Jump to content

AR-3a / AR-91 Combo Build


Recommended Posts

This project came together slowly, until enough parts had become available so as to reach a critical mass, and actually cause me to pick up a soldering iron.

Some might recall the AR-91 rebuild from a couple of years ago that began with a pair of beat-up cabinets; while the resultant sonics were very pleasing, there was just something about the size & shape of the '91 cabinets that didn't blend with the room's aesthetic, and when I bounced it off my Significant Other, I was pleased to learn that she was open to a more "vintage" look for the speakers in this space.

After A-B testing the AR-91 systems against a pair of borrowed 3a's, we decided on the AR-3a as a starting point; but I knew that I wasn't interested in restoring a straight-up '3a...while I've owned, repaired, and rebuilt various AR-3 and 3a systems over the years, I've found that the AR-9 series provides a more satisfying listening experience. This is a purely subjective observation, and certainly not meant as any sort of criticism of these older systems, which are superb. That said, I know this acoustic space, our musical idiosyncracies, and the specifics of my own listening biases, and felt that the more modern drivers would better suit these requirements.

After acquiring a pair of low-cost 3a's with rough, but decent cabinets and shot drivers - the previous owner had painted them! - I set about nabbing the upper drivers from eBay, and had immediate success with a pair of AR-91 mids, and 200029-1 tweeters from a parted-out AR-9. I already owned a pair of AR-91 crossovers, and two extra pairs of 12" woofers - the standard 200003-1 woofer from the AR-9, as well as a pair of Tonegen replacements - from the AR-91 project. The AR-9 woofers had been re-foamed by Millersound, and the Tonegen drivers were in original, but excellent condition.

Based upon previous experience with the AR-9, AR-90, and AR-91 systems in this room, I knew that I would not need to incorporate level-control switches in the crossover. Again, this choice is strictly due to personal preference and the confidence born of being really familiar with the acoustics in this space. To be honest, I cannot recall ever preferring an AR system at any setting below maximum mid & tweeter output; it would seem that I tend toward a "hot" response, but this was certainly not the case with loudspeakers from other manufacturers. The speakers in our main music system are run with a resistor in place to slightly diminish the ribbon tweeter's output, otherwise they would seem a tad bright.

Even without the level switches, duplicating the AR-91 crossover - actually, it's an AR-58S crossover when the switches are omitted - in the space of an AR-3a cabinet was a no-go; the offset area for the binding posts and distance between the panel braces makes things too confining. And so I decided to build the new crossover on two 1/2" plywood boards, with the woofer board being located in the space previously occupied by the 3a's two large electrolytics. All of the crossover coils were salvaged from the AR-91 crossovers, as they were in excellent physical condition. The original coils had been held in place by riveted aluminum posts, speed nuts and hot glue, and came apart after the glue seal had been carefully broken. On the new boards, the coils were secured with solid brass screws, nylon washers, and GC Electronics silicone sealant/adhesive; the metal speed nuts were again used to provide a nice tight grip between the screw head and the nylon washer sitting atop each coil.

The capacitors are held in place with zip ties to mounts that were secured with a single screw, with the caps set in a bed of the GC adhesive. Note here that this is a neutral cure product, without the vinegar smell or undesirable properties of standard silicone adhesive, and it provides an additional degree of stability along with the zip ties. All of the connections are soldered, from the binding posts through to the spade connectors for the drivers. The original 1 and 2 ohm resistors were replaced with 25 watt Erse resistors that measured spot-on.
I went with Erse PulseX polypropylene and PEx metallized polyester capacitors for the mid & tweeter, and PulseX for the woofer crossover; this was after seriously considering the much less expensive NPE alternative for the 100 and 40 mfd caps, and building a test board for evaluation.

During this time, both woofer types were employed, and it became clear that neither sounded as they should with the NPE caps. After replacing the NPE's with polypropylene capacitors, both woofer types sounded significantly better - tight, but with excellent extension. After a few weeks of listening, the decision was made to go with the 200003-1 woofers, which were slightly less "plump" than the Tonegen versions.

