SIDE BAR
I was going to discuss this at the end of the project, but now seems to be as good a time as any: Performance of the stock design using an 8 ohm load. In the post containing the base line test results, I provided power output levels for both 4 and 8 ohm loading conditions. Frankly, if you go back and look, there isn't a whole lot of difference in actual power between the two. In fact, at the power bandwidth extremes of 40 Hz and 10 kHz, power output at 8 ohms actually ticks upwards slightly which would be expected in this scenario, since there are less transformer losses at these frequencies when using an 8 ohm load, versus using a 4 ohm.
The relatively unchanged power output level between the two load scenarios stands in stark difference with what happened when base line testing a stock 8600 chassis with an 8 ohm load in the project with that amplifier. Both amplifiers are rated for a 4 ohm output, but unlike with the 9300 project here, when the 8600 amplifier was loaded with 8 ohms, power output was basically cut in half. Both designs in stock mode are grossly over biased, which is a significant cause of the performance issues noted in both designs. But there are other differences to consider.
In a properly designed single ended (SE) amplifier, the quiescent operating point is placed smack dab in the middle of the load line to produce Class A operation, this then allowing for an equal excursion either way from that point, generating maximum power, and minimum distortion from the stage. With the stock 8600 design so over biased however (operating point set too low on the load line), the output waveform was already showing early clipping on negative going grid excursions with a 4 ohm load, so the low side of the load line defined the maximum possible plate voltage swing that could be achieved on either side of the operating point, before clipping commenced. This is the point at which maximum undistorted power output was measured and offered as a base line specification for the stock design when operating into a 4 ohm load. You can see then that with the maximum plate voltage swing already achieved using a 4 ohm load, then substituting an 8 ohm load generated no more voltage swing. The 8 ohm load just cut the current flowing in the load in half, cutting the power output in half with it.
With the 9300 however, it is a different story. As a push-pull Class AB1 design, the operating point for each tube is intentionally set very low on the load line, so that it doesn't take much negative grid swing to cut off either tube. But that's OK as in this design (theoretically), when one tube has reached cut off, the other tube is pulling hard and strong, with plenty of excursion left on its load line, so the combined output from both tubes at the secondary of the OPT shows a complete, undistorted waveform, of more power output than the SE Class A stage can generate in the 8600. Except that using cathode bias in the 9300 design really compromises its performance capabilities.
Based on the plate-to-plate load offered by the OPTs in the 9300 (7600 ohms), the use of cathode bias would require operating the tubes very hot under quiescent condition to allow them to pull enough current at maximum power output to deliver low distortion in the process. Remember that with cathode bias, the bias applied to the stage is a product of current flowing through it. Therefore, as power output increases, so does the bias. The use of a common cathode connection and cathode bypass capacitor will act to smooth much of bias variations out. But since this is a Class AB amplifier, there will still be an increase in average current draw by the stage as power output is increased throughout the Class B portion of the operating cycle. Therefore, there will be a corresponding increase in the bias voltage applied to the stage as well. As a result, the quiescent bias must be set quite hot so that as increasing power output acts to increase bias voltage, it doesn't end up over-bias the stage to the point of causing distortion as full power is approached.
Now the normal way of dealing with these cathode bias concerns is to use a higher plate-to-plate load impedance as mentioned in a previous post -- this to limit the peak current draw and therefore minimize changes in bias -- and accept lower power output as a result, but still achieve it with acceptably low distortion. Except that's not what Magnavox did. They chose to use a load impedance appropriate for fixed bias operation (to maximize power output), but use cathode bias instead for matters of economy, and strap the cathode circuits of both channels together to help minimize the bias shifts that exist under dynamic conditions. Then, they biased the tubes much cooler than would normally be required for the load impedance used (to enhance tube life), and counted on a low average power output requirement so that all the distortion this approach produced would never be produced.
Based on the number of units sold, it was clearly a strategy that worked, as you can find these things seemingly under every rock. However, the approach also helped to (in part) give birth to a whole class of economy consoles defined by the sound they produced. This is not meant to criticize that sound, but simply to say that when such a design is laid out bare for what it is today, the many warts that make it up then become quickly evident. However........
If you've followed along this far, then you realize that the central points of the discussion center on the fact that based on the load impedance offered by the OPT used, the bias system used was inappropriate. That for cathode bias to work acceptably in Class AB designs, the load impedance must be elevated over that used for fixed bias operation to prevent distortion. Which brings us to operating the stock design into an 8 ohm load, which of course then raises the load impedance offered to the tubes. With this loading scenario, power output drops slightly, but not that much. Why? Because with a 4 ohm load, the peak output tube current developed causes enough bias shift to prevent the tubes from even being able to reach the full excursion of their normally available load line. One the other hand, with an 8 ohm load, peak currents are reduced, bias shifts are minimized, and the tubes can then make full use of the altered load line. As a result, power output is maintained, and the crossover distortion is completely eliminated. In effect then, operating the bone stock amplifier into an 8 ohm load become an effective corrective measure for an otherwise compromised design. Realize however that in utilizing this "fix", the down side is lost power output -- as is the normal outcome when cathode bias is used.
When operated into 8 ohms, the stock amplifier should be considered as a 9 watt RMS per channel amplifier, with both channels driven. If the amplifier is modified to include EFB, power output will increase slightly to 11.0 watts per channel into 8 ohms, but while the crossover distortion is gone with both of these scenarios, general THD still rises due to the less than ideal loading conditions this creates. It's a darn side better in the stock design than when using a 4 ohm load, but the lowest overall distortion performance will still be had (by a considerable margin) when the amplifier is properly loaded with a 4 ohm load, and EFB is used to bias the output stage. Under those conditions, the full 15.5 watts will be available per channel with both channels driven with mid-band THD well under 1%. Finally, it should be noted that EFB will still allow for the lowest quiescent current to be used regardless of whether a 4 or 8 ohm load is used with the stock OPT.
This then gives owners an option: Consider the stock amplifier as an 8 ohm 9 watt per channel amplifier -- greatly improved over the 4 ohm (10 watt) performance it was designed to deliver -- OR -- add EFB for the lowest possible distortion and quiescent current requirement for either 4 or 8 ohm operation, with the best overall performance had with that combination operating into a 4 ohm load. While the current effort is to make the most of what this unit is for what it is, it also just screams out for a better OPT (for many reasons) that would at least be tapped for 4 and 8 ohm operation. But that will be reserved for a future effort. For now, the next episode will tackle the phase inverter circuits.
Dave
