AFTER BUILDING SEVERAL different active-filter noise-reduction devices, I found myself in need of a standard source of white noise for comparing the units. Having also acquired an octave-band equalizer for my system, I also wanted an accurate pink-noise source that I could use to adjust the bandpass sliders in con junction with a real-time analyzer.
After reviewing the characteristics of broadband noise, I built a broadband digital noise generator that satisfies both my requirements.
White Noise, Pink Noise
Noise sources are useful for audio testing because they provide a continuous signal source across the audible spectrum. That means the ideal noise source generates all the individual frequencies simultaneously in a random mix. In contrast, an audio oscillator generates only one frequency at a time.
Of the many different types of electronic noise, the two most useful for audio work are white and pink noise. White noise contains constant energy per spectrum line, while pink noise contains constant energy per spectrum octave. Expressed as a voltage, white noise has the same voltage level at all frequencies. Pink noise rolls off at a rate of -3dB/octave, as illustrated in Fig. 1.
Real-time spectrum analyzers for home audio systems are constant percentage bandwidth instruments with an analysis window whose width increases as the frequency rises. Because of this, these units display a flat response to pink-noise in put and a rising +3dB/octave response to white noise, as shown in Fig. 2. Figures 1 and 2 are equivalent.
Audible white noise is a “hissing” sound (similar to the noise between FM stations), whereas pink noise is more of a “rushing ” sound, due to its preponderance of low-frequency components.
The author's digital noise generator
is an accurate pink and white-noise source for testing audio devices.
Practical Noise Generators
Any practical noise generator circuit includes a broadband white-noise source, a pinking filter, amplifying and output buffering components, and a suitable power supply. The critical component, of course, is the noise source itself.
Common electronic components such as resistors and semiconductors are known for their ability to intro duce unwanted noise in audio circuits. Since all semiconductor junctions generate some random noise, you can assemble a simple noise generator using a forward-biased diode.
The emitter-base junction of a bipolar transistor also becomes a noise source when reverse-biased. The problem with junction noise sources is that they tend to be unpredictable in the quality and quantity of their noise output. It seems that when you want a noisy transistor, you can't find one! Because the output varies from device to device, you must either select the semiconductor specially or tailor the circuit gain to the device. I tried to avoid these hassles by finding a noise source that would always function predictably in a generic circuit. Fortunately, the folks at National Semiconductor have provided a solution.
The MM5837 Chip
The MM5837 (or alternate designation S2688) is an eight-pin PMOS broadband digital white-noise source whose internal circuit consists of a 17-stage pseudorandom shift register driven by an internal clock. National's literature describes the chip as having a “very uniform noise quality and output amplitude.” This is at least partially true.
Compared to semiconductor noise sources, the MM5837 is unquestionably a major improvement. The un loaded noise output with a 15V supply is around 11V , much higher than any junction noise source. The quality of the noise is uniform, but The first is the internal clocks frequency. Although the clock is not externally adjustable, it does vary with the supply voltage, reaching a peak of around 40kHz at about 13V. Since the peak clock frequency results in the most broadband noise, a 13V supply will provide optimum-quality noise for audio use.
The second and more serious factor involves the design philosophy of the chip. It is a pseudorandom rather than a random noise source. Pseudo random noise is produced by a rapidly clocked string of shift registers in a repeating cycle. When the end of the string is reached, the cycle is repeated, with the cycle frequency being around 0.5 to 1Hz, depending on clock frequency and supply voltage.
At the end of each cycle, a low-frequency pulse or series of pulses is generated, which produces an audible beat heard over the noise level.
The beat pattern varies from chip to chip, is always low frequency and is always very regular. This “bumpety bump-bump ” beat is rather irritating and tends to interfere with audio measurements in the bottom octaves. To make matters worse, the beat tones are enhanced when the output is passed through a pinking filter.
This pseudorandom noise problem baffled me. I could not filter out the beat tones without seriously altering the frequency response of the generated noise. I could, however, try to smooth out the overall output by blending the outputs from several chips. After some experimentation, this approach yielded satisfactory results. I then constructed the circuit shown in Fig. 3.
The Circuit
In Fig. 3, you can see that IC1-IC4 serve as the white-noise source. Coupling capacitors C9-C12 block the DC bias present at the chip outputs, while R6-R9 form part of the summing circuit, which blends the four separate noise outputs. Opamp IC5 performs the summing operation with some gain. One output of IC5 is coupled without further gain through C21 to output buffer IC6a, which provides the white-noise out put. The second output of IC5 feeds the pinking filter, which consists of R14-R17 and C15-C18. This filter provides the -3dB/octave roll-off required to produce pink noise. It is a common-value composite of pinking-filter circuits collected from several sources.
