Unbaffled – A Tale of Three

In this installment of Unbaffled, I test three different dipole low-frequency configurations.

Up first is a 4 x 15″ driver configuration.  The woofers are value priced models from Goldwood.  You can see something on the cones – that is silicon that I added to them in order to lower the resonance frequency from 30 Hz down to 24 Hz.  (As a side note, I would recommend using a black silicon instead of clear.  The clear silicon doesn’t have a very attractive appearance.)

2016-03-25 15.18.17

I also built an H-dipole recommended by Linkwitz.  This was really easy to build, OK to move, and much stiffer than the 81″ tall panel version above.

2016-03-25 15.05.40

Finally, six of the 12″ carver hex woofers from the original Carver Amazing loudspeaker.  Notice that half of the drivers are reversed – and connected electrically out-of-phase – so that some of the even order non-linearities would cancel out.  The drivers are connected 3 in series / 2 groups in parallel.  The panel is about 15″ wide at the bottom as I was not able to find the 24″ wide board that I used to make the 4 x 15″ dipoles.

2016-03-25 15.47.56

Here are some measurements adjusted to be equal in output.  All measurements are smoothed by 1/24th octave filtering to help reduce the noise from the wind.

  • Green = H-dipole response measured with the ground plane method at 2 meters.
  • Blue = the 4×15″ dipole panel with the microphone at 1 meter, 41″ off the ground (which is the center of the 4 drivers).  The level is reduced by -3 dB relative to the H-dipole.
  • Red = the 6×12″ dipole panel with the microphone at 1 meter, 41″ off the ground.

I was surprised at how close the H-dipole configuration behaved compared to the panel.  I was expecting the H-dipole to show more peaks and dips and otherwise be a poorer performer.  Not so!  In fact, I like the fact that the H-dipole forms a naturally well braced cabinet.  The two panel designs both bend like a reed in the wind and flop around at low frequencies (4 Hz is especially violent).

Comparison_NoEQNext up are some distortion plots.  An important caveat is in order here – the wind and automotive traffic were rather problematic during testing.  The plots are in the same order as introduced above.  The H-dipole plot was the least windy and shows the best results.

Dipole_4x15_distortionH_dipole_distortionDipole_6x12_distortionAnother note is in order… the 6 x 12″ dipole panel was damaged because it fell over and cracked.  Moreover, I had to swap out two of the drivers as they now buzzed.  (I had just enough spares.)  You can see that the panel isn’t quite straight at the top.  The  whole thing fell apart when I finally removed the drivers for the next prototype.

2016-03-25 15.48.15 HDR

Unbaffled – NOS Carver Drivers

I have been purchasing used Carver amazing drivers for some time.  Today I put my collection of original hex cone drivers to the test in a full six driver dipole configuration.  The drivers are connected in two sets of three wired in series, the two sets then wired in parallel.  Each driver has a DCR of 3.5 ohms, so the net impedance is a reasonable 5.3 ohms.

2016-03-05 6x12 2mGPTest configuration is a ground plane measurement at 2 meters.  I also did a similar measurement with the 4×15″ driver dipole panel I built with Goldwood drivers.

Compare_0dBWI know what you are thinking, these are the same speaker measured twice.  Nope!  I have added some silicon to the cones of the 15″ drivers to decrease their resonance to 24 Hz / raise their Qts to about 3.5.  They are very close in specs to the original Carver drivers at 24 Hz / Qts = 3.0.  Note the baffles are even different – the baffle is 24″ wide for the 4×15 and 15″ wide for the 6×12 – so maybe that compensates a little for the small difference in Qts.

2016-03-05 4x15 close micClose_mic_15For reference, consider a close mic measurement of one of the 15″ drivers (blue) compared to the array 2 meter ground plane measurement (green).  You can see that the dipole bass cancelation is -15 dB at 24 Hz.  You can also see multiple dips at 35, 57, 121, and 141 Hz.  I had the speakers facing into the garage, so I am wondering if these dips were due to the finite baffle, or a garage Helmholtz resonance.

So another test I tried was to put the 6×12″ on its side and compare to the standing array.

