input voltage is expected to be in the order of 10's of uVolts, so I need ~10^5 voltage gain.
I intend to make it with high frequency opamps. For the gain-control part I will probably use some digital potentiometer between gain stages(or is there a better choise?).
1. I am not experienced with noise-sensitive things, but I guess that the common mode rejection ratio mentioned everywhere is an important part in this device to suppress common noise at inputs.
I could make the first stage with a transformer(to serve as unbalanced->balanced converter) followed by a low noise high speed opamp, like one of these:
Another option is to eliminate the transformer and take a nice instrumentation amplifier
One more piece of advice I have seen in the Internet is to go fully differential from input to ADC (Once I had a look at a professional Imagenex sonar board, and they did exactly that with fully-differential opamps ). But I am not sure how much sense it makes in my case if I am using a single-ended ADC.
Also one more general thing I would like to clarify about opamp circuit design is this: if I only care about high frequency part of spectrum, should I care much about opamp specs like offset voltage and input current? As I understand it makes huge difference only if I were to amplify DC signals.
What's the bandwidth? I presume it's not necessary to go from DC to 240kHz? I imagine the bandwidth is very narrow.
I don't have any experience with current feedback or transconductance amps yet, but I will see what I can do with them in a simulator.
As for the transformer, as I understand I must use it to match source resistance to input resistance for optimal noise. But I'm not sure if I will ever know it, unfortunately.
It is only not clear for me why there is noise increase in case of differential inputs. I thought that differential inputs are meant for decrease of noise.
Not all ADCs have high bandwidth sample and hold. Whether you go fully differential will depend on the ADC you pick, if it's single ended go single ended after the first stage.
Be sure to include over-voltage protection for the input amplifier. Piezoelectric hydrophones can generate large voltages when excited by vibrations or other mechanical disturbances. Static charge can also build up due to temperature changes. In sonar systems, it is good practice to place back to back diodes across the input for this purpose.
Note that amplifiers used in Naval sonar systems also have to be protected against the voltages generated by explosives such as depth charges.
Operational transconductance amplifiers are designed to operate without feedback and have a separate current input which controls their transconductance making them useful as a variable gain amplifier. The Linear Technology (now Analog Devices) LT1228, Burr-Brown (now Texas Instruments) OPA860/OPA861, and old slow National LM13700 are examples of these. They usually convert their current output into a voltage using a resistive load for flat bandwidth response but they can also drive a tuned load directly for a bandpass response.
The STM32 ADC can work up to 2.4 MSPS, while I'm going to use it at 192 kHz. Also it has configurable sample time, so i think it's going to be fine. Ar am I missing something which will prevent the ADC from working well in undersampling mode?
Now I think I will stick with transformer input design, to adjust its winding for noise if needed, as you suggested. Also, if I connect the input transformer between the + input of the CFB Opamp and the ground, does anything prevent me from also using the CFB opamp for the first stage?
Why use a digital potentiometer instead of a PGA (PGA112 for instance).
I intended to use the protection circuit below(1st attachment) at the amplifier input. It's an inherited variant used with other sonars here. I didn't add the 1k resistors to the simulation, but it turns out that a 1k resistor generates 4 nV/sqrt(Hz) of noise(and two of them give sqrt(2) times more), about the same amount as the op amp itself, assuming a 1:1 transformer. It kind of nullifies our efforts to make a low noise input stage.
What I don't quite understand is why is their amplifier's input impedance so low(220 Ohm) while we typically want higher input impedanse for such a sensitive device.
The diode method uses no resistors at all but is pretty esoteric and may produce too much noise.
PS. the input impedance of the starfish inputs is 1220 Ohm, not 220 Ohm. The 1K is simply in series with the 220 Ohm resistor as far as the signal is concerned. Don't let the coupling capacitors and diodes fool you.
It is possible to make low impedance switches and overdrive protection circuits using diodes. Perhaps easier in this case is to put a pair of depletion mode power MOSFETs back to back with a source resistor to set a current limit in place of those 1k resistors. Below the current limit, the resistance is just the source resistor and MOSFET channel resistances. Above the current limit, the channel resistances increases with voltage to product a constant current.
Does the piezoelectric hydrophone's connection really need to be differential or could one side be grounded? I have never messed with one.
In the example schematic there is another reason; the amplifier will be very unhappy with a capacitive source on its inverting inputs (two of them for the shunt feedback differential amplifier shown) without series resistance.