The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> Noise figure Temperature
https://designers-guide.org/forum/YaBB.pl?num=1192525133

Message started by aaron_do on Oct 16^th, 2007, 1:58am

Title: Noise figure Temperature
Post by aaron_do on Oct 16^th, 2007, 1:58am

Hi all,

by definition, NF must be quoted at 290K however, most receivers are designed to operate over a broad temperature range. So what is the convention for defining NF at different temperatures? I'm asking because i was writing a paper and one of the reviewers pointed out that it is incorrect to define NF away from 290K...he didn't offer any alternative however...

thanks,
Aaron

Title: Re: Noise figure Temperature
Post by Frank Wiedmann on Oct 16^th, 2007, 2:20am

You might be able to use the difference between effective noise temperature and ambient temperature (see http://en.wikipedia.org/wiki/Noise_figure).

Title: Re: Noise figure Temperature
Post by aaron_do on Oct 16^th, 2007, 8:09am

Thanks for the link

that makes me question though, in spectre, if we simulate NF versus temperature, is the source temperature taken as 290K or the operating temperature when calculating NF?

thanks,
Aaron

Title: Re: Noise figure Temperature
Post by Frank Wiedmann on Oct 16^th, 2007, 8:27am

If you read the definition in the standard (mentioned in the link), you will find that the temperature of the input termination is always assumed to be standard noise temperature (usually 290 K) when noise figure is calculated.

Title: Re: Noise figure Temperature
Post by pancho_hideboo on Oct 16^th, 2007, 8:37am

See http://www.designers-guide.org/Forum/YaBB.pl?num=1190112138

Title: Re: Noise figure Temperature
Post by aaron_do on Oct 16^th, 2007, 5:50pm

Ok thanks,

does that seem like the wrong way to measure NF though? From my understanding the choice of 290K is arbitrary and is only used so people have a point of reference. i.e. the actual source noise depends on temperature.

Aaron

Title: Re: Noise figure Temperature
Post by pancho_hideboo on Oct 16^th, 2007, 10:57pm

Hi.

ENR table is defined as noise ratio of Tcold(290K) and Thot.
But actual temeperature(Tambient) is not 290[K].
So user have to compensate of temperature difference from 290[K].
Default value of Tambient is 296.5[K] not 290[K] in actual instruments such as
Agilent N8974A or N8975A Noise Figure Analyzer.
These NF analyzer have a temperature sensor built in a noise source, so Tcold can be auto corrected.

In legacy NF meter such as Agilent 8970B, there is no auto correction for Tcold.
Unless user enter temperature of Tambient, NF meter assume Tambient is 290[K].

Anyway actual NF meter with correct setting calculates NF as T0=290K regardless of Tambient.
If this is corresponded to simulation(e.g. Cadence Spectre), port temperature is 290[K],
analysis temperature is Tambient.

http://cp.literature.agilent.com/litweb/pdf/5952-3706E.pdf

Title: Re: Noise figure Temperature
Post by Frank Wiedmann on Oct 16^th, 2007, 11:42pm

aaron_do wrote on Oct 16^th, 2007, 5:50pm:

Yes, of course the actual source noise depends on the temperature. However, the official definition of noise figure is just what it is and it's not very useful to argue about it.

This is why I suggested that you use the effective noise temperature together with the ambient temperature to characterize your results. For example, you could define a quantity 1+T_e/T_phys (just don't call it noise figure, see http://en.wikipedia.org/wiki/Noise_figure for the definition of the quantities in the formula), which for an attentuator with attenuation L is always equal to L, independent of the ambient temperature T_phys.

Title: Re: Noise figure Temperature
Post by Ken Kundert on Oct 18^th, 2007, 9:48am

It is also important to realize that the temperature that is needed for noise figure is the temperature of the source not the temperature of the circuit. The temperature of the source is the temperature of the channel, and 290 is probably a reasonable approximation to that.

