Question about DIY sub crossover

[rothwellaudio]

Regardless, it’s all opinion.

I hear this a lot on this forum, particularly with regard to speakers. My view is that some some equipment is better than others, and some components are better than others. Capacitor distortion is something which can be measured, and it's objective, not subjective.

As for motor run capacitors, I have no idea if they're better or worse than similar alternative capacitors, so won't comment on that.

 

[Mr Boochie]

I am not sure where you deduced 20mh for 100hz crossover at 6db per octave?

12mh is the correct value for 100hz at 8 ohms.

Also air core inductors are more linear compared to iron core inductors.

http://www.1728.org/resfreq.htm

You maybe using a calculator that determines both mH and capacitance. If you use 177 mfd cap, you need a 12mH choke. But, I used a 100mfd cap. So, prolly needed a bit more inductance, to get same crossover point.

https://goodcalculators.com/crossover-calculator/

 

[Retrovert]

I’m sorry you got beat up in school by motor run cap bullies. Enjoy your audio grade caps and I will enjoy my motor runs.

As an electrical engineer I was taught to read datasheets, use test equipment, and contact manufacturers with questions.

I have clearly set forth the issues with motor-run capacitors and why these low-cost capacitors are not likely the optimal choice for crossovers.

 

[Retrovert]

I am not sure where you deduced 20mh for 100hz crossover at 6db per octave? 12mh is the correct value for 100hz at 8 ohms.

Yes, this is absolutely correct for a first-order crossover at 8Ω, at least for standard inductor values. The exact frequency would be 12.7 mH but that value requires two inductors, and it isn't that critical.

You maybe using a calculator that determines both mH and capacitance. If you use 177 mfd cap, you need a 12mH choke. But, I used a 100mfd cap. So, prolly needed a bit more inductance, to get same crossover point.

Ummm, no, that's incorrect.

JimPA is completely correct.

Consider the formulas (cut-and-paste from a posting I elsewhere made):

Given:
R is Nominal Speaker Impedance (Ω)
C is Capacitance (Farad)
L is Inductance (Henry)
f is Frequency (Hertz)

NB:
L is inductance in Henry.
To convert from Henry (H) to milliHenry (mH) multiply by 1,000.
To convert from milliHenry (mH) to Henry (H) divide by 1,000.

C is capacitance in Farad.
To convert from Farad (F)to microFarad (µF) multiply by 1,000,000.
To convert from microFarad (µF) to Farad (F) divide by 1,000,000 or multiply by 0.000001.

Inductor Formulas:
L = R / (2π × f) (Henry)
f = R / (2π × L) (Hertz)
R = 2π × L × f (Ω)

Capacitor Formulas:
C = 1 / (2π × f × R) (Farad)
f = 1 / (2π × C × R) (Hertz)
R = 1 / (2π × C × f) (Ω)​

Running the math for an 8 Ω speaker clearly yields the correct values as above set forth by both JimPA (in No. 19) and by myself (in No. 11):

L = R / (2π × f) (Henry)
= 8 Ω / (2π × 100 Hz)
= 12.7 milliHenry

C = 1 / (2π × f × R) (Farad)
= 1 / (2π × 100 Hz × 8 Ω)
= 199 µF
​

Random internet calculators may yield different values, of course, but one can always verify the math as per the above and see that any other value would be incorrect.

So unless your speaker is 16 Ω, and I looked up the JBL 136A which is listed as 8 Ω, your crossover points are incorrect, leaving a large hole: twice the frequency for the 100 µF capacitor (now 199 Hz), and half for the 20 mH inductor (now 64 Hz). Halving the capacitor doubles the frequency, while doubling the inductor halves the frequency. This is why tweeter capacitors are small and woofer inductors are large. I suspect you misremembered the values, as you would have noticed such a hole in your frequency response.

 

[Mr Boochie]

Yes, this is absolutely correct for a first-order crossover at 8Ω, at least for standard inductor values. The exact frequency would be 12.7 mH but that value requires two inductors, and it isn't that critical.

Ummm, no, that's incorrect.

JimPA is completely correct.

Thanks, for this. I must have forgotten the exact mH value of the chokes I wound. As I made this crossover a long time ago (summer between HS & college). And I followed a friend's calculations, him being an Physics major at the time. I followed all his numbers, specific for the 136A. Internal box volume, as well as speaker placement and crossover design. It must be fine, as the subs have always seamlessly blended into the L300's response (which uses 136A woofers, also). You cannot "hear" the subs, separate from the L300's. Until you turn off their amp. Then, there is a big hole in the bass. Esp on how even the sound is, across the room.

