Fullscreen HTML5Med Res (??MB)HTML4

National Semicondu – a new super-low distortion T he LM4562 dual high performance audio op amp is featured in the preamp module for the Class-A amplifier, elsewhere in this issue. It has ultra-low distortion, low noise, high slew rate and high gain-bandwidth product. In fact, its total harmonic distortion + noise (THD+N) is so low you cannot measure it directly with current equipment such as the Audio Precision Test Set! Even without its super-low distortion figures, the LM4562 is an impressive op amp. Its typical input noise density is 2.7nV/ Hz while slew rate is quoted at ±20V/ms. Gain-bandwidth product is 55MHz and open-loop gain is 140dB for a 600W load. In addition, it has excellent figures of 120dB for PSRR (power supply rejection ratio) and CMRR (common mode rejection ratio). But it is the harmonic distortion performance that made us pick it as the only suitable op amp to be teamed with the new 20W Class-A amplifier current being featured in the magazine. We wanted an op amp that would not degrade the amplifier’s performance in any way. The LM4562 turned out to be the right choice but we have to admit that the preamplifier module featured in this month’s issue was our third attempt to actually extract that fine performance. As in any low distortion design, PC board layout is critical. Having obtained a great performance figure, typically less than .0005% from the module, we had two problems: One, that distortion is about the same as the residual distortion of the Audio Precision test set and Two, the THD+N of the op amp is more than an order of magnitude (ie, ten times) better again, at 0.00003%. So how does National Semiconductor manage to quote a value that cannot be measured? Well, there is always a way! Although the THD+N cannot be measured directly, a simple set-up enables the ultra-low distortion to be measured indirectly. The circuit is as in Fig.1 and is taken directly from the National Semiconductor data sheet. Fig. 1: The suggested circuit from the datasheet used to measure the ultra low distortion of the LM4562. R1 should be low, typically 10W. 36  Silicon Chip The op amp is connected as a unity-gain buffer but with a low resistance R1 (around 10W) placed between the inverting and non-inverting inputs and a higher resistance R2 placed in the feedback loop. The result is a closed loop gain of 1 but a noise gain of G=1+(R2/R1). This is because R1 and R2 effectively form a voltage divider. This means that the error (ie, harmonic distortion) and noise signal is amplified by this factor and this allows the measurement to be made on currently available equipment! Another way of looking at it is to regard the op amp as having its open loop gain reduced to a figure of 1 + (R2/R1) and this means that much less feedback is available to reduce the circuit non-linearity. Either way, we decided to try this method to verify the typical ultra low THD+N figure given in the datasheet (0.00003%). This will vary according to the voltage of the test signal, its frequency and the impedance of the load, among other factors. We connected the op amp as in Fig. 1. For R1 and R2 we used 10W and 1.2kW 5% resistors, which we measured using a digital multimeter to be 9.9W and 1202W respectively. You could use 1% resistors instead but since we were measuring the resistors with a multimeter, this would not make very much difference to the result. Fig. 2: graph of THD+N vs. output voltage from the LM4562 datasheet. The test load is 600W and the supply voltage is ±15V. At 3V RMS, the THD+N value is 0.00003%. Compare this with our measured value! siliconchip.com.au uctor’s LM4562 op amp By Mauro Grassi The distortion gain, using the formula 1+(R2/R1), was therefore G=1+(1202/9.9)=122.41 A 1kHz 3V RMS test signal from the Audio Precision set is applied to pin 5 (the non-inverting input) and the output of the op amp at pin 7 is 3V (because the gain is unity). We used a ±15V regulated supply that we measured to be within 0.06% of this value. However such small variations in the positive and negative rails as well as asymmetrical rails should not make any significant difference to the result since the LM4562 has a PSRR (power supply rejection ratio) of -120dB. We selected a simulated load of 600W, to match the data sheet. We made sure to connect the unused op amp as a buffer with grounded non-inverting input, to prevent it from oscillating and contributing to the noise measurement via the common supply. To measure the THD+N, we used the same technique as described on pages 28-30 of the June 2007 issue. It involves using an oscilloscope with averaging to eliminate the noise on the distortion signal. This was found to be necessary because the breadboard version of the circuit in Fig.1 was quite prone to noise pickup. So, if we let DV denote the RMS voltage of the residual distortion signal and AV denote the RMS voltage of its averaged version, we obtain the scaling factor F=AV/DV. If we let MD denote the THD+N reading from the Audio Precision Test Set, which is made in the audio range 20Hz to 22kHz. Since the distortion gain is G, the true measurement for the THD+N should be: F x MD /G % Fig. 3: Screen grab from of the Le Croy WaveJet showing the test signal in yellow, a 1kHz 3V RMS sine wave and the distortion signal at pin 7 of the LM4562 in cyan, showing the RMS value of DV=59.3mV. siliconchip.com.au Now the LeCroy WaveJet 324 scope we used cannot display the distortion and its averaged-out version simultaneously (as can the LeCroy WaveRunner) so we had to switch between the waveforms. What we did was to take typical values as guides. Figs. 3 and 4 show screen grabs from the oscilloscope. Let’s run through a typical calculation, with values as shown in Figs. 3 and 4. From these screen grabs, we have AD=11.45mV and DV=59.3mV, being the RMS values of the averaged distortion signal and the normally sampled distortion signal, respectively. The ratio F is therefore F=AD/DV=0.1931. Since the reading given by the Audio Precision Test Set was typically 0.013%, its true value (using the averaged residual distortion) was 0.1931 x 0.013 or .0025% Now since G was 122.41, the calculated value for the distortion is .0025 / G = 0.00002%, lower than the typical distortion given by the data sheet! This technique of varying the distortion gain can be applied whenever the measurement to be made is too small for the available instrument. SC Fig. 4: Screenshot of the Le Croy WaveJet showing the test signal in yellow, a 1kHz 3V RMS sine wave and the averaged distortion signal at pin 7 of the LM4562 in cyan, showing an average RMS value of AD=11.45mV. August 2007  37