Is noise analysis necessary for all analog IC designs?

[Thomasee]

Hello Everyone,

I am working on an analog IC chip whose noise issue is not important for the design specification at all, so I've not done much noise analysis till now.

Two week ago I attended an interview, after I've explained my project the interviewer kept on asking me about noise and noise sources for my chip although I already said the noise issue is not important for my design at all. It was so embarrassing for me. It also seemed to me like he was not satisfied with my answer at all.

Is thorough noise analysis a must for all the circuits in companies?
If not, for example like my project where at least I think the noise problem is not important, how can people give a convincing answer/explaination in such case?

 

[erikl]

Is thorough noise analysis a must for all the circuits in companies?

Not for all, but for quite a lot: Noise is important

• for all circuits, where the input voltage lies in the order of the input related noise (LNAs, opAmps for low signal sensors, high resolution converters, etc.)
• for autonomous circuits, because it determines phase noise, phase/frequency shift, and jitter

So for an analog design engineer interview I think it is important to know about noise, how to simulate or measure it, and know methods how to try and minimize it.

If not, for example like my project where at least I think the noise problem is not important, how can people give a convincing answer/explaination in such case?

By succeeding to explain and convince them, that

• your min. input signal by far exceeds the input related noise - in this case you should be able to explain how you calculated or simulated the latter value
• phase noise, phase/frequency shift, and jitter of your oscillator/VCO/PLL/DLL is far below the specs - but for this you probably must be able to explain how you calculated/measured this.

 

[Thomasee]

The unit of noise is measured in power
But my signal input is measured with voltage(mV) or current

How to compare them to say that the noise is much smaller than the voltage or current input?

Additionally, does offset have correlation to noise, if the offset of one circuit is large, then I can say that its noise is also large?

 

[dick_freebird]

Interviewers ask what they want to ask, looking for whatever.

In over 25 years of designing ICs, about 2/3 of that in analog,
I have never, ever done a noise analysis. Except for jitter,
which still was more about deterministic couplings than any
random effects and did not involve any traditional noise
analysis.

It's important to some people but in my experience I've always
had much more important things to worry about.

 

[strennor]

noise is quite critical, it depends on the application of your circuits, depends on what you are doing

power relates with voltage through current

offset can be reduced by larger transistor size, flicker noise is smaller for larger transistor size.


noise is not important for some cases, maybe includes your case. it is true if you have done correct things to make it smaller enough.

 

[Braski]

The unit of noise is measured in power
But my signal input is measured with voltage(mV) or current

How to compare them to say that the noise is much smaller than the voltage or current input?

Additionally, does offset have correlation to noise, if the offset of one circuit is large, then I can say that its noise is also large?

After the calculation of the noise input-referred power-spectral-density, you have to calculate the rms-voltage (or current) from your band specifications. Then you can compare it with your input voltage, to obtain the dynamic range of your circuit.

Offset is not really correlated to noise. Offset is (generally, but not only) due to the mismatch. However, it can be showed that we can minimize offset increasing the area occupation of the devices, just as for the flicker noise. Offset could be thought as a Dirac delta of noise at the zero-frequency (DC) (this way to see the thing is useful to understand offset cancellation techniques).

 

[oermens]

i think noise analysis isn't just important to quantify the noise in a circuit, but to gain insight into where the noise is coming from (contribution from the devices in the circuit). actually sitting down and calculating the noise in a circuit is usually a tedious exercise left for undergrad students.