The output of the pinking filter, whose amplitude is greatly reduced relative to the white-noise input, is coupled through C14 to output buffer IC6b, with sufficient gain to bring the pink noise up to the white-noise output level. C13, C19 and C22 provide radio-frequency (RF) stabilization for the op amps, while R11 and R22 provide current limiting to prevent gross distortion and instability when driving low-impedance loads.
The power supply has some special features. Regulated and filtered supply voltages of +15V, +13V and -15V are provided, even though transformer T1 lacks a grounded center tap. Diodes D1-D4 rectify T1's output on both sides of the out put waveform. This results in efficient use of the available power by creating two separate supplies with floating grounds. Each of these sup plies is individually filtered and zener regulated. A common circuit ground is made by connecting the negative side of the upper supply to the positive side of the lower supply.
FIGURE 1: Basic characteristics of white and pink-noise spectra. (Graphics and photos by the author.)
FIGURE 2
-------------------
Parts List
Diodes & ICs 1000, 1W 0.02xF, 600V ceramic D1-D4 1N4003, 1A, 200 PIV 220 0.014F, 600V ceramic D5, D6 1N47448B, 15V, 1W, 5% zener 820 1,000xF, 35V radial electrolytic D7-D9 1N914 silicon signal diode 2.2k, AW 2204F, 16V axial electrolytic D10 LED, XC556R or equivalent 100k 0.1 uF, 50V ceramic IC1-iC4 MM5837/52688 digital noise 5600 220uF, 16V radial electrolytic source 100k 10uF, 25V radial electrolytic IC5 TLO81 BIFET op amp 4.7k 5pF ceramic or mica IC6 LF353 dual BIFET op amp 3000, 2% 10 uF, 16V radial electrolytic 1k, 2% 0.033 uF, 100V Mylar, 5% Misc.
3k, 2% 0.1 uF, 100V Mylar, 5% $1 SPST mini toggle switch 6.8k, 2% 0.33uF, 100V Mylar, 5% S2 SPDT mini toggle switch 220k 1uF, 100V Mylar, 5% m 25V, 300mA transformer 100k 5pF ceramic or mica F1 1/2A fast-blow fuse 5600 0.14F, 50V ceramic VR1,VR2 1M panel-mount pot 3.3k 47uF, 35V radial electrolytic 5pF ceramic or mica metal chassis and mounting hardware, hookup wire, AC line cord and strain relief, LED holder, panel-mount RCA jacks, fuse holder, knobs, chip sockets.
All resistors are 2 or 0.5W, 5% or 10% tolerance, carbon film or composition, unless otherwise noted.
--------------------
FIGURE 3: Schematic diagram of the digital noise generator.
NOTES
1. ALL CAPS EXPRESSED AS nF/VOLTAGE, EXCEPT AS NOTED.
2. ALL RESISTORS ARE FIXED COMPOSITION, 5% OR 10%, 2 OR ½ W, EXCEPT AS NOTED.
---------------------
This forces the floating negative side of the lower supply to become the negative rail of the common bipolar supply.
FIGURE 4; FIGURE 5: Stuffing guide for the author's unit.
FIGURE 6: According to Mr. Poehland, his prototype has worked well for
some time and has been a valuable addition to his test bench.
D7-D9, R23 and C8 provide the + 13V for the noise chips. The value of R23 is not critical. You can raise or lower it as required or even eliminate it if the + 15V is a little low. Its purpose is to pull a little more current through the dropping diodes to en sure an extra 100mV or so drop (be yond the nominal 1.8V caused by the diodes). This yields a final value within 200mV of the desired +13V.
Finally, C1 and C2 provide transient suppression, while C7 and C20 provide RF bypassing for the opamps.
Parts and Construction
One of the things I like about this project is that, with the exception of the pinking-filter components, component quality and tolerances are not critical. I built my prototype using 10 percent composition resistors salvaged from all the junk I have lying around.
The pinking-filter components should have tight tolerances, though, because the filter's accuracy is directly related to the parts' values and stability. I recommend film resistors and low-loss film caps. I selected the op amps for their high slew rate, wide bandwidth and availability. Lower quality units such as the 741 and 1458 will function in the circuit, but their poor slewing performance will reduce the noise bandwidth and distort the output waveform.