We made them into sealed boxes. About 5 cubic feet internal volume each. Unequal sided box (to avoid standing waves, IIRC). Filled with 4-inch square fiberglass fill ( about 1 cu-ft worth total) , also to avoid standing waves in audible frequency range. The speakers were placed, slightly lower than dead center of the front face. I kinda remember this as a strength and practice to avoid some type of induce resonance issue with box. Three-quarter inch MDF, sides anchored together by 3/8-inch bolts & nuts. You can see the putty I put over each nut tightening access hole. Heads of bolts are countersunk into the top of the boards.

The friend ran one speaker through some equipment at JBL (he had summertime school related job with them, at the time). Seemed to pass the tests he tried. I forgot what they were and unfortunately lost track of him. He's off in Europe doing some research project (last I heard).

Anyhow, these have really withstood the test of time. Plus, four new-house moves They are OLD (close to 40 years-old now). I've only had to re-do the surrounds twice. Just re-capped and checked the wiring on the DIY crossover for the first time, last year.
.

 

[Mr Boochie]

Most interesting thing about the subs. Imaging is much better, particularly at lower volumes ----> with the subs on, both stereo and mono. Not sure how that works?!? However, much the placement of highs and lows are better focused.

 

[toxcrusadr]

Yes, this is absolutely correct for a first-order crossover at 8Ω, at least for standard inductor values. The exact frequency would be 12.7 mH but that value requires two inductors, and it isn't that critical.



Ummm, no, that's incorrect.

JimPA is completely correct.

Consider the formulas (cut-and-paste from a posting I elsewhere made):

Given:
R is Nominal Speaker Impedance (Ω)
C is Capacitance (Farad)
L is Inductance (Henry)
f is Frequency (Hertz)

NB:
L is inductance in Henry.
To convert from Henry (H) to milliHenry (mH) multiply by 1,000.
To convert from milliHenry (mH) to Henry (H) divide by 1,000.

C is capacitance in Farad.
To convert from Farad (F)to microFarad (µF) multiply by 1,000,000.
To convert from microFarad (µF) to Farad (F) divide by 1,000,000 or multiply by 0.000001.

Inductor Formulas:
L = R / (2π × f) (Henry)
f = R / (2π × L) (Hertz)
R = 2π × L × f (Ω)

Capacitor Formulas:
C = 1 / (2π × f × R) (Farad)
f = 1 / (2π × C × R) (Hertz)
R = 1 / (2π × C × f) (Ω)​

Running the math for an 8 Ω speaker clearly yields the correct values as above set forth by both JimPA (in No. 19) and by myself (in No. 11):

L = R / (2π × f) (Henry)
= 8 Ω / (2π × 100 Hz)
= 12.7 milliHenry

C = 1 / (2π × f × R) (Farad)
= 1 / (2π × 100 Hz × 8 Ω)
= 199 µF
​

Random internet calculators may yield different values, of course, but one can always verify the math as per the above and see that any other value would be incorrect.

So unless your speaker is 16 Ω, and I looked up the JBL 136A which is listed as 8 Ω, your crossover points are incorrect, leaving a large hole: twice the frequency for the 100 µF capacitor (now 199 Hz), and half for the 20 mH inductor (now 64 Hz). Halving the capacitor doubles the frequency, while doubling the inductor halves the frequency. This is why tweeter capacitors are small and woofer inductors are large. I suspect you misremembered the values, as you would have noticed such a hole in your frequency response.

Interesting. I used the 2d order Butterworth at DIYAudioandVideo, https://www.diyaudioandvideo.com/Calculator/SpeakerCrossover/

I used 150 Hz, 8 ohms for both speakers, and got 12.01 mH and 93.75 uf. This is why I went with the 12 mH inductors I found on sale at PE, and a 100 uF cap. Strange that it came up with almost the same inductance, but less than half the capacitance of your calculations, for a much higher crossover point.

 

[Retrovert]

That's a second-order crossover, not a first-order. Use the dropdown for a first-order filter and it has the same values I set forth.

Here's a very brief explanation which has not been extensively edited. I'm not even sure how many people read what I write on these technical subjects.

Filter design can be complicated, so I'll try to keep this simple at the cost of breaking off some of the complex bits (transfer function) to stuff it into that box. No quibbling.

A first order is a low-pass (inductor) for the woofer, a band-pass (capacitor and inductor) for the midrange, and a high-pass (capacitor) for the tweeter.

