speaker dave

I'm guessing the fan blade is tied to the core pole. We did a similar heatsink on a 10" woofer used in a couple of Snell models. It did lower the VC temp a little but we had to deal with air noise issues since you were essentially loosing the dustcap.

As a young speaker engineer I wrote to many companies looking for work. Roy was the only one who took the time to write back and offer some personal advice and a friendly note.. A true gentleman. Rest in Peace Roy

That looks like a very clean and professional repair ot me. I wouldn't worry about having a half roll where the previous surround was flat. Considering the alternative (dead speakers) it seems like a good choice. Clearly AR knew that the excursion required was nil and so gave it a flat surround, but foam surounds are fairly light and I'd bet the response is little changed. David S.

Here is a nice PDF drawing of the final modified network, courtesy of Richard Lowe (thanks Richard!). I need to explain that there is a little controversy over the woofer inductor. I've measured mine twice and get 1mh (or 0.99mH) with a digital meter. The AR lists show a number 4 inductor as 0.88 and the #5 as 1.2mH. I believe mine is a number 4 but the value is not to spec. If you are going to do the mod and have a #4 measuring 0.88, then don't worry about the exact value or adding inductance to up the value to 1.0, at least until we can see if there is a consensus about a possible third value. My system serial numbers are FX141819 and FX157677. It would be nice if others could measure their inductors and see if any serial numbers close to mine are 1.0 or 0.88. Those with a 1.2 can do the full mods as well but optimally the 50uf, 6.8 ohm combo might need a little adjusting (the 6.8R would probably increase). Hard for me to say without having the systems to try. (Zilch, what is your woofer and inductor stockpile like??) Anyhow, here it is and have at it. David AR4x_Mods.pdf

Here is a schematic of the final network. Pardon the hand drawing, perhaps someone can do up a proper drawing? In the end I have put the variable resistor back into the circuit as an aid to those with unequal tweeter sensitivities (a seemingly common problem) and also to let you "adjust to taste". Early on I found that just a series R (rather than potentiometer connection) suited the response shape I was trying to get, so the pot is connected differently than the standard AR configuration. The 10 ohm resistor around the pot reduces its total range and puts the expected 4 ohm value about 1/2 way along its travel. From my adjusting I know that most of the component values are non critical, with the exception of the tweeter inductor. When I get a chance I'll measure the DCR of that choke and add that to the info. It was a fairly fine wire over one of the compressed dust cores, nothing special. Note that my speaker pair have the 1.0 mH AR inductor. If you have the 0.8mH inductor you should add 0.2 in series. Also note that the tweeter polarity is reversed from stock. Good quality low loss electrolytics will work fine. Please don't do anything silly like buying a film 50uF (just so you can put 6.8 ohms in series with it!). You can substitute 47 for the 50 and 20 for the 22, if that is what you find locally. Let us know if you mod a pair and how it comes out! David

Yes, that is the floor bounce dip. The mic was about 3' 10" in front of the speakers. Tall stands had the systems about 3ft off the floor, so the dip frequency could be calculated from that. I tried the mic back at the listening position and at that resolution (1/6th octave) the dip is gone, replaced by lots of room effect wiggles that average out. I didn't notice it in the sound (actually it didn't appear, as shown, at my position) but I did notice that the sound was a little "phasey" when I leaned back in the couch and a little clearer when I leaned forward. These were spatial effects and didn't seem to "overwrite" the frequency effects. David

