Audiophile Musings: January 2013

Thursday, January 31, 2013

Bozak Rebuild Project - Part 6

Old drivers like the Bozaks I use in my project have very little technical information available; the Thiele-Small driver parameters readily available today did not exist until 15 years after their manufacture. But with the help of the newly acquired XTZ Real Time Analyzer, some of this parameter mystery is moving out of the darkness and coming into light.

All dynamic drivers have an impedance plot that resembles the one shown next. Where the measurement peaks (the resonant frequency), how wide the peak is (the ratio of the bandwidth at -3dB to the frequency at the height of the peak is its Q), and how low it plunges (the minimum impedance) tells you a lot about the behavior of the driver.

Typical Impedance Plot of a Dynamic Driver

Even without knowing the impedance plot you can estimate the resonant frequency by what you observe the driver does from a SPL analysis. For example, the Bozak B-209A midrange driver I am working with in this project was measured to have the following output within its enclosure and attached to its stock network (pink noise, all other driver loads removed from the network).

Near Field SPL Graph of the Bozak B-209A Driver

You would assume that the sonic energy suppressed by the network design (an 800Hz first-order Butterworth High-Pass Filter) should predictably suppress the sonic energy below 800Hz, and it does to some degree. However, as you can see in this graph, the resonance of the driver is somewhere around 315Hz. Making a system measurement as shown below reveals the effect of this sonic contribution as the acoustic energy begins to rise significantly at this resonant point. The result is that the resonance of the driver is measurably coloring the sound.

RTA Measurement of the First-Order Network

This RTA measurement tells you what the driver is doing as opposed to what the mathematical model predicts it should be doing. It is the non-linear impedance of the driver that alters how it behaves as opposed to how you predict it will behave. In the case of first-order networks such as the one used in the stock Bozak design, the tradeoff gained by good phase relationships between the woofer and midrange drivers is lost by the naturally occurring rising resonant impedance of the midrange driver below the crossover frequency. What happens acoustically is that both drivers produce sound in an overlapping region and creates louder sound in this region.

As you can see, the sound pressure rises at about the 315Hz resonant frequency of the midrange driver. But notice that the sound pressure also dips at about one and two octaves above the driver resonance (630Hz and 1.26KHz). Because of this harmonic recurrence, this indicates some sort of destructive acoustic or electrical phase interference at these points. All of these problems in the uniformity of sound pressure can be attributed to the resonance of a driver.

So if you assume that the drivers are creating sound in their assigned bandwidths, here you would be wrong. Simple mathematical calculations and even mathematical models do not accurately predict reality.

Changing the first-order Butterworth HPF to a second-order Bessel network on this midrange driver changed the behavior drastically as shown below. The sonic contributions of the impedance resonance are now adequately suppressed and the resulting sound pressure is considerably more uniform.

Second-Order Bessel HPF Network

The results of this test show that the nonlinear sound pressure was indeed being contributed by this midrange resonance. The 315Hz peak is now greatly reduced as is the 630Hz dip. But as it is with most problems, fixing one can create another and the 1.26KHz dip has now become a 1.26KHz peak. A resonant tank circuit may be necessary to completely control this sonic contribution of this driver’s impedance peak but only time will tell if the phase shifts of such a circuit will create more problems than it solves.

The results of changing the network design show a huge gain not only in the observed uniform sound pressure but also confirmed in subjective listening tests. What I heard most definitely was a reduction in the lower mid-bass region particularly noticeable in the human voice. Before there was breathiness to the vocal content that now sounds considerably more natural. Overall, the system sounds more like the real thing where acoustic guitars now produce a fundamental string component that can be heard resonating in the guitar body. Small inner detailing of fret fingering is more in the forefront and the hollowness of a clarinet in its lower registers is much more representative of the sonic truth.

