Signal Disappears After C8
With one issue solved, we saw another problem slightly further down the signal path. After capacitor C8 (see below), the signal line feeding into the variable gain amplifier was just noise. This did not seem like a design issue, so we looked back at the circuit layout. It turns out that in the layout an obvious error had escaped our notice. Under capacitor C9, the two pads were connected with a trace, effectively shorting the VGA input to ground. No wonder it looked like noise! It took us a while to find this issue because we started out looking for abnormalities on the physical circuit board. That trace was hidden under C9, so it wasn’t actually visible. After a minute or two with an knife, the short was gone.
Still No Input
We weren’t done debugging the VGA circuit yet. It turned out that there was a cold solder joint on the input pin, which kept it from feeding into the chip. Also, we were still getting noise at the input despite having fixed the short at C9. The thought came to mind that the issue might be due to the lack of termination on the input of the INH pin. We had discussed terminating it with active impedance matching, which is a special feature of the VGA chip, during the design process. However, we were warned by Dr. Richard that the active impedance match altered the frequency response at high frequencies. That made us decide to not do it. Still stymied by the problem of just getting noise at the input, we decided to try out the active impedance matching. Low and behold, it worked! This was surprising given that the feature is shown as optional in the datasheet. After that, we were getting sinusoidal outputs from the VGA with sinusoidal input.
Wrong Common-Mode Bias for Differential Signaling
By this point, we thought we had the problems figured out with the VGA, but that was not the case. The VCM pin, which sets a common-mode voltage for the differential output, was at a DC level of 2.5V despite the fact that we were driving the pin with a resistive divider. The intended setting was 1.65V, which happens to be the ADC’s center reference. This was puzzling at first, but looking back at the datasheet for the VGA, we noticed that that pin needs to be driven by an output impedance of less than 30 ohms and a current greater than 2mA. Our resistive divider was not getting close to either of those requirements. Thankfully, we had a few extra 33 ohm resistors to remake the bias circuit with lower output impedance and a much higher current.