AR4x with 3rd order network vs. Snell K5 Just to throw another system into the mix I did a final comparison between the fully modded AR4x and a Snell K5 bookshelf system. This was a fairly expensive unit with Peerless 6 ½” woofer and metal dome tweeter, a damped laminated baffle and sophisticated crossover. (Also a beautiful Rosewood cabinet made in our factory in Haverhill Massachusetts.) Of the Snell products we did in that era this was fairly successful and something I would offer up as a well balanced system. From the notes on the fully modded AR4x #2: “A richness in the lower 100s. Nice overall balance. More ‘eh’ sound to treble.” On the K5: A little more mid emphasis.” “Reminiscent of previous comparison (somewhat like the 3rd order AR4 without the woofer additons).” “Piano slightly dull. More ‘ow’ sound.” And finally a pink noise comparison: “More upper mid/lower treble” (on full mods AR4x) and More “polite” or recessed for the K5. Note that the K5 has more bass extension but it gave up nearly 3dB of sensitivity to get it. The AR4 is not as insensitive as we would guess. In truth AR speakers were only low sensitivity relative to the 50s and 60s systems that preceded them. The curves for the two speakers are surprisingly similar, with no clear trends to suggest either is superior. The flip flop of levels between 800Hz and 2kHz is probably what leads to the balance judgments but we are getting to the limits of what our curves can tell us. Conclusions? As I stated before, I have to admit to a heavy amount of bias in this test in that I had modified an AR4x to measure flatter and was likely to prefer it no matter what! Still, I believe the descriptions of the sonic qualities and the magnitude of differences that I have described are accurate and proportionate, and that other experienced listeners would have similar observations. I might tweak the 1kHz level up a tad and the 3kHz down a smoosh, but otherwise I think it is about right. I think the end result is a very good speaker that I could happily listen to any type of music over without the system ever getting in the way. These are not exotic drivers by today’s standards yet the end result doesn’t have anything to apologize for. It really is competitive with modern systems of similar size and bigger systems would only surpass in areas of bass extension and output capabilities. So it meets my initial goal of being based on AR parts, still looking classic, but not sounding dated. For those that have stuck with the thread this far: Bravo! Hopefully you have picked up a bit about the design process as some of us practice it. It is a bit science and a bit art, but no real magic is involved. Its time for some others out there to try their own modifications and see if they like the results. Regards, David

3rd order without woofer shelving vs. full 3rd order network For the third test I had connected my second 3rd order tweeter network to the #1 system, but not the extra 2 components that convert the woofer to a damped 2nd order network. By this point the differences are becoming minor and a preference is less consistent. From the notes: “System 1 (all mods but woofer network) darker balance, rich”. “Piano a little dull, sax nice, voice a little “shouty”. The fully modded AR4: “sax thin, better balance between fundamental and harmonics on Soprano.” At this point the differences were still fairly obvious in a direct comparison but either system would probably sound fine in isolation. If you like a little bit of the classic AR balance then this version (#1,no woofer mod) would probably suit you well. It was really nice on symphonic music and had a nicely rounded quality to it. I thought the fully modded system sounded a little more neutral, but might be accused of having a bit of treble edge. The curve shows the difference the second order woofer network addition makes. Orange curve is tweeter net mods only. The green curve has the woofer net mods as well. It pulls down the region from 800 to 2kHz or so by about 2dB. I wonder if pulling this down too much has left the 3kHz region a bit exposed (“Fizz on Sax”)

Revised first order network vs. 3rd order network I had previously found that a first order network worked a little better with the 20uf cap reduced to 10uf and tweeter changed to in-phase with the woofer rather than reverse phase. I adjusted the #1 system to that configuration and compared it to the 3rd order modified system. The listening notes again concentrate on the characteristics of the system with the simple network. “Piano: dull, Soprano: dull, hollow. Saxophone projects” (3rd order system: “fizz on saxophone”). Symphonic music was a better match to this simply modified 4: “Strings in upper range good. Slight mid emphasis.” “Oboe bad, “aw” sound.” By comparison the heavily modified 4 had “more edge to voice” but I noted that, in general, the differences between the 2 systems had been "cut in half" relative to the stark differences of the first pairing. Curves for this pairing show much less 1200Hz bump but still a quick drop above that frequency and a low general level for lower treble. Instruments that straddle this transition are clearly going to be effected. This relative difference between mid level and treble level led to the mid emphasis and dull sound.