The finished boards were screwed & yellow-glued into place, with a very thick seal of the GC adhesive around the circular cutout and perimeter, and given a week to set up. The binding posts are the excellent Parts Express BPA-38NI posts in brushed nickel, which is a bit of an homage to the 3a's original knurled nuts. It was also necessary to remove the cabinet's bottom brace in order to mount the woofer board. The brace was screwed & glued back into place after the woofer board was installed, and the two nailed-in side braces at woofer level were also better-secured with new wood screws.

The cabinets were hand-sanded up to 400 grit level, and then finished with a Minwax oil combination that I've used on earlier projects with pleasing results. Five coats of Minwax spray clear satin poly was used as a sealant, and buffed with 3M final finishing pads until very smooth. Except for a light sanding, the backs of the cabinets haven't been touched; I want to try the spatter-coat effect used on the rear of AR speakers built in the UK, but I ran out of warm weather.

Stuffing is the factory AR-9 fill, in the same amount as used in the AR-91 cabinet. A word of caution: when handled, this material will put particles into the air, and is a mild skin irritant, so use a mask and gloves!

And so, long story somewhat shorter, I could not be happier with the result; we think they sound terrific, and look very much classic alongside the Mac amplifier and Technics SL-1210. I'd also like to thank some of the members for inspiration from their own projects, as well as their detailed photos and the methods used; the knowledge base of this group is profoundly deep, and centered in experience & practicality. Lakecat's and Roger's recent builds provided a nice perspective on how methods of refinishing could differ, and still provide an outstanding final appearance, and Roy C's information on the variations of the 12" woofer, as well as the AR-3a "Limited" was extremely useful.

For reference, here's a list of the capacitors &resistors used in this crossover, with their unit cost and measured value:

4uF 250 volt Metallized Polyester Mylar Film Capacitor ($1.23) (measured 4.0uF)
8uF 250 volt Metallized Polyester Mylar Film Capacitor ($2.12) (measured 8.03uF)
24µF 250Vdc PulseX Capacitor ($6.70) (measured 24.19µF)
39µF 250Vdc PulseX Capacitor ($10.47) (measured 39.13µF)
100µF 250Vdc PulseX Capacitor ($27.01) (measured 100µF)
1.0Ω 25W wire wound resistor. 5% tolerance ($.60) (measured 1.0Ω)
2.0Ω 25W wire wound resistor. 5% tolerance ($.60) (measured 2.0Ω)

Again, I'm hoping to spray the backs come spring, and attach some sort of identification - I'm thinking perhaps a laser-etched replica of the modified AR-3a label shown below. ^_^





















Link to comment
Share on other sites

This is a great project, with terrific documentation. Great pics and text. I have been thinking about a similar but different variation of this idea for some time but have not yet begun any assembly. Maybe some day it will get off the ground.

Like others have said, I'll need to read through this more closely. Thx for sharing. 

Link to comment
Share on other sites

I love this kind of experimenting and those came out excellent!....kudos on a lot of great work. You changed the woofers after the pic? I do like your refinishing job......did you sand the old finish off? Would love to hear those beasts! Your ears are so much better then mine as I could never for the life of me hear a difference in new caps and their approach to the driver. It might be from a lack of experience also.

Just curious.....but do you have an idea of the hours you have in them? I like your label idea also.

Link to comment
Share on other sites

Yes, cabs came out great. Hard to believe they had been painted previously. I will be keeping an eye out for empty 3a cabs to transfer my 58s drivers into at some point though I'm not going to sweat it either way since I don't have to maintain any particular esthetic.

Found your choice of caps interesting although you robbed us of the benefit of your thought process.

Overall, five stars, need some matching stands now though ;)


Link to comment
Share on other sites

Well, thanks everyone, and Merry Christmas, Happy Hanukkah, and Fantastic Festivus to all!

These were fun to build, and I'm happy for the nice compliments & observations.

I'll try to answer the questions, as sometimes my writing can seem inside-out, and the point I was trying to make gets lost. ^_^

The cabinets had not been previously painted, but the drivers had; I never saw anything like it in my life. Who paints a tweeter?  Unfortunately, the '3a woofers had been annihilated, and were beyond repair, leaving me with the choice of using either the AR-9/91 woofers (200003-1), or the newer Tonegen versions that I had on hand. Testing the woofer circuit with electrolytic caps, and then the poly caps left me with a more favorable impression of the AR-9/91 woofer and the polys. The Tonegen woofer is a terrific driver, with an excellent appearance, and I've used it in the past with an AR-9 restoration, but in this case, the other woofer seemed preferable.