-Ken

Title: Re: Noise figure Temperature
Post by Frank Wiedmann on Oct 19^th, 2007, 1:10am

Thanks for emphasizing this, Ken. I had (perhaps incorrectly) assumed that the noise of the source is equal to kT_physB.

Title: Re: Noise figure Temperature
Post by pancho_hideboo on Oct 19^th, 2007, 6:43am

Hi.

"1+Te/Tphys, the ambient temperature Tphys" is wrong.

Again see my append. :(
The followings are summary.

From NF definition, NF is defined as NF=1+Te/T0,
where T0 is reference temperature(=port temperature)
and Te is equivalent noise temperature of DUT
(Te is dependent on analysis temperature(=Tambient)).
Standard T0 is 290K.

In sp analysis of ADS, ADS treats port temperature as 290[K] even if you set any value as port temperature.
While in Spectre, user can set any tempereture value in port but it's default(blank) value is 290K.
So in Spectre, port temperature should be 290K or blank.
In both ADS and Spectre, user can set any value as analysis temperature(Tambient).
http://www.designers-guide.org/Forum/YaBB.pl?num=1190112138

Actual NF meter with correct setting also calculates NF as T0=290K regardless of Tambient.
http://cp.literature.agilent.com/litweb/pdf/5952-3706E.pdf

[Conclusion]
The following three are all matched, that is, NF=1+Te/T0, T0=290[K] regardless of analysis temperature(Tambient).
(1) Actual NF meter with correct setting
(2) Agilent ADS NF
(3) Cadence Spectre NF with 290K or blank as port temperature

In simulation, analysis temperature means temperature of DUT.
In actual measurement, analysis temperature means ambient temperature which is almost same as temperature of DUT.

Title: Re: Noise figure Temperature
Post by aaron_do on Oct 20^th, 2007, 7:52am

Thanks for all the replies.

I read that 290K is used for satellites which are aimed directly at the earth such that the beams width is completely covered by the earth. When the beam covers space, then space temeperature is used which is as low as 3K. So it seems that if your channel temperature is not 290K, then the NF is not exactly the best way to measure noise performance...

cheers,
Aaron

Title: Re: Noise figure Temperature
Post by didac on Oct 20^th, 2007, 8:18am

Hi,
I think that all noise due to the channel it's just putted in the antenna temperature, To=290 it's an standard used since Friis paper I think, it's like talking about why a meter is a meter. NF is a convention so all the designs can be compared in equality.

Title: Re: Noise figure Temperature
Post by aaron_do on Oct 20^th, 2007, 9:15am

Hi,

I understand the logic of having a basis for comparison, but from a designer's point of view we still need to know what are the design requirements in terms of thermal noise...

Aaron

Title: Re: Noise figure Temperature
Post by didac on Oct 20^th, 2007, 10:54am

Hi,
The system requirements are dictated by the link budget given a modulation,necessary Eb/No for a given BER is obtained,the you can calculate the SNR at the input of the demodulator, from here you can obtain NF of the receiver. You must fullfill the requirement over the expected range of temperature, as pancho said:

Quote:

From NF definition, NF is defined as NF=1+Te/T0,
where T0 is reference temperature(=port temperature)
and Te is equivalent noise temperature of DUT
(Te is dependent on analysis temperature(=Tambient)).

,the DUT noise obviously will vary with temperature,but To it's standarized to 290.

Title: Re: Noise figure Temperature
Post by aaron_do on Oct 21^st, 2007, 6:58pm

Hi didac,

So to sum up, NF measures the degradation in SNR due to the receiver and is a standard for comparing between other designs. But if you're trying to find the degradation in SNR, NF can only be used accurately at a port T of 290K right?

thanks,
Aaron

Title: Re: Noise figure Temperature
Post by pancho_hideboo on Oct 22^nd, 2007, 1:06am

Hi, Aaron.

You can't apply straight Fris equation to cascaded systems when DUT's temperature isn't 290[K].
Straight Fris equation is valid when temperature of source and DUT are all 290[K].
Of course you can modify equation to deal with temperature difference.

See http://www.designers-guide.org/Forum/YaBB.pl?num=1190112138

Run the following netlist using Cadence Spectre.

