AnsweredAssumed Answered

wrong noise figure in SpectraSys

Question asked by jphua on Sep 29, 2004
Latest reply on Nov 14, 2006 by rulonv

Like • Show 0 Likes0
Comment • 3

I have a RX chain whose cascaded noise figure increases, then decreases after a mixer (which is not possible?!) before increasing down the chain. The final cascaded noise figure reported is 1 dB lower than one calculated by spreadsheet. The help files mentioned that I check the cascaded gain, which I did, and it is correct. I also increased the no of noise points, with didn't help. Are there other factors that I should have taken care of? What is happening here?

No one else has this question

Outcomes

3 Replies

JERRY_MARKETOS Oct 1, 2004 2:01 AM
Mark Correct
Correct Answer
I recommend that you read this post:

https://edasupportweb.soco.agilent.com/pls/portal/support.show_topic?id=1028&highlight=noise+decreases

If that doesn't solve the question, feel free to send a copy of the workspace (along with brief statment of the question) to support@eagleware.com

When emailing support@eagleware.com, please first select
Genesys main menu HELP, ABOUT... and indicate what is the
serial number of your hardware key and what is the version
of Genesys you are running. Please also tell your operating system.
Like • Show 0 Likes0
Actions
hhoffm Nov 10, 2006 7:56 AM
Mark Correct
Correct Answer
I have been looking into this decrease in NF as well--it also occurs with ADS if one uses non-deal VSWR. I had assumed this effect was due to a noise power calculation similar to the one Jerry Marketos recommended.
While this makes sense for Noise Power, I don't see how it can be correct for Noise Figure.

The definition for Noise Figure (see Pozar, for example) is:
SNR(in)/SNR(out).
It is non-physical to improve SNR through a poor match, which is what the recommended equation predicts.
To be fair, the academic definition says that the match is assumed perfect, but it still does not make any sense that one may improve noise figure due to a reflection, since the Signal part of SNR is reduced along with the Noise.

Does anyone have a good paper on how to treat NF for mismatched systems properly?
Thanks,
Jim
Like • Show 0 Likes0
Actions
rulonv Nov 14, 2006 12:50 AM
Mark Correct
Correct Answer
As you know from the prior link specified by Jerry the cascaded noise figure calculation in SPECTRASYS is based on power not voltage. Generally for IC design extreme mismatches occur (i.e. 1000 ohms in and 10 ohms out) and calculating cascaded noise figure using a power based approach makes no sense. We are planning on adding this in the future.

As you know NF measurement test equipment measure noise voltage at a point that is swept across frequency much like a network analyzer. SPECTRASYS uses integrated power across the channel. A drop in the cascaded noise figure means that the noise power change and cascaded gain didn't change at the same rate due to VSWR. However, this is a result of this particular measurement. As you know any simulator can give non-physical results and the user must use their judgment to understand how the tool is being used and limitation of the measurement configuration.

When I have seen this occur in the simulator it is generally because the user specifies a wide bandwidth for the channel which has significant VSWR across the channel. In this case it is best to use a voltage based approach spoken of earlier. If this is the case please use a narrow bandwidth like 1 Hz for example.

Please understand that the power based approach used in SPECTRASYS is far superior to traditional cascaded noise figure equations which completely ignore VSWR, image noise, multiple paths, and compression effects.

Please see, 'Practical RF System Design', William F Egan, Chapter 3 for more information.
Like • Show 0 Likes0
Actions

wrong noise figure in SpectraSys

Outcomes

Related Content

Recommended Content