A Test Set For Audio Noise Measurement |
![]() Some time in the early months of 2011 I decided that I really needed an accurate and versatile ac voltmeter for my audio test bench. Something capable of making true R.M.S measurements of signals up to at least 20-kHz, from a few tens of volts down to (preferably) micro volts. After searching through some of the numerous categorized depositories of service manuals for popular makes of obsolete and/or vintage electronic equipment that are accessible on the internet, and searching for examples of various select items on eBay, it eventually became evident that there wasn't really anything particularly affordable and readily available that fulfilled my hoped for specification. Around about the same time I started to play around with low noise audio circuit design and realized that what I could really do with is not just a wideband, high sensitivity R.M.S. voltmeter, but one specifically tailored for audio signal noise measurement with a host of switchable measurement-bandwidth filters. Pulling out the LTC1968 integrated circuit datasheet that had been sitting in my drawer for past two years or so, I decided to build my own. The instrument detailed in full here is what I eventually came up with. Internally, the instrument is comprised of five individual printed circuit boards, or modules; these being the R.M.S. voltmeter itself, a low noise signal amplifier, a filter board, a signal generator and finally a regulated power supply. The interconnection and arrangement of the modules is detailed in full in the Wiring Diagram. The voltmeter module itself is based on the LTC1968 wide band R.M.S. detector chip and has a switchable, by means of the "Range" switch on the front panel, full scale input sensitivity of either 10-mV, 20-mV or 50-mV R.M.S. A highly accurate DC voltage reference circuit is incorporated on board, for the purpose of calibration. As a consequence no fancy external test or signal reference laboratory equipment is required at all to accurately calibrate the wideband R.M.S. detector. Calibration of the detector is a simple "DC, two point" [1] affair, for both offset error and gain error, achieved simply by setting jumpers on the P.C.B. as specified and adjusting a pair or trimpots in turn for firstly a voltage null between two test points (measured with a high impedance voltmeter), and then for a specified indication on the meter movement. The input signal to the R.M.S. voltmeter board is switchable between two sources, the first being the low noise signal amplifier board via the filter board, for the measurement of low level noise voltage signals, and the second being the "External Input" jack on the front panel, via a four-position switched attenuator. This switched attenuator, labeled "Volts/F.S.D.", provides attenuations of either 0-dB, 20-dB, 40-dB or 60-dB. The switch-selectable voltmeter sensitivities via the "External Input" are thus from from 10-mV to 50-V R.M.S. full scale in twelve 1, 2, 5 steps, as selected by combination of the four position "Volts/F.S.D." switch and the three position "Range" switch. The "External Input" is explicitly intended for general purpose R.M.S. voltage measurements of high level ac signals from 10-Hz to substantially beyond 20-kHz. With the "Source" selector switch on the front panel turned to the "Filter" position, the input of the voltmeter board is connected to receive signals amplified by the low noise signal amplifier (L.N.A.) module, as bandwidth limited by the filter board. The L.N.A. has a switchable voltage gain of either 20, 40, 60 or 80-dB. The switch-selectable voltmeter sensitivities via the "L.N.A. Input" are thus from from 1-uV to 5-mV R.M.S. full scale in twelve 1, 2, 5 steps, as selected by combination of the four position "L.N.A. Gain (dB)" switch and the three position "Range" switch. The "L.N.A. Input" is explicitly intended for R.M.S. voltage and noise measurements of low level ac signals from 10-Hz to substantially beyond 20-kHz. Noise voltage measurements for audio equipment only have meaning when the measurement bandwidth is defined and specified, and this is where the versatility of the filter board comes into play. There a four bandwidth-limited modes of measurement, as selected by the "Mode" switch on the front panel. In the "Bypass" position, the voltmeter input is essentially connected directly to the L.N.A. output, with the measurement bandwidth being defined solely by the frequency response of the L.N.A. itself, which is in the order of 200-kHz. In the "20-kHz" position of the "Mode" switch, a single pole bandpass filter with a pass band gain of unity and a bandwidth of 20-Hz to 12.73-kHz is switched in between the L.N.A. and the voltmeter module. This filter setting is used for general purpose signal-to-noise voltage measurements of audio equipment in the full audio bandwidth; this filter has a noise bandwidth of 20-kHz, which is the figure used when computing the input-referred noise on this filter setting - more on this further on.