The power transformer, line cord, op amps, output jacks, switches, chip sockets, chassis, hardware and most of the resistors, diodes and capacitors are stock items at Radio Shack. The MM5837 digital white noise source IC is available from Jameco Electronics ( 1355 Shoreway Rd., Belmont, CA 94002) and Digi Key (Highway 32 South, PO Box 677, Thief River Falls, MN 56701). Digi Key also carries a complete line of precision resistors and film capacitors, as does Old Colony.
Construction is simplified by mounting the components on a circuit card, using the foil pattern and stuffing guide in Figures 4 and 5. First install the jumper wires, resistors and diodes. Then mount the chip sockets and capacitors, adding the transformer last. Observe the correct polarities of the diodes, ICs and electrolytic capacitors. When mounting the power transformer, slip a flat washer under the transformer mounting tabs to avoid bowing the board.
Also, ensure good metal contact with the tab by placing a lockwasher un der the head of the screw that connects the left tab to circuit ground on the foil side. I have provided ample space at the comers of the board for the mounting holes. The exact location of these is up to you. Note that the noise chips are mounted in 16 pin DIP sockets to conserve space and that the two pairs of chips face in opposite directions. Sockets are a must for all the ICs.
When construction is complete, install the op amps in their sockets.
Do not install the noise chips yet.
You must test them first. Apply power to the circuit without the noise chips in place. Notice that the LED glows. After a few minutes, check the supply voltages at TP-1 and TP-2 against the values in Table 1. Cut power to the circuit and connect either a speaker, headphones or an oscilloscope probe to any of the out put jacks, with the gain controls up full.
Plug one of the MM5837/52688 noise chips into the socket closest to D7 (observe polarity) and apply power to the circuit. If the chip is okay, you will hear the noise or see the complex waveform on the scope. Cut power, remove the chip, plug in the next one and check the remaining chips in this way. If they are all good, install them in their sockets and ap ply power to the circuit.
Allow at least ten minutes for stabilization, then measure the voltage at TP-3. It should be as specified in Table 1. A bad chip is rare but difficult to detect once they are all in place. If none of the noise chips tests functional, the problem could be in one of the op amp ICs or the output
TABLE 1 -- Measured with digital multimeter, input Z equals 10M relative to common circuit ground.
Performance
The pinking filter in my prototype (Fig. 6) exhibited a linear roll-off slope of -2.93dB/octave from 20Hz to 20kHz. Below 20Hz and above 40kHz, the roll-off was steeper. I evaluated the two noise outputs with a BSR SX-100 real-time spectrum analyzer. The bandwidths (Figs. 7 and 8) agree closely with the ideal spectrum slopes displayed in Fig. 2.
The circuit has a lot of gain and will easily drive the line-level inputs of any amp or receiver to overload.
(Keep all level controls at minimum when making connections.) With the level controls up full, the generator will drive low-impedance loads (such as loudspeakers) directly, but at reduced volume.
FIGURE 8: Real-time spectrum analysis of the white-noise output (36dB
scale).
The outputs are short-circuit protected. During extended use, the power-supply components, especially T1, R1 and R2, become noticeably warm, but after a 24-hour burn-in period with both outputs shorted, my prototype still functioned normally. The output configurations are flexible, with two fixed outputs on either side and a switchable one in the middle. This permits direct in put/output comparisons or switching from one type of noise to another without having to disconnect any thing.
I have been using my digital noise generator for some time. It meets or surpasses all my requirements for uses such as equalizing my sound system, testing audio filters and crossovers, and quick-checking speaker systems. It is a useful audio accessory and versatile addition to the test bench.
ABOUT THE AUTHOR
Mr. Poehland has a B.S. in chemistry and works in the pharmaceutical industry doing antibiotic drug research. He has published several scientific papers on liquid chromatography and has been involved in “roll-your own ” audio for seven years.
REFERENCES
1. Petzold, John A., “A White Noise Generator and Pink Filter, ” TAA (3/76), p. 3.
2. “Pink Noise Generator,' Audio/Radio Handbook, National Semiconductor Corp. ( Santa Clara, CA), 1980, pp. 2-62.
3. Mims, Forrest M., 'Experimenting with Noise, ” Popular Electronics (Volume 17, Number 3), March 1980, p. 81.
4. Mims, Forrest M., “Experimenter's Corner: Reader Letters, ” Popular Electronics (Volume 19, Number 2), February 1981, p. 98.
5. “CMOS Noise Generator, ” Elektor (Volume 4, Number 1), January 1978, p. 1-05.
6. “S2688 Digital Noise Generator, ” Archer Semiconductor Replacement Guide, Radio Shack Division, Tandy Corp. ( Fort Worth, TX), 1981, p. 42.
--------------
Benjamin L. Poehland
-------------------
Also see: