Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Register Log in

1/f noise in diff amp

Status
Not open for further replies.
nandu

nandu

Member level 2
Joined
May 12, 2005
Messages
46
Helped
1
Reputation
2
Reaction score
0
Trophy points
1,286
Location
Timbuktu
Activity points
1,630
i was simulating the noise performance of a mos diffamp, and i found that the tail mos current source contributes significantly to the 1/f noise measured at the differential output. why is this happening? i always thought the tail mos doesnt contribute to overall noise as its noise current gets evenly distributed between the 2 branches and hence gets canclled off when we view the differential output.
 

Perhaps the way you make the noise analysis introduces some sort of imbalance to the circuit making the circuit not trully differential...

Can you send a schematic?
 

Hey firstly, noise is always additive. You cannot take two noise sources and tie them to a wire and get a differntial noise. Remember in both the correlated and uncorrelated noise cases, the vnoise,eff = sq.root(v1^2+V2^2 and another term). Hence you cannot get a difference of noise. Remember that noise is random and is not definate like any other voltage, etc.

1/f noise is also dependent on the DC offset. Hence try and increase the gm of your input transitors and see that 1/f noise to be reducing. Try reading the Razavi Noise chapter
 

nandu said:
i was simulating the noise performance of a mos diffamp, and i found that the tail mos current source contributes significantly to the 1/f noise measured at the differential output. why is this happening? i always thought the tail mos doesnt contribute to overall noise as its noise current gets evenly distributed between the 2 branches and hence gets canclled off when we view the differential output.
Click to expand...

Hi Nandu,
Your argument that the noise from the tail current source is common mode and hence should be rejected at the output is correct, but this rejection is based on the common mode rejection of the differentil pair, which I observe is bad because your tail current source output resistance is low - you are using short channel transistor.

Increase both the W and L of the tail current sources by a factor of 4, while keeping the W/L ratio constant. Two effects will be seen :

i) Flicker noise is inversely proportional to area, so you are reducing the amount of flicker noise contribution by the tail current source.
ii) You also improve the output resistance of the tail current transistor, so you should observe better common mode rejection.

Bharath
 

Thanx Bharath. I increased the dimensions of my tail mos and the noise dipped.

But I also recently came across a paper on ring oscillators wherein the author avoids using a tail mos in his delay cell for he says that would "increase the output noise via up-conversion". What does this mean? by-the-way the paper is by Yalcin A. Eken and John P. Uyemura, "THE DESIGN OF A 14GHZ I/Q RING OSCILLATOR IN 0.18μMCMOS", ISCAS 2004.
 

tsb_nph said:
Hi Nandu,
Your argument that the noise from the tail current source is common mode and hence should be rejected at the output is correct, but this rejection is based on the common mode rejection of the differentil pair, which I observe is bad because your tail current source output resistance is low - you are using short channel transistor.
Bharath
Click to expand...
I see no reason why differential output noise should be a function of current source impedance. Any current in the tail should split evenly between the two transistors in the diff pair (unless they are mismatched). The output noise should be all common mode, independent of the current source impedance.
 

10kangstroms said:
I see no reason why differential output noise should be a function of current source impedance. Any current in the tail should split evenly between the two transistors in the diff pair (unless they are mismatched). The output noise should be all common mode, independent of the current source impedance.
Click to expand...

The same effect is what user nandu has asked in the first message of the post and you are repeating it again. You are quoting a trivial first order effect that common mode noise should be cancelled if there is no mismatch and if Vin,diff = 0 (circuit is fully symmetric). If there is a slight differential input offset, then the reasoning does not hold good.
The question is , if first order effect cannot explain it, what is the second order effect that is causing the output in simulations? Cadence simulations do not take mismatch into account (only Monte Carlo sim take mismatch into account), so you can discount that effect - then, why is there a flicker noise contribution of the tail current source observed at the output? Assuming fully balanced input, the only other non-ideality is the finite output resistance of the current source that could be causing this problem. The reasoning here was that noise uses rms values - in that case signals which have +x and -x is same as + x and +x in terms of rms, maybe common mode gain comes picture, which is affected by the output resistance of the current source. Can you think of any other second order effect that could be causing this problem?
 

OK, I think I understand after looking at it for a while. Here's my simple-minded analysis (which may be wrong). :|
First, the lower the current source impedance, the more of the noise generated by the individual transistors in the differential pair will show up on the outputs. I was only looking at the noise generated by the current source, assuming the diff pair trannies were ideal.:cry: The current source noise will be split unevenly between the outputs due to the noise (looks like random offset voltage) from the diff pair trannies. It's not clear to me how current source impedance affects current source noise.
If you know this to be wrong, please correct me.
Tsb_nph, thanks for motivating me to take another look at that.
 

You can increase the width and length of the tail transistor
 

what is noise up-conversion and how is it relevant to this diff-amp circuit?
 

Noise up-conversion is a totally different effect. It takes place when the diff-pair is switched ( as in oscillators ). In this case the diff pair would act as a mixer, upconverting baseband flicker noise of the tail source into the carrier frequency.

In your circuit, this effect should not take place ( it's a small-signal circuit ). In addition, to see such an effect you need to run pss + pnoise analysis ( it won't be predicted by the simple ac-analysis )
 

excuse me !!!
how to simualte noise in spectre??
 

hello elbadry

i do actually intend to design a voltage controlled ring oscillator by cascading these diff amps. so could you explain to me in a little more detail how and why this 1/f noise gets upconverted.

Thanx in advance
 

what is your application?
 

im hoping to design a 3-stage volt controlled ring oscillator running at 3.125G (based on XAUI standard), using 0.18u tech. noise specs arent very tight but still i would like to keep it low. and i hope to optimize it for low power, of about 1mW or less.
 

nandu said:
i was simulating the noise performance of a mos diffamp, and i found that the tail mos current source contributes significantly to the 1/f noise measured at the differential output. why is this happening? i always thought the tail mos doesnt contribute to overall noise as its noise current gets evenly distributed between the 2 branches and hence gets canclled off when we view the differential output.
Click to expand...
may i ask you how can you use cadence to run analysis about 1/f noise
 

lily1981216 said:
nandu said:
i was simulating the noise performance of a mos diffamp, and i found that the tail mos current source contributes significantly to the 1/f noise measured at the differential output. why is this happening? i always thought the tail mos doesnt contribute to overall noise as its noise current gets evenly distributed between the 2 branches and hence gets canclled off when we view the differential output.
Click to expand...
may i ask you how can you use cadence to run analysis about 1/f noise
Click to expand...

First, check that you have the AF /KF parameters or equivalent.
Many old or simple models don't, and 1/f default parameters are 0.
 

nandu said:
what is noise up-conversion and how is it relevant to this diff-amp circuit?
Click to expand...


we usually use chopper stabilization technique to reduce the DC offset. 1/f noise also belong to the dc offset of the amplifier.
in chopper stabilization technique, the mixer is used to transfer the low frequency 1/f noise to the high frequency and filter it.
 

The output noise power spectral due to the tail current source would be (2*(1/4)*In^2*R^2 ). As long as you have noise current in the tail, it will be transferred to the output. When you check the noise contribution, you will find this component. To reduce this noise component, I think you can increase your tail current, use large size transistor (large length), and make sure this transistor in saturation always.

nandu said:
i was simulating the noise performance of a mos diffamp, and i found that the tail mos current source contributes significantly to the 1/f noise measured at the differential output. why is this happening? i always thought the tail mos doesnt contribute to overall noise as its noise current gets evenly distributed between the 2 branches and hence gets canclled off when we view the differential output.
Click to expand...
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top