2016-03-05 6x12 on side

A little bit of gain, but otherwise the same response.


Speaker IV – In-Cabinet Speaker Padding

I finally got around to a final modification to the “built-in” speakers of the entertainment center (AKA Speaker IV).  The way the speakers sit in the cabinet leaves a gap of about an inch between the baffle and the door.  That gap is an acoustic mass that has to interact with the speaker output, creating resonances.

Solution: add some acoustic foam around the front of the speaker to seal the gap.  So the foam cut-outs for the drivers are something of a hatchet job.  It works.  Well.  The mid-range, such as voices, are so much clearer, more intelligible.  Little details that were masked before are now discernible.

2016-01-18 17.25.57Drivers are a 2″ G2Si ribbon tweeter and Peerless 6.5″ HDS series mid-bass.  Cross-over is acoustic 4th order LR at 2.5 kHz.  The mid-bass is in a sealed cabinet tuned to 80 Hz.  Why so high?  The intention is for these to be crossed over to a subwoofer at 80 Hz as “small” speakers.

Response tested sitting about 32″ off the floor – pre foam.  (I’ll do an update when I do a post-foam test.)  Test level is 1 watt, measurement distance 18″.  Measurement window is about 3.5 msec, so the response is smoothed.

S4 right zoomedThe distortion levels are respectable.  The distortion above 1 kHz is mostly 3rd harmonic (green) which suggests the ribbon tweeters would sound best crossed over above 4 kHz, as well as the mid-bass could be crossed over lower as part of a 3-way design.  (Will have to confirm with additional measurements of the individual drivers.)  Alas, not a big enough opening in the entertainment center for 3 drivers.

S4 Right Distortion

Seas 22TAF/G (H1283)

I used these in my brother’s speakers, Speaker III.  I’ve always had a soft spot for 3/4″ dome drivers.  I used them in my first design, Speaker I, listened to them for years.  When I switched out to a pair of Polk RTi4’s with 1″ tweeters, one of my first impressions was the very top end lost some air, some imaging.  I could localize symbols to the speakers, which I didn’t recall with my own design.  I have always – and possibly erroneously – attributed that to smaller diameter tweeter.

2014-04-20 19.25.43 Spkr3 II front cropped

This Seas 3/4″ uses an aluminum/magnesium alloy for the dome, fabric (Sonomex) surround, and ferro fluid for additional power handling.  Resonance is 1100 Hz, and sensitivity rated at 92 dB.  Testing was therefore done at -2 dBW.

Frequency response is very smooth, and quite extended.

Seas 22TAFG -2dBW freqz

Distortion isn’t as a good news story.  Performance here is generally average, with the good news that its primarily second order distortion.  I believe the distortion suggests that a 3 kHz or higher cross-over is the best choice, and as low as 2 kHz would likely work (my Speaker III crossover is 2.5 kHz).

Seas 22TAFG -2dBW distortion

Vifa (Peerless) D27TG-05-06

The D27TG-05-06 is a conventional 1″ fabric dome tweeter with ferro fluid (pictured below on the left).  Resonance is a respectable 1 kHz with a Qts of 0.77, sensitivity is advertised at 92 dB.  Testing is done at -5 dBW (should have done at -2 or -3 dBW).

2016-01-04 20.19.32

The frequency response is quite smooth, with a few dB of gradual roll-off on the top end.

Vifa D27TG05-06 -5dBW freqz

What is even more impressive is the distortion performance – this unit is 6 years old, been sitting around in a box!  Distortion is at -50 dB from the fundamental through most of the band above 2 kHz.  The third harmonic only rears its head at 4-5 kHz.  This tweeter could easily cross over at 2 kHz at 2nd order, and maybe 1.8 kHz at 4th order if you wanted to push it.  That explains why this tweeter has had such a long life and is still available.

Vifa D27TG05-06 -5dBW distortion

Dayton Audio Horn

Dayton Audio produces a 1″ polyamide compression driver and several threaded horns.

