The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> A conflict between noise and gain of my filter
https://designers-guide.org/forum/YaBB.pl?num=1187680737

Message started by hanm on Aug 21^st, 2007, 12:18am

Title: A conflict between noise and gain of my filter
Post by hanm on Aug 21^st, 2007, 12:18am

Hi,

I' m designing a complex filter, I found the voltage gain is 3dB. When I do the noise simulation, I integrated respective output noise and input reference noise over a frequency band and find the integrated input reference noise is bigger than the integrated ouput one.

I use a ideal voltage source as the input, and the ouput of the filter is floating.

I don't know what is the reason.

hanm

Title: Re: A conflict between noise and gain of my filter
Post by ACWWong on Aug 21^st, 2007, 4:12am

noise analysis calculates the output noise. It can also backcalculate the input referred noise, for which it uses the small signal gain at each swept frequency point as the transfer function. So output noise V² at a frequency is divided by the Av² at that frequency to get the input referred noise.
So I guess you must look carefully at your gain vs frequency plot over your noise band-of-interest, and check your calculations.

Title: Re: A conflict between noise and gain of my filter
Post by hanm on Aug 21^st, 2007, 6:28am

Thanks ACWWong,

I known the relations. That's why I speak the conflict!

One thing I forgot to point out. My filter is a complex one, which has 4 inputs( I+,I-,Q+ and Q-) and corresponding 4 outputs. While I setting the noise simulation eviroment. In the output cyclic field, I select voltage and select just one of 4 paths signal and let the minus one is gnd. In the input cyclic field, I do the same thing.

I wonder to know if the input reference noise is caculated to the only voltage I refered while the output noise is counted all of the 4 paths?

I want to use just one input port and split it into 4 paths signal(I/Q) and group the output four paths into one port to do noise analysis. But I don't known how to do it :-[

Title: Re: A conflict between noise and gain of my filter
Post by ACWWong on Aug 21^st, 2007, 6:48am

I have successfully designed several complex filters in the past, and haven't seen this conflict you speak of.
I use differential output I and differential input I for the noise and ac response (I also use the technique shown on this site http://www.designers-guide.org/Analysis/diff.pdf so you can look at common-mode effects as well).
Doing differential Q should yield the same gain noise response for the pass band.

Title: Re: A conflict between noise and gain of my filter
Post by hanm on Aug 21^st, 2007, 7:06am

ACWWong,

Please help me to check this ideal:

"If the input reference noise is caculated to the only voltage I refered while the output noise is counted all of the 4 paths? "

I will check my design again. It's hard to understand!

Thanks!

hanm

Title: Re: A conflict between noise and gain of my filter
Post by ACWWong on Aug 21^st, 2007, 8:27am

but you have two channels, I and Q.

So you have two input differential sources I and Q (for ac simulation you can have magnitude for each is say 1, but the phase are offset 90^o, then output on I or Q will show the complex filter magnitude response, i.e non symmetrical +ve/-ve frequency, but Q has again a 90^o offset from I)
So the noise on the I differential output can be referred to the I input. Likewise for Q, and the results will be the same.

Its best for circuit design to specify the filter figures of merit such as gain and noise for I or Q channels (which will be the same). I don't see why you would want to refer BOTH I and Q outputs to only a single I OR Q inputs ?

Title: Re: A conflict between noise and gain of my filter
Post by aaron_do on Aug 21^st, 2007, 8:28am

Maybe the output voltage due to that particular voltage source is not 3 dB higher than the source voltage. i.e. maybe 1/4 of the output voltage at that node is from that voltage source (the one for noise) and the other 3/4 is from the other three sources. So when you calculate the TF from noise analysis you get much less than 3 dB but when you do AC analysis you'll get 3 dB gain...just a possibility.

You could use a balun to get differential but i'm not sure if there's anything in the rfLib to get quadrature...

Aaron

Title: Re: A conflict between noise and gain of my filter
Post by ACWWong on Aug 21^st, 2007, 8:52am

aaron_do wrote on Aug 21^st, 2007, 8:28am:

You could use a balun to get differential but i'm not sure if there's anything in the rfLib to get quadrature...

you just enter the small signal stimulus phase to get quadrature as stated in my last post.

i still suggest the filter is treated as two channels I and Q, the IQ cross coupling is only there to linearly shift the response such that it becomes complex. I'm pretty sure the system design is treating the them as I and Q channels...