(1) Tambient=-250[degC], T0=16.85[degC]
(2) Tambient=-250[degC], T0=-250[degC]

(1) gives NF=4.00182[dB]
(2) gives NF=13.0000[dB]

Do you think which results express DUT's thermal noise effect reasonably ?

=============================================================
simulator lang=spectre
global 0

aho options temp=-250 // Analysis Temperature in [degC], (Tambient)
parameters T0_degC=temp // Reference Temperature in [degC]

parameters att=13 // Attenuation in [dB]
parameters k=10**(att/20)
parameters z0=50 // Reference Impedance [ohm]
parameters Rp=z0*(k+1)/(k-1)
parameters Rs=0.5*z0*(k*k-1)/k

R1 (in 0) resistor r=Rp
R2 (out 0) resistor r=Rp
R3 (in out) resistor r=Rs

port1 (in 0) port r=z0 num=1 type=dc
//+ noisetemp=T0_degC

port2 (out 0) port r=z0 num=2 type=dc
//+ noisetemp=T0_degC

boke options reltol=1e-3 vabstol=1e-6 iabstol=1e-12 \
tnom=27 scalem=1.0 scale=1u gmin=1.0e-15 rforce=1 maxnotes=5 \
maxwarns=5 digits=5 cols=80 pivrel=1e-3 ckptclock=1800 \
sensfile="../psf/sens.output" rawfmt=psfascii

sp sp values=[100M] donoise=yes oprobe=port2 iprobe=port1
+ annotate=status

Title: Re: Noise figure Temperature
Post by didac on Oct 22^nd, 2007, 1:31am

Hi,
Aaron I think that you are confusing ambient temperature with equivalent noise temperature, To it's a reference temperature for measurement(290 was proposed because it was a round number in a time of slide calculators). The NF can measure the degradation of SNR at any temperature, the effect of ambient temperature reflects on the equivalent noise temperature of the DUT(Te).

Title: Re: Noise figure Temperature
Post by aaron_do on Oct 22^nd, 2007, 1:51am

Hi pancho_hideboo,

Sorry i'm not familiar with how to simulate from the netlist like that, but i know what you've done and i agree with your result. I was only trying to say that the NF should depend on the targeted application. For example, using NF = 1+Te/T0, we know that the output SNR of the receiver is,

Psig/(kT0B(1+Te/T0) = Psig/(k(Te+T0)B

so the final SNR depends on T0.

On the other hand i also agree with the use of a standard temperature (290K) since this makes it easy to calculate the noise added by the DUT without having to know the channel conditions. I think this is your point right?

cheers,
Aaron

Title: Re: Noise figure Temperature
Post by aaron_do on Oct 22^nd, 2007, 1:57am

Hi didac,

just read your post...now i am confused. Is the equation i just inserted wrong? How does NF measure degradation of SNR when T0 is not 290K? As pointed out by Ken, T0 is the temperature of the channel...

thanks,
Aaron

Title: Re: Noise figure Temperature
Post by pancho_hideboo on Oct 22^nd, 2007, 2:10am

Hi, Aaron.

Assume this netlist name is "pi_att.scs".
You can run this netlist like following.

% spectre -E +log pi_att.log pi_att.scs

You can see simulation results in "pi_att.raw" directory.
logFile sp.noise.sp sp.sp
These are ascii file.

Title: Re: Noise figure Temperature
Post by didac on Oct 22^nd, 2007, 2:23am

Hi,
Noise due to the channel is putted in the antenna temperature(like you said in a previous post satellite links pointing to the sky see low antenna temperature), channel noise doesn't have any effect on the NF of the DUT, ambient temperature translates in an increment of Te if you warm a resistor or a decrease of Te if you cool it with ice(to say something), To it's just a reference temperature.

Title: Re: Noise figure Temperature
Post by aaron_do on Oct 22^nd, 2007, 2:46am

Yeah it seems you're right...its just a reference temperature. Thanks for the info about running the netlist too,

Aaron