Friday nights, I play bass in a jazz trio. The sound of my upright acoustic bass easily fills our small club, so I constantly have to think about the level of noise I'm making, and that brings me to the subject of this column.
I recently concluded some power-supply immunity testing on an IC. The test fixture directly injects intentional noise into the power terminal of the DUT (device under test) to see how much it can take without failing.
Ordinarily, I place an impedance in series with the power terminal of the DUT, severely weakening the power system at that point and making it easy to inject substantial quantities of intentional noise into the chip's power terminal. This time a special requirement complicated my test: that the power system feeding the DUT remain unaltered during testing. Making matters even worse, the power system for this layout was hefty. Analysis of the design revealed a power-to-ground impedance of about 0.02Ω at all frequencies as high as 1 GHz.
Under these conditions, to induce ±200 mV of noise on the power plane feeding my DUT, the test apparatus must pump at least ±10A into the power system. To generate that much current, a 50Ω laboratory signal source needs an open-circuit voltage of (10A)×(50Ω)= ±500V and dissipates 2500W rms when clamped to VCC by this stiff power system. I've seen bass amps that pump out that much power, but my lab signal source—nor any source I can afford to rent—does not cut it.
How about using a transformer to overcome the mismatch between the source impedance (50Ω) and the load impedance (0.02Ω)? The impedance ratio is 2500-to-1, so a transformer with a 50-to-1 turns ratio should, in theory, reduce the required source current to a level of (±10V)/(50Ω)=200 mA. The required source voltage would then be (200 mA)×(50Ω)= ±10V—an achievable figure. Unfortunately, I was unable to find, or figure out how to construct, any 1-GHz transformer suitable for such a low load impedance. Friday afternoon rolled by, and I still had no way to properly stimulate my system.
Over the weekend, thumbing through the pages of Bass Player magazine, I came across a picture of a young woman standing in front of a wall of bass speaker cabinets. Her setup arrayed 32 10-in. speakers in parallel, producing an impregnable wall of sound. That picture triggered a crucial thought in my brain: A massive array of sources creates a huge amount of noise.
My test card now has the DUT on one side, with its power system. On the other side of the card, using a separate power system but shared ground planes, sits a huge FPGA with 400 outputs. Each output pumps noise through a dc blocking capacitor directly onto the VCC plane that the DUT uses. Each FPGA output is programmed as a 50Ω source powered by 3.3V. This circuit easily drives the required ±200 mV onto my DUT power plane. By adjusting the number of active outputs, I control the level of noise.
If you build a similar circuit, don't let it run very long. The source, like good jazz, gets really hot.