June 2008

Audiolics Anonymous Chapter 104
Flat Is Not Always Best Marchand XM-44 Crossover
Article by Bill Gaw

  Welcome to the June meeting for Insatiable Tweaks. Mud Season is finally over here, so its time to get out the tractor and start doing a little yard work. For those not interested in tweaking, and think all amplifiers, wire, speakers, etc., sound the same, you may want to stop now and go out and clean your yards, as today's topic is setting up and voicing multi-driver home-built speakers.

Thankfully, as tweaking time is short this time of year, in the past two months I've been able to complete most of the project which was begun back in Chapter 102,(please review) when Wally Malewicz, of WAM Engineering, came over, did some measurements with his CLIO program, found that my tube preamps with built-in active crossovers had become defective. He had to leave the next morning, which left me with the task of deciding how to rectify the situation, primarily by experimenting on the best implementation of active crossovers for the various drivers.

Over several hours of travail learning how to use the WINAudio MLS analyzer program, I blew three tweeters by shear stupidity (see below), but finally got the hang of the program, and the necessity of placing a volume control between the computer and speaker. Each individual horn and driver for all seven speakers was measured for frequency and impulse response using both MLS and Chirps (rapid frequency sweeps that try to keep ahead of room modes which affect speaker response measurements) to get approximations as to where the best crossover points would lie.

Interestingly, probably due to room position and driver differences, each unit produced somewhat different frequency curves, making it imperative that for best reproduction, each be given its own frequency range using separate active crossovers and amplifiers. (Let's see, 7 speakers x 4 drivers each = 28 amplifier and 7 active crossover units). Thus two more active crossovers and four more stereo amplifiers would be needed.

For the past couple of years I've been using Behringer Super-Pro 3400 2/3 way stereo active crossovers on the center and rear speakers. While for their price ( as little as $50-100 for used units on eBay) they do an excellent job, they do A/D and D/A conversion to allow them to work in the digital domain, which gives them the ability to also do time alignment and phase reversal, but that double conversion does do some slight damage to the signal. So I settled on getting two more for those channels. As I had a set of four MOSfet monoblock 100 watt amplifiers from years ago, one of which would function perfectly with the center woofer, and the recently repaired Crown Macro-Reference stereo amp for the extra subwoofers, the system was set for amplification. Unhappily, while all this was going on, one of my Vacuum State tube amplifiers went on the Fritz, which required that I use two of the MOSfets on the main woofers.

I wanted the ability to go analog from my turntable to my left and right main speakers, and had been very happy with the several different types of the active analog crossovers from Marchand Electronics but had been stupid enough to sell them off several years ago as other revue products came along. Mr. Marchand was contacted and consented to build one of his four-way XM-44 units, previously reviewed at AA 43, with his top-of-the-line parts, including 1/2 dB step volume pots, in balanced mode. In addition to acting as a crossover, it also has the ability to do notch or boost filtering, baffle step compensation, delay of individual drivers and bass boost.

As the unit changes its crossover points by using removable cards (see the six greenish cards in the photo), I had no idea where to set the crossover points. He was kind enough to send along extra cards with values surrounding the ranges I thought would be best. The unit arrived in less time than he had agreed to, and needless to say, functioned flawlessly and sounded as good as I remembered. Because I had asked for a four-way unit, there was no ability to add a volume control for each channel, but that would be taken care of by my Lexicon pre-pro. 

The process consisted of setting up a measuring microphone first within a foot of the mouth (opening) of each horn and measuring the output of each separately with a full range MLS or Chirp signal from my computer's sound card using the WinAudio MLS program. By measuring only the first few milliseconds of the signal, room modes would be removed from the results. Luckily, all of the driver's frequency ranges overlapped by at least 1/2 octave before their natural 12 dB per octave horn rolloffs, which allowed me to use a range of crossover points to allow the flattest frequency response. Unhappily, the mid-horn only had a frequency range of about 800Hz to 7000 Hz, which required the woofer horn to be used up to that point. Happily, the woofer was actually very flat to 1200 Hz, so this wasn't a problem.

Through stupidity, this is where I got in trouble with my tweeters, as they don't like to have any low frequency information passed through them, which caused several to self-destruct. You'd think I'd have realized this by the second mistake, but it took me three burned out units. Oh well, such is the life of a tweaker. Luckily Madisound Speaker Components where the units had been originally purchased, had extra diaphragms on hand, and for a really decent fee for parts and service, replaced the diaphragms and tested the units to confirm they were up to spec.

Once each driver had been measured, the microphone was placed exactly at my listening position at ear height, and the measuring signal was run through the crossover to each pair of horn drivers to get the best crossover point and volume output. Finally the signal was used on all four drivers of each speaker to make sure the curve was as close as possible to flat. This was repeated on all seven horns until they matched each other as close as possible.

