sense several microvolt,who can help me .3q

[lghhust]

I want to sense a volt with only several microvolt difference,ordinary op sucks,who can give me some advice ,
thanks very much

 

[FvM]

several microvolt difference to what? A reference voltage existing in your circuit?

 

[jwfan]

If you can give the application of this design, it may help us a lot. several microvolt is too small, even larger than the device noise.

 

[FvM]

several microvolt is too small

Not generally, it's a function of the bandwidth (respectively measurement time) and possible long term stability requirements.

If the µV per V is intended as absolute measurement, a reference with below ppm/K drift would be required, achievable only with a thermostatic controlled benchtop calibrator or a similar high precision instrument.

 

[lghhust]

several microvolt difference to what? A reference voltage existing in your circuit?

It's a part of clock generator.However,the forward stage only provide a voltage difference about several microvolts,it's differential,thus no voltage reference is needed.
The block in front of me is how to find out a circuit which can solve this tough problem.

 

[FvM]

This is called amplifier in electronics, usually. In many applications, uV and sub-uV voltages are amplified. It may be also possible in your case. If you give some substantial information about it, EDAboard members can help.

 

[lghhust]

This is called amplifier in electronics, usually. In many applications, uV and sub-uV voltages are amplified. It may be also possible in your case. If you give some substantial information about it, EDAboard members can help.

yes,that's right.What I need to do is to amplify a uv signal to a logic voltage.The first stage of course is used to generate a uv difference(sorry for further detail ),then amplifier is needed to amplify this signal ,and followed by a shaping circuit,which turn a weak pulse to a clk.

 

[zopeon]

It could be done if ur input signal has freq in KHz range, then u could use a simple 2 stage OPam with CS as second stage. Use Higher W/L for input pmos and Have L of load abt 20 times L input.
U wull get input reffered noise of arnd less than 100nV level but if ur freq is low u will probably have to use BiCMOS , use BJT input as it has lower noise,
Read Low noise Opam from CMOS Analog Circuit Design from Allan Hollberg

 

[lghhust]

thanks.
I need to sense a difference of several uv,and turn it to "1" or "0", thus , I need an op with more than 100dB,to gain low offset and better noise performance , I have to chose bjt say npn or pnp as input pairs,and add additional stages.I use the form of simple 2 stages cmos op,but exchange it's input pairs with pnp,obtaining about less than 40dB GAIN,I notice that all transistors and bjts are working in proper region,could you give me some addvices.The current through pnp is about 40uA each input pair at it's emitter.

 

[Siniša]

You failed to say what frequency of clock you have. If it is not at fixed frequency what is the range of frequencies.

 

[FvM]

There have been pretty strong hints. I'm tired of the discussion.

 

[lghhust]

You failed to say what frequency of clock you have. If it is not at fixed frequency what is the range of frequencies.

it is about 50 meg;
I don't regard this important ,because my problem is to amplify a uv volt to about logic 1 or 0.Of course, this op is followed by some turning circuit.

 

[Siniša]

It is about 50 meg;
I don't regard this important

It makes huge difference, as FVM said:

...several microvolt is too small...
Not generally, it's a function of the bandwidth (respectively measurement time) and possible long term stability requirements.

Noise is function of bandwidth so if your signal is about50MHz, noise is going to be function of about

Since you have signal in µV range, you have to reduce bandwidth to absolute minimum to improve signal/noise ratio. Solution to your problem looks something about like that.

 

[FvM]

In return to that new jigsaw part, I can contribute some basical thoughts. Consider your amplifier has a noise density of only 1 nV/√Hz, that's equivalent to the noise generated by a 50 ohm resistor (a small value, but not impossible to achieve) and you're processing the input signal with 100 MHz bandwidth, you get 10 µV RMS respectively 50 µVpp noise. It would be 1/10 of this amount, if you're able to reduce the amplifier bandwith to 1 MHz, with implies a Q=50 bandpass, only applicable for a signal of fixed and known frequency.

 

[zopeon]

All right what was by FvM is what u have to do reduce the bandwidth and also u dont need to go in for a BJT based amp if ur operating freq is in Mhz to get low noise. Flicker noise need not be considered in ur case (Freq) u have to consider white noise.
Also if u cannot reduce ur BW u could try seeing if chopper can help u. but it wld mean using higher chopping fereq than ur current!