The Vivi-Voice Project

This circuit is basically a low-power version of
Lynn Olson's Amity, with some
simplifications and changes. (schematic)

The input uses the Lundahl LL1676 wired in
2:1+1 mode for phase splitting. With 4.7k
load resistors, the input impedance of the
amp is about 10k, which is high enough if
you've got a preamp with low output
impedance like the Odilon. The input is
balanced, but can be driven from
unbalanced sources. Driven from the Odilon
in my system (photo), the Vivi-Voice is
completely free of noise, and sounds good
on very efficient speakers.

The Lundahl LL1660s/PP interstage is driven
from a 6CG7 biased to 10mA per side, in
order to lower its rp' to 7k, so as to not limit
HF response due to interwinding capacitance.

Both driver and power tubes use individual
cathode resistors, ala Mullard, to minimize
imbalance with mismatched triodes.

The output transformer is the Magnequest B18, which has an 8k primary. My model has nickel pinstriping, by special request. The
Hammond 1608 or 1609 would also work in this circuit, with slightly more ringing and jagged frequency response than the B18 - but
then, this amp is not exactly an expression of cost-savings.

I'm using the Ei EL84EG in my final version. They are affordable, well made, and biased correctly in the amp. The set of RCA's I
own only biased to 10.4 volts, and sounded a bit cold too. Ei also makes a copy of the 6FQ7 (even though they call it a 6CG7). I
haven't tried it yet because I have some NOS GE and Sylvania stock left. The 5V4 is not being made anymore, so you'll have to go
with old stock. Depending on your local line voltage, you may need to lower the 6.3V heater winding voltage. I use a 0.1 ohm, 3W
1% Dale wirewound resistor in each leg of the winding.

Test Results
Voltage Response: -1dB@30Hz and 41kHz. (graph)
Sensitivity: 3 watts RMS at 0.825V RMS input. 3.78 watts RMS at visible clipping, 0.98V RMS input.
Total Harmonic Distortion + Noise, at 3 watts output: *
20Hz <2.4%
200Hz <0.52%
2kHz <0.6%
20kHz <3.3%
Hum and Noise: inaudible on 98dB speakers with input jack open.

*The signal generator was a Protek B-850 with an additional low pass filter added. Generator THD+N is
0.6%@20Hz, 0.19%@200Hz, 0.14%@2kHz, 0.13%@20kHz.

1kHz square wave.
10kHz square wave. (Notice that the primary is ringing! I think the parallel windings are taking their toll.)

UPDATE: January 8, 2006
Loading the LL1660s secondaries with 47k per side reduces the resonant peak at 70kHz to -3.74dB. (graph) After much probing, it's
obvious that the lumpy frequency response is largely due to the input transformer. A zobel network on the output transformer
secondary did nothing, and the LL1660s seems to have good waveform fidelity now that it's loaded. I may swap out the LL1676 for
a Jensen JT11P1 and see how things look. (schematic)