A second order is a hybrid creation, essentially augmenting what would be considered the poor filtering of a first order with a second stage in parallel with the driver to further remove any components making it through the first stage. Because it has multiple stages damping control is now possible. But because it has multiple stages energy is stored and released which creates entirely new signal which is added to the original signal; that is distortion. Damping is where it all becomes a little tricky.

Because filters are not ideal — being either calculated results (DSP code) or physical approximations to mathematical ideals (capacitors and inductors) — energy is stored and released on each waveform cycle. (Easily seen with capacitors and inductors.) The release causes the filter to have ripple or ringing (the two meanings are not exactly identical, but close enough), i.e. overshoot and undershoot, which over time gradually reduces in magnitude and damps out. That ripple, an error term aka distortion, may exist in the pass-band, the stop-band, or both the pass-band and stop-band. Different types of filters (essentially different formulas or different combinations of physical components, i.e. inductors and capacitors) have different types of ripple. (Opamps, BTW, is an analog means to simulate tuned circuits without using inductors, and this is called a gyrator, commonly used in equalizers.) Ripple is why one application might select between (to pick from the four main filters) Butterworth, Chebyshev, Bessel, or Cauer. Compare, for example, a Butterworth versus a Chebyshev,. The Butterworth has a highly accurate pass-band with lower ripple, but very slow rolloff. The Chebyshev, to contrast, has a lower accuracy pass-band with higher ripple but very fast rolloff.

A second-order filter has design parameters including rolloff rate and ripple characteristics. It has a damping factor for the ripple, indicated by the symbol ζ (Zeta) which is the inverse of Q. Remember, Q indicates how lossy a filter is, so lower Q is higher losses and higher Q is lower losses. ζ is the mathematical inverse of Q, so lower ζ is lower losses and higher ζ is higher losses. A system may be underdamped (think of saloon doors which oscillate back and forth, gradually stopping at rest) with ζ < √2, overdamped (the saloon's door closer very slooooowly closes when released, but without overshoot) with ζ > √2, or critically damped (minimal oscillation) with ζ = √2. Butterworth is ζ = √2 ≈ 0.707. Damping is directly related to transient response, so maximal damping will very poorly respond to rapid changes in the input signal whereas minimal damping is going to rapidly respond but will have lots of ripple.

A second-order Butterworth is tinkering with the values to reduce ripple and create a more even response. This is important for a crossover. But (handwavy explanation for pedagogical purposes) it might be less important for, say, removing aliasing errors from the output of an oversampled sigma-delta DAC as the filtering occurs far above the audio band so a little extra ripple in the pass-band matters less than ensuring the offending signal is removed.

The short explanation is higher-order filters add group delay, phase shift, and ripple, and are consequently more distorting than first order. Some argue humans can't hear this. Which is why the fourth-order Linkwitz-Riley (double Butterworth) is sometimes used, even though its group delay is far greater than a first-order. The general thinking for crossovers has been that less is more, which is why most of the famous designers used first-order filters. Some disagree. Joseph Audio (see the patent if you care), for example, used very high order filters, something like 96 dB / Octave which, as far as I am aware, is the highest order in any commercial speaker.

Barbie says, "Filter Design is Hard!"

 

[Retrovert]

Most interesting thing about the subs. Imaging is much better, particularly at lower volumes ----> with the subs on, both stereo and mono. Not sure how that works?!? However, much the placement of highs and lows are better focused.

Earl Geddes wrote his PhD on multiple subwoofers and I recommend his papers on the subject.

See, for example:

http://www.gedlee.com/Papers/multiple subs.pdf​

His website is http://www.gedlee.com

 

[toxcrusadr]

Oh, duh, I read right over where it said first-order at the top of your math post. My bad.

It wasn't expressly stated in your last post, are you suggesting first order is superior in this application then?

 

[Retrovert]

Oh, duh, I read right over where it said first-order at the top of your math post. My bad.

No worries.

I routinely see identical misunderstanding about filters continually resurrected and reposted, and thought I should write up an explanation and then cut-and-paste it in as needed to other discussions. Several generations never learned how to use the actual formulas with a calculator, that the math won't lie, unlike the interwebs, and that what the calculators do behind the scenes is crucially important to being able to understand the effects of components and working the math will provide a deeper understanding and simplify the design process.

It wasn't expressly stated in your last post, are you suggesting first order is superior in this application then?

First, consider the nature of the speakers being blended and what happens to out-of-band signal. Is the subwoofer more or less mechanically limited, such that any signal above the pass-band (say above 100 Hz in this case) is not well reproduced, so whatever makes it through the first-order filter is insignificant? Ahhh, that depends upon the subwoofer. The same question can be posed for the speakers shifting reproduction to the subwoofer: are they distorting at the low end below 100 Hz, or is the mechanical rolloff sufficiently large that any signal in that area is negligible and non-distorting? Most speakers do not well reproduce lower frequencies. So a second order with a steeper slope may not be delivering the expected benefit of reducing out-of-band signal.

