MJR7-Mk5 Test Results
Most of the test results are for a 4V output voltage and a 7R5 load. At lower levels operation may be limited to class-A, and at higher levels the signal spends a lower proportion of its time in the crossover region, so for this particular amplifier 4V appears to be a good level to check for the effects of crossover distortion. Results for higher levels and lower impedance loads are also included.
For all the distortion tests the test signal is partly nulled using the circuit shown on the Distortion Testing page, and the -10dB level is calibrated by a 100mV rms signal. The distortion relative to the 4V output level is therefore given by adding 22dB to the scales. The 1kHz 2nd harmonic distortion is shown next for the latest Mk5 (with 100p compensation capacitor), and here the 2nd harmonic is under -120dB (0.0001%), and the higher harmonics are significantly lower. The load used was 7R5, and the output stage quiescent current was set to 100mA.
To give some idea how far below audibility the measured distortion levels are I have marked the 300uV level on the first harmonic distortion trace to show the lowest signal level I found to be audible to someone with excellent hearing (not myself) at 3kHz at 1 metre from one of my Mordaunt-Short MS20 speakers. Sensitivity is 87dB per watt at 1m. so 300uV is theoretically a 7dB sound level, which is about the lowest level we could expect to hear with masking from typical domestic background noise. At much higher or lower frequencies, or with the masking effect of the music, or at a normal listening distance, even distortion at this level would be inaudible.
Distortion tests using music driving a speaker are not included here yet, but were done for my earlier MJR6 design, without revealing anything unexpected, so the better MJR7 should have a wide safety margin.
Next is the 20kHz distortion, again at 4V rms output and 7R5 load:
The 2nd and 3rd harmonics are plotted as a function of test frequency next:
Intermodulation distortion at 1kHz produced by 19kHz plus 20kHz, each at 2V rms, was measured at -126dB. (This figure is the ratio of rms amplitude of distortion product and the 20kHz component of the test signal. If stated relative to the rms level of the whole test signal the result would be 3dB better.) There are higher order components at 1kHz intervals either side of the test frequencies (which are actually 19.3kHz and 20.3kHz because my fixed frequency '19kHz' oscillator has drifted and needs recalibrating). These components are evidence of odd order distortion (3rd, 5th and 7th), but the levels are higher simply because there is far less feedback loop gain at those frequencies to reduce these components. Increasing quiescent current can reduce the 3rd order products, but then the higher order products increase a little, so I am happy to stick with 100mA quiescent current. The 3rd order products are the highest at about -108dB. As usual -10dB is calibrated at 100mV rms and the test signal is partly nulled.
The 1kHz crosstalk is shown next for a 10V output, so for levels relative to this add -30dB, giving 1kHz crosstalk -98dB, 2kHz crosstalk distortion -134dB.
Next is an easy test, a square wave at 1kHz with a resistive load, most amplifiers can produce a good result here.
Next are square waves at 10kHz into 7R5 resistance with parallel 2u2 capacitance. First is the result across the load, where the ringing is clearly visible. The second is before the output inductor where almost no effect can be seen, showing that the ringing is just a resonance effect of the inductor plus capacitance, there is very little effect from the amplifier itself. Using 1uF or less capacitance there is no obvious ringing either before or after the inductor. The level of 'ringing' is also affected by the amplifier input filter, so this is not a reliable way to compare different designs. It is also not the best way to test for stability problems, for this I prefer looking at clipping with various loads.
Here is the result of clipping at 1kHz, and to check for problems at higher frequencies the second image is at 100kHz, where nothing serious is found. The peak to peak outputs are about the same, 44V, at both frequencies, demonstrating that slew rate is more than adequate, at least 10 times greater than the peak level I found was needed for music on CD. My signal generator only goes up to 100kHz but there is really no point testing any higher. Since this test the compensation capacitor has been reduced from 220pF to 100pF so the maximum slew rate should be even higher.