Both cabinets were sanded to remove the original finish, and smooth out minor dings; since the '3a cabinets are so small, I stuck with using a sanding block instead of a power sander. The speakers are atop sand-filled steel stands, with a block of Pennsylvania bluestone as a base. The same dense stone is used under the turntable. I'm uncertain about the amount of time put into these, as it was stretched out over months, and working when I felt like it.

This modified speaker sounds very, very much like an AR-91, which is what I was shooting for; I would personally give it a slight edge over the '91 in low end reproduction, where the 3/a mod seems to be missing a very slight "bump" at around 100-120 Hz. I can't explain why this is - it could be a function of the cabinets, the difference in the distance from the floor of both woofers when placed on the same stands, or maybe something to do with the Dayton caps used in the AR-91, and the Erse caps used in the 3a/91 mod.

In the spirit of frugality, I'd really hoped that the inexpensive Erse electrolytic caps would do the job in the woofer circuit, but I was disappointed in their variance from spec, and when placed on the woofer test board, the polys seemed the much better choice. I'm left wondering if the original electrolytics from the era of the AR-3a and AR-91 were in some way superior to what's available now? I did check the values on the 24 mfd and 100 mfd cans removed from the AR-91 crossovers, and only one of them had drifted far beyond spec in 40 years, so who knows?

I prefer the AR-91 and this 3a/91 mod to the restored, but original AR-3a that we had borrowed for comparison;  I believe this has more to do with those AR-9 upper range drivers. Speaking personally, I feel they were the best mids & tweeters to ever come from AR, and they really suit my listening preference. It doesn't hurt that they're so rubust, and readily available, either. The McIntosh amplifier used with this system is rated at 200 watts/channel from its autoformer taps of 2, 4, or 8 ohms, and seems to bring out the best in vintage acoustic suspension speakers. I especially enjoy its ability to provide a balanced midrange & high frequency presentation through speakers that have been unfairly criticized by some as being reticent, or too laid-back.

Not discounting the fact that we significantly prefer the appearance of the AR-3a over most other AR systems - although the LST is in a class of its own - the sound of this modification is exactly what I'd wished. They're sealed up for good at this point, and after a period of getting used to them, I'd like to see about adding a super-tweeter to the mix, having read some interesting reviews of the uber-expensive Sopranino, among others. This would be a means to the end of reproducing a greater sense of air from the dynamic drivers in the AR-3a box; something that seems desirable from my point of listening.

Link to comment
Share on other sites

  • 2 months later...

OK, I've been listening to these for a couple of months, with a very favorable impression. That said, I've also been able to detect a certain mudiness in the low end. This is primarily noticeable on acoustic bass, as in a jazz trio; it's like a slight loss of detail, with individual low bass notes smearing, and a very slight tendency to accentuate one note over another.

Are there any opinions on how this may be addressed? Could it be an issue with stuffing? The 3a cabinets were stuffed with the amount of polyfill that AR had originally used with the AR-91; Tim Holl's paper on their design mentions that the cabinet had a "dense packing of internal damping material", and the 3a cabinet is certainly well-stuffed. Could this be too much for the 3a box? Ideas sought! ^_^


Link to comment
Share on other sites

1 hour ago, genek said:

I'd try putting the original amount of AR-3a fiberglass in one cabinet and see what it sounds like.

That was my thought as well. The two speaker models originally used different stuffing material, and from what I can glean, the internal volumes were different as well: 46.8L for the AR-91; and 48.2L for the AR-3a. 

Link to comment
Share on other sites

Fiberglass. Dammit.

Disregarding the angled upper sides of the AR-91 and the internal bracing of the two boxes, I believe the AR-91 is the slightly larger enclosure.

AR-91 external dimensions = 31x14x9.25

AR-3a external dimensions (minus front moulding) = 25x14x11

Subtracting the 3/4" construction of each cabinet:

AR-91 2857.8 ci (1.65 cubic feet)

AR-3a 2790.6 ci (1.61 cubic feet)

The original 3a fiberglass is long gone; I'd need to replicate its stuffing with AR-9 vintage polyfill.



