On Monday 11th August, the Section held its AGM via Zoom.
(full details at https://aesmelbourne.org.au/2025-agm/).
Following the brief AGM, audio hardware designer Paul Messick presented on the topic of:
Low Noise Design – a How To
– Photo Rodney Staples
In his introduction, Section Chairman Graeme Huon outlined Paul’s wide experience in his field, names like Dallas Semiconductor, JPL, M Audio/Digidesign, where he was head of engineering. We heard how he later founded Avermetrics (production line and benchtop audio analysers), and since moving to Australia, has founded Glassworks Audio which specialises in high-end professional audio equipment. Paul started his presentation with an apology and a warning that “there will be mathematics”.
After providing an overview of the topics to be covered, Paul listed the types of noise and their characteristics.
He zeroed in on Johnson Noise, which is inherent in all resistors, and is proportional to the resistance, noting that a 1 kilohm resistor (in a 20kHz bandwidth) exhibits an inherent 0.29uV of noise, whereas a 100 kilohm one has 2.9uV. He went on to describe the excess noise present in different resistor types, from the noisiest being the old carbon composition type and the least noisy (but potentially inductive) the wirewound construction, noting that thin film and metal film varieties tend to be a common choice.
He took us through the formula for calculating the noise voltage for a specific value of resistance.
Paul also covered the mechanisms for generating resistor distortion, indicating that thin film resistors are the sweet spot for minimum distortion, and thick film types should be avoided. He commented that the distortion inherent in resistors only becomes important with very low-distortion designs, -110dB and below, where it might make one or two dB difference.
Paul then took a short detour into Shot Noise in transistors, noting that a balance was necessary between high collector current (lower voltage noise but more current noise) and lower collector current (with less current noise but higher voltage noise) depending on the signal source.
He then discussed EIN (Equivalent Input Noise), describing it as the equivalent noise present at the input of a perfectly noiseless amplifier to achieve the noise level found at its output. He then discussed the case of needing to insert a pad into an amplifier to adapt the gain for differing input conditions. Using maths, he showed that putting the pad after the gain stage gave superior noise performance.
Paul then spoke of parts selection for practical circuits, starting with the TI OPA1612 opamp, describing it as an almost ideal opamp, but expensive at around US$5 in small quantities. He compared it to the older classic NE5532 priced at less than US$1 in similar quantities. He pointed out that, paradoxically, the one specification where the 5532 is better than the OPA1612 is Input Current Noise Density (0.7 picoAmps per root Hertz vs 1.7 picoAmps per root Hertz ), making the cheaper part possibly a better choice for high impedance sources where current noise will dominate.
He then started to outline the design process for a microphone preamplifier. First, he discussed self-noise and output impedance of a typical high-quality microphone (4dB and 150 ohms). Then he went through the target specifications for high gain and low noise. The numbers he came up with were a target EIN as close to the noise of a 150 ohm resistor (-133dBV with 20kHz bandwidth), a gain of 0-60dB, a bandwidth of at least 50kHz at 60dB gain, a maximum output level of +22dBu and a maximum input level of +22dBu (at 0dB gain). He also mentioned the challenge of getting the gain down to 0dB because the gain-setting pot has a finite maximum value, hence the inclusion of a switchable 20dB pad in many designs.
Paul then presented his basic design using three opamps. He cautioned that, as presented, it did not include “niceties” like input protection, RFI filtering, or phantom powering. He commented that without some of these, the design may not survive exposure to the real world.
He noted that Rg, the gain-setting resistor (potentiometer), would need to be infinity for 0dB gain, so with real-world pots, if you need to get down to 0dB, a pad would be needed. He also mentioned that often a large-value capacitor would be added in series with Rg to suppress wiper noise.
He then took us through the noise calculation, arriving at the conclusion that the design’s target EIN is below the -133dBV of a 150ohm source microphone.
Paul then presented the measured performance of the breadboarded circuit. Taking real-world effects into account, the measurement came extremely close to the calculated design.
After presenting his design Paul took us on a side-tour of engineering vs marketing with regard to 32bit DACs and ADCs – showing with maths that because of real-world noise performance only 20 to 22 bits can be used, and even if full 32-bit utilisation was possible the resistance values for those noise levels would be impossibly small, being 16 x 10 -3 ohms at room temperature. Even liquid nitrogen cooling to a 40K temperature only brings the target resistance to a still-not-practical 0.12 ohms. He posited that even the heat death of the Universe, at 1 Kelvin, would only increase that value to about 5 ohms – so it’s just not practical to achieve those sorts of noise levels as long as we continue to exist.
An informal Q&A occurred at the end of the presentation, with many quite specific questions being raised by a very engaged audience. Topics like choices of alternative devices to the OPA1612, improving Power Supply Rejection Ratio, and esoteric discrete designs versus monolithic IC solutions.
One very distinguished member expressed an appreciation for Paul’s reliance on maths and then demonstrating the calculations necessary. There was also a discussion on how to measure noise and distortion down to these very low levels in an electrically noisy environment. Paul told us that he tries to design circuits and boards for noise immunity, but sometimes it is necessary to turn off specific lights or other equipment.
Paul was extremely generous with his time, carrying on the conversation and answering questions long after the allotted time.
The Melbourne Section thanks Paul for the time he put into the preparation and presentation of this important and interesting topic.
Here’s an edited video of the Zoom session:
This video can be viewed directly on YouTube at:
https://youtu.be/4s-CLXnMExI
A PDF of Paul’s slides is available at:
https://aesmelbourne.org.au/wp-content/media/Low-Noise-Design.pdf
Author: Peter Smerdon (Secretary)