The 4V signal level used for most of the distortion tests is chosen to reveal crossover distortion. At much lower levels the output stage can remain in class-A and so have no significant crossover effects, and at much higher levels the crossover region is a relatively small proportion of the output and here also crossover effects may have less effect. Lateral mosfets don't have the fall in gain common to bipolar transistors at high output current, and have far less increase in nonlinear capacitance than vertical mosfets near to clipping, so a great increase in distortion at higher output is not expected. I checked the 2nd, 3rd, 4th and 5th harmonic levels as a function of output voltage into a 7R5 load at 10kHz, and these are plotted next. The only surprise is that the 2nd harmonic percentage stays almost constant as output is increased from 4V up to 13V.
Reducing the load impedance can be expected to increase distortion, and the next tests are 1kHz at 4V and 10V rms with a 3R load. Many high harmonics are revealed, but note that the averaging function has been used, and without this few of the higher harmonics would be visible above the noise. At 10V (33W) only the 2nd, 3rd and 5th harmonics are above -120dB (0.0001%).
The maximum output for just visible clipping seen using an an oscilloscope was checked, this will be somewhere near the 1% distortion level which is a widely used standard for specifying maximum amplifier power. With 7R5 load the power is 31W with 59V supply voltage, and with 3R load power is 50W with 57V supply. This corresponds to a peak current 5.8A, so is well under the 7A maximum current specification of the mosfets. To avoid exceeding this limit the load impedance should not be less than 2R5. The power specifications are 'continuous average sinewave power' often referred to as 'RMS power'. For a complete power amplifier the continuous power with both channels driven should also be specified, but the only reason why this is usually lower is that the supply voltage falls due to the high current demand, so the result depends on the power supply used.
Next is the result of clipping at 1kHz with a 2u2 capacitive load in parallel with 7R5, which looks almost the same as clipping with a resistive load shown earlier. Amplifiers with only conditional stability may have bursts of oscillation when coming out of clipping with capacitive loads, so this is a useful test.
Next is a non-standard test, related to the DIM-30 and DIM-100 tests. I have written somewhere that I think these tests are pointless, and have little if anything to do with amplifying music signals, but I was still curious to see how well the MJR7 deals with this sort of test. Unfortunately I have only a variable frequency generator to produce the required 3.18kHz square wave and a fixed frequency 19.3kHz signal instead of the usual 15kHz, and also the amplifier has its own low-pass input filter with -3dB at 120kHz instead of the 30kHz or 100kHz for the two standard tests. The result is therefore an almost entirely non-standard 'DIM-120' test, and so there is nothing to compare the results with. As usual the test signal is partly nulled and in this case the distortion is specified relative to the 19kHz component. The 19kHz output is at 1V rms and the square wave is 4V. Only two components are visible under 20kHz and the highest of these is around -100dB. I guess this is reasonably good. The top spectrum is the square wave alone, plus a small spike from a nearby tv. The test signal is not very good, it includes even harmonics which should be absent from a square wave. The second spectrum is with the 19.3kHz added, and the new intermodulation components introduced are marked by white dots.
Measuring distortion components at such low levels does not require expensive test equipment, my signal generator is one I made myself many years ago, and has distortion not much better than 0.01%. The signal nulling method also nulls much of the signal generator distortion, and adding a suitable capacitor across the generator's 600R output also reduces the harmonics. The only 'expensive' item is my E-MU 1820M sound card, but even this I got cheap on eBay because one of the sockets is damaged. The test setup is shown in the next photo. The results are displayed using a PC spectrum analyser called 'OscilloMeter' which is available in a free version with a 15sec limited running time per measurement.
There has been occasional disbelief that such a simple power amplifier circuit could have very low distortion figures, and maybe my unconventional distortion measuring technique using signal nulling is partly to blame, so I am pleased to have been given a link to a page of measurements using more conventional test methods.
L'amplificateur de Mike Renardson à transistors Mosfet by Forr.
The article is in French, but for non-French-speakers most of the test results are still easily understood. A not entirely accurate translation can be obtained by copying and pasting sections of the text into Google Translate.
The tests include inductive and capacitive loads, revealing that there is no serious effect on distortion levels. I was surprised to see the test with a 16.8uF load, I never tried more than 4.4uF and had expected instability with higher values. Checking a simulation I find that the feedback loop phase shift does then go well past 180 deg. but the loop still has gain far enough above unity to remain at least conditionally stable.
Many thanks to Forr for some impressive work.