The point is this: there is more to building a quality loudspeaker than stuffing drivers into a box and calculating a crossover network, much more. Getting them to work in harmony and create the accurate sound you want without coloring it too much is where the real art lies in design. Untold hours of testing, retesting, rethinking, redesigning, and reflection on results is where the sheep are separated from the goats. Properly interpreting what your measurements tell you is where true art enters into the picture.

I hope to better eliminate the remaining small problems introduced by the midrange driver resonance after which I will consider its design finalized. The sound this system produces today is light years ahead of the sound it created at its start just over 13 months ago. I would encourage you to hack into your speakers after making similar measurements and see if you can squeeze more out of your system than available from its original design. This experiment was fruitful and yours may differ, but the fun you have and the experience you gain is worth the effort alone. If all else fails, you can go back to the original design so what do you have to lose besides a little time?

So I hope you enjoyed this dissertation and can benefit from what I have tried. I will make the network pretty once I stop tweaking it and maybe add a final picture of it in Part 7 (although this may never happen, but then again I said that Part 6 would not happen either but here it is). As I said, I am still writing and I have a life. Besides audio I also love the beach, astronomy, and bicycling with my wife. I have another novel and a self-help book in the works so time allocated to my blog is very limited.

Bear with me, keep checking back for updates, and you will see I have much more to share. I have my eyes and thoughts on the tonearm of my turntable…

See also Part 1, Part 2, Part 3, Part 4, and Part 5

Yours for higher fidelity,
Philip Rastocny

I do not use ads in this blog to help support my efforts. If you like what you are reading, please tell your friends and if you are so inclined, buy some of my work at Amazon.com. My titles include:

Tuesday, January 29, 2013

Bozak Rebuild Project - Part 5

Driver materials totally impact the characteristic sonic signature of a dynamic driver. For example, harder materials like aluminum or titanium dome tweeters have a sonic signature that can be audibly recognized once your ear is trained to listen for that signature. The Audax TW025A28 I am using in this rebuild project is no exception to this phenomenon.

The Audax is a gold-sputtered titanium dome with a sonic signature reminiscent to that of titanium compression horns. Like these horns, there is a huge benefit of very fast attack and decay times at the expense of an inherent resonance at about 6KHz. Taming this resonance can prove to be a problem but one way to do it is to cross it over at a pretty high frequency with a second order or higher network. The results are quite tolerable compared to a lower crossover point or a first-order crossover network.

In working with this super tweeter, I kept running into the sonic signature when trying to cross it at about 8KHz and finally gave up and disconnected the driver. I preferred listening to no top octave at all instead of one highly colored in the 6KHz region. Of course that decision did not last long and after borrowing a friend’s XTZ Room Analyzer, I discovered more of what was going on behind the scene.

I made near-field (less than ½” away) measurements of all drivers in the satellite box one at a time. Resting the microphone on a tripod and removing the physical presence of myself and the laptop from the measurements as far as I could, I tested each driver with the same amplifier volume level and moved the tripod and microphone only. I also disconnected the leads on the other satellite drivers not under test leaving the network components but not the load to the amplifier. These 5 measurements revealed several things: each of the six B-200Y tweeters had its own contour, and the midrange and super tweeters were very well matched. You may notice the polyester pad at the rear of the satellite used to absorb wrap-around energy at the rear surface of the box.

Near Field Measurements with the XTZ

The Bozak tweeters are 35 years old and it is amazing that they still work at all. I understand that the variation in output is attributed to their age and accept it as a compromise in this design. I want to retain as much of the original components so that other can try what I have done to improve their own Bozak systems.

The first thing I did was to change the super tweeter crossover point from 8.3KHz to 12.5KHz since the acoustic energy contributed below the crossover point was not only significant but also highly objectionable. Using the two Russian 0.56uF polystyrene capacitors as a starting point and the measured impedance of the tweeter at 12.5KHz (6.5 ohms), I wound a 0.14mH choke and installed the Bessel-2 components into the super tweeter HPF. I then modified the tweeter LPF network in a similar manner by adding a hand-wound 0.31mH choke to its network configuration (BW-1, 8.3KHz).

