The LX55 modification required depends on which LX55

My investigation of the LX55 responses under stock and two different modifications has shown:

1) Some, perhaps most, but not all, LX55's benefit from the tweeter phase (or polarity) "correction," that means changing the connections (at the crossover) so that the tweeters are nominally in phase with the woofers, instead of nominally out-of-phase as the crossovers were originally wired.  I first saw this clearly after applying the phase correction to unit #1.  This usually makes a big difference in high frequencies above 2kHz-10kHz, with the wrong phase connection having very depressed highs which sounds very dark, recessed, etc:

I found 2 exceptions.  In one case (unit #6) the high frequency response is already so flat that  changing the phase could not have possibly helped.   I already hung the unit to the wall before getting into this experimentation, and I decided to leave it alone.  Here's the final response of all units, with two phase inverted (stock) and two phase corrected units--in each case with the flattest response.  Unit #6 (inverted) tracks Unit #3 (corrected) above all else until the highest frequencies.  Then the other two units (one inverted and one corrected) more or less track each other at a slightly lower level.  This shows that not all the variation among unmodified units comes from tweeter phase misidentification, though the graphs above and below show that a lot of it can.

Final response of all LX55 in use, May 2019

In another case, "correcting" the tweeter phase actually reduced output 3-7kHz, in contrast to what it did for other units.  This was the shock after I had modified unit #5, the high frequency response got sucked out even more than before (it was a bit lower than the others which had already been fixed).  I had to revert unit #5 to the original inverted connection of the tweeter to get decent highs.  This would probably also have been the case  for unit 6, for which I did not make any changes, because it was already the best.

Stock (after restoration) is green, phase "corrected" is purple

2) In the two LX55's that were sold to me with choke removed, I found response considerably flatter with the choke restored.  With no choke,  those two units there was an annoying upper midrange boost that tended to obscure the highs above, giving the speakers a kind of "car radio" sound.  If the car radio salesman has his hand on the volume control, he could sell you either one by not correcting the level difference correctly.

Now, one thing that "simpler is always better" fans might not guess, is that the step response WITH the choke is far better than without it.  Without the choke, the step has an undesireable notch right around 1mS--corresponding to the beginning of the midrange bulge at 1000 Hz.  With the choke, the step response smoothly tapers down through that region as it should, and correspondingly the frequency response is relatively flat.

In unit #1, it's hard to say, the speaker sounds so dark with the choke having no choke might be better.  Here is #1 in all 3 conditions.  I'd still go for the midrange-flatter choke-plus-phase correction blue line.  That's actually similar to units #3 and #5, in their final configuration, which I decided to use in preference to unit #1.

3) The higher value choke from the LX5 may not be an advantage.  It may cause some dropoff between 3-4k, and still does not depress the 5k peak much.

Green unit #2 with 0.9 mH choke, Yellow #3 with 1.0 mH choke

With the tweeter phase connected so as to provide maximum output 3-10kHz, and with the original factory chokes these are very nice sounding speakers, though with more than desirable variance.  I've arranged them so the highest HF output units are at the sides (the on-axis gives less tweeter output with the Lineums) and next best HF output unit is in the corner (which depresses HF response the most).  From right to left, the units are 6, 5, 3, 2.

In listening, the surround engineering on Wish You Were Here (SACD multichannel version) and Islands (King Crimson 40th anniversary edition multichannel DVD-Audio) finally makes audible sense to me for the first time because the surrounds are sounding close enough to the virtually flat Revel M20's in the front.  Until this month, they were wildly different.

For 5 channel discs, I'm now simply duplicating the signal from the sides in the backs.  This seems to add a desireably "rearness" to the 5 channel sound, which otherwise is right at your sides.  I found the SSI unit I had been using was chopping out the middle (for center channel) and the sides (for steering) leaving me with only phase noise in the back.

Precisely tuned antenna

This is really for a TV antenna, but the same principles would apply to FM or whatever.

In the past I've always started from twinlead folded dipoles, the kind that used to be boxed with every FM receiver in the 1970's, with 300 ohm connection.

It seemed obvious that if you want a higher frequency than these are made for (the twinlead folded dipoles for FM are generally tuned to the middle of the FM band) you simply cut it as needed, Pretty much according to the wavelength of the broadcast frequency (more on that assumption later).

But some time ago, some guy who knew a lot more than me about radio, questioned as to why I was starting from a twinlead folded dipole, which I could in perhaps even less space use the genuine article: a halfwave dipole, made from simple wire instead of twinlead.