Link to comment
Share on other sites

1 hour ago, ar_pro said:

I believe the AR-91 is the slightly larger enclosure.

Your calcs seem based on very good assumptions and are most likely accurate - - - I was not nearly so industrious and my comment was based solely on product literature which, of course, has been occasionally refuted on these pages in the past. Apologies for very difficult to read 91 specs, but you are spot-on - - it says 46.8L or 1.65 cf. The 3a shows a larger internal volume, but we cannot vouch for how this was calculated. New "modern" fiberglass batts are really not so bad to handle - - far less itchy and scratchy than the older variety that are often shredded in smaller pieces.  

AR-91 and 3a cab.vol..jpg

Link to comment
Share on other sites

BTW, how much polyfill did you put in the cabinets? The assembly drawing for the AR-11 says to use either 20oz of fiberglass or 10oz of poly. The problem with that spec is that it doesn't provide any specs for the poly itself, such as density, and we know there are different kinds of poly with different characteristics. Fiberglass, on the other hand, doesn't seem to have that kind of variation.

Link to comment
Share on other sites

You boys are geniuses.

I gloved & masked-up, and pulled the vintage polyfill, replacing it with approx. 20 ounces of fluffed-out fiberglass, and it's like night & day!

I'd originally used polyfill saved in bags from a pair of AR-91's, and it completely filled the AR-3a cabinet to a dense packing, as per Tim Holl's description for the AR-91. The measured weight turned out to be a whopping 27 and 28 ounces. Also, this vintage polyfill is degrading - I could see particles floating into the air as it was pulled out; certainly to a greater extent than I'd ever noticed before.

Thanks for the input, fellows. Sometimes you can't see the forest for the trees. ^_^


Link to comment
Share on other sites

David's amazement is gleefully shared by me, but then I can't really pretend to understand fully how the different stuffing materials have such an immense impact on a speaker's characteristics. It may just be my own peculiar intuitive reaction, but I've never liked poly fill as speaker stuffing - - it feels ultra-cheap, wads up unevenly, has virtually no density, and is unpleasant to handle. That last statement will strike some as amusing since others have scoffed at my practice of often re-using original mineral wool found in Classic AR's from the early 60's, but so be it. 

It's already been said, but I really like the idea and execution of the project shown in this thread, and I'm very pleased to hear that the situation has been greatly improved with the change in stuffing. Per the attached parts list from the AR-91 assembly drawing (showing 20 oz. poly), it appears ar_pro's initial stuffing effort to replicate the AR-91 was perhaps 35 to 40% overinflated. Also, genek's references to various assembly drawings for the AR-11 are correct, but I am inclined to believe that the 10 oz. poly citation is a drawing notational error, since drawings for several speaker models with cabinets of similar volume (AR-91, 92 and 915) all reflect consistent use of 20 oz. poly fill.

This discussion prompted me to review chapter and verse in the AR-3a restoration manual, and section 4.3 provides some illuminating statements. Of course, this document is all based on the use of AR-3a drivers, but it does state that the later 3a employed 20 oz. of fiberglass, and that poly stuffing will not provide the required low frequency resonance. The next paragraph goes on to emphatically recommend the use of fiberglass, and then further describe the adverse affects of using poly fill in the 3a speaker model. For anyone who has followed this thread this far, it is worth taking another moment to re-read that portion of the 3a restoration document.

Great project, glad it has all worked out.  :)         

AR-91 stuffing.jpg

Link to comment
Share on other sites

Amazing that a stuffing material would make a difference.

this thread makes me want to try swapping out the original polyfill for some fiberglass....

I know, right?!

Good catch with the AR-91 blueprinet, ra.ra, it never occurred to me to actually weigh the fill that had been removed from the old AR-91 boxes. I'd love to know if changes had been made during the model's run, or if the Tim Holl white paper's approach to stuffing had ever been implemented.

I'd pretty much just assumed that the way to go with the later-model drivers would be polyfill from an AR-91, but with as even a distribution as possible, the AR-3a cabinet was still just crammed with the stuff. And as genek has mentioned, who knows to what standard polyfill was/is manufactured? That, and the fact that the vintage fill appears to be breaking down could certainly be factors in the difference in sound, but a difference there absolutely is!