The first picture shows the individual driver outputs of the left front speaker. 
Orange = subwoofer Green = woofer Red = mid Blue = tweeter

The second picture shows the right speaker individual driver and both the Mlssa and Chirp output of all drivers together.

Orange = sub Green = woofer Red = mid Yellow = tweeter Brown = total Mlss
Blue = total Chirp

The first thing to note is the 5 dB higher result with the Mlss of the subwoofer compared to the Chirp. this is due to a 60 Hz. room mode of about 15 dB. which has been eliminated by the Lexicon pre-pro. Also there is a suckout at about 180 Hz. Second one can not a gradual decline of about 2 dB. from 20 Hz to 20kHz. due to my voicing the speakers (discussed below) to what I perceive is best for my listening preferences. Third, the RAAL tweeter is flat to beyond 20 kHz. with the rolloff beyond that point due to the measuring instruments, as is the rapid rise in the Chirp curve beyond 15 kHz.

Next, using my ears and a Radio Shack dB meter, white and pink noise was played through first the woofers, then the mids, the tweeters, then the whole speaker, which were adjusted until each pair of speakers had the same volume output and the sound would appear to be centered between each speaker as with a mono signal.

Except for room modes which gave +/- 10 to 15 dB peaks and troughs at several frequencies, each speaker's output was flattened to +/- 4 dB. I know, they should be +/-1 dB for audiophile specs, but its almost impossible to get perfect measurements with 12 to 15 dB room modes. As my Lexicon MC-12B preamp-processor can eliminate the room peaks, using the same measuring microphone as above, this was done for each speaker using its automated program.

So how did the system sound after this mountain of travail. In a word, with classical orchestral music: STERILE. Maybe like Avery Fisher Hall in New York at its worst. Whether I had just become used to the mellow sound of my original setup, or because of my deep association with Symphony Hall and the Musikverein, flat frequency response, while possibly good for jazz or rock, on orchestral music which is played in large concert halls where the space and ambiance eats up the high frequencies, but where miking directly over the front of the orchestra leads to brighter than natural recordings, the sound to my ears was somewhat harsh on most recordings.

On the other hand, imaging and ambiance retrieval were the best I've heard on my system. Instruments were more three-dimensional, and the space between them and the hall walls was more filled in. On SACDs of the Boston Symphony recorded in Symphony Hall made from Golden Era RCA tapes one could hear the traffic noise to the right of the right speakers out on Mass. Ave. I know, that's not music but boy does it add to the realism.

On movies, the sound space was more alive and movements of sounds between front center and back speakers were more natural. On present day movies, voice which is almost always stuck in the center channel was clearer and more precise, and on Ben Hur, a movie from 1959 which had a soundtrack with five front channels, the voices panned nicely across the soundstage and followed the actors. Thus, I had produced a great system for movies, and possibly rock, pop, and jazz, but not my cup of tea for classical recordings.

Whether the changes, both positive and negative was due to the new voicing of the speakers, or their better matching, or the removal of the defective preamp, or the quality of the Marchand or a combination of the above is indecipherable.

By judiciously adjusting the outputs of the different drivers, which forced me to keep getting off of my butt every few minutes for several evenings, using the volume pots for the various crossovers, and doing the previously discussed steps over and over again, each speaker was tailored to my listening preference.

I know, flat frequency response is supposed to be correct, but there are so many reasons why it is not:

First, the further back one sits in a concert hall, the more the high frequency information is attenuated and the more diffuse the low frequency information becomes. Thus unless one is in the middle of the orchestra, there is no such thing as flat response.

Second, the majority of recordings today are just too bright. One of the reasons the Golden Era recordings sound so good today is the judicious use of equalization that was used by the great recording engineers to tailor the sound to match what they heard in the concert hall. The other reason is that today's microphones are either flat or tipped up in the high end, and are placed right on top of the instrument, which magnifies the upper registers and overtones.

Third, digital anomalies such as jitter and downsampling seem to lose bass quality while amplifying or distorting high frequency information.

As it turned out, for all of the speakers, optimal sound for my taste was obtained by decreasing the output of the mids and tweeters by 0.5 dB., and increasing the output of the subs and woofers by the same amount. That's all it took. The above positive qualities were maintained while the harshness disappeared. I can now see why some speaker manufacturers, after using programs, anechoic chambers, etc. to produce their speakers, will take hours of listening time to voice them. Even 0.5 dB. changes over narrow frequency ranges, especially in the mids where our ears are most sensitive, can make all the difference in the sound of a speaker that distinguishes it from others.

On the other hand, by keeping a record of how the system sounded with various adjustments, I can now tailor each individual recording to either flat or how I want it to sound. That's truly a tweaker's dream (or nightmare).