Well, I had simply started from what I knew worked.  I wasn't sure if a "regular dipole" needed to be that same length, or perhaps twice as much.

So now I was informed, that except for some esoteric factors (the velocity) I could just use a wire instead of twinlead, and it would be about the same length (not double etc).  AND, there'd be another big advantage.  Unlike the twinlead dipole, which is matched to 300 ohms, a half wave dipole is nearly a perfect match for standard 75 ohm coax, which is now the more common connection for tv's and tuners, and actual coax cable doesn't have to be run in  a straight line free of obstructions.

There is one tiny difference in the performance of an antenna made with wire and that of twinlead.  Because the twinlead is thicker, primarily, being two wires instead of just one, it has a wider bandwidth.  So this is what you'd want if you want to cover the FM band, say, the purpose for which these "indoor antennas" were included with tuners and receivers.  But if you only want some specific frequency, and you can tune for it precisely, you can get better tuning with the narrower bandwidth wire.  But really, for reception at least, a wire can cover quite a bit of bandwidth as well.

Anyway, here's a guy who discusses making a TV tuner out of twinlead, sort of how I used to do it.  The overall length of the antenna is a half wave, making each side a quarter wave.  That part is easy enough if you know a few simple things.   The wavelength of 300 Mhz being 1 meter is a good place to start.  If you ultimately prefer to use inches, to 3 digits you can conveniently use 39.4 inches.  The actual antenna that works best for reception is in either case a "half wave dipole antenna", with each side being a quarter wavelength.  That works well intuitively because each end is at the opposite displacement at any time, aka 180 degrees apart. (Full wave dipoles are fundamentally more resonant, but difficult to match to lead-in cables and generally not used.)

The trick is knowing the frequency of the channel of interest.  The link above gives some formulae for doing so for TV channels.  You can also just ask your search engine.  However, this assumes you know the "actual broadcast channel" which for TV stations may not be the number used anymore.  Once again the "real channel" or the plain old "real frequency" of any actual channel can be fairly easily looked up.  For my channel of interest, the real Channel 9, the frequence in the middle of the channel is 189 mhz.  That was their original VHF broadcast channel in the analog era and they stuck with it.

So then the wavelength of 189 Mhz is

300 million meters/sec  /  189 million cycles/sec

300/189 = 1.587 meters = 62.5 inches

1/2 wavelength is 31.25 inches

1/4 wavelength  is 15.6 inches

OK, so then I could just cut the sides of my dipole to that length.  What I'm actually doing in my fashion is buying a ready made "dipole" for FM, made with insulated single wire instead of twinlead. The pre-attached lead in is very fine twisted pair (for low-cost-to-the-manufacturer and semi-invisible appearance, otherwise coax would have been better for performance, but all the prebuilt ones for the FM band I could find were exactly like this--none used actual coax) terminated with a push-on F connector.  This is very convenient and not balun is required as I've needed to use for my twinlead antennas for decades.  Eliminating the balun saves space, cost, installation difficulty, and a potential reliability issue as the ends of the balun can work loose and if not sufficiently or constantly retightened can cause drop outs.  Reaching back behind the TV to tighten the connections of the balun is a chore I've long gotten tired of.  And the balun itself has loss, so eliminating the need for it might actually improve performance, depending on how well the lead-in is matched (with real coax, it would be better than twinlead overall).

I'm cutting this "ready made FM band antenna" to the required length for Channel 9, in such a way as to re-attach the terminations (with nail holes) at the correct length.  As I learned before, the wires must be spliced and soldered well and nicely insulated.  Unsoldered bare copper twisted stranded wires generate noise at RF frequencies in RF antennas.

NOW, the complicating factor is that the correct length is not as I just described, it is adjusted by a Velocity factor.  The link above gives the velocity factor for a twinlead dipole working in non-differential mode as 0.95.  That is higher than the normal propagation velocity cited for twinlead because that normally assumed differential model, as the author explains.  But other than simply saying that it's higher than differential, he does not derive the 0.95 or give further sources (I think it can be derived, from Maxwell equations, if you know every minute detail of the construction and materials in the twinlead, and other factors such as mounting height, ground density, air humidity, etc, etc, and how they are all used in the equations, which is no simple task).

I decided to take the advice of another old ham.  Start your dipole longer than needed and cut it bit by bit until you hit the target resonance.  He said never, ever, did he cut a dipole correctly just from calculations.