Stock network vs. 3rd order network The stock tweeter network is a 20uf capacitor leading to the 15 ohm potentiometer, with the tweeter reverse phased from the woofer. My notes ascribe a “shouty” megaphone quality, especially on the Joe Williams track, to the stock AR4x. On switching from the modified 4, especially on the piano piece, there was a strong nasal quality. Some times colorations can be described as contributing vowel sound qualities and the stock AR4 had an “aw” vowel quality. Massed strings playing in their upper registers sounded okay but Clarinet and woodwinds shared the nasal quality. After some minutes of listening to the stock 4 the balance was more palatable but the midrange prominence and a general treble dullness was always there. Immediately upon changeover, though, the nasalness was striking. The modified 4 was clearly flatter in the midrange and had more upper harmonics more of an “eh” lower treble vowel sound, at least in comparison. My notes say “voice is sharp but not resonant.” “Piano has a slight split with the lower register warm and the upper register light”. I brought my wife in to see if she would notice the differences and how she might describe them. All I said was “here are two systems, based on similar parts but both of which had been modified. Can you hear a difference?” We switched back numerous times between the two while listening to the Soprano singing Loewe. Finally she said: “I prefer the one on the left (the 3rd order network). It sounds crisper and wider.” “The one on the right sounds like a room with no drapes.” That seems like a pretty apt description of the resonant quality of the stock system. The curves show the comparison of the two. The orange curve is stock and the green curve is the fully modified system.

Listening test: preamble In the last weeks I’ve worked on an increasingly complex crossover revision for the venerable AR4x. From the series capacitor (1st order) stock network I’ve worked up to a third order tweeter network and even added an order to the woofer network. A number of you have asked about the sonic differences and so, at your requests, I’ve just finished a long session of listening to most of the variations that I had previously measured. In most cases I compared my # 2 system with the crossover internal, to the #1 system with various crossover iterations attached via external wiring. Finally I compared the fully modded #2 system to a Snell K5, a similar sized 2 way bookshelf unit that I designed 7 or 8 years ago. Am I biased? Of course I’m biased. I know that the 3rd order network measures considerably flatter than the stock network. I can remember the approximate response curves of each variant. So it wouldn’t be too hard to predict what the sonic balances of the various iterations are. You’ll just have to trust that my descriptions are “fair and balanced”. I will do my best to give useful descriptions of what is heard, in a manner that anyone can relate to. No nebulous “speed and timing” type descriptions but more how each speaker’s balance impacts a variety of instruments or the human voice. I’ll include a real time analyzer curve taken of each paring taken after I listened to the particular pair. To keep things simple I picked one sampler disc that had a variety of classical and jazz tracks, all from the Delos catalog and all recorded by John Eargle (much missed). Four tracks were principally used for every comparison. For a symphonic work: Prokofiev, Three dances from Romeo and Juliet. A Haydn Piano Concerto for the sound of piano, a female soprano singing “Before I Gaze at You Again” by Loewe and Joe Williams a “Mean Old World” for a Jazz piece. I listen to plenty of rock music, at least 60s and 70 folk and geezer rock, but these particular tracks covered the bases well. Rock music would have gotten me more focused on the frequency extremes and output level issues. Rather, my focus in listening was for matters of balance and midrange neutrality, as this is what you would be striving for while finalizing a crossover. The speakers were on tall stands and well away from the walls, driven by a Scott integrated amp of about the same vintage. I sat about 2.5 to 3 meters away on a listening axis about even with the top of the woofer. At that position the bass was smooth if a little light. With the chosen pieces bass was not a big factor and of course the crossover differences wouldn’t impact bass. Please realize that the lowly stock AR4x will suffer some criticism, but we love it just the same since it will prove to have more potential then ever we could have guessed! I’d also note that the losers in these comparisons are frequently speakers that we would be perfectly happy to listen to music through. Comparison tests magnify the differences, while long listening sessions let us get used to the “personality” of a system.

Thanks John, for all of your useful information. Sounds like three variables here: FG stuffing, woofer inductor, and tweeter level. As to woofer stuffing that is independent of the crossover and people can choose whether they want tighter base from more FG or not. It won't impact the crossover range or any notion of flattest midrange balance, since there are a number of octaves between the woofer corner and the midrange we are playing with. The inductor value, since it appears to be a revoicing (rather than a woofer correction) would need to match the value in my speakers if anyone wanted to build their own clone pair. I'll measure mine and see which it is. If it is the lower value #4 then someone wanting to achieve the same response could take off turns. Alternativley they might add turns if it goes the other way. It might be possible to trade the woofer external cap and resistor mods (at least their particular values) for the difference in inductance but someone else would have to do that experiment. I am modding the pair I have and can only guess what changes would be needed for a different starting point. I think you are right about the need for a tweeter pot. I am happy to fix the treble level on a pair for myself, based on measuring and listening, but tastes will vary pretty widely on this and we are seeing a great range of variability in tweeter sensitivity. Not sure what you meant by your last paragraph. Obviously the crossover order change is effecting the character there, that being the main intention. If you are worrying about crossover frequency shift with tweeter level, one of the curves above shows the effect to be largely a shelving down of response for the low sensitivity tweets, with no other side effects. Regards, David