2016-01-04 20.19.59

The combination compression driver with horn is very efficient.  I estimate about 105 dB at 2.83 vRMS / 1 meter based on the measurements, which matches the manufacture’s specs.  Thus all testing is performed at -15 dBW.  The frequency response has a gentle roll-off of about 2 dB / octave until 12 kHz.  Then it drops off more quickly, and shows resonant behaviors.

Dayton horn freqz -15dBw

The distortion curve is quite surprising.  It is completely dominated by second order harmonic distortion!  Reportedly second order harmonic distortion is the least offensive.  I’ll need to build something from this and find out how it sounds.  Recommended cross-over is 1.6 kHz, and that looks about right provided the slope is 2nd order or better.

Dayton horn distortion -15dBW

Tang-Bang TW25-1744S

This is a ceramic dome tweeter with a neodymium magnet.  I picked these up based on the manufacturer’s claims of low distortion, extended response (to 30 kHz), and the compact mounting dimensions.  A plastic mesh grid covers and protects the dome.  Its the center tweeter shown below.

2016-01-04 20.19.32

Here’s the first unit frequency response

Tang-bang freqz 10 msec

Very smooth, and extended to 700 Hz, but a combination null / peak at the top end.

The distortion measurement

Tang-bang distortion

A little higher than expected.  Other tweeters and systems that I will be showing have the harmonic distortion as low as 50 dB below the fundamental.  The distortion is rising below 2 kHz which limits the crossover frequency and slope to be 2.5 kHz at 3rd order.

Tweeter Testing Methodology

Finally getting comfortable enough with the Room Equalization Wizard software to start some serious testing.  Another audiophile has set the bar high – Zaph Audio.  He has done large tests of tweeters, mid-ranges, and woofers.  One of his claims is harmonic distortion is a key discriminator in the sound of speakers and their drivers, and is an important factor in setting the cross-over frequency.  So my goal is to mimic his methodology.  Part of the Zaph Audio approach is consistency – testing all of the drivers using the same process.  So even if it turns out the process has some hidden defects, the ranking between drivers are likely to remain valid.

I started by building a large baffle with a replaceable 12″x12″ panel to which drivers can be mounted.  Flanking on the left/right are two of the GIK Acoustics panels borrowed from the living room.

2015-12-30 11.07.17

The problem with this setup is the reflection from the ground.  The driver center is about 30″ above the ground, and the microphone is 18″ away.  Doing some math, you can show that the round trip travel time will be about 3.85 msec for the ground reflection compared to the main impulse response.  Here’s an example.  Notice the “blip” at 3.85 msec?

Example Impulse Response with Reflection

Why is the extra blip a problem (as well some other smaller, extra blips)?  It distorts the frequency response calculation.  Show below is the result with the full 10 msec of data:

Dayton horn impulse 10 msec

Next, with the data limited to a 3 msec data record:

Dayton horn impulse 3 msec

So how to get a longer, clean data record?  How about laying the board down so that the driver faces up?  The ceiling is 10′ high in the room.  There is a lot of equipment in the room which will generate reflections.

2016-01-04 20.18.23

Here’s the resulting impulse response – clean for a full 10 msec!  (Alright, not perfectly clean, but significantly better.)

Example Impulse Response NO reflection

The resulting frequency response, compared with the upright measurement with a 10 msec data record.  The imperfections in the curve – likely the result of the finite test baffle – are sharper and more exaggerated with the reflection present.

Dayton horn freqz compare 10 vs 10

For completeness, repeat the comparison, 10 msec data record on the floor vs. 3 msec data record upright with reflection.

Dayton horn freqz compare 10 vs 3

What else needs to be controlled?  For the distortion measurements to be comparable, the input power should be set so the driver output is at a consistent value – I’ll use the same value of Zaph Audio of 90 dB at 1 meter.

NHT Classic 2 Center Channel

On occasion I buy speakers because I can’t make them for the price.  A good example is the NHT Classic 2 center channel.  I purchased a pair of these to act as the left/right speakers in the living room during the NHT holiday sales.  Price was $299 each.  The NHT Classic line, particularly the Classic 3, have received a lot of very positive magazine reviews.  The family lineage caries over to the Classic 2 center channel.  I really like these speakers, recommend them in general, and on value.  You can spend a lot more money and get a lot less speaker.

