Magnavox 8300/8600 Single-Ended Triode (SET) 12BH7 Conversion
This page contains some resources for converting a Magnavox 8300 or 8600 series amplifier to run 12BH7 dual triodes as output tubes. It complements my YouTube video (above).
The Magnavox 8300 and 8600 run 6BQ4/EL84 outputs and have a single 4 ohm secondary winding on their output transformers. Because the majority of high-sensitivity speakers are 8 ohms, this poses an issue for many users. Audiokarma tube guru dcgillespie recommends this problem be solved by replacing the original transformers with new 8 ohm ones.
However, it is my philosophical belief that output transformers are the soul of a tube amplifier, so I generally avoid substituting them. Luckily, a 12BH7 (with its triodes wired in parallel) requires about double the load of a 6BQ4, which means it will happily drive an 8 ohm load via the original output transformers. This conversion was proposed by Audiokarma user FlaCharlie in a thread here, and I decided to put my own spin on the design.
For reference, here is FlaCharlie’s schematic:
My schematic is available below as a PDF:
The majority of my revisions pertain to the first stage, a 12AX7 (or 6EU7) double triode. Magnavox used an interesting trick to limit degeneration here — a voltage divider from the B+ supply feeds positive voltage to the first stage’s cathodes, which allows for a smaller cathode bias resistor.
There are three downsides to this arrangement. 1. Because cathodes are non-inverting inputs, ripple and noise applied to the cathodes will appear at the anodes. 2. This approach still requires a moderately-sized cathode bias resistor to limit the power wasted by the voltage divider. 3. The voltage divider draws substantial current from the first stage’s B+ supply, which has a high output impedance. Thus, it lowers this supply’s voltage substantially.
Why is 3. an issue? Well, the loadlines below demonstrate that the first stage will perform better if its B+ voltage is closer to the main B+ rail.
Removing the cathode bias voltage divider helps achieve this, as the divider pulls substantial current through the high-value decoupling resistor. In addition, I halved the value of the decoupling resistor from 20k to 10k ohms. These two modifications give us a first stage B+ voltage that is about 10 volts below the main B+ rail, as opposed to the ~50 volt differential specified on FlaCharlie’s schematic.
Of course, removing this divider means finding a different way to bias the first stage while minimizing degeneration. I chose a partially-bypassed cathode arrangement, with a 100 ohm resistor in series with a bypassed 2.7k ohm resistor. This helps maximize gain while also enabling the application of negative feedback. In theory, the addition of another time constant could be problematic. But, in practice, the capacitor does not seem to cause any ill effects.
As documented in the above video, I built an 8605 to the above specification and it performs quite well.
Some tips: if your Magnavox has a spare heater winding, you should use it for bucking the line voltage (as I did), or your heater voltage will probably be too high. If you do not have a spare heater winding, you may need to either use an external bucking transformer or place dropping resistors in series with your heaters.
In their original configuration, these amps use the chassis to carry ground current, and components connect to ground at various points. This violates the conventional wisdom that tube amplifier grounds should either be starred or should use a low-resistance bus which connects to chassis ground (and thus mains earth) at the input only.
However, I dislike star grounding, and I did not wish to install a ground bus, so I decided to utilize the chassis — with some tweaks. I connected all grounds for the input tube directly to the ground tabs on the input RCA jacks, and I consolidated all grounds for the 12BH7s at the chassis grounding post located near the tag strip. I also wired the speaker jack grounds directly to the output transformer’s ground leads. As in the original amplifier, I left the HV secondary center tap and can capacitor grounded directly to the chassis. The measurements below demonstrate that this grounding approach works quite well. Both 60 hz and 120 hz noise levels are close to background. I was even able to install a proper three-prong line cord with no apparent noise or ground loop issues.
First up: frequency response at half a watt output. Unsurprisingly, the small output transformers struggle a bit with low frequencies, but the high end is basically flat to 20 khz.
Next, I tested each channel’s distortion at 1/10, 1/2, and 1 watt output. The variation in performance between channels is largely attributable to my NOS 12BH7A output tubes — swapping the tubes causes the distortion figures to swap channels as well. Thus, although I label the graphs left and right channel, what they really show is the difference between my 12BH7s.
1/10 watt L:
1/10 watt R:
1/2 watt L:
1/2 watt R:
1 watt L:
1 watt R:
Lastly, here is the 10 khz square wave response. I took this screenshot during the test phase — the blue trace has no compensation and the yellow trace has a 4700 pf step network capacitor across the negative feedback resistor. The above tests were performed with 4700 pf on both channels, but a smaller value would be acceptable as well.
When built according to my schematic, voltage gain is ~10 dB and there is ~14 dB of negative feedback. Despite the low voltage gain, a modern source should have no problem driving this amplifier.
If anyone wants a BOM for this project, I would be happy to draft one. The amplifier performs quite well, with no apparent noise or instability, and I am quite happy with how the project turned out.