Lowest noise amps

It's all about power. Goose up some monolithic designs and you can do just as well noise-wise. But I've done as good as 1nV/rt-Hz in squeezed current situations.

...Jim Thompson

Check this out. I found 0.1pV/rt-hz gain of 500 transformer preamplifier. The datasheet shows 0.1pV/rt-hz for both transformer and buffer, it says you can bypass the buffer for a gain of only 100. I wonder what the input noise is without the buffer! You can probably get even better cores than that. I've heard that cobalt core materials have nearly zero Barkhausen noise.

Paul

Even better yet, supposedly this guy built a 65pV/rt-hz 5Hz-100kHz amplifier:

J. Lepaisant, M. Lam Chok Sing, D. Bloyet Low-noise preamplifier with input and feedback transformers for low source resistance sensors Rev. Sci Instrum. 63(3), March 1992, p2089

Is that just a low noise transformer?

Paul

This may be close to being true in high-Z circuits as long as there is no high sound pressure level, shock or vibration.

But remember that tubes are microphonic and any physically induced noise level in the audio range can be way higher than the thermal noise, even leading to oscillation if a preamp tube is close to a speaker.

The noise figure is 1 dB, which is not all that good. It is only great at 10 Hz to 10 kHz It also distorts and causes harmonics at low level signals, because of non-linear magnetic, and they do not list a spec for it either.

No - but good toobs are actually surprisingly quiet.

Discretes will outperform even the best op-amps. Best noise I've seen is

Graham

I've only just skimmed it, but it looks like a typical non-inverting buffer, where some of the output is coupled back to the input in antiphase, to give negative feedback.

Instead of setting the gain with resistors like in a normal op-amp would, with gain.

G = 1+R1/R2

it is done with transformers. So the gain must be set by the turns ratio of the transformers.

The actual circuit has FETs in front of the op-amp.

The Stanford data sheet someone pointed to earlier shows the noise figure to be 1dB at optimal frequencies. As an RF engineer, I think in noise figures and temperatures.

A noise figure of 1dB is equivalent to an internal noise generated by the preamp equal to a resistor at 75 Kelvin.

A decent UHF (500MHz) pre-amp can achieve noise figures of around 0.3dB, or 21K, so produce between a third and quarter of the noise of that Stanford device.

Those UHF premaps I talk of usually have less noise at lower frequencies, but they might be next to impossible to keep stable at audio frequencies. With gains to well over 30GHz, they tend to oscillate somewhere I suspect.

In article , Pooh Bear wrote: [...]

The Interfet will do about 0.25 if you run selected ones at about 10mA.

I've often wanted to see what the noise of a great big power MOSFET would be when it is running at a modestly high current.

Input impedance of 0.5 ohms won't be any use for audio ! Rubbish frequency response. This is designed for entirely different applications.

Remember you asked about *voltage* amplifiers. These usually have high input Z.

Graham

And RF practice is also entirely different to audio.

Graham

Funnily enough, the device I had in mind was from Interfet. Someone here posted some info about them so I went and got the online data. Looks very interesting.

As you say, you need to use some quite serious current to get those low noise figures. Typical audio mic preamps using discretes run the input devices at around 2-3 mA each to get better than 1dB noise figure re: 200 ohm source ( typical microphone impedance ).

Graham

I just purchased 100 Toshiba 2sk170, and measured samples at 1 and 3mA: under 1nV. I like that their input capacitance is lower than many others with similar or lower e_n. I purchased mine from MCM.

It's easy to devise circuits that eliminate any linearity effects from the JFET's gate capacitance changing with drain voltage.

-Snip-

And the 1/f Noise of those semiconductors goes skyrocking at "near DC" (In means of Audio Frequencies)

Jorgen

Depends on the device a bit. Some are better.

I've never known the parts that make it into pro-audio to be troublesome in practice.

Graham

In article , Pooh Bear wrote: [...]

Watch out in the capacitive micro-pre-amp. The gate capacitance of the very low noise JFETs is enough to effect the linearity.

At 2mA, I think I'd look at the LSK170 for my low noise FET.

Vacuum tubes have more noise because they're hotter. kT/q is twice what it is for solid state.

Norm Strong

This is not the distortion I was refering to. In a capacitive microphone, the sound moves one plate of a capacitor by the amount I'll call X. For small spacings between plates, the capacitance of the capacitor varies as

You end up with 1/(1/X) or simply a voltage that varies with X.

If the gate capacitance is high, you end up with 1/(Cgs + 1/X) and hence distortion.

Using a high-gain source-feedback circuit effectively bootstraps the JFET's source, eliminating most of its high Ciss gate-source capacitance, with an effectiveness approaching a follower circuit if enough loop gain is present. A cascode configuration reduces the Crss gate-drain capacitance, and bootstrapping the cascode off the source drives that already small capacitance toward zero. All that's left from the JFET's capacitance is high-frequency e_n-Cin noise, which is not canceled out by configuration or the feedback circuits (this means one musn't get carried away in choosing too large a JFET, just because it has a nice low voltage-noise rating).

The 2sk170 presents a nice compromise for many applications. For example, I'm using them in a sensitive capacitance position gauge.

The 2sk170 with its under-1nV spot voltage-noise level isn't best employed in a follower circuit, because this low noise level would force one to follow the JFET follower with an unusually-low-noise BJT amplifier, such as one using Rohm's 0.55nV 2sd786 transistors. That's not appealing, compared to a properly-designed common-source JFET amplifier that provides all the advantages without the pain.

Although in a 'follower' circuit, the effect of the gate-to-source capacitance is considerably reduced. (Gate-to-ground capacitance is not affected unless you bootstrap the surrounding metalwork)