So I took that advice, and I now have a SARK-1 antenna analyzer I needed to justify getting.  I started longer than half wavelength total and ultimately cut it to about 12.25 inches as it crept to 189 mhz notch.  As installed on the wall, 7 feet high, with larger nails (and not stainless) than I would have liked (because, smallest thing I could find on hand) it still measures a VSWR notch at 184 mHz, which is remarkably close to target, within about 3%, probably far closer than I've ever gotten by calculation.  The reverse calculated "velocity factor" is .79.

The resulting antenna works extremely well and not only do I get channel 9,  I get the other VHF channel 12 and I get nearly every UHF channel as good as I get them in the Kitchen, which has two antennas I spent a month colocating for best performance.  Notably on the twinlead dipole in the kitchen I could not get all the UHF channels as well, but that might be as much a factor of location.

This from very cheap-to-make antenna which actually uses just the two wires in zip form as the "cable" from the dipole, and thence to an F connector.  Using a real RG6 might be a bit better, but far uglier.

Why the Radio Shack LX5 and derivatives have inverted tweeter polarity

There's no obvious reason for it.  The crossover is first order, which normally doesn't need it.

I have a cynical explanation.

Actually, it's quite curious that the nominal crossover points for the reactive elements in the crossover are 3300 for the tweeter and 1300 for the woofer.  That works because the woofer has a lot of resonance above 1300 Hz.  Even with the choke that large (1.0 mH for the LX5 and it seems 0.9 mH for the LX55) there is still a tendency toward bulging at 3kHz or so.

I think Radio Shack tried to lower the cost as much as possible, and rather than using a slightly larger choke, say 1.2mH or so, to get flat response at 3kHz, they instead reversed the tweeter polarity, which  also flattens the 3kHz peak (it lowers in the range 3kHz-8kHz).  They were less concerned about the loss at higher frequencies.

So, it was an attempt to make the speaker cheaper than doing it the right way, with a larger choke.

Now, this is just a guess, it could alternatively be the Linneum tweeter requires that inversion for some reason related to its polar response, which is so hard to understand anything is possible.

#1 Thing Subjectivists Do Wrong

Failure to do proper level matching.

That's precisely why the AudioGeorge LX55 modification doesn't work for me (or, I believe, anyone actually committed to "high fidelity").

Actually, both measurements and listening tests confirm that George's first modification, the tweeter polarity "correction" (corrected from the factory inverse polarity connection, which has no obvious explanation but was consistent) actually may improve matters, both in frequency and impulse response.  It certainly sounds better than the full modification, I've confirmed that in listening tests.  (I have not yet consistently compared with the factory version(s) in listening tests.)

Now all my listening tests are level matched...in this case using the weighted pink noise my Yamaha receiver puts out...for level matching purposes.  I believe it is identical or close to C weighted pink noise, what is recommended by objectivist experts and gray hats such as John Atkinson.

George refused to use such things.  He would only level match by ear, on the fly, adjusting level after any change to what sounded best to him.  And he insisted this was the ONLY way to do this.  We had a long argument about it.  To me, this was endless cheating, he could make anything sound better or worse than the last at whim.  To him, he was playing music to satisfy himself, so he ought to see how best each change COULD satisfy himself, by setting the level to what sounded best to him each time.

That position is hard to refute in it's own terms, I admit.  The problem lies in insufficient data, and believing that your auditory perception is unbiased and uninfluenced by other irrelevant factors, such as how far your wrist is turning.

In principle, George is right, the system exists to satisfy him, so every change ought to be measured as to how well it satisfies him.  The error is that in a quick twist of the knob, he has found the one and only, ultimate, and always most satisfying playback level given the state of the system.

The way the satisfaction test would really have to be done...would be to test each and every possible 0.1dB increment, from faint to blasting, in random order, under both conditions, to find the highest ranking satisfactions, and under each condition where such satisfaction ranking might vary (such as early in the morning, and after coming in from heavy traffic).

Failing to do that, he (and most subjectivists) have a strong tendency to favor the louder, because hearing bandwidth is increased for louder signals.  And the tendency, when adjusting volume, is to return to the same electrical level (same rotation of the knob, etc). So if any change makes the system louder, that is immediately what is preferred.

Anyway, that explains why he finds that removing the choke makes the speaker sound better.  With the same power input, it now sounds louder.  Nothing in the output has been restricted by this change, only more has been opened up.

When you do the level matching, however, the part that has been opened up now dominates the level more than it did before.  So, at matched level, everything else becomes LESS.

This is how, with level matching, a change which impedes nothing in the high frequency response...removing the choke on the woofer...can make the system now sound dark and univolving.  Removing the choke creates a mid frequency peak, relative to which everything else is now at a lower level, including the higher frequencies.

