By Trevor Smouter | Apr 11, 2014
Nuvation Engineering recently developed a product for a client that was a high-quality audio recording device with an integrated microphone. Our client needed their device to have a high dynamic range and low noise, but also a low production cost. During the design phase we used engineering best practices to maximize the dynamic range and minimize the noise, and we selected the most cost-effective components that would meet the client’s product requirements.
We could get a ballpark estimate of a the system’s performance using datasheet specifications and engineering analysis, but to accurately assess whether the design met dynamic range and noise specifications required complete system testing. This type of testing can be performed in an anechoic chamber, but for this situation renting and transporting all the equipment was cost prohibitive and too time-consuming.
Outside of an anechoic chamber it is difficult to test a recording device where the ambient noise affecting the mic is lower than the noise figure of the analog front end (AFE). One way to evaluate the AFE is to remove the mic and run experiments on the bench with the input shorted (AFE noise figure), or while driving the input with a signal generator (AFE signal to noise ratio). Unfortunately both these tests eliminate the microphone and the microphone bias circuits from the equation, and do not represent the true operating conditions.
We needed a way to evaluate the noise with the mic in circuit. Our client’s technique was to perform testing in a large soundproof chamber at night, very quietly, with the ventilation system shut down; we simply didn’t have that option. Initial brainstorming ideas for a low-sound environment revolved around making a box filled with sound-absorbing materials. But, we quickly realized that this effort could turn into an exercise of making better and better soundproof boxes, and we would never really know whether performance measurements were from the electrical noise we were trying to measure, or were ambient noise leaking.
One proposed alternative was to test the recorder in a vacuum. This made sense; removing the air from around the recorder which would significantly reduce the ambient sound getting to the mic. An inexpensive vacuum pump, called a water aspirator, can be connected to a cold water tap to generate enough vacuum (29inHg) to make water boil at room temperature. Luckily I happened to have a recirculating water aspirator at home so we didn’t need to waste too much water.
We needed a vessel to put the recorder in, and decided to use a brand new paint can with a hole drilled into the lid and a brass hose barb connecting the vacuum line. We liberally coated the lid in Vaseline to maintain the vacuum seal. Would that work? Well…no. Fortunately we had the good sense to test the idea without the device first.
The concept was good, as the Vaseline maintained a seal and the hose barb worked to connect it to the vacuum line. Unfortunately the paint can was no match for the crushing weight of Earth’s atmosphere at surface level.
Using a Pressure Cooker
We needed a much stronger vessel. We considered making a new paint can with a steel pipe insert to maintain the can’s structural integrity, but realized that can would likely only sustain a single vacuum application, since any deformation would break the seal.
A pressure cooker was suggested, and I realized that I had just the device to do it: an old lab-grade autoclave. By replacing the pressure gauge with a vacuum gauge and replacing the safety valve with a hose barb (a la Home Depot) we were able to setup a decent, almost professional, vacuum chamber. We added Vaseline to the tapered aluminum seal, since the pressure wedging it closed was going to be high and we wanted it to still be able to open the lid afterwards. We tried a vacuum test with the new setup; with thoughts of imploding paint cans in the front of our minds we took a step back and watched the vacuum gauge rise. The system was able to get down to 27.5 inHg, which is a respectable vacuum level. It took about 25 minutes to get to that vacuum. Knowing the random motion of particles in near-vacuum requires the gas molecules to bounce around down to the end of the 10 foot hose by happenstance before the aspirator can eject it from the party, 25 minutes isn’t that bad.
To ensure vibrations did not affect the recording device the recorder was suspended from an elastic band within the chamber and the chamber was placed on a padded chair.
To test the vacuum system for sound insulation the recorder was placed in the vacuum chamber and a sound system was placed up against the chamber. The radio was played at a very loud volume and the chamber was pumped down. As can be seen in the picture above a valve was inserted in the vacuum line to ensure so that once the vacuum had been established the valve could be closed and the pump turned off. The sound recording is shown below.
The beginning of the file shows a large response while the chamber is being closed. There is a low level of noise until the vacuum pump is turned on which is really quite loud. As the air is pumped out of the chamber the sound of the pump gets harder and harder to hear. The spikes that can be seen in the response during the pumping down are from tapping on the vacuum gauge to get the proper pressure reading. Close to the end of the recording the valve is closed and there is basically no sound heard even though the radio is very loud right up against the chamber. Finally when the vacuum is released there is a tremendous sound in the chamber.
All in all, the vacuum testing was a complete success. We got the data we needed and only a single paint can was destroyed in the process!