Low Noise Amplifiers, just what exactly is it?

[jubjub7]

I'm looking for a low noise amplifier to amplify some uV signals from 300kHz-1Mhz. It's for an NMR application. My questions are:

1. What exactly is inside a low noise amplifier?
2. Is it just a few transistors in CE, Cascode, etc topologies that have been "tweaked"?
3. Is it something that I could build on my own on a breadboard or dead-bugged?
4. If I found a low noise, precision op-amp that met the noise, frequency, and gain specs that I need, would that work just as well?

Thanks in advance!

[EEVblog]

3. Yes, you could, but you'd usually only resort to this if you can't find something off-the-shelf that meets your specs. Be prepared for much tweaking.
4. Yes, it should work.

A lot of the manufacturers have their own "secret sauce" that goes into the real high performance low noise products.
But some are a lot better than others and can be designed for different ends of the spectrum. For instance, low frequency 1/f noise that you wouldn't be concerned with.

Dave.

[gregariz]

1. A device with a low level of internally generated noise
2. Typically an amplifier is based around a good device as found in 1. In my experience its less about any 'topology' than it is about the device.

When selecting a device, you need to scour the datasheets. For an RF amp, typically transistors or integrated amplifiers will be spec'd with a noise figure table over the range of frequencies for which the device was designed to have a low noise. Low noise is achieved by mismatching the transistor. Device makers spend alot of time trying to get the level of mismatch vs noise figure as minimal as possible. These days devices are pretty good... in the old days you needed to cryo-cool the devices to lower the internally generated noise. For op-amp based low noise amps, the devices typically are spec'd for an equivalent internal noise voltage and internal noise current. You need to scour the datasheets to find the ones which are lowest. You can convert these voltages and currents to Noise figure as is given in the transistor sheets. You can use a Noise figure source/analyzer to then tweak the built circuit when you have it on the bench. Simulated design won't be enough to get you there IME. Sorry.. thats it in a nutshell but this is a complicated subject with alot of gotcha's in it. If you have never done it .. you are probably going to need some help. Good luck!

[ejeffrey]

The important thing with low noise amplifiers is not just the choice of transistors and topology, but that the noise contributed by your passive components like gain setting resistors can dominate your noise.  If you need a 50 ohm input impedance the johnson noise of the terminating resistor contributes 0.9 nanovolts / Hz^.5.  A 5 kohm feedback network would contribute 10 times that much.  So keep your resistors small, or use reactive circuit elements instead of resistors.

At < 1 MHz you can be well served by opamps.  These days there are some really nice fast + low noise opamps.  The LT1028 for instance has < 1 nanovolt/Hz.^2 and a GBW of 75.  Analog devices and TI have similar parts, and even faster devices exist.  If you need a differential output (say to drive a twisted pair) some of the TI fully differential amplifiers are also quite handy.

[qno]

There are two kinds of noise, current noise and voltage noise.
From your spec's I assume you refer to the voltage noise of the amp circuit.

Standard techniques are

Keep the input impedance as low as possible.
Keep the bandwidth of the amp as small as possible.

Temperature is also a factor. Keep your low noise circuits away from power dissipating devices.

In low noise opamps the bias current of the input stage is tuned for low noise.
In a low noise transistor there is usually a noise vs collector current graph so you can tune the bias current of the transistor.

[ejeffrey]

One more thing:  Low voltage noise usually means bipolar transistors with relatively large input bias currents and current input noise.  These parasitic currents will be injected into your pickup coil.  If you are doing low temperature NMR the bias current can cause unacceptable heating, and for low field/low noise NMR the noise current will cause a fluctuating field that can contribute to the noise in your measurement.

Capacitive coupling or transformer coupling can keep the bias current from flowing through your sensor.  With transformer coupling you can also pick the turns ratio to change the effective impedance of your source to match the optimal impedance for your amplifier -- in particular you might be able to use a FET amplifier with an essentially zero noise current but higher voltage noise than a corresponding BJT.

It is hard to say what you actually need without more details of your setup and resolution goals.  Can you post any more information?