But if you don't match the level acoustically, more more like electrically, the highs stay the same and now the midrange is WOW.

Now, the difference is easy to see, but perhaps not entire unambigous in the frequency response graph.  Actually, with no choke, the highs are boosted past 10kHz!  But the boosting above 3kHz is far less than the boosting at 3kHz, with the audible result I've described, level matched.

What I might not have expected was the deterioration in the step response.  The step response is greatly messed up within the first 1 msec by removing the choke.  So much for "simplicity" and all that.  I suspect the high frequency output of the woofer is highly frequency variable, and without the choke it dominates the relatively nice tweeter output.

Listening to LX55's

I can listen to either the modified or unmodified pairs of LX55's now in my surround sound by selecting "7 channel stereo" mode, turning off the Hafler 9300, and disconnecting the other pair at the quite accessible rear to my Yahamah HTR-5790.

Notwithstanding he sonic differences may have something unavoidably to do with location and orientation, the modified units being in back atop the cabinets, and the unmodified units being to the sides  just over 90 degrees and with the left atop a cabinet, and the right on a wall bracket at the same distance from the ceiling and pointed downwards to help compensate it being a much sharper angle down to the main listening position, ..., I believe these differences I'm hearing are there in themselves as I hear them anyway, with some qualifications.

The modified rears have a somewhat beguiling quality in some ways.  They "don't sound hifi."  There's no electronic sound whatsoever, perhaps in part because there seem to be no highs above 5kHz.  (BTW, this is NOT what the measurements show...even in relative terms they are only slightly down from the unmodified sides at any particular frequency, with the exception of to the stellar Unit #6.)

They don't actually have a honky sound, merely kind of rolled off, exactly like a one driver car speaker.  And in fact that's about a similar grade 5" speaker.  Somehow, the lineaum tweeter just seems to disappear entirely, masked by the now extended range woofer.

That being said, I had to listen very carefully to be sure it wasn't just my previous attachment to what you could call Hi Fi, full frequency range reproduction, that was unfairly coloring my opinions.

And on careful listening, there did indeed seem to be ... added ... coloration of various kinds.  Listening to Dusty Springfield singing Look of Love, she takes on the sound of a 300 pound 4 foot 8 black singer.  As much as Dusty does unbelievably good Black Woman sound--her trademark--she doesn't sound quite like that, I have to believe.  Beguiling in some ways, but not real, not high fidelity.

I similarly didn't find the reproduction of Rebecca Pigeon to be quite on.  But then you could say I've never heard either woman in person, etc., I'm just imagining what they sound like, and that's true.

But really I find that instrumental music really lights up the differences better.  All the high register orchestral instruments sound wrong, along with the David Sanborn playing sax right after Dusty sings.  On the modified speakers, the sax sounds like an unidentifiable instrument I've never heard before.  On the unmodified speakers, it's the one I've always heard David Sanborn play.

Everything is way back, too, as if in another room.

Now, all this being said, I could raise complaints about the unmodified sides...starting with that "hifi" sound.  Perhaps things are too close up.  And there is far more metallic sound...but then that's what comes from reproducing metallic instruments including strings, winds, and percussion!

I could possibly even find the unmodified sides harsh at times, moreso than the backs.  But a lot of this may simply come from the side mounted position.  Overall, I just don't think I could live with the rolled off sounding backs.  The unmodified sides light up THIS room with symphonic music, the modified speakers seem to light up some distant room, with a lot of clarity lost in between.  If I wasn't listening to my stellar M20's (now with appropriate sub setup...a long neglected detail in the kitchen) I might occasionally confuse the unmodified speakers with them...when walking in from another room.  No such chance with the modified speakers, they are just different, like speakers from the 30's which were in fact designed like that (no crossover chokes), the classic chokeless sound.

This is a pretty good synopsis of what I heard 15 years ago after bringing these speakers home.  Somehow the AudioGeorge mods didn't seem to have the same magic they had at his house.  They seemed like a step backwards into car radio mid fi.  Was it has magnetic personality, or his anti-magnetic house, that made them sound so good there???

So now that I've made my mind up about that again, I think I am going to separate the two modifications and see how they pan out.  Perhaps just reversing (correcting) the tweeter phase by itself is a good mod, for example.

I'm considering units 2,3,4,5 having essentlially the same driver quality, only differing in crossover.  But anyway I now and finally again (since the death of the fiji card in my Win 95 LAUD setup, thank you planned obsolescence) I have a consistent way of measuring speakers, and can even plot on the same graphy (I finally figured out how to do that) so I can see the change wrought by each difference.