One of the aspects of the AR 12" woofer that I really enjoy is its ability to properly define lower-bass notes; whether from an acoustic bass, or the bottom octave of a piano, you could always count on this attribute. But the more I listened to this AR-91/3a pair, the less confident I was that this was happening. The LF extension was certainly there, as was an overall tautness, but a solo bass run could not be as easily resolved, and it was beginning to get aggravating, since a lot of the music I like is acoustic jazz.

The 20 oz of fiberglass has done the trick; there's no indication that anything of a negative nature has been made with the switch, everything remains in balance - I still prefer the midrange & tweeter run hot, without level controls - but the presentation of the low bass is much improved!

Link to comment
Share on other sites

If 10oz in the AR-11 drawing is an error, it's one that persisted, because the library contains 13 revisions of that drawing from 1974-1978 that all say the same thing.

I have a pair of empty AR-11b cabs here awaiting use. The traces of stuffing left inside are clearly yellow fiberglass.

Link to comment
Share on other sites

On 3/2/2017 at 6:48 PM, Stimpy said:

The late Brian Cheney, of VMPS, swore by fiberglass stuffing for a bass cabinet.  He felt that fiberglass was better able to control bass response, and respond lower in frequency than polyfill was capable.  So, maybe he was right?

Yes. Our friend John O'Hanlon, armed with an entire science lab with electron microscope etc. determined that fiberglass and polyfill had very different characteristics and advised against replacing FG with poly, if memory serves.

Link to comment
Share on other sites

This is a bit after the fact, but I came across these words from AR guru Tom Tyson comparing the 91 to the 3a. I think it is relevant to the topic here of building a 91 in a 3a box and what should be considered:


Others have commented on the AR-91 (and this applies to the AR-92 as well), but there are some engineering highlights that have not been mentioned.

When the AR-9 was introduced, AR paid careful attention to the “boundary effect” (originally identified and characterized by Roy Allison), whereby the woofers were mounted on each side, close to the floor-wall intersection when the speaker was placed, as intended, back-against-the-wall. Unfortunately, AR did not give credit to Roy Allison for this development.

The AR-91 (and AR-92) also attacked the problem of boundary interference, and each enclosure placed the woofer close to the floor with the speaker systems also intended for against-the-wall placement. The compromise was that the AR-91 woofer was front-mounted, but the boundary-interference effect was somewhat reduced. The AR-91’s crossover also was modified to further minimize the effect. The AR-91’s cabinet volume was somewhat larger than the AR-3a’s, and system resonance was dropped to 40Hz (-3dB was 35Hz) with the “Q” set to 0.56. The result was a slightly overdamped system offset somewhat by the crossover and by the fact that the woofer was placed relatively close to the floor.

The AR-91 was also a “vertical” speaker, and by vertically stacking the woofer, midrange and tweeter, the interference effects were shifted into a vertical plane, rather than horizontal, and thus the frequency response from both speakers, as heard in the listening position, was more stable. The AR-91 tweeter was identical to that used in the AR-9 and AR-90, but the midrange unit was modified. This unit has a different part number as well. The 1-1/2-inch midrange unit used in the AR-9 and 90 used Ferro Fluid on *both* sides of the voice coil, and sealed off the cavity under the dome, thus raising the resonance frequency of the dome; but in the AR-9 and 90, the crossover was set quite high and this was not an issue. In the AR-91 the crossover was 700Hz, so the voice coil in the midrange used Ferro Fluid on the *inside* of the voice coil only, and thus the cavity beneath the dome was larger, lowering the resonance frequency of the dome. It could then operate within the 700Hz crossover range more effectively. The small “semi-horn” appendage on the front did not affect the lower cutoff frequency of the dome, as it was designed to help maintain efficiency in the upper level of the operating range. It did nothing below 3kHz according to Tim Holl.

The result of all this was an improvement in many ways over earlier 3-way AR speakers, and a speaker not always recognized for its fine performance. As Steve F remarked, it was somehow overlooked in the audio press.

--Tom Tyson


Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Create New...