A white noise generator is a surprisingly useful tool in an audio lab. It allows you to quickly check the frequency response of a preamplifier or power amplifier, highlight resonances and colorations, verify the effectiveness of a filter or tone control, calibrate a simple RTA analyzer and, above all, “stress” a signal path in a repeatable way to track down hum, hiss and grounding problems. In this article I describe a small, fully transistor-based white noise generator, designed to be inexpensive, easy to build and suitable for direct connection to a line-level input. The output is buffered, the level is compatible with Hi-Fi electronics and the power supply is stabilized and filtered so that the unit itself does not become a source of interference.
Having a white noise generator can be very useful for performing various tests on audio electronics, so I designed a small transistor-based generator capable of delivering white noise with sufficient bandwidth for testing audio equipment and an output level compatible with a line input. The basic idea is to create a source that is “dirty in the right way”, meaning wideband and statistically random noise, but with a controllable level and an output capable of driving the typical impedance of a line input without trouble, even when it drops to relatively low values.
The circuit is deliberately simple and uses common components. The noise source is the EB junction of a BC337 reverse biased. Under these conditions the junction operates in breakdown and generates wideband random noise, a raw noise that is then amplified. This solution is practical because it does not require special Zener diodes or selected transistors, and in most cases already provides more than enough noise level for audio test applications.
The signal generated this way is very small and must be brought up to a usable level. The two following BC337 transistors therefore work as audio amplification stages, with a high overall gain, on the order of about 60 dB. Such a high gain is convenient because it makes it easy to obtain a line-level output, but it is also the reason why construction must be careful. Any disturbance physically picked up by the circuit, typically 50 Hz mains hum, would be amplified together with the noise, ruining the result and making the generator unusable for serious measurements.
To make the output stable and less influenced by the load, I added a BF256 JFET buffer used as a decoupling stage. In practice, the transistor stages generate and amplify the noise, while the JFET isolates the circuit from the outside world, providing a low output impedance and reducing sensitivity to load variations. This is important because, in the lab, the generator may be connected to different inputs, long cables, attenuators, or equipment with impedances that are not always ideal.
R8 and R5 form the feedback network whose purpose is precisely to compensate for the effect of the load on the output. The goal is not to achieve any “magic” distortion figure, which is irrelevant here since we are generating noise, but to maintain a reasonably constant level and predictable behavior when the input impedance of the connected device drops. In this project the output remains usable even with loads down to 22kohm, a value that covers many real line inputs and quite a few passive attenuators.
The power supply is another critical point, often underestimated in small bench generators. The circuit is powered at 15volt stabilized using an LM317L, more than sufficient to supply the roughly 7mA drawn by the generator. I deliberately chose a simple solution, a small salvaged external wall-wart power supply providing 24volt DC, so that everything directly connected to the mains stays outside the enclosure. This reduces risk and, if the wiring is done properly, also helps limit the coupling of mains-related interference into the high-gain circuit.
To prevent any noise or ripple from the external power supply from entering the generator, the RC cell R9/C6 acts as an additional filter upstream of the most sensitive section. In practice it is a “second wall” that attenuates unwanted components before they can reach the amplification stages. This is particularly useful when using inexpensive switching supplies, which can introduce high-frequency spurs, and when working with long cables or imperfect outlets.
Because of the high gain, it is essential to shield the circuit properly and set up grounding and wiring correctly. If left unshielded or mounted in a plastic enclosure, the generator would easily pick up the 50 Hz field and ambient interference, amplifying it excessively. A metal enclosure properly connected to ground, short connections, a well thought-out ground point that avoids loops, and an output connector firmly fixed make the difference between a clean generator and one that produces more hum than white noise.
In practical use, this generator can be employed in many ways: as a source to measure frequency response with a sound card and RTA software, to compare two signal paths, for example before and after a modification, to quickly check whether one channel “sounds” different from the other, or to verify the effectiveness of a noise filter or a shielding intervention. Even without sophisticated instrumentation, white noise is useful because it immediately highlights anomalies, such as superimposed hum, excessive hiss, a bandwidth that rolls off, one channel being more attenuated, or a contact or grounding problem.
If you want to be precise, you can also adjust the output level so that it is consistent with real sources, for example a few hundred millivolts RMS into a line input, using a simple attenuator or calibrating the level with a true RMS multimeter or via a sound card measurement. This is not strictly necessary, but it helps to repeat tests consistently and to avoid overdriving the input stages of the device under test. Here is the schematic, click to enlarge.
Looking at the schematic, the “lab-oriented” approach is immediately clear. A simple noise source, two gain stages to bring the level up to usable values, and a final buffer to make the output robust. It is a project that favors practicality and reliability, and it can be built on perfboard or on a home-etched board without difficulty. With neat assembly and good shielding, the result is a compact generator that becomes very convenient to keep on the bench. And here is my build.
In the practical build I tried to keep connections as short as possible in the high-gain areas, physically separating the power supply and filtering section from the amplification stages. Component placement also matters. The more compact the signal path, the less “antenna” area is exposed to ambient electromagnetic fields. If you plan to replicate this circuit, my advice is to think about the layout first and only then solder, because with 60 dB of gain an improvised layout often leads to disappointing results.
Here the compactness of the circuit is even clearer. If you use a metal enclosure, it is worth connecting the main ground point to the enclosure at a single point, near the output connector, to reduce the likelihood of loops. Alternatively, if the generator is connected to equipment that is already well grounded, it may make sense to include a small resistor or decoupling network between signal ground and chassis, but in most cases clean wiring is already sufficient.
Once completed, this small generator becomes a “Swiss army knife” for testing. You connect it, set the level, and in a few minutes you can tell if one channel is noisier, if an input has a problem, if a filter works as expected, or if the shielding of a device is insufficient. With very little expense and few components, it is one of those projects worth building because it turns out to be useful far more often than you might think.