Surround Expansion not Working

It has become clear that my 5.1 channel to 7.1 channel expansion system based on an early surround processor is not working as I expected.  When either left or right surround is playing by itself, BOTH back surrounds play at a somewhat reduced level.  So the back surrounds are no longer stereo, as I intended.

I tried every control on my SSI System 1000 II processor, and nothing could change this.

I'm not actually sure I ever tested discrete 5.1 signals this way.  I simply plopped in a 5.1 disc, and played it through the processor, not at all, or unprocessed, and decided it sounded "better" overall through the processor in a quick subjective test.

What may be happening is one of two things.  Either the processor, now 35 years old, isn't working correctly.  OR this is correct operation, and what I'm seeing is actually the result of "steering", the function that keeps a sound localized to a single speaker if it seems to be coming from one.

Steering is very appropriate if you're deriving side surrounds from a 2 channel signal, encoded or not.  But I don't think it is really appropriate to deriving back surrounds from side surrounds.  The backs should be pretty close to the same signal, only with a tad of blending...less separation than the sides, but not mono either.  And possibly with a bit more delay.  That's what I had hoped to get.

What I really need is some more interesting test material, that moves a virtual source from right side surround to left side surround, and then see what happens in the backs with some surround deriving process.

For now, but this doesn't seem quite right either, I'm duplicating the sides in the back for 5.1 material.  I tried duplicating and not playing the backs at all, and decided I liked it with the backs if I set the backs to a much reduced level compared with the sides.  This only seems to widen and pull the sides slightly to the rear from their roughly 90 degree position appropriate to a 7.1 system--the inherent incompatibility between 5.1 and 7.1 I discussed a few months ago thereby being somewhat fixed.

However, I'm using my SSI 1000 to adjust the level, from stereo inputs to front outputs.  This is unavoidably removing the center information from the backs...which I'm not sure is a good idea either.  A simple buffer with volume control might be better.

Measuring ALL the LX55's

Today, I decided, was the big day.  I measured all 6 of my LX55 speakers in a consistent, hopefully repeatable way.  Unlike the way I measured 3 speakers yesterday.

I adjusted an Athena speaker stand I haven't used in years to the maximum height.  That turned out to be what was needed to align the bottom of the speaker with the UMIK-1 measurement microphone in my microphone stand.  Fortunately I had previously obtained the correct adapter to attach the UMIK-1 microphone coupler to the stand (along with a few others).  This was a bit over 3 feet from the floor.

I actually adjusted the microphone to be at the height of the center of the woofer, believing that to be a pretty good spot for these speakers.  Actually, in my near ceiling surround positions, I'm way below that axis, which I'm pretty sure is suboptimal for these speakers.  But I think they are supposed to be on fairly high stands so you are not too far ABOVE the tweeters.

The microphone and speaker were adjusted to be 1M (39 inches) appart, plus or minus an inch or two.  This is far better than the 10 inches yesterday.

I put the microphone toward the back corner to be out of the soundpath of the refrigerator, and the speaker in the middle of the room facing towards it.

I've never done a series of speaker measurements with such a serious set up.  And now, including a calibrated microphone and the industry standard program REW.

One big surprise was that the improvements I made to the right surround speaker, called previously and still unit number 1, still didn't seem to have made it a very flat speaker.  It's too much work now to go back and restore the factory out-of-phase tweeter, but previous measurements showed that to be horrifically cut off above 5kHz.  But it's still pretty much shelved down, starting at 3kHz, making this speaker fairly useless without further modification:

I had to go back and double check that it was this speaker I had actually modified.

Meanwhile, I hadn't for a long time measured the two speakers with the complete AudioGeorge (tm) modification, which I bought directly from George himself.  This modification consists of correcting the tweeter polarity (as I just did for unit #1), and also bypassing the choke which normally rolls off the woofer, as well as damping materials on the tweeter and straws on the port.  (I have sealed the ports on all my surround speakers, which is a more drastic measure, to help compensate for near ceiling corner locations.)  I had previously disliked the choke bypassing, feeling it produced some hump in the upper midrange around 2kHz.  In fact, it does, but it also seems to contribute to a generally flatter or at least more extended upper frequency range from 1kHz to 10kHz.  Though we can't tell exactly yet how much of this may be due to drivers with better response than in unit #1.  Here is LX55 number 2, which has these modifications.

This looks greatly preferable to the partly modified #1.