Your pair looks like my pair (she said with a wink). I was thinking of fixed resistors, and probably will keep that with what I build up. But for the sake of the experimenters that want to try their own, it looks like some variable resistance to be set by ear will be called for. If I remember right the shape worked well with series resistance only (rather than a potentiometer connection) so the 15 ohm pot with another resistor around it might be the best answer. Second topic: I was looking at some of your threads at AK and other threads here and saw the inductor #4, inductor #5 issue. I'll have to check to see what I have but this adds another complication for those wanting to duplicate this. I wonder if that was a conscious re-voicing or to compensate for a woofer change. Anyone? David

I tried changing the first cap (was 22uF) and the second cap (was 10 uF) and in the end even changing both a significant amount didn't get me quite to parity. This all makes you think about restorers that are matching their parts to 1% or better. Good idea if you drivers are really close, but what does it get you when tweeters are 3dB off? This curve shows the front 22uF cap with +5.6uF and the 10 + 4uF. We are now about within 1dB of the other system but I've run out of time for today. I'll tweek a little more when I get a chance and, I promise, next time I'll do a proper listening comparison! David

To get the treble end of both systems to match I'll probably have to adjust more than one component. Lets start with the padding resistor. If I put 3.3 ohms in parallel with the current 4 ohms that will give an effective pad of about 1.8 ohms. This is a pretty big change but we have a long way to go. As an asside, if you were a speaker company you would most likely buy drivers that were guaranteed to have broad band sensitivity held to about +- 1.5dB from an approved sample (bigger deviations at spot frequencies would usually be allowed as well). You can do better but you will usually have to pay a premium for it. The smaller resistor value pulled up the top end of the tweeter nicely, but we are still pretty far off at lower treble frequencies. Orange is sytem #2 with no change, blue is the smaller total resistance, and the top burgundy curve is system #1 that we are trying to match. David

With two boards done I could hang one outside the system I had been working on and place the second inside the other cabinet. The #2 modified system will be listened to with the new network and the first one (with its connection points outside the box) modified back to AR's original network. A quick measurement of system #1 showed the tweeter reverse phase from what I expected. In fact I'm now totally confused about its current phasing vs. the factory network. I had thought that AR had the tweeter out of phase with the woofer. Earlier measurements showed most midrange output with that phasing, although a smaller tweeter cap and tweeter and woofer in phase did give a little better response IMHO. The third order network pretty clearly needs the woofer and tweeter in the same phase. Usually a battery to the tweeter terminals will reveal which is the plus terminal but I just couldn't see enough motion to tell. I'll keep digging and see if I can clear up the confusion. Anyhow, with the right polarity the new board mounted network looked great for system number one. For number two, it was immediately obvious that the second tweeter wasn't a great match for the first one. Sensitivity was 2 to 3dB down all across the board. Response shape was good and the driver isn't defective (it sweeps clean) but I'll have to compensate for the lack of sensitivity. The woofers are quite close, within a dB, and good enough for me. The orange curve is a built up network for system#1. Blue is system #2.

Part 4, Bumps in the Road So far we have one prototype that looks very promising with significantly flatter response and good on-axis summing. In actuality it was a mess of wires and alligator clips and dangling parts. What I need to do is build it up in a more permanent basis. At the same time, in order to do a listening comparison I'll need to get one speaker working with the revised crossover and one to the original. I decided to mount the tweeter crossover parts on a board and the 2 woofer circuit elements directly to the woofer terminals. I had a small double crossover (passive subwoofer crossover) board with solder terminals and a symmetrical layout, so it could be cut in half and used for both crossovers. Parts were wired point to point and then glued down with a silicon caulk. For no particular reason I wired it from back to front (right to left). The drawn schematic is left to right in the normal way.