After a quick audition, this revision proved to be moving in the right direction. Sibilance and vocal colorations essentially disappeared and I knew that I would not have to toss my efforts with the Audax driver and start over (as I had throught as recently as yesterday). Smoothness returned to this spectral region and although there is still some lingering metallic sonic signature to the super tweeter, the level is adequately suppressed with the new network design.

Next was the tweeter work. Presently the three tweeters are attenuated with a T-pad to -3dB. I ordered new components and installed the -5.5dB attenuator to the tweeter network. The levels of the tweeter and super tweeter were now perfectly matched.

I decided to also change the crossover design for the midrange LPF to a Bessel-2 using the stock choke. Moving the tap to the N102 terminal uses the stock 1.1mH choke. Coupled with a 4.7uF shunt, the midrange energy above the crossover point was again controlled and the results were highly satisfying. After listening, I needed to address the individual driver phasing issue and did so this morning. Polarities are now W+, M+, T+, ST- and the resulting RTA measurement of the entire system is shown below.

RTA Analysis of the New Network

I have not had a chance to do much prolonged listening but from my initial tests this is a really good thing. Everything above 1500Hz is fantastic with incredible inner detailing and superb accuracy. Strings are delicate and revealing and vocals are smooth with a natural presence. The sound stage is well defined and now reaches deep into the rear corners. The speakers completely disappear and a wall of sound is now produced from about four feet into the rear wall of the room. This psychoacoustic illusion is only present in the best of the audio playback systems I have ever heard.

My wife is still asleep and will be so for another hour or so. With the volume of the system at a very quiet level, the smoothness of the Native American flutes I hear now are stunningly accurate, the drums dynamic and revealing, and the voices gentle and sweet. There is a slight exaggeration in the lower mid-bass region as also observed in the RTA measurement and I will address that issue next. But for now, the raspy 6KHz region is tamed and tonal balance has returned to my listening room.

So I hope you enjoyed this dissertation and can benefit from what I have tried. I will make the network pretty once I stop tweaking it and maybe add a final picture of it in Part 6 (although this may never happen, but then again I said that Part 5 would not happen either but here it is). As I said, I am still writing and I have a life. Besides audio I also love the beach, astronomy, and bicycling with my wife. I have another novel and a self-help book in the works so time allocated to my blog is very limited.

Bear with me, keep checking back for updates, and you will see I have much more to share. Hmmm…what if I do this to the tonearm of the turntable…

See also Part 1, Part 2, Part 3, and Part 4

Yours for higher fidelity,
Philip Rastocny

I do not use ads in this blog to help support my efforts. If you like what you are reading, please tell your friends and if you are so inclined, buy some of my work at Amazon.com. My titles include:

Saturday, January 26, 2013

Bozak Rebuild Project - Part 4

Working with inexpensive tools can only provide data to the level of the accuracy of those budget tools. While relative changes between two measurements can be useful, absolute levels are essential for fine tuning. The RTA Pro app I have been using on my cell phone is now falling into the relative measurement category and I must find a better piece of equipment to continue my tuning efforts.

After contacting the developer, a text file calibrating the contour of the feeble cell phone microphone is forthcoming in a new release of this fine software. Until such a time that this new version becomes available, my search is on for a suitable or temporary substitute. In other words, I can no longer use the cell phone app to tweak the crossover network until I am confident that what it tells me is actually what I am hearing. But work in areas other than this can continue.

Once completing the steps in Part 3 of this series, I discovered that the sibilance problem inherent with the B-200Y tweeters and the addition of a supertweeter was going to be a complicated and prolonged issue to resolve. Individual driver and crossover component measurements revealed many things that - assumed to be handled in the original 1958 Bozak design - were unfortunately not addressed.