Meanwhile, #3, another AudioGeorge unit, looks pretty similar to #2, which is a good thing, maybe even more extended and better in the highs, before the seemingly inevitable 10kHz roll off.

Unit #4, which is a stock unit whose woofer I damaged in January (the woofer surround was so flaky a mere touch of a screwdriver caused it to break into pieces).   This still looks far better than unit #1 even after having done the phase correction on #1.  Which begs the question, has unit #4 had the phase reversal mod?  Not as far as i know.

Unit #5, had been in storage, climate controlled storage, until I damaged the woofer in unit #4.  Then I brought #5 out of storage to replace #4 as the left side surround.  This is another stock unit, I think, but measures the best of all the stock units I've had for over a decade.

Now, I recently acquired a sixth unit, with a mind to ultimately expanding to a "9.1" system, and also doing lots of tests, experiments, and modifications.  This new unit #6 is clearly the best measuring unit of all, the most extended high frequency response anyway, with only about 5dB of decline at 10kHz, and only 12dB at 20kHz (at which point, it might pick up, according to the wider bandwidth LX5 measurements by John Atkinson).

Well, except for that hump in the bass, which I had not dealt with until today.  After stuffing the port with the same polyurethane foam backing rods I used for the other speakers, I got flatter bass:

Now for unit #6, I don't imagine the phase reversal to have a positive effect.  The response 200Hz to 5kHz is virtually flat!  Changing the tweeter polarity and especially bypassing the woofer coil is going to cause elevation in the 3-5kHz response.

So if something ain't broke, don't fix it, I decided.  I decided to use Unit #6 as s drop in replacement for the funky unit #1, which can't seem to achieve extended flat response regardless of tweeter phase, as the right side surround.  Meanwhile, the here-to-fore hardly used unit #5 is the best of the remaining unmodified units, on the left side surround.  And with the existing AudioGeorge units in back, I have a full set of pretty good LX55's.

That solves the immediate problem, and right away the surround setup no longer has a dark sound like it did before (just one dark sounding speaker can do that).  I had to readjust the level for the right side surrond a little, and it is very reassuring the speaker now sounds pretty much like the others.

As to what is going on here, one thing seems to be pretty obvious.  There is high variability in the drivers is these inexpensive speakers, just as John Atkinson noted, which can cause excessive rolloff in the highs above 3kHz.   In some cases, but not all, it might improve matters greatly to provide the AudioGeorge modifications, which tend to boost the highs.   But not always, and that's all I can say now (though I suspect speaker aging, and the woofer dustcap resonances, are critical factors).

It may be, that the now useless unit #1 could be salvaged with the AudioGeorge modifications.  It's worth pursuing, but now that I have a great set of 4 for my "7.1" system that are all pretty good, I feel that the important work is done here for now.

It may be more interesting to see what the modification does for unit #4, since #4 has virtually identical response to the now-in-use #5.  It's a decent way of doing the experiment without messing up my now better-than-ever (as of Friday morning, all channels leveled within 1dB, something I'd never been decently able to do before) surround sound system.  #4 is currently more rolled off than the new #6, could the AudioGeorge modification bring them both to the same level?  I'm currently disinclined to mess with #6 for many reasons, including how much of a pain it is to get back on the wall again, and also it just seems to work best of all right out of the box.  But if I get favorable results with two other speakers first, that could be the third.  If.

The long term plan for #4 and #1 is woofer and even tweeter replacement experiments.  At least #4 has a messed up woofer that needs replacing anyway.  It could be the tweeter on #1 is defective somehow.

But meanwhile, also, I'm thinking of a more serious, ear level surround system in the kitchen.  That would require a new set of 4 Gallo A'Diva SE speakers, in Stainless, which will depend on my cash situation later this year.  The current surrounds would then become "ceiling" surrounds...which would require a processor, as would the intended front width and height speakers...which should probably be SE's as well.  My ultimate planned LX55box + special 5" woofer + Infinity Walsh tweeter system will be the ultimate side height surrounds.

For now, the improvement in my surround system (with fixed right surround--the new one) is comparable in magnitude to the discovery of how to enable true 7.1 from discs, which was just two months ago--that was when I spun the wheel and did the the final Oppo update on my BDP-95, which seems not to have adversely affected anything, including my later ability to use "the ISO" disc to make it region free, and the update made the rear surround outputs work for the first time in 10 years.  It hadn't been an issue until this year anyway, after I added the permanent rear speakers, so I finally decided to take the risk of update.