All those pretty curves and you want me to listen to it too? Geez! I'm workin on it! I'm workin on it! David

First off, I don't want to sound critical of what was a very successful design for a long period of time. We can only speculate while looking at the system several decades later. We do know that the AR design preference at the time was to optimize the individual driver and crossover bandpasses in a reverberent chamber. This would emphasize the power response of each driver while downplaying the axial response. When looking at the woofer and its crossover, and the tweeter and its crossover, in a reverb room, the primary difference would be the woofer's response. As any woofer rolls off considerably off axis, it would effectively have a lower crossover point in a reverb room. This would downplay the overlap issues and might suggest a larger capacitor value as optimum. AR also prefered a balance that followed measurements of concert halls (room R converted to a response curve). This, along with optimizing woofers to be flattest in 2 pi (half space), is the reason why every early generation product seems to have a family curve that crowns at mid frequencies when the system is measured in a free field. The 2pi-4pi difference gives an uphill trend in the woofer range and the down-tilt in the treble matches the concert hall trend. None of these are matters of "right philosophy, wrong philosophy", but are typical af a general industry-wide evolution in design approach over the years. We certainly saw the same thing within AR: the AR 9 is designed to a flatter free field response and has key features such as the acoustic blanket, that only impact the direct response. Its also worthwile pointing out the contribution of test equipment over the years. Much of the Linkwitz/Reilly approach to crossover design, which is the basis of my discussion of "adding an order to get the phase more in line" would have been, as a practical matter, impossible in the 60's. The essential component of measuring phase response couldn't be done on a frequency sweep basis. The first tool for doing that that I recall was the B&K phase meter of about 1980. It still didn't allow removal of excess phase, the 1 or 2 meters of air path delay, so it wasn't a great help. (My rusty recollection.) It really took MLSSA, TEF and FFT methods to give easily made phase curves. I don't think capacitor cost was a factor. But I'm sure that a more complex crossover, certainly a third order network, would have been rejected for reasons of cost. This is again typical of the times. JBL made a lot of money selling L100's with 2 caps and 2 L-Pads, nothing else. An inductor on the woofer was already an extravegence on AR's part, in a budget system. Regards, David

I want to build up a pair of crossovers and see if the values are a "best fit" to both tweeters and woofers that I have. Then document the inductor resistances and confirm polarities, etc. (Hey, there's a lot of pressure to get these things right!) But don't let me stop you from doing your own mods. You are more than capable! David

I ran across this curve on the Stereophile website. It illustrates a point made way back in post #2 about the problem of a first order network on the tweeter. While first order (a series capacitor) should give a 6dB per octave electrical rolloff, this only works well when the load is a resistor. When the load is a tweeter and its realworld impedance curve, then things won't be so tidy. At the tweeter's resonance its impedance will go quite high. This is a "light load" on the network and the voltage will rise also (in the graph the bump at 500Hz). At frequencies above that, the tweeter impedance will drop to near its DC resistance and voltage at the tweeter terminals will fall. Finally, at highest frequencies the inductance will bring the tweeter impedance up again and the voltage will rise again. One solution for this would be a Ferrofluid of the heavy damping type. It would drop the impedance rise at resonance (lower Qm) and give a more classic highpass. For these reasons text book crossovers are never very successful. A software based approach must always bring driver impedance into the calculations as the load on any proposed network. David

Will do. Zilch is already helping on this. David

Absolutely, but this is a hobby, an interesting waste of time by definition. In the end I'll have some totally stock looking AR4's with performance as good as most any 2-way system built in the 60's up through the modern era. In Hot Rod terms: a Resto-Mod!

Almost forgot... Carl had asked for a nearfield woofer curve. I don't have an impedance curve but it looks like roughly a Q of 1 or so at 65 - 70Hz. The woofer and network inductance tends to obscure the exact Q.

And finally a comparison of the original network (red), the best 2nd order network (blue), and the new 3rd order network (green). we've been able to further reduce the variation and can now fit most of the curve into a 4 to 5 dB window. Note that I did the latest measurements in a different room and position, so the low frequency room effects (especially from 100 to 200Hz) are different. Perhaps its a more complex network than absolutely necessary, but there are many speakers on the market that would pursue this complexity in order to achieve the best overall result. So far lll of this has been done by measurement, but there will be a listening phase next. I like to get the sections working and blending as well as possible and then fine adjust a few key components by listening test. Certainly the tweeter resistor and the new woofer damping resistor would both nicely control level for a couple of octaves in their respective bands. They will be tweaked if needed.