For example, the rising impedance of any dynamic tweeter is normal and to be expected, hence the typical use of Zobel networks to prevent a driver from contributing acoustic energy above the crossover point. The B-200Y tweeter and the B-209A midrange drivers are no exception to this rising impedance characteristic. What I measured explained part of the problem I heard. The minimum Z ot the tweeter, presumed to be close to the rated Z of 16 ohms, was 1/3 that of the Z at 20KHz (yikes!). And the Z at the crossover point was over 2x the minimum I measured. Careful measurements helped me to correct for this issue although internal resonances of the sealed tweeter combined with its age are proving to be a challenge to address.

I disassembled one of the blown tweeters to see what the heck was going on with the measured 4KHz peak. Once disassembled, I noticed that the tiny holes observed from the front of the aluminum diaphragm were indeed holes completely through the diaphragm and not delaminated layers of the aluminum-to-acrylic cone. This means that any tweeter with these tiny specs will have tiny holes in it. Of the six drivers in my system, three have many, many specs (holes).

Disassembling A Tweeter Revealed Internal Problems

I also found that the rear of the diaphragm had fused itself to the sound absorbing material used to control internal resonances behind the aluminum cone. This means that any of the six tweeters I am using could have this same problem and totally change the way they performed today as compared to that of new.

Diaphragm Fused to Damping Material

Also stated in Part 3, I believed Rudy Bozak matched the sound pressures of two 12” woofers to one 6 ½” midrange but this also proved to be an erroneous assumption. After further measurements, a -5.5dB attenuation of the midrange driver better matches the acoustic level of the 12” woofer in the stock B-302A cabinet.

The last modification still ongoing is to minimize first reflections back to the tweeter and supertweeter. After building the satellite box, I realized that I forgot to build into the design the recess for the tweeter and supertweeter frames (duh!). What I heard lacking was the deep and wide sound stage from previous designs and attributed this defecit to my oversight. Setting out to test my theory, I fashioned a smooth transition between the front baffle board and the frame edges with painters tape. The results were astounding!

Testing First Reflection Theory

Not only did the tape improve the depth and width of the sound stage, it also helped reduce the 4KHz peak in the system. I was pleased from the results of this test so I removed the tape and treated the edges with a liberal coating of black RTV.

Adding RTV

So I hope you enjoyed this dissertation and can benefit from what I have tried. I will make the network pretty once I stop tweaking it and maybe add a final picture of it in Part 5 (although this may never happen, but then again I said that Part 4 would not happen either but here it is). As I said, I am still writing and I have a life. Besides audio I also love the beach, astronomy, and bicycling with my wife. I have another novel and a self-help book in the works so time allocated to my blog is very limited.

Bear with me, keep checking back for updates, and you will see I have much more to share. Hmmm…what if I do this to the tonearm of the turntable…

See also Part 1, Part 2, and Part 3

Yours for higher fidelity,
Philip Rastocny

I do not use ads in this blog to help support my efforts. If you like what you are reading, please tell your friends and if you are so inclined, buy some of my work at Amazon.com. My titles include:

Sunday, January 13, 2013

Bozak Rebuild Project - Part 3

The next steps are to address the sibilance inherent with the B-200Y tweeters and then add a supertweeter to do the job right. The entire B-200 series tweeters are not uniform SPL drivers and have an annoying rise between 4KHz and 7KHz that drives one nuts after moderate listening to vocal material.

Rudy Bozak matched the sound pressures of two 12” woofers to one 6 ½” midrange pretty well. Each 12” woofer realistically needs about 8 cubic foot of box volume to operate optimally and anything smaller restricts the deep bass response (for the B-302A, this is about 50Hz at –3dB). My system uses two 12” woofers in an isobaric arrangement to lower the box volume requirement to 4 cubic feet and it is ported to enhance the deep bass. To better match the midrange to the isobaric woofer, I first attenuated the midrange driver about -3dB. The results were extremely satisfying smoothing out the minor dips and peaks.

