MiniDisc - Mic Preamplifier Noise Measurements
To establish if the mic preamp of a Sony MZ NH700 minidisc recorder significantly impairs the noise performance of a Sennheiser MKH40 mic, a typical example of the performance of mics used in nature sound recording.
A typical microphone of the type used in wildlife sound recording has a low output impedance such as the Sennheiser MKH40 has a nominal impedance of 150Ω. This was approximated by the closest value meta-film resistor to hand, 180Ω. The baseline theoretical noise level is therefore set by this output impedance. A perfect 180Ω resistor generates a noise voltage of
Vnoise(rms)=√(4KTRB)=253nV [-130dBu] 
- K is Boltzmann's constant = 1.3806503×10-23
- B is the system bandwith (half the sample rate) = 22.05kHz
- T is the absolute temperate, approximately 290K
- R is the resistance = 180Ω
The microphone is, however, an active device with a noise contribution of its own. The Sennheiser MKH40 is quoted at a sensitivity of 25mV/Pa and a noise level of 21dB (CCIR 468-3)/12dBA. For the sake of this estimation I will use the 12dBA value and model is as flat white noise across the band. This flatters the mic spec slightly as A-weighting may decrease the level anywhere from 3-6 dB depending upon the nature of the noise  The sensitivity means at the reference level of 1 Pa (=94dB SPL) the output voltage will be 25mV [-30dBu]. With zero sound level input the rest noise would correspond to a a noiseless microphone is a noise field of 12dB SPL, 82dB down on the nominal level of 94dB SPL that generates 25mV. The rest noise is therefore
10(-82÷20)×25 mV = 2μV(rms) [-112dBu]
This is 18dB greater than the thermal noise of the test resistance, but is a signal equivalent to the output of the microphone with no sound impinging on it. The difference between this signal and the HiMD noise floor indicates how much the MD preamps compromise the signal to noise ratio of the microphone in the presence of weak signals.
To realise the 96dB potential of a 16-bit recording with this microphone, the loudest sound level should be at least 12 + 96 dBSPL = 108dB SPL. This is a fairly hefty sound level typical of a rock concert, 14dB above the 1Pa level, so the output will be
10(14÷20)×25 mV = 125mV(rms) [-16dBu]
This sound level is far above anything the wildlife sound recordist will hopefully meet in the field. If we take the 60dB SPL level of conversational speech as a typical sound we will meet, this is 34dB below the reference level, and will give us an output level at the mic of
10(-34÷20)×25 mV = 0.5mV(rms) [-64dBu]
The MZ-NH700 preamplifiers clip on the high sensitivity range at 18mVrms, and at 200mVrms on the low-sensitivity range. It is therefore safe to set the mic sensitivity to high sensitivity for this level. To analyse whether the preamps of the HiMD recorder impair the performance of the microphone, we can adjust the HiMD to record, reading full-scale with an 1kHz tone input of 0.5mV(rms) injected via the 22kΩ resistor and 4.7μF 35V DC blocking capacitor. The 5kΩ MD input impedance shunts the 180Ω resistor bringing the input impedance down to 174Ω, the signal level required at the oscillator is therefore (22000+174) ÷ 174 × 0.5mV = 64mV. The tone can then be turned off, and a further measurement made without input.
Fig.1: Test setup - the DC blocking capacitor is 4.7μF 35V + to D.U.T. to prevent excessive loading of the plug-in-power system.
A 1kHz tone was set up on the oscillator, a Farnell LFM-4 Wien-bridge analogue sinewave oscillator, at 191mV p-p measured with an oscilloscope and 10x probe. An AVO M2007 rms reading digital multimeter was used to confirm a 64mV rms value.
The test was run, the HiMD was set to manual gain, PCM recording, Mic sensitivity High and a record level of 28 corresponded to peak signal level before clipping. 20 seconds of this was recorded, then the deck was paused. The experiment was repeated with the oscillator, meter and scope turned off. Another 20 seconds was recorded, and the two files were downloaded using SonicStage. This was then analysed using Rightmark audio analyzer, set to a 22Hz measurement bandwidth, with the following result:
Fig.2: Spectrum with the 1kHz tone -
this ageing oscillator still has an acceptable spectral purity.
The plot with no tone shows the noise floor and is shown in white on the plot below. It now remains to represent the noise floor of the microphone. 60dB SPL is the level of conversational speech, and the microphone is specified at a 12dBA noise level - 48dB below conversational speech. CoolEdit was used to generate white noise and adjust level to read -48dB on the metering. This was then taken through RightMark with the same settings and overlayed on the tone result - it is the horizontal green line on the plot below.
Fig.3: comparison with the mic and HiMD - white plot is the hiMD noise floor,
green horizontal line is the noise corresponding to the mic noise level
It can be seen that the HiMD noise floor is about 4-5dB below the mic above 1kHz, and is less than the mic noise performance, except at low frequencies. This low-frequency noise lift is probably an artifact caused by the 4.7μF coupling capacitor, which has a reactance equal to the 180Ω resistor at 188Hz which means the internal amplifier noise is damped less by the source impedance as frequency falls below about 200Hz.
Here it can be seen that the HiMD may just be limiting the ultimate noise performance of the microphone. Although its noise floor is still about 4dB below the simulated mic noise floor, as a rule of thumb the following amplifier should have a 10dB lower noise than the incoming signal to have a negligible effect on the system noise. The microphone may well have a lower noise in the crucial midrange and a significantly higher noise at the extremes of the spectrum, so it is possible that the mic noise will be slightly lower at some mid frequencies than the MD preamp. However, the effect is still small and at the very limit of sensitivity - preamp noise is not going to be an order of magnitude worse than the mic in this case.
The HiMD wildlife sound recordist can still justify using microphones of the quality of the Sennheiser MKH series without feeling the low noise performance of these mics is destroyed by the limitations of the preamplifiers, even on the lowly MZ-NH700! The input noise of the MD set to the gain used in this test is about -116dBu
This conclusion can be verified in the listening test of a HiMD/Phantom PSU in comparison to a Sound Devices 722 and a M-Audio Microtrack with low noise mics and low SPLs
It is stretching a point to say an audio mic amplifier that falls over 10dB short of the theoretical performance at the given input impedance is low noise. The MD input stage noise is about 1.2μV(rms) [-116dBu], if the LF lift from the coupling capacitor is eliminated, 13dB worse than the thermal noise of the source impedance. However, in the particular application to a typical premium capacitor microphone this amplifier does not greatly impair the system noise performance. There are some combinations where this would not hold - a microphone with a lower sensitivity than 25mV/Pa but still a low equivalent noise would start to be compromised. Equally, a mic with the same sensitivity but a significantly lower noise level would be compromised (in suitably quiet locations!) The Rode NT1A has the same sensitivity as the MKH40 but the equivalent noise is 7dB better at 5dBA SPL, so the MD input would have a little more noise than the mic noise of -119dBu and therefore compromise the signal to noise ratio a little. A ME66/K6, with twice the sensitivity as a MKH40 (50mV/Pa) and a slightly lower noise (10dBA) would have a higher electrical noise floor due to the higher sensitivity at 3.2μV(rms) [-107dBu], clear of the -116dBu MD noise threshold. The MD would be a rotten match to a dynamic mic of a couple of hundred ohms impedance - at the very least a step-up transformer should be considered, with a blocking capacitor to keep the PIP power out.
It is worth noting that the overload margin of HiMD preamps is poor. This, and what you need to know to avoid limiting in the input stages, is described in a separate article