Second, consider the phase shifts introduced by each filter order: 90°. So first order is 90° while second order is 180°. That is considerable phase shift so the subwoofer is now delayed in time compared to the speakers. SInce delay is not polarity, inversion does not solve this. (See my writings elsewhere on AK about that very subject.) Group delay is audible and can cause artifacting.

This is an inexpensive experiment to try, just alligator-clip the inductors and capacitors and return the unwanted parts.

Edit: Fixed Typos

 

[toxcrusadr]

Yeah, I'll put this thing together on a prototyping board and see how it does, starting with first order (inductor on the sub and cap on the mains).

 

[Mr Boochie]

Yeah, I'll put this thing together on a prototyping board and see how it does, starting with first order (inductor on the sub and cap on the mains).

You are going to love them. Will they be self-powered?

 

[toxcrusadr]

The M&K is a piece of history so I didn't really want to modify it, and it's a passive sub so I'm just going to run it with an external amp.

The DBE was used to master Steely Dan's Pretzel Logic album.



Mine has M&K badged woofers but they are not the originals which had square magnets. Anyhoo, each woofer has two 8-ohm voice coils and two sets of inputs per driver, with no labels, so the back panel is confusing at first glance until you learn that there is no internal crossover and you're just looking at dual VC inputs. I planned to use one driver for each channel, rather than summing, and use a stereo amp. The VC's in parallel would be 4 ohms. The original square magnet woofers apparently had 4 ohm VC's that could be run in series for 8 ohms per driver. They also had a single-driver version of this sub.

This thread has a little info on it if you're interested:
https://audiokarma.org/forums/index.php?threads/m-k-passive-sub-any-specs.593927/

I had it running for awhile off the sub output from my Denon HT receiver and it sounded fantastic with pipe organ and whatever else I could find with bass. But for some reason the Denon just quit outputting sub signal. It's a nice receiver but I bought it new in '98 and I've noticed some issues with buttons and relays, so it may just be a break in the signal path. Anyway that's why I'm having to gin up a crossover.

BTW Ken Kreisel himself recommended to me a couple years back when I asked for info (see the linked thread) that a 24 db Linkwitz-Reilly was a good crossover for this, and he said that's what they use in a lot of receivers.

 

[Retrovert]

BTW Ken Kreisel himself recommended to me a couple years back when I asked for info (see the linked thread) that a 24 db Linkwitz-Reilly was a good crossover for this, and he said that's what they use in a lot of receivers.

The driver's frequency response and the cabinet's response determines the crossover order. No hard-and-fast rule exists for all drivers. So comments from the designer of a particular subwoofer system may not apply to other subwoofers.

 

[Mr Boochie]

The M&K is a piece of history so I didn't really want to modify it, and it's a passive sub so I'm just going to run it with an external amp.

The DBE was used to master Steely Dan's Pretzel Logic album.

Mine has M&K badged woofers but they are not the originals which had square magnets. Anyhoo, each woofer has two 8-ohm voice coils and two sets of inputs per driver, with no labels, so the back panel is confusing at first glance until you learn that there is no internal crossover and you're just looking at dual VC inputs. I planned to use one driver for each channel, rather than summing, and use a stereo amp. The VC's in parallel would be 4 ohms. The original square magnet woofers apparently had 4 ohm VC's that could be run in series for 8 ohms per driver. They also had a single-driver version of this sub.

FWIW, I use a separate Anthem amp to drive my pair of subs. They make current stable amps. MCA 225 seems like a very good amp. I'm using the PVA2. Though only 105 watts per channel RMS, it's is more than adequate for the sub cabs and 136A speakers which I'm using. They are pretty efficient. The subs are not overwhelmed by the main L300 (driven with a 150 watt/channel amp). Or are lost in the L300/L166 mix. Arranging the speakers in the room, I can get pretty good imaging, esp with mono records. Sound dead center and clear focus.

Like to hear your setup results!

 

[toxcrusadr]

I think this M&K is pretty sensitive, it did not seem to take a lot of oomph to drive it.

 

[Mr Boochie]

In my experience, current stable amp is the trick. For consistency in bass, throughout the volume range and transients.

 

[Pete B]

I think that I saw two different answers but will each woofer present 4 ohms to
each crossover (8||8 ohm coils)?

 

[toxcrusadr]

Yes.