Stats

  • Locked Locked
  • Replies 5
  • Subscribers 124
  • Views 4005
  • Members are here 0
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

why is the input referred noise not proportional to the output noise?

MarcoL

Hi all,

     Recently I found the input referred noise is not equal to the output noise divided by the ac gain in specter noise analysis, a little bit confusing.

    Such a testbench is built. Generally it 's a common-source amplifier with an ideal resistor as the load of a transistor. A DC voltage source, of which the AC voltage is set to 1V is connected to the gate, and the output noise is obtained at the drain using the expression:

    getData("out" ?result"noise").

The input referred noise is obtained at the gate, using the expression:

   getData("in" ?result"noise").

Besides, the gain is obtained using :

  db(v("/Vout") ? result "ac") ), where Vout denotes the drain voltage.

  Thus I expected to get two curves illustrating the foregoing noises, of which the input referred noise should be equal to the output noise divided by the gain. It turned out, however, while the ratio of two noises seems fixed, the value deviates from the gain. For instance,

    vn,o= 255.19 nV/sqrt(Hz), vn,i = 49.32 nV/sqrt(Hz), the ratio = 5.18 or 14.29dB at f= 1M Hz, and this ratio is almost fixed before 3dB bandwidth.

   gain = 15.82 dB until f reaches beyond 500M Hz.

   So why is that? where does the 1.53dB difference come from? In my understanding, the noise analysis is based on the ac model, so the gain for ac analysis and noise analysis should not be different. Can anyone explain that?

Thanks a lot,

Marco

2016/11/18

  • Andrew Beckett

    Marco,

    noise analysis is actually based on xf analysis (not ac analysis), although the results should be similar. The gain should also be output by the noise analysis - worth looking at that - this is the gain from the specified input source to the output (rather than ac analysis which only gives the gain if it's an ideal input source and the source value is 1).

    If you can't get consistency on this, we'll need to see the circuit - it's not obvious to me from your description what might be wrong. 

    Regards,

    Andrew.

  • MarcoL

    Andrew,

    Thanks a lot for your kind reply.

    According to your last note, I checked the xf analysis results. Yes, it turned out the gain in the xf analysis is 14.3dB while ac gain is 15.82dB, as mentioned above. So the question now turns into: where does the difference come from? I supposed the transfer functions from the input voltage source to the output net in these two analysis should be the same in my test bench.Well, due to information security policies of the company, I have been blocked from uploading anything. Forgive me to describe the test bench with the net list as follows.

    Mosfet1 vout vin gnd gnd W=*** L=**

    Resistor1 vout gnd 10G

    Resistor2 vdd vout 30K

    Vsource1 vdd gnd 1.2V

    Vsource2 vin gnd ac=1V xf =1V

    The ac gain is obtained directly from the magnitude of net vout, which is also specified as the output node in xf analysis. Vsource2 is specified as the input source.

    So what's wrong?Anything related with my test bench?

    By the way, I am using IC6.1.6-64b.500.13.2 and spectre 15.1.0.345.isr2 .

    Thanks in advance,

    Marco

    2016/11/22

  • MarcoL
    Oh,forget to mention that the components are all noiseless except Mosfet1.
  • Andrew Beckett

    Hi Marco,

    I think you'll need to contact customer support so that we can take a look. I suspect there's a mistake in your setup somehow. I took what you described (and added a made up model with pretty much every parameter as default):

    // noise gain versus ac output voltage discrepancy test

    model nch bsim4 type=n

    Mosfet1 (vout vin 0 0) nch w=1u l=0.2u
    Resistor1 (vout 0) resistor r=10G isnoisy=no
    Resistor2 (vdd vout) resistor r=30K isnoisy=no
    Vsource1 (vdd 0) vsource dc=1.2
    Vsource2 (vin 0) vsource type=dc mag=1 //xfmag=1

    noise (vout 0) noise start=1 stop=10G iprobe=Vsource2
    ac ac start=1 stop=10G

    If I plot vout from the ac analysis and gain from the noise analysis, they are identical. There's no need to specify xfmag on the vsource; it is a scaling factor which defaults to 1 anyway.

    I just ran the netlist above with "spectre forum.scs" and then opened the forum.raw results in ViVA using the results browser. If I plot the two gains in db20 too, it still shows they are identical across the frequency range, as I'd expect.

    Regards,

    Andrew.

  • MarcoL
    Hi Andrew,

    Many thanks for your help. You have helped much more than I expected.
    Aye, now I suppose it's really something in my setup. Fortunately we have got some of your FAE guys who pay regular visits here, twice a week I heard. I think it's better to consult them rather than bother you sir, so I am gonna turning to these guys with a bunch of questions:)
    Anyway, Thanks again.

    Regards,
    Marco.
    2016/11/24

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.