The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> time-domain NF
https://designers-guide.org/forum/YaBB.pl?num=1168331076

Message started by aaron_do on Jan 9^th, 2007, 12:24am

Title: time-domain NF
Post by aaron_do on Jan 9^th, 2007, 12:24am

Hi all,

when i simulate the NF of my circuit, I get the answer versus frequency or at a certain frequency. For proper design however, I must consider the NF integrated over the entire spectrum. Is there an easy way to simulate NF or SNR in the time domain? I understand that since it is random fluctuations it probably needs to be averaged over a few cycles. Anyone have any insight?

thanks,
Aaron

Title: Re: time-domain NF
Post by Geoffrey_Coram on Jan 9^th, 2007, 6:31am

Why do you want to do it in the time domain? As opposed to simply integrating the noise as a function of frequency?

Few simulators support true time-domain noise. Several have the built-in ability to integrate noise vs frequency.

Title: Re: time-domain NF
Post by aaron_do on Jan 9^th, 2007, 5:31pm

thanks for the reply. Yeah I could also integrate noise over frequency. Should I simply plot the output noise and use the calculator in cadence to integrate?

Also, I guess the definition of NF would be a bit hazy since the bandwidth at the input is much larger than the output after all the filtering. I guess we should take the NF over the signal bandwidth at the input and over the passband at the output right?

thanks,
Aaron

Title: Re: time-domain NF
Post by ACWWong on Jan 10^th, 2007, 2:43am

I typically do what Geoffrey suggested, ie integrate the NF over the signal bandwidth. But there are a couple of points I would like to add:
1) make sure you integrate linearly in frequency, you might think it a nobrainer, but its all too simple in some tools to get it wrong be selecting a log freq NF plot and attempt to integrate.
2) Depending on your modulation scheme/burst lengths etc. noise at different parts of the signal bandwidth might be more or less important. An example would be how much flicker noise effects the SNR in a nominally 0Hz~3dBBW signal spectrum Direct sonversion receiver. The lower cutoff frequency in which to integrate the noise may be related to something like 1/burst time. For a modulation scheme like 2FSK with datarate=freq deviation, there is a null is signal at 0Hz, so even if you had loads of flicker noise it would really effect your likelihood for bit errors/impact SNR... Anaylsis of this kind of stuff is probably best done in matlab or similar.

Title: Re: time-domain NF
Post by aaron_do on Jan 10^th, 2007, 11:51pm

thanks for the reply.

I'll try and do as you suggested. Regarding your point about the different NF at different frequencies, I guess that's why a lot of people over-design :D

Aaron

Title: Re: time-domain NF
Post by ACWWong on Jan 11^th, 2007, 2:51am

ACWWong wrote on Jan 10^th, 2007, 2:43am:

..For a modulation scheme like 2FSK with datarate=freq deviation, there is a null is signal at 0Hz, so even if you had loads of flicker noise it would really effect your likelihood for bit errors..

Slight typo in my last post... This sentence should read:
For a modulation scheme like 2FSK with datarate=freq deviation, there is a null is signal at 0Hz, so even if you had loads of flicker noise it wouldn't really effect your likelihood for bit errors...

Title: Re: time-domain NF
Post by aaron_do on Jan 11^th, 2007, 9:46pm

Hi aw,

can you elaborate on your last point? I thought since sampling occurs in the time-domain, it is basically frequency independant. i.e. if you must consider all noise within the filter bandwidth.

In my design, the IF is 2 MHz and the signal bandwidth is 1 MHz. The channel spacing is 5 MHz, so I originally thought that I could simply use a low pass filter for channel selection. This is because none of the unused channels will mix down to a frequency below the IF. After rethinking it I decided that the low frequency noise must be rejected due to the time-domain sampling but what you are saying seems to refute this.

One more thing. Based on the required BER, we can calculate the pre-ADC SNRmin. However the DSP can provide brickwall filtering. Will this DSP filter remove out of band noise and therefore relax SNRmin, or does SNRmin strictly apply to the circuits preceeding the ADC?

thanks for your help,
Aaron

Title: Re: time-domain NF
Post by ACWWong on Jan 12^th, 2007, 3:52am

Hi aaron,
yes you consider all the noise within the filter bandwidth. My point is that the modulated signal spectrum (or noise or both) might NOT be uniform over the filter/signal bandwidth. For example in your 1MHz wide channel you would get a certain bit-error rate from your ADC if your noise was unifrom across the band and integrated to say a value x. A different noise profile (vs freq) across that 1MHz band could also integrate to x, but give rise to a different no. of bit errors (depending on signal psd, demodulation process etc.). This may or may not be relevant for you.
cheers
aw

Title: Re: time-domain NF
Post by aaron_do on Jan 12^th, 2007, 5:54pm

Hi aw,

sorry to be so much trouble. I think I understand what you're saying but I want to get it correct. Do you mean that if we use a particular demodulation process and there's no signal power from f1 to f2, it doesn't matter whether or not we filter the noise from f1 to f2? If this is right, do you know anything I can read to get more info?

thank you so much, I really want to make sure I understand this correctly,

Aaron

Title: Re: time-domain NF
Post by ACWWong on Jan 13^th, 2007, 4:23am

yes, if there is no signal power between f1 & f2 then in some demodulation circumstances noise between f1 & f2 might not impact bit errors.
I make this point based from my own experience with simulation work in matlab. Given a particular signal modulation (and hence psd) & demodulator process, different noise profile within the signal band caused different error rates (although the SNR in the signal band is constant). This might not be important in your system (my system was at zeroIF so flicker noise profile), but a system simulation would tell you.

Title: Re: time-domain NF
Post by aaron_do on Jan 14^th, 2007, 5:14pm

thanks,
Aaron