Low noise current source

On Aug 6, 2:59=A0pm, MooseFET

Is it possible to filter the noise thru a low-pass filter ? A cap in parallel to R3 and another between drain and GND ?

Looks just like the diode laser current driver I use. (Except the current is sourced to ground.) I get a few hundred mA of DC current with < 50nA (rms) of current noise. I used 50 ohms for R3 which, as you say, used up most of the voltage I had available. I never needed anything with less noise, but can you reduce the bandwidth somehow? Make R3 look like a large impdeance at high frequencies??? (Is that right?) I do put a biggish (10-100uH?) inductor in the current path.

George H.

What are the numbers?

John

Hello Moosefet,

Are you considering noise above the frequency where you have significant loop gain or not? When your frequency band of interest is above, you also should include the noise parameters of the mosfet.

When you did the entire noise math (I think you did), and current noise of R3 is still dominant, George's idea (increasing the impedance with a series inductor) may help (depending on frequency band of interest). I once used a parallel tuned circuit in an MF application to reduce the noise on a spot frequency (455 kHz).

Best regards,

Wim PA3DJS

MooseFET a écrit :

How much current, voltage compliance and noise do you target and in which BW?

And a tenfold improvement above what? U1 noise/R3 or your actual bench measurements?

In the first case, you've got a big pb.

In the second case, one pb with the config you show is that the mosfet noise voltage, which you know isn't excellent to say the least, isn't corrected above the R2-C1 corner frequency. Also there's no point in reducing R1 well bellow the mosfet noise, even less the opamp noise.

If DC accuracy allows you're better to go with a BJT. If the mosfet is for power/voltage compliance, then go for a BJT/mosfet cascode.

If you don't need the ref for any other purpose I used what follows in my 200pV/rtHz preamp, a few years back:

| | | ||-+ ||------| '-----|+/ |\\ |/ |

the zener being your floated filtered reference. Get the details at

You might consider L-R combination instead of R3; the response could be accounted for. Anyway it would be interesting to know the performance requirements and the values of the circuit.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Room temperature resistors are really a pain. You can get a factor of 2 by using a string of diode-connected transistors instead of a resistor--they have a noise temperature of T_j/2.

An inductor might work if your impedance levels are really low, but the really low noise thing in electronics is the beta of a BJT. If there were a good way to synthesize a floating resistor that had the noise of a transistor, that would be something.

Cheers

Phil Hobbs

I have a zero noise resistor in my PSpice repertoire... G-source ;-)

...Jim Thompson

I have a whole bunch--crappy op amp macromodels.

Cheers

Phil Hobbs

Reminds me... I want to hone my modeling capabilities. Suggest an OpAmp you'd like modeled.

...Jim Thompson

Analog Devices ADA4898 and ADA4817, please, sir. My two current faves for a lot of things.

Cheers

Phil Hobbs

I'll take a look. Do you have any information beyond the data sheets... undocumented quirks, etc.?

...Jim Thompson

Not so far, but I'll let you know. The 4898 is what the LT1028 should have been, except that it has only two gain stages.

Thanks

Phil Hobbs

As we both understand, two stages compensate delightfully, three stages are almost guaranteed squirrels ;-)

...Jim Thompson

Both of those suggestions make the noise worse. The best place for the capacitor in that area is across the load. A capacitor on R3 also could lead to instability.

The current is 0 to 500mA. The noise needs to be down by about 10^8 or 9 from there.

I have the option of killing the high frequencies with a capacitor on the load.

The op-amp and R3 are about even on who makes the most noise. The op- amp wins by a nose.

I care about frequencies below 10Hz so the inductor would have to be huge.

My band is basically DC to 10KHz.

How about both. My existing circuit seems to be limited by the combination of R3 and U1. From the numbers it looks like the fundamental limit is just about what I see on the bench.

I can kill the high end with a capacitor on the load.

I have a switched resistor divider on the reference. I also have the reference in one corner and the power devices way over in the other corner of the box. Without that the heating the reference was causing a variation.

=A0 =A0 =A0|

=A0 =A0 =A0|

=A0 =A0 =A0|

=A0 ||-+

=A0 || =A0 =A0 =A0 =A0 =A0 =A0 .-----+-------------------||-+

=A0 =A0|

=A0.-.

=A0| |

=A0| |

=A0 =A0'-'

=A0 =A0 |

=A0|

=A0 =A0|