Updates Friday May 3

Actually, John Atkinson recommended having the speakers on low rather than high stands, because below the tweeter axis a suckout quickly develops at 3kHz...  I was measuring on the woofer axis, but at least consistently on all speakers.

I couldn't resist the desire to see what would happen if I did the full AudioGeorge crossover modification, eliminating the choke on the woofer as well as reversing the tweeter phase.

And the result is, yes, the combination of both of these changes does boost the highs to be pretty amazingly close to an unmodified unit that "works" (but not the amazing new NOS unit).  Also, not quite as far as my two actual AudioGeorge modified units.  So it looks like those units are intrinsically among the "better, but not best" category of LX55 drivers, and Unit #1 is intrinsically a bummer of some kind, suggesting there are at least 3 cateogries of LX55 drivers, #1 being the worst, and all but #6 being he middling kind.

Unit #1 with both crossover mods

Unit #5, Stock

While #1 with the double crossover mod is similar to #5, there are telltale differences, such as the apparent bulge right around 2kHz, where #5 is stock.  That seems to be a telltale sign of removing the choke, it adds a lot more output just below the crossover frequency.

Carefully studying the #1 before and after the second modification graph, it appears that unchoking the woofer adds output all the way up to about 7kHz.

Unit 1 with both mods

Unit 1 wih only tweeter polarity change

Now I haven't listened to #1, but I have listened to the fully modified units #2 and #3, and generally I find the bump at 2kHz adds a somewhat honky quality, whatever the benefits of greater extension provide.  But for unit #1, it was so rolled off and dark sounding that now at least it should have credible highs.

In both cases, however, notice that unit #1 actually has response extending to 20kHz at the same level as 10kHz.  On the actual tweeter itself, it's pretty level from 5kHz to 20kHz.  This suggests to me that the tweeter is at a too-low level compared with the woofer, in effect a "low output" lineaum, but one that turns out to have more extended response than the others, at the lower output that it does have.

An additional test on unit 1 showed that bypassing the light bulb "fuse" did not make a significant difference.

Testing the defective LX55

It was clear that my right surround LX55 is defective, it always sounds dark or recessed compared with the other speakers.

Using REW, and my UMIK-1 at a distance of about 10 inches with attempted (but not achieved) consistency in placement, I tested 3 LX55's, the defective one (1), another unit with damaged woofer surround (2, actually I call this "unit 4" because it was the original left surround until I damaged the surround of the woofer slightly), and a NOS LX55 I recently acquired (3).

The defective unit showed a large drop above 5kHz, from there fairly flat to 10kHz, then down to zilch.

Number two showed a more gradual from 3-10khz, but 10kHz was equally down.

Number 3 showed only a minor loss around 6kHz, otherwise fairly flat to 10kHz, then sharply down.

At first I figured the tweeter in unit 1 may have been fried, what I was seeing was the woofer response.  I tested that theory by disconnecting the woofer.  Now the tweeter response was clearly visible, and, strangely, fairly flat from 5kHz to 10kHz.

Next I tried reversing the tweeter polarity on unit 1.  This made it somewhat similar to unit 2 (making me suspect unit 2 already has the tweeter phase reversal modification).  But it did nothing the bring the upper highs back up around 10kHz, and therefore still showing considerable rolloff compared to unit 3.

This looks like the difference between units 1 and 2 might be tweeter phase reversal.  It's possible that unit 2 already got this modification.  But the extension to 10kHz is apparently something different.  The NOS unit 3 has that because, possibly, it hasn't been used as much.  Or perhaps there is just considerable unit-to-unit variation, and unit 3 is the lucky one that came out best.

As good as unit 3 is, I wonder if it would be even better with the tweeter polarity change.  I suspect it would fill in the 6kHz suckout, but it would also probably add more bulge around 3kHz.  I'm almost inclined just to take unit 3 to replace unit 1, with or without the phase reversal, but then how would it compare with my other units???

THURSDAY UPDATE

Interpreting these graphs, despite their "objectivity", is full of ambiguities.  Despite using REW for the first time, which was a major accomplishment for me (requiring the new DAC, etc), I was unable to control exact microphone and speaker positions.  I started out planning to mark positions with masking tape, but as all was being done on on a placement on the marble coffee table, the placement pulled up and all was lost.  Everything was unavoidably moved on each trial, and with a mere 10 inches from microphone to speaker, which itself is not good at all, but it also made it difficult to assess all the angles precisely eatch time.  So I plan on doing a full set of greatly improved precisely positioned measurements of all my LX55 speakers today (Thursday).

