10W 6L6GC Power Amplifier
15-65kHz (-3dB), 1kHz input, at 3W RMS output
across 16 ohms.
10W RMS for 700mV RMS input, with a THD+Noise
1x 6J5 (input), 1x 6SN7 (differential phase
inverter), 2x 6L6GC (output), 1x 5R4GYB (rectifier)
Around $375 per mono amplifier, without cover.
When the opportunity presented itself to possibly
lay my hands on my father’s set of inefficient AR-
3a speakers, I needed to put together an amp
with some reasonable wattage for once. I haven’t
got the speakers yet (if you’re married you know
why), but the amp is ready. At ten watts it may
be still a little light for AR’s, but it was a
successful project nonetheless.
A look at the schematic shows us a pair of push-pull 6L6’s run as pentodes. Pentodes?!? Eeek!
The output tubes are biased into class A using the manual values of Ec –14V, Eb 250V, Ec2 250V. At these low plate and screen voltages,
any of the 6L6 versions could be used: 6L6 (metal), 6L6G, 6L6GB, or 6L6GC, etc. The book value output transformer primary at this
operating point is 5k p-p, but I chose 3.4k deliberately to lower higher order harmonic distortion. The Radiotron Designer’s Handbook, 4th
Ed., contains a graph showing just such a decrease as the load resistance is lowered for a 6L6 at our operating point. The second harmonic
climbs to over 12% as you do this, but is cancelled due to the push-pull operation. Meanwhile the 3rd harmonic has fallen from 3% at 2.5k
to under 2% at a 1.7k load. This seems to give a good compromise between lowering the distortion, and the drop in output power.
The Output Transformer
The Hammond 1650K is a 7 lb. oversized monster for our current and wattage demands, but it has a 3.4k primary, and the extra iron
can’t hurt . With a correctly designed circuit excellent performance is possible from this unit.
Leaving the bypass capacitor off the input tube cathode bias resistor gives about 4dB of negative current feedback. 11.5dB of negative
voltage feedback is brought from the 16 ohm tap back to the 6J5 cathode. This leaves a 6dB stability margin before oscillation occurs.
(note: 8 ohm speaker owners will have to adjust the 47k feedback resistor value down)
A 1MEG resistor between each 6SN7 and 6L6 plate adds only a small amount of feedback, but cleaned up a good deal of the residual ringing
on transients. The amp's high frequency 3dB point drops to 65kHz with these in circuit, as opposed to 70kHz with them out.
The output tubes are balanced for DC by adjustment of the 100 ohm wire-wound pot. This controls the positive bias applied to each grid.
The 10k wire-wound pot in the 6SN7 plate circuit is adjusts the balance of the AC signal delivered to the output stage.
There are many opinions on how to best balance a push-pull amplifier. Some people recommend making the currents in the output tubes
equal - even going so far as to install extensive dc metering systems, and servo amplifiers; some balance the AC at the grids using pots,
and use a "matched" set of power tubes. All the usual methods make numerous assumptions - that the DC resistance of both halves of the
primary are equal, that the gain of both output tubes is equal, that the AC current balance point is the same as the DC current balance
point, just for starters.
It's strange that distortion measurement is almost never mentioned in conjunction with balancing the output stage. If the amp were
perfectly balanced, all the 2nd harmonic distortion would be completely cancelled from the output. So, what I do first is adjust the AC
balance so it's equal on both grids, then I adjust the output tubes' bias pot until I seem to find a minimal point on my HP 334A harmonic
distortion analyzer. At this point I readjust the AC balance and see if this can get the reading any lower. If not, I'm done.
If you don't have a THD meter, you can follow any of the standard methods and still get great sound - with only slightly higher distortion.
Under test the amplifier remained stable with a resistive 16 ohm load, with a 16 ohm load in parallel with a 1uF paper capacitor, with a 50mH
choke in series with the load, and with open terminals at inputs of 500Hz and higher. Clipping was symmetrical, and very smooth.
Hooked up to a real 2-way speaker with crossover, square waves showed a slight triangular overshoot, but nothing excessive.
Figure 1 is a 1kHz and 10kHz square wave with a 16 ohm load.
Figure 2 is a 1kHz and 10kHz square wave with a 16 ohm load in parallel with a 1uF capacitor. The bottom picture is on the grid of one of
THD+Noise is 0.29%, +-0.15% (the residual of the signal generator), at 10W RMS output.
This is a good sounding amp, as the test results would imply. The excellent sound quality is the reason this became the first project
described on my website. On my Altec 848A enclosures the bass response has no hint of tubbiness. Piano and guitar recordings seemed to
have punch and clarity. Sopranos and treble registers had no grain or bite. With no input, hum and noise is inaudible at any distance away
from the speaker cone.
Definitely worth the time and expense to put two of these together. I hope you enjoy them!
©2005, Joel Tunnah