The design exercise was interesting in itself, and when I realized the contestants would be using John's measurement files, I really wanted to try it. The chance to see and use the measurement files generated by a member of the DIY community with this much measurement experience was very interesting. I would be able to compare John's ZA14 measurements to the measurements I made for my satellites (this was helpful and quite worth the effort). I had enough sense to stay out of the 3-way competition realizing that it was more than I could handle in the allotted time. I'm glad I did, Dan Neubecker's winning design trounced my simple thoughts for a 3-way.
After the entry date had passed, John announced that he was going to extend the prizes in the contest to include three runners up in the 2-way side of the contest. The results are posted in John's blog and in a design contest link at the bottom of his main page.
Javier Huerta's 1600 Hz LR4 XO design won.
Alexandre Chamagne's 1100 Hz LR6 design was listed as first runner up.
Jay Kim's 2 kHz LR2 design was listed as second runner up.
My design was listed as third runner up.
It was very informative to see that the more experienced designers all opted to use the SB29's extended low frequency response. Either for lower XO frequencies as Javier and Alexandre did, or as Jay Kim did. Jay's design uses the SB 29's extended low frequency response to cope with the low XO slope in an LR2 design. What I was afraid to try due to my lack of experience was revealed by the more experienced designers.
My contest submission is attached below if you're curious, and the design and results are shown just below John's measurements. The thoughts driving my design choices are detailed in the last pages of the submission.
John's Measurements:
I compared my measurements of the ZA14 with John's and found some interesting differences. As usual, they were things I was doing which were causing the anomalies.
My choice to use the E-MU 02 sound card and m1616 dock led me to think I could conduct my measurements with the input gains set to 0 dB (see this review at Liberty Audio). While this does work, and always applies the same (small) amount of gain to the input signals, unfortunately, it puts the signal level very close to the noise floor of the E-MU system. This became apparent when I compared my response curve to John's.
June 2010
Back in April John Krutke (Zaph) announced a crossover design contest. Objective? Design either a 2-way XO for a ZA14-8 and SB Acoustics SB29RDCN-C000-4 tweeter using the enclosure for his ZA5.2 system, or a 3-way XO for a system using the woofer of your choice and the same ZA14/SB29 pair of drivers as the midrange and tweeter with no restrictions on the enclosure design. The winners were to get a pair of ZA14s and a pair SB29s.
I'm amazed... As a design exercise, I entered the contest and actually got things right. I entered a straight-forward design and John awarded me as a runner-up for my "Conservative LR4" approach. My thanks to John for the contest and for all of his efforts over the past years. I have a lot of respect for his technical insights and no-nonsense approach. Thanks again for the drivers John! I plan to put them to good use in a 3-way MTM system for my bedroom using RS270-8s for woofers. I hope I can do them justice.
John's plot of his ZA14 driver mounted in his enclosure.
Plot of my measurement of ZA14 in a 0.25 cubic foot enclosure.
HD plot of E-MU system. This was a simple swept sine analog measurement with the E-MU connected for a two channel loop-back test.
The above response measurements were both taken with the woofers mounted in their enclosures. Both measurements were made using SoundEasy and calibrated ECM8000 measurement mics. The big differences in equipment are the sound card/preamp/power amps used. The frequency response plots actually agree pretty well. Mine runs from 5 Hz to 50 kHz, John's runs from 10 Hz to 50 kHz. This makes it hard to do an overlay, but the important features are there, occurring at the same frequencies and very close to the same magnitudes. Also, it looks like John uses less smoothing than I did. Other differences are the baffle width, John's enclosure is 8" wide, mine is the quarter cubic foot Parts Express enclosure, which is 7 1/2" wide. Also, the ZA14's position was slightly different in the enclosures; centered vertically and horizontally in John's box, offset 3/8" left horizontally and 1 5/8" down vertically in my box.
The two anomalies that bother me are the rise in the low frequency response in my measurement, and lack of steadily falling response from about 700 Hz down to 100 Hz.
The rising response below 23 Hz looks to be noise from the system, i.e., the signal level was so low, it shows the noise floor. Not an issue for a 2-way, but it looks like it would play hob with a 3-way design unless it is corrected.
The lack of an accurate baffle step measurement is a bigger problem, and probably threw off the design of my satellites some. Working with only my files and no external base line for comparison, I didn't catch the problem. While the low bass distortion/noise in the E-MU probably elevated the F-R plot below 100 Hz, there is something going on between 100 Hz and 500 Hz that definitely isn't right. I honestly don't remember, but I thought I applied baffle step correction to the near field measurement, which is spliced to the 1 meter measurement at 500 Hz, but it sure doesn't look like it.
I tailed my measurement, but didn't take the last bit of phase error out during the tailing process. Not a design killer, but the phase alignment in my satellite XO design isn't as good as it could be.
The LF rise in my measurement also affects the phase response of the ZA14, and the level response from 200 Hz to 600 Hz just about guarantees that the phase response near the XO frequency (2.3 kHz) used for the ZA14 during the XO design is incorrect. The surprising thing is that the resulting XO still works pretty well.
Now I'm eager to re-measure and try a corrected design to see just how much difference it will make in the sound of the speakers.
My Design for the Contest:
XO Parts List:
High Pass components from Parts Express:
C12 = 5.6 µF poly (±10%) PE # - 027-425 $2.50 ea
L10 = 0.37 mH air coil 18 ga (DCR = 0.28Ω) PE # - 255-222 $5.35 ea
C14 = 40.0 µF poly (±10%) PE # - 027-442 $10.24 ea
R8 = 6.0 Ω Mills PE # - 005-6 $3.75 ea
R9 = 3.0 Ω Mills PE # - 005-3 $3.75 ea
Low Pass components from Parts Express:
L0 = 3.0 mH iron core 18 ga (DCR = 0.33 Ω) PE # - 266-558 $9.16 ea
C7 = 0.1 ΩF poly (±1%) PE # - 027-200 $1.23 ea
C2 = 11.0 µF poly (±10%), (10 µF || 1 µF)
10.0 µF PE # - 027-428 $4.12 ea
1.0 µF PE # - 027-410 $1.12 ea
L5 = 0.50 mH air coil, 18 ga, (DCR = 0.34 Ω) PE # - 255-230 $5.8 ea
C13 = 3.3 µF poly (± 10%) PE # - 027-420 $1.69 ea
Total for HP network $25.59 ea
Total for LP network $23.12 ea
Total for one XO network $48.71
Notes
C14 improves the ripple through the upper half of the XO region only very slightly (due to improved phase response). However, it tunes the phase of the tweeter to match the phase of the woofer at FC quite handily. Actually, it adds a pole, (its F3 if it were used as a single crossover element) below the tweeter's pass band so far out of the tweeter’s range that it doesn’t show on the frequency response plots.
C7 across L0 forms a notch filter lowering the W14’s response in the range of the breakup node @ 9 kHz.
The modeled frequency response is ± 1.4 dB from 125 to 10 kHz. The actual XO frequency is about 1820 Hz.
I'm kinda proud of the modeled phase response of the system. Good agreement over a very broad range.
The system presents a very benign load, suitable for any/all solid state amps.
Jay R. Taylor
June 26, 2010