However one pattern seems pretty clear.  And that is that the new LX55 has much more extended response, almost flat at 10kHz, whereas the other two speakers are down 10dB at that point, and this was true regardless of the tweeter phase reversal on unit 1.  That reversal did fix the 5-10kHz suckout, but exchanged it for a gradual rolloff down to 10kHz.

John Atkinson measured several LX5 units, and also found big differences similar to what I show (loss of about 10dB at 10kHz in the badly performing unit).  He determined however (via unspecified means) that all drivers were connected the same in both speakers regardless of measurment: the woofers were in polarity and the tweeters out of polarity.  Thus, in his evaluation, miswired polarity did not play a role in these differences.  Instead, he suggested quality control of the tweeter was an issue.

I have often been skeptical, thinking perhaps he merely checked the wires in the speaker, and perhaps the wires themselves were reversed within the tweeters themselves.  But the truth is he does not say how he did this driver polarity assessment, and he could have used a dirac or similar test signal, which someone like him (and me now actually) would have right at hand.  (For me it's not just the test signal, it's getting one of my scopes to trigger properly on it...)

However the findings shown here suggest that Atkinson nailed it, at least that it is a driver quality control issue.  While the defective unit is somewhat better wired in different polarity, it does not fully make it like the good unit.  Notably, the total loss at 10khz is unchanged.

But thinking about this over and over, and looking about my pretty good measurment of the tweeter itself, I'm thinking perhaps the tweeter is not the source of these differences.

Perhaps they are largely determined by the dustcap resonance of the woofers.  That resonance may change as a result of the history of heat and humidity it has experienced.  In some it has a broad 5-10kHz resonance, and in others the resonance is mostly gone.  And, following this theory, it is the better measuring speakers that have the most dustcap resonance.  Further, following this theory the tweeter itself has fairly flat response...but it is adjusted fairly low in level so that the woofer dominates the response up to 10kHz.

But perhaps it only appears this way because of sloppy measurements.  Testing the theory fully requires a fair amount of soldering and resoldering, which I try to avoid, and I want to assay all my surround speakers too.

XtremPro X1 DAC

I need some kind of fairly decent audio output from a laptop for REW.  The Built-in audio on my Asus laptop computer (the i3 "SonicMaster" model from a few years back) is horrible, I can hardly get it below 3% THD, and it's full of spikes and noise.  Then I was unable to get my old Emu 0404 to work at all on this Windows 10 computer, even with the last drivers from Emu (which were beta drivers for Windows 7).  So I ordered the new cheapie winner recommended AudioScienceReview.com (I'm a fan of Amir who does the best measurements of all, and finds the cheapie gems as soon as they appear, not being hung up on the dross from the usual sources, though I think he exaggerates the importance of various kinds of technical performance, but then so do I) the XtremePro X1.

In this case, I plugged it in, and Windows immediately recognized and installed some driver for it.  Sometimes Windows immediately makes it the new default "Speaker" output when it is plugged in, other times I have to use the Windows Mixer to select the XtremePro, it helps if no audio application is running to change from built-in Realtek audio to the USB Dac.

I find the little black cable that comes with the XtremePro isn't necessary to make the DAC work, but it IS necessary to get the DAC to work reasonably well.  Distortion at 1kHz was above 1% without the little black cable, and 0.2-0.33% with the cable.  I suspect the cable reduces some HF noise on the power line or isolates the ground.  I am using ungrounded and full balanced input into my 1700B analyzer.

The lower distortion 0.2% is at approximately 1V output, -9dB on the REW generator control panel.  Distortion rises to 0.33% at the full -3dB.

Frequency response falls by 0.1dB at 20Hz and 20kHz relative to 1kHz, and about 0.3dB at 10Hz.

Not as good as I'd like, perhaps related to the USB on this little laptop, but certainly good enough for acoustic measurements within the standard audio range 20-20kHz, and with some care down into the infrasonic as well.  This notebook would not be useable for audio experiements without something at least this good (despite the name SonicMaster...).  I suspect it would be good into the near ultrasonic as well, but I didn't notice the sampling rate adjust on REW until after I'd shut down the analyzer.

(Note: It's still possible that performance defects are due to the laptop I'm using and the tests were done with power connected...a further issue and a retest is in order.  However, I just broke the input terminals on my 1700B !!!)

While the Dac is good for the money, and drivers were installed automatically by Windows 10, the "Quick Installation Guide" is quite funny.  It's a single page, double sided, in multiple languages, which actually says nothing at all about installation.  The short sections are entitled  Capabilities, Features, Environmental Information, Declaration of Conformity.