Here is the RTA measurement at this point of the modifications. The green lines show the crossover points and the yellow line shows the point at which sibilance begins (4-6KHz). 7KHz is where I plan to cross in a super tweeter:

Stock Tweeter reversed polarity, Midrange Attenuated

I was encouraged from what the graphs showed and what I heard with my ears matched what I saw. I was delighted to hear that the transient response was still very snappy, something I suspected may suffer. While extreme dynamics were a little compressed from the addition of the second order network to the midrange (I no longer jump out of my chair when an unexpected explosion appears on a video), the system will still impress the heck out of you. The system still handsomely demonstrates what low-mass drivers can do right and high-mass drivers cannot (the old F=MA law of physics rears its head here).

TWEETER WORK

As you can see in the above RTA analysis, the double peak of the tweeter was there as anticipated. So the next step was to tame it with a first-order choke hoping that the choke would also eliminate the peak without impacting the lower frequencies. I tried several different values from those calculated and settled on a 0.31mH. BUT the addition of the choke introduced insertion losses to that part of the network and reduced the SPL output of the tweeter. So I removed the extra –3dB attenuation returning the tweeter array attenuation to -12dB. Also remember that my system has three Bozak B-200Y tweeters in each channel so the attenuation used by Mr. Tobin in his modifications should work fine for a 2-tweeter B-302A.

B-200Y Tweeter with LPF

THE SUPER TWEETER

As you can see, the sibilance peaks although much better are just not going to go away and the tweeters should be replaced with something a little more linear. Finding drivers this efficient is the problem. There are a few around and the super tweeter I found was one that also fit in the existing hole (I did not want to build another cabinet). So I moved on to adding a super tweeter to bring twinkles to the top octave. I bought some Russian polystyrene capacitors for this job. I like the way they sound at the very high end of the audio spectrum and played with a few values trying to add sparkle without adding sibilance. 1.12uF turned out to be a good option and the resulting RTA is shown next.

Adding the Super Tweeter

I suspect that the microphone, although “calibrated,” is not telling the whole story in these RTA images. It appears that at the extremes below 150Hz and above 6KHz, the microphone is just – well lacking. But RELATIVE differences between RTA measurements can be perceived and this is where I focused my attention. One day I will get a better unit…after the trip to Colorado and after the new telescope and after the OPPO Blu-Ray player and…well you get the idea. The point is that you CAN do a lot of good to your system with even modest tools and a very limited budget if you are patient and interpret your results based on the limitations of your equipment.

So this uniformity of sound pressure was getting pretty close to what I was hoping for at the start of this project back in January of 2012. In that 12 month period, I took almost one hundred RTA measurements of different network configurations and tried many approaches to solving problems. I succeeded at one point and then took two steps back at another, but in the end persistence prevailed and I believe Rudy Bozak would approve of what I’ve done.

The sound of instruments, vocals, and synthesizers is even smoother than ever and the characteristic Bozak dynamic sonic signature remains. However, they do sound quite different from the stock B-302A in that the bass is MUCH deeper and the sound is much smoother. The sibilance of the Bozak B-200Y tweeter is tamed but not completely gone but I am not a believer in notch filters at a place where your ear is most sensitive to them.

Below is a picture of the externally-mounted crossover network on these speakers. As you can see, the system is bi-wired from the amplifier.

The Current Network

Presently I am in a prolonged listening and evaluation period to the system to see what my ears tell me I should do next. It sounds really good but as it is with all audiophile gear there is probably something else that you can do to make it better. Here, it would be to replace the tweeters and super tweeter with a really good planar driver that is also very efficient. This is almost an oxymoron but there are a few out there if you search long and hard.

So I hope you enjoyed this dissertation and can benefit from what I have tried. I will make the network pretty once I stop tweaking it and maybe add a picture of it in Part 4 (although this may never happen). As I said, I am still writing and I have a life. Besides audio I also love the beach, astronomy, and bicycling with my wife. I have another novel and a self-help book in the works so time allocated to my blog is very limited.

Bear with me, keep checking back for updates, and you will see I have much more to share. Hmmm…what if I do this to the turntable…and that to the preamp…

The Satellite