2016-02-27 17.26.28Yes, you read right.  I’m using a true center channel for a left and right speaker.  Here’s my reasoning.  The TV in the living room is above the fireplace, leaving just less than 8″ of height for speakers below it.  I looked at several loudspeaker bars and really did not like most of them.  The crossovers are too high, and the mid-bass units too small to really go down to 80 Hz.  Not all, just the vast majority.  The few audiophile quality designs I liked were quite pricey.

2016-02-27 17.26.58

The NHT Classic 2 is a 3-way design.  A tweeter, a small mid-range, and a pair of mid-bass units.  The advantage of the small mid-range is the cross-over to the mid-bass units can be low enough that a horizontal layout doesn’t beam in the upper mid-range.  And the Classic 2 is short enough to fit underneath the living room TV.

First up, the frequency response.  Measurement conditions are on a box 32″ off the ground, mic at about 18″, in the front third of a 14′ x 17′ room.  the measurement window is only 3.5 msec, so the response is limited to 220 Hz on the low frequency end.  3 dBW was required to reach 90 dB at 1 meter.  The response is reasonaby flat up to 10 kHz.

NHT Classic2 FreqWhat I am really happy with is the distortion levels at 90 dB.  0.4% or less at frequencies over 1 kHz.  Distortion falls to 0.2% between 300 – 800 Hz.  This is excellent, beating anything else I own or have built !!!  Distortion is also mostly second order, the nice distortion, except the 300 – 800 Hz range, where the third harmonic rises.

NHT Classic2 DistortionThe sound is very detailed, even with cable TV.  Vocals are so clear.  I watched the movie The Man from Uncle and was blown away.  This movie has several parts where the movie plays music – just an amazing, immersive experience.  I attribute this sound quality to the very even distortion characteristics.

New Sub – The Crown is Back

In my previous installment I compared the results of equalization with the SVS AS-EQ1 vs. my Onkyo TX-NR545 receiver.  The SVS AS-EQ1 notched the resonance peak at 40 Hz.  However, it did not do anything to extend the response below 40 Hz.  Bass equalization algorithms like Audessey have to work with a lot of different subwoofers with varying capabilities.  So it isn’t a surprise to me that the algorithm does not attempt to extend the bass – it would be easy to damage lesser subwoofers.

How to fix that?  I know that the 18″ driver can handle a lot more – equalization is part of my design approach.  The Crown XTi-1002 showed up today – back from repair.  One of the nice features of this Crown is it includes a built-in digital processor complete with parametric equalization, crossovers, shelving circuits, and a limiter.

Step 1: Turn off all equalization (SVS AS-EQ1, Onkyo TX-NR545).

Step 2: Experiment with shelving circuits.  The goal is to extend the bass from 67 Hz to around 35 Hz.  I used two shelving circuits centered at 45 Hz.  The adds +6 dB to the low frequencies, the second -6 dB to the high frequencies.

2x shelving

Step 3: Looking at the result suggests that too much attenuation is being applied to the second shelving circuit.  Eliminate the second shelving circuit, leaving just +6 dB on the low-frequencies.  Shown in red, the peak is very symmetric.

1x shelving

Step 4: Notch out the room’s resonance peak at 43 Hz.  The notch is set to -10.5 dB with a Q = 5.7.  Result show in light green – nice isn’t it!

1x shelf with notch

Step 5: Re-run the autoeq function of the SVS AS-EQ1 and compare to my handy-work.  Light green is my shelf + notch, blue is the new equalization result.  Pretty sweet!



  • Subwoofer is in the left corner.
  • Onkyo receiver equalization is OFF.
  • All measurements are at my primary listening position.
  • SVS Audessey calibration performed with 5 measurement positions centered around the primary listening position.  One measurement at head level, the other 4 within 18″ of this position to the front, left, and right.
  • The room has 3 bass traps, one each of the available corners (other corner is a set of stairs).  All 3 traps are 2′ x 4′ and set diagonally in each corner.

Home Theater Left Home Theater Front