Hello,
I built a low noise amplifier as a learning experience for more complex circuits but have encountered unexpected behaviour that I can't explain with my level of understanding. The design was based on the parallel op-amp concept of the LNA in Jaromir's design from the nanovoltmeter contest (
https://www.eevblog.com/forum/metrology/how-i-made-my-nanovoltmeter/). The goal was to simplify things by only making a modified amplifier that I could extend the capabilities of other instruments with. While I was expecting to see issues with noise and offset, what I ended up with is an extreme power draw level and voltages throughout the design that I can't explain.
I kept the basic structure of a non-inverting amplifier in the first stage, followed by summing amplifiers in the second and third stages. The number of op-amps in the first stage were kept at 10, but I reduced the number of second stage groups to just 2. In each stage I switched the op amps to the OPA210 since the noise levels looked good (90nV for 0.1 to 10 Hz), removed the ferrites, and lowered the resistor values. The ferrites and resistor values were a mistake caused by my lack of understanding of how noise adds in resistor dividers and the need for filtering/protection. While the board was built with little protection, I did add some diodes to the second stage input to prevent damage if the first stage maxed out.
The with the attached schematics I did not populate the input current compensation part (for input bias current), the first stage ground current injection was setup. Power is from two groups of 3 LiFePO4 cells (600 mAh) to form a dual supply. Chassis ground is through the 1M resistor. The board is set up with a ground plane on the back and a "shield plane" connected to ground on the second layer between the power plane on the third layer and the components on the top layer. To minimize thermal EMF I used 81Sn-19Bi solder based on Vishay Technote 26, thin film resistors, the input connectors to the board are pure copper and were polished before being wire wrapped with SPC wire and connected to Pomona 3770 TeCu binding posts. All testing was with the LNA set at x1000 gain via the rotary switch with the 500Ω resistance verified across the switch.
Now to the issue. When I turned on the power for the first time with nothing connected I immediately saw and smelled smoke. That appears to be from some flux I missed, but I didn't expect to see that much heat without any voltage input. All the op-amps (all three stages) get quite hot (beyond what is comfortable to touch for even a moment) right away and the batteries all get warm. I tried shorting the inputs with a piece of SPC wire but it didn't change any results. The voltage at the output is somewhat inconsistent, it spikes very negative when power is applied (I have seen it approach the negative power supply voltage on some tests) then settles at a more mildly negative voltage. I have seen settled voltages in -500 mV several times, +120ish mV once when the batteries were dying (it drains the batteries in minutes, though I never allow it to run for longer than ~15 seconds), then -236 mV in the latest test. When measuring between ground and the second stage inverting input I am seeing ~0.4 V, suggesting that the diodes might be conducting to relieve the voltage and current at the second stage op amp is maxed out at 25 mA. That indicates that the first stage op-amps should be putting out above 2.5 mA, meaning at least an output of 675 mV (400 mV + 275 mV). given the x11 gain of the first stage, that would suggest an input of ~60 mV. The datasheet for the OPA210 gives a max input offset of 50 µV and even assuming a temperature differential of 60˚C and thermal EMF of 3 µV/˚C I don't get anywhere near that error. I have check and can't find any solder bridges, both first stage groups behave similarly.
I assume I made a beginner's mistake somewhere that is leading to excessive power consumption or the high offset voltage but I can't figure out at all what it is.
Home
Help
Search
Login
Register
