Output Capacitor In Feedback Loop
The idea to include the output coupling capacitor inside the overall feedback loop dates back at least to one of the first 'transformerless' solid state designs, published in 1956 by H.C.Lin. The same approach appears in a number of early designs, including one published in Wireless World, Oct 1969, p.456-457 Low-cost 15-W Amplifier by Hardcastle and Lane. Commercial designs include the Sansui AU-101 from 1973.
Most modern designs use direct coupled outputs, but as I previously mentioned 'the only unique achievement of direct-coupling is the ability to drive DC into the speaker, which we then need to prevent with speaker protection circuits'.There are a few variations, in most cases it is also necessary to include separate DC feedback to control the output stage DC operating level. Here I will just consider two variations, one is my own MJR7 circuit and the other is maybe the most obvious circuit, shown next;
The gain at low frequencies is now determined primarily by the input coupling capacitor, the value of the output capacitor no longer need be the limiting factor. The DC feedback is via the 200k resistor, the AC feedback is via the 10k and 500R, but there is also AC feedback via the 200k and the combined effect sets the closed loop gain close to 20. The important signal level we need to worry about is that at the amplifier output prior to the capacitor, and if we extend the low frequency response too far using a high value input capacitor then there can be an alarming peak in this voltage, with danger of clipping. This is demonstrated next showing the voltages before and after the output capacitor in red and green, showing the extended frequency response with 4u7 input capacitor compared to 2u2. The voltages prior to the capacitor in red show the price paid for this response extension. Fortunately the rise in the 4u7 case is mostly below 10Hz, reaching a peak almost +6dB at 0.6Hz. This is unlikely to be a serious problem in practice, few audio sources are going to have significant output under 10Hz, though vinyl enthusiasts should consider using a 2nd order or more rumble filter, or just sticking to a 2u2 input. Other component values also affect the peak level of V4, so for any variation a simulation is advisable to ensure there is no serious problem. Also the speaker impedance has some effect, I just used 6R as a 'typical' low frequency impedance, at very low frequencies the DC resistance may be around this value.
Next is the result with my MJR7-Mk5, again with 2.2u and 4.7u input capacitors. Again with 4u7 there is a peak in the voltage ahead of the output capacitor, a little under +6dB at 0.7Hz. My original 2u2 for the input capacitor still looks like a good choice. Using 4u7 only reduced the fall in gain at 20Hz from 0.5dB to 0.1dB, which is highly unlikely to be noticeable. Again vinyl enthusiasts are advised to stick to 2u2.
There are alternative ways to add a small bass boost while keeping the 2u2 input capacitor, one example has now been added near the end of my MJR7-Mk5 modifications page.
I also tried a simple equivalent circuit for a speaker load, and again there was a peak at very low frequencies with a 4u7 input capacitor, but much better with 2u2 as expected.
One more feature of this circuit technique is that the output impedance of the amplifier can become a slightly negative resistance at a small range of low frequencies. For the MJR7 I measured this as -0.01ohm at 15Hz, and a maximum of -0.06ohm at 10Hz. This should do no harm, typically it will just cancel some of the speaker cable resistance. It is however something needing to be checked when designing circuits using this technique.