The third position of the "Mode" switch connects into circuit the A-weighting filter, which is a standardized psychoacoustic weighting filter for audio equipment noise measurement. This filter provides a small amplitude boost to mid range frequencies while attenuating the low and high frequencies at a specified rate, which supposedly relates to the average persons perception of noise loudness. This perception peaks at approximately 2-kHz but falls off appreciably at either extreme of the audio frequency spectrum. This filter has an equivalent noise bandwidth of 13.5-kHz. The final position of the "Mode" switch is the "Bandpass" setting. In this position the input of the voltmeter module is connected to the L.N.A. output via either one of eleven individual bandpass filters, as selected by the 11-position switch labeled "Bandpass Fc (Hz)". The bandpass filters have center frequencies ranging from 10-Hz to 20-kHz in 1, 2, 5 steps. Each filter has a quality factor (Q) of 3.14, which returns an equivalent noise bandwidth equal to half the center frequency (Fc), which is a convenient and straightforward value for use in noise calculations. These bandpass filters are specifically intended for profiling the flicker noise characteristic, typically called "1/f noise" of the piece of audio equipment under test. Technically, noise spectral density is the noise in a 1-Hz noise bandwidth, but, in practice, somewhat fortunately, one does not need to construct a 1-Hz bandwidth filter to conduct a spot noise measurement at any audio frequency. This is covered comprehensively by Motchenbacher and Connelly [2], who compute an additional measurement error, in a system dominated by 1/f noise, of only 0.22% for a bandpass filter bandwidth equal to 30% of Fc. The final module of the complete test set (except for the power supply) is the signal generator. The signal generator is specifically intended to be used, in conjunction with the instruments voltage metering function, for making transfer gain measurements on the device under test (D.U.T.) by what is known as the "sine wave method" [3]. It is necessary, for computing the input referred noise of a piece of equipment being measured for noise, to know not only the amplitude of the noise signal voltage, in the specified measurement bandwidth, as self-generated by the D.U.T. and present at its signal output, but its transfer voltage gain as well. When using the bandpass measurement filters to profile the 1/f noise of a D.U.T. having a non-flat frequency response characteristic, it is additionally necessary to know the actual voltage gain at each specific frequency of spot noise measurement. The operating frequency of the signal generator is thus switch-selectable in twelve fixed 1, 2, 5 steps from 10-Hz to 50-kHz. The sinusoidal waveform is generated by a Direct Digital Synthesis IC, so the frequency accuracy and stability of oscillation is essentially crystal-accurate. The generator provides a calibrated and fixed output voltage of 1-V R.M.S. terminated into 100 ohms, (2-V R.M.S. un-terminated). An external padding-type attenuator, housed in a screened diecast box, is used in conjunction with the signal generator output to provide excitation signal amplitudes in the range of 10-uV to 1-V R.M.S in switched 20-dB steps. The attenuator has an output impedance of 50-ohms.
|
|
Low Noise Amplifier
|
This scope photo is of the L.N.A's shorted-input, self-generated noise, as taken from the "Noise Output" jack with the A-weighting filter switched in. The horizontal timebase is set to 1-mS/division and the vertical scale is 50-mV/division. The L.N.A. gain is set to 80-dB. There is an additional 20-dB of gain provided by the metering board prior to the "Noise Output" jack. Total gain is therefore 100-dB; equal to 100,000 Av, giving an actual vertical sensitivity of 500-nV per division. This noise floor measures approximately 150-nV R.M.S. on the instruments meter movement. factoring in the 13.5-kHz equivalent noise bandwidth of the A-weighting filter, this computes to an input referred noise of approximately 1.3-nV sqrt/Hz; or -136 dBv. |
|
Filter Board
|
|
R.M.S. Voltmeter Board
|
|
Generator Board
|
|
Power Supply Board
|