose "Par", though I'm not exactly sure what it implies.

Vrc, Lrc, Fr, Tal, Vtc, and Atc are set to zero.

I chose "Offset driver horn", hit calculate, chose "combined response" in the tools section, and then entered a distance of -28,6cm since the port of the horn is on the backside.

Attached you'll find what the program computed. Does that make any sense? What can we learn from it?
 

Attachments

  • W6_623C_SPL_1st_sim_01.pdf
    15.2 KB · Views: 38
  • W6_623C_Imp_1st_sim_01.pdf
    15.1 KB · Views: 33
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.