Pocket-Size White Noise Generator for Quickly Testing Circuit Signal Response

However, CX can be optimized for flatness or bandwidth, with typical values ​​being CX = 30 nF to 50 nF. For wider bandwidth and higher peaking, use a smaller CX. For a faster decaying response, use a larger CX.

The key IC parameter is related to the op amp supply current; a low supply current device may require a slightly larger CX, while a high supply current device will most likely require less than 30 nF while achieving a wider flat bandwidth.

The curve here highlights how the CX value affects the closed-loop frequency response.

Measurement

The relationship between output noise density and CX (RS = 10 kΩ, ±2.5 V supply) is shown in Figure 4. The output RC filter can effectively remove the clock noise. The figure shows the relationship between output and frequency when CX = 0 and CX = 2.2 nF/10 nF/47 nF/68 nF.

Figure 4. Output noise density of the design shown in Figure 1.

CX = 2.2 nF shows slight peaking, which is strongest at CX = 10 nF and then gradually decreases as CX increases. The trace for CX = 68 nF shows no peaking, but the flat bandwidth is significantly lower. The best results are for CX around 47 nF; the clock noise is three orders of magnitude below the signal level. Due to limited vertical resolution, it is not possible to accurately judge the output amplitude flatness vs. frequency. This plot was produced using ±2.5 V battery supplies, but the design allows for the use of two coin-cell batteries (approximately ±1.5 V).

The Y-axis of Figure 5 shows the flatness after zooming in. For many applications, a flatness within 1 dB is sufficient, and <0.5 dB is typical. Here, CX = 50 nF is optimal (RS = 10 kΩ, VSUPPLY ±1.5 V); CX = 45 nF, but 55 nF is also acceptable.

Figure 5. Zoomed-in view of the output noise density of the design shown in Figure 1.

High-resolution flatness measurements take time; for this plot (10 Hz to 1 kHz, 1000 averages), it took about 20 minutes per trace. The standard solution uses CX = 50 nF. The 43nF, 47nF, and 56nF traces shown (all CS < 0.1% tolerance) show small but noticeable deviations from optimal flatness. The orange curve with CX = 0 was added to show that peaking improves the flat bandwidth (from 230 Hz to 380 Hz for Δ = 0.5 dB).

For exactly 50 nF, 2 × 0.1 μF C0G in series may be the simplest solution. 0.1 μF C0G 5% 1206 is readily available from Murata, TDK, and Kemet. Another option is 47 nF C0G (1206 or 0805); this part is smaller but may not be as common. As mentioned earlier, the optimal CX varies with the actual IC parameters.

We also checked the flatness vs. supply voltage, see Figure 6. The standard circuit is ±1.5 V. There is a small change in peaking and a small change in flatness when changing the supply voltage to ±1.0 V or ±2.5 V (because VN varies with the supply and thermal noise dominates). The change in peaking and flatness is about 0.2 dB over the entire supply voltage range. This curve shows that the amplitude stability and flatness are good when the circuit is powered by two small batteries.

Figure 6. Output noise density for various supply voltages.

For this prototype, the flatness is within 0.5 dB up to about 380 Hz for ±1.5 V supplies. At ±1.0 V supplies, there is a slight increase in flatness and peaking. There is no significant change in output level for ±1.5 V to ±2.5 V supply voltages. The total V pp (or V rms) output level is determined by the fixed 10 μV/√Hz density as well as the bandwidth. The output signal for this prototype is about 1.5 mV pp. At some very low frequencies (MHz range), the noise density may exceed the specified 10 μV/√Hz. For this prototype, it has been verified that the noise density remains at 10 μV/√Hz at 0.1 Hz.

In terms of stability over temperature, thermal noise dominates, so for T = 22 (±6) °C the amplitude varies by ±1%, which is barely visible on the plot.

EMI Considerations

The prototype uses a small copper foil with a polyimide insulation as a shield. This foil or flap is wrapped around the input element (10 M + 22 pF) and soldered to ground on the back side of the PCB. Changing the position of the flap has a significant effect on EMI sensitivity and the risk of low frequency (LF) spurs. Experiments have shown that the occasional LF spurs are caused by EMI that can be prevented with very good shielding. Using the flap, the prototype has a clean response in the lab without any additional mu metal shielding. No mains noise or other spurs show up on the spectrum analyzer. If excessive noise appears on the signal, additional EMI shielding may be needed.

When using an external power supply instead of batteries, common mode currents can easily add to the signal. It is recommended to connect the instrument ground with a solid conductor and use a CM choke in the generator supply wire.

limit

There are always applications that require more bandwidth, such as the full audio range or the ultrasonic range. Higher bandwidths are not practical at a supply current of a few μA. With a flat bandwidth of about 300 Hz to 400 Hz, the LTC2063 resistor noise-based circuit can be used to test the 50 Hz/60 Hz mains frequencies of some instruments, such as geophone applications. This range is suitable for testing various VLF applications, such as sensor systems, because the frequency range is down to less than 0.1 Hz.

The output signal level is low (<2 mV pp). The subsequent LTC2063 configured as a non-inverting amplifier with a gain of 5, plus another RC output filter, provides the same controlled 300 Hz flat broadband noise output, but with greater amplitude. In cases where the closed-loop frequency range cannot be maximized, the capacitor across the feedback resistor can reduce the overall bandwidth. In this case, the effects of RS and CX are small or even negligible at the edges of the closed-loop response.

Conclusion

The white noise generator described in this article is a small but important tool. As measurement time increases, a standard instrument for low-frequency applications—a simple, reliable, portable device that can measure circuit characteristics almost instantaneously—is a welcome addition to the engineer's toolbox. Unlike complex instruments with numerous settings, this generator does not require a user manual. The special design has a low supply current, which is critical for battery-powered operation in long-duration VLF application measurements. When the supply current is very low, no switching is required. The battery-operated generator also prevents common-mode currents.

The LTC2063 low power zero-drift op amp used in this design was key to meeting the project constraints. It enables the use of noise generating resistors amplified by a simple non-inverting op amp circuit.

Aaron Schultz [aaron.schultz@analog.com] is an Applications Engineering Manager in the LPS business unit. He has held various positions in design and applications systems engineering, working on a wide range of topics including battery management, photovoltaics, dimmable LED driver circuits, low voltage and high current DC-DC conversion, high speed fiber optic communications, advanced DDR3 memory development, custom tool development, verification, basic analog circuits, etc. He has spent half of his career in the power conversion field. He graduated from Carnegie Mellon University in 1993 and MIT in 1995. In the evening, he enjoys playing jazz piano.

Peter started working in electronics development in 1986. Since 1993 he has been working as an independent consultant in sensors and instrumentation. He has worked with many different clients, from small businesses to large companies and scientific institutions. Peter lives in the center of 's-Hertogenbosch next to St. John's Cathedral in the Netherlands and has seven HP3562A signal analyzers in his apartment.