I am using AD8655 as LNA for piezo transducer signal conditioning. I am getting a strange output frequency spectrum, lots of spikes every 2200Hz, roughly.
The low-pass and high-pass roll-off is accomplished further down the signal chain. The above spectrum was captured with J8 shorted (see schematic snippets below).
I directly shorted the TR_NETWORK net side of R7, and got the following spectrum:
Any ideas what could be causing the noise in the first spectrum and how to remove it? Please let me know what other details would be helpful.
I have tried changing R7 to 200ohm, adding 100pF capacitor to make low-pass RC on non-inverting input of AD8655, but that did not do too much. Some colleagues had applied a full 0.22uF to form LPF with R7=10ohm, but that I think was just hiding the problem by creating a capacitive divider with C155 (2200pF), which just knocked the gain down tremendously.
Is there a way that the AD8655 can be isolated? If it is possible, it would be good to ground the input of the amplifier through R7 and do a measurement to check whether the noise is still present. This will allow you to at least isolate whether the noise is coming from the amplifier circuit or from the input interface.
Also, what is providing the +/-3.3V rails?
Grounding R7 removes the noise. +/-3.3V rails provided by LDO. I realized AD8655 has max recommended input supply of 5.5V, which I'm violating. However, I fixed this on LDOs and replaced AD8655, but noise still present. I'm changing the design to use AD8421, which I have used in similar setup with none of these issues.