MJR7-Mk4 HARMONIC DISTORTION.
The low levels of distortion expected for the MJR7 amplifiers make measurement difficult. My signal generator has second harmonic at -87dB at 1kHz, so this would be useless if I just applied this to the amplifier input and measured the output. There are ways to reduce the generator distortion, ranging from a single capacitor across its output to more complex filters to reject the distortion frequencies.
The solution I have always chosen is to use a signal nulling circuit to reject as much as possible of the test signal and its distortion. One circuit I have been using to test the Mk4 amplifier is shown next. This is suitable only for testing an inverting amplifier. A more general method is described at great length here.
The circuit is similar to one which appeared in Wireless World, Nov 1977, page 65, in an article by Peter Baxandall.
The MJR7 has voltage gain about -18, so with an input Vin, if it adds distortion D, its output is -18Vin + D.
The test circuit feedback amplifier has two inputs, one is the original test signal Vin, taken via a gain and phase adjustment network, the other is the MJR7 output, which is taken via 56k so that the test amplifier with its 56k feedback resistor gives gain -1. With careful adjustment of the variable resistor and capacitor the Vin and -18Vin can be made to cancel, leaving just the distortion D. In practice accurate nulling of Vin is difficult, but reduction by 40dB is enough to make accurate measurement possible.
Checking the gain confirmed a very accurate unity gain for the distortion. Two diodes are included at the output to protect the soundcard from excessive voltage. I have no specification for maximum input to my soundcard, so I don't like to connect direct to any amplifier output without some protection. Signal level needs to be kept below about 100mV to avoid the diodes adding significant distortion.
INITIAL TEST RESULTS.
Initial results for my first version of the MJR7-Mk4 are not quite as good as previous results for the Mk3, but there is some doubt about the quality of transistors used, so some improvement may be achieved when I have resolved that problem. As before the second harmonic is the highest, the figures at 4V output level are -110dB (0.0003%) at 1kHz input, and -91dB (0.0028%) at 20kHz.
Further investigation revealed a few surprises. I had assumed that increasing quiescent current would reduce all distortion, but what I found was that second and other even harmonics were almost entirely unaffected by small increases in Iq, but third and other odd harmonics changed sharply with just a small increase. Increasing Iq by just 12mA the 3rd harmonic of 20kHz fell by 17dB. At different frequencies and for different harmonics the optimum Iq was different, so there seems to be no single ideal value. Increasing Iq further the second harmonic did eventually start to fall, and reached 0.001% (-100dB) at Iq = 350mA.
I had assumed most of the second harmonic was caused by the differing gm of the two mosfets, but another possibility was the different capacitances. Reducing the gate resistor to the p-channel 2SJ162 should reduce the differing effects of the gate-source capacitances, but trying this had no effect.
Changing the position and spacing of the supply lines produced variations in second harmonic, which could be reduced by up to 10dB, now better than my 0.001% target with Iq only 100mA. It would be wrong to conclude that supply loop pickup is the cause of the 2nd harmonic, all this may be doing is adding a second harmonic component from the supply current loop to cancel the second harmonic caused by amplifier nonlinearity. This may give an impressive specification, but this sort of cancellation is unlikely to work so effectively for a music signal. With both channels driven by a stereo signal the supply distortion will be higher and different, so we don't want to intentionally feed that into the circuit.
Here is the distortion with 4V output at 1kHz with Iq = 120mA. The dB levels are relative to 316mV, so the 4V amplifier output at 1kHz is at +22dB, so 22dB must be added to the scale for levels relative to 4V. The 1kHz test signal has been nulled by 45dB by the test circuit, so it appears at -23dB. The 2nd harmonic is at -110dB, and the 3rd harmonic is somewhere below -135dB. A few higher harmonics are visible, up to the 7th at around -130dB. The 3rd harmonic is only so low at a precise Iq setting and signal level.
To give some indication of how this relates to audible levels there is a small white line at -60dB, equivalent to a signal level of 300uV. Using my MS-20 speakers and listening at 1 metre distance, using a test subject with excellent hearing (not me), this was the lowest audible level for a sinewave signal around 3kHz, which is the range where our ears are most sensitive.
The same 1kHz 4V output test was done with the Renesas mosfets replaced by Exicon 10N16 and 10P16. Having heard reports that the Exicons have better matching I expected lower even order distortion, but the result was almost identical.
Again small changes in Iq produced big changes in the odd harmonic levels, for example the 7th harmonic could be reduced below the noise by reducing Iq to about 55mA, but then other odd harmonics increased. Comparing even harmonics is therefore more informative, and here the Exicons are slightly worse. Testing at 20kHz also gave very similar results for the two types. Without trying many different samples of both types it is impossible to say with any certainty which is a better choice, but my own preference is for Renesas because of the internal gate protection zeners, so I returned to using these for all the following tests.
Next a not entirely meaningful result at 20kHz, showing second harmonic at 0.0008%. This was with carefully adjusted supply leads, giving a cancellation effect. A figure of 0.003% is closer to the truth. For this measurement I switched off the averaging function, which is why the noise looks worse.
To see how the distortion changes with signal level I tried the more direct measurement method mentioned at the top of this page, with my signal generator driving the amplifier, and using a simple filter to reduce the higher order generator distortion. The 2nd harmonic is still too high, but for measurements at high signal levels I was expecting higher distortion. To protect the soundcard I used a 20dB attenuator at the amplifier output.
The distortion measured by this method was found to remain lower than I had expected at higher output levels. The 5th harmonic was the only one to increase consistently relative to the output as level was increased, and even that only reaches -110dB in the next spectrum relative to 12V rms output. (12V rms = -10dB in this plot). The 2nd harmonic is primarily from the signal generator. The small peaks around 20kHz are interference not related to the test signal.
The single capacitor low-pass filter can be improved, there are many possible circuits. Using an active filter the most important requirements are low noise and distortion. To avoid common-mode distortion an inverting amplifier is a good starting point. For low noise we then need a high source resistance and a low impedance feedback network, for which a twin-T notch filter is the obvious choice. A notch in the feedback network becomes a peak in the closed-loop gain, but to get a predictable gain at the pass frequency we need a bypass resistor so that the feedback is not completely nulled. Passing the 1kHz generator output through the following circuit could reduce its 2nd harmonic by a factor of 30 or more, and also reduce its low frequency supply effects. Of course the inverting amplifier needs to have even lower distortion than the amplifiers we want to test with the output signal.