When I first hooked up my A-3s (I'm driving them with homemade OTLs similar to the older, 12-tube version of the Atma-Sphere MA-1s) I adjusted the equalization according to the suggested settings in the owner's manual: LOWs at 0, MIDs at 0, and BRILLIANCE at 2 o'clock. While I was astounded by the increase in resolution over my Magneplanar MG-IIIas, I noted that the midrange seemed a bit reticent in comparison. The vocals on one recording in particular, "The Weavers: Reunion at Carnegie Hall, 1963" (Analogue Productions APF005), with which I was very familiar, were not as immediate as I was used to. The first thing I tried was to move the MIDs to +3. This was much better, but now I found that my amps were clipping on the another record I use for voicing my system, "Marni Nixon Sings Gershwin" (Reference Recordings RR-19), something that had never happened on the Maggies.
While the A-3s were about 2dB less efficient than the Maggies, I thought surely they should be easier to drive, given that the Maggies range between 3.5 and 4 ohms and electrostatics are known for being synergistic with OTLs (translation: a high impedance load). To help understand what was going on, I measured the A-3 impedance, and much to my surprise, with the LOW/MID settings at 0/+3, it dropped to 2.4 ohms at 1kHz! (Yikes)
When I removed one of the backplates to see how the equalization was accomplished, I found that the LOWs were adjusted by selecting one of four different taps on the primary of the bass transformer, and the MIDs were controlled by placing different values of inductance in series with the primary of the bass transformer. The "0" setting used 5mH and the "+3" setting connected the bass transformer directly to the input (no inductor). The high-frequency toroidal transformer was connected to the inputs through a single-pole high-pass filter where the series C=48uF (two 24uF polyproylenes in parallel) and the shunt R=5 ohms (four 5 ohm 10W in series/parallel), for a corner frequency of 660Hz. This explained a lot, since the presence of the 5 ohm resistor meant that the overall impedance could be no higher than 5 ohms above the corner frequency, so I knew my measurements were not in error.
When I asked Roger West why he used such a low impedance, he explained that,
To better understand what was happening when I moved the MIDs from "0" to "+3", I disconnected the toroid so I could see how the impedance looked with just the bass transformer. I found that, by itself, the bass transformer was also a tough load. At low frequencies with the 0dB tap, it was purely capacitive, measuring 40 ohms at 20Hz, which is equivalent to a capacitance of about 200uF. When the MID tap is at "0" there is a 5mH inductor in series with the bass transformer, so at 100Hz the impedance becomes inductive and starts to rise. But when I moved to the "+3" MID tap, the impedance stays capacitive and kept on dropping to below 4 ohms before it started to rise.
At this point I realized that needed modify the crossover network somehow to raise the minimum impedance. I knew I needed some inductance in series with bass transformer, but that 5mH was too much, since it started rolling off the bass transformer too soon (160Hz), given that the toriod didn't kick in until 660Hz. So here's what I did:
These changes filled in the midrange to the point where the tonal balance of the modified A-3s was very similar to MG-IIIas. But there the similarity ended, as the Sound Labs had far better resolution throughout the bass and midrange. These changes also brought the minimum impedance up from 2.4 ohms to 3.5 ohms, which was much easier for my OTLs to drive, but still on the low side.
I cannot over emphasize the improvement the polystyrenes made. These speakers really deserve better caps. I had been running with the BRILLIANCE control at 2 o'clock, since going beyond that made the highs somewhat brittle sounding. But with the 'styrenes I was running wide open, so
At this point I was thrilled with every apsect of the A-3s but one: dynamics. When I really cranked the volume my amps were still straining a bit, even though there was no longer any apparent clipping. I considered bigger amps but instead I wired the backplates for bi-amping, as I just happened to have another pair of OTLs hanging around since I had previously bi-amped the Maggies.
With two 125W OTLs per speaker, the A-3s really came to life. When I shared my modifications with Roger, he told me that early Sound Labs had provisions for bi-amping, but that some customers damaged their amplifiers when they accidentally had one of the inputs out of phase. According to Roger, "This apparently cancels the flux in the audio transformers in the region of the cross-over frequency and thereby makes it a near-short-circuit. For this reason we stopped doing it." I suggest you heed Roger's warning and approach biamping with caution, but I have had no problems.
In fact, the amps are loafing now that the highs and lows are split between two amps. I'm feeding each amp with a full-range signal and relying on the A-3s internal crossovers. I considered placing a high-pass filter in front of the treble amps so I could bypass the filter ahead of the toroids, but I rejected the idea because any DC offset would surely cause them to saturate.
In retrospect, I could have lived with my system with a single amp per speaker, but the A-3s really needed more like 200W. If you're like me and can't afford a big 200W OTL, you might want to consider bi-amping with two smaller amps if you feel you need more power.