Noise figure measurement

gobruins <chunmei.kang@gmail.com> wrote:

>3. we can cascade two RF to increase the gain, but then the measurement
>become the noise figure for RF+RF+ADC+DDC, from this measurement how can I
>estimate the actual receiver noise figure?

Google on "cascaded noise figure" and the equation to
do this calculation will appear.

You usually want a measurement mode in which the front-end
noise shows up as a reasonable-level signal at the ADC.
This may require a gain stage that has no other purpose.

Steve

On Tue, 31 Mar 2009 11:52:50 -0500, gobruins wrote:
(top posting fixed)
>>gobruins wrote:
>>
>>> 1. when we talk about noise figure for a digital receiver, it is the
>>> RF+ADC, or the DDC will also contribute to the noise figure even
>>> though
> it
>>> is all digital? If DDC contributes, in what manner?
>>
>>When you talk about noise figure, you mean the SNR impairment due to
>>whatever reasons. Let's say your system needs the minimal signal of X to
> 
>>operate normally, whereas the theoretical limit is Y. The noise figure
>>is X/Y.
>>
>>
>>> 2. the noise floor at the DDC output is too low to see, is there any
> other
>>> experiment setup method that I can get around this?
>>
>>If you can't see the noise floor after the DDC, that means your
>>sensitivity is limited by DDC, not by RF part. Setup the DDC properly.
>>
>>> 3. we can cascade two RF to increase the gain, but then the
> measurement
>>> become the noise figure for RF+RF+ADC+DDC, from this measurement how
> can I
>>> estimate the actual receiver noise figure?
>>
>>The question does not make sense.
>>
>
> 'theoretical limit is Y', how can I compute this number?
> 
By finding the noise temperature of the signal environment and turning it 
into a noise power spectral density.  Your -174dBm/Hz figure is the noise 
power spectral density for a temperature of 300K -- what is the noise 
temperature of the region of space to which you are pointing your antenna?

BTW: "Noise Figure" is often measured w.r.t. a resistor at 300K; 
microwavers often point their antennas out to space which has a much 
lower noise temperature unless you're pointing at the Sun or (full) Moon 
or some such -- so for them a more meaningful measure is the noise 
temperature of the receiver.

-- 
http://www.wescottdesign.com

On Tue, 31 Mar 2009 10:42:04 -0500, gobruins wrote:

> we are trying to measure the noise figure for our receiver, but seems
> the receiver output noise power is too low to measure.
> 
> Here is the receiver configuration: RF+ADC+DDC(digital down converter)
> the total RF gain is 40dB(that is the maximum gain we can get), and the
> RF noise figure is about 5dB
> the ADC is 14 bit
> the DDC decimation rate is about 1600, and the dynamic range is 90dB.
> 
> Here is our problem
> the receiver input noise power is -174dBm/Hz after RF the noise power
> becomes -174dBm/Hz+40dB+5dB=-139dBm/Hz. the bandwidth is about 30MHz
> so after ADC we get around +-40 counts. but then since the DDC has a
> huge reduce to the bandwidth, after DDC the noise due to the analog
> device is too small to show on the spectrum.
> 
> Here is my question:
> 1. when we talk about noise figure for a digital receiver, it is the
> RF+ADC, or the DDC will also contribute to the noise figure even though
> it is all digital? If DDC contributes, in what manner?
> 
> 2. the noise floor at the DDC output is too low to see, is there any
> other experiment setup method that I can get around this?
> 
> 3. we can cascade two RF to increase the gain, but then the measurement
> become the noise figure for RF+RF+ADC+DDC, from this measurement how can
> I estimate the actual receiver noise figure?
> 
> Thanks a lot.

If after the DDC step the noise falls out, then your biggest contributer 
to "noise" is the DDC's quantization noise.  Why in heavens name did you 
give your DDC too-narrow data paths?

To measure the noise figure as-is you may consider injecting a sine wave 
into your system, at a frequency that falls into some easy-to-deal-with 
portion of your DDC's output spectrum and an amplitude that's high enough 
so that the DDC output amplitude cleanly follows changes in amplitude of 
the input.  Then do spectral analysis on the DDC output for intended 
signal power vs. noise power, or signal vs. signal+noise, or some such.  
The sinusoid will give you a signal that will "carry" the noise (most of 
it quantization noise in your DDC), which should give you a setup that 
you can measure with.

Note that you'll have a noise figure that's much worse than it should be: 
your noise figure in this case is going to be largely a consequence of 
quantization noise in the DDC, which should be the easiest part of the 
design to control.

-- 
http://www.wescottdesign.com

On Tue, 31 Mar 2009 14:50:49 -0500, Tim Wescott wrote:

> On Tue, 31 Mar 2009 10:42:04 -0500, gobruins wrote:
> 
>> we are trying to measure the noise figure for our receiver, but seems
>> the receiver output noise power is too low to measure.
>> 
>> Here is the receiver configuration: RF+ADC+DDC(digital down converter)
>> the total RF gain is 40dB(that is the maximum gain we can get), and the
>> RF noise figure is about 5dB
>> the ADC is 14 bit
>> the DDC decimation rate is about 1600, and the dynamic range is 90dB.
>> 
>> Here is our problem
>> the receiver input noise power is -174dBm/Hz after RF the noise power
>> becomes -174dBm/Hz+40dB+5dB=-139dBm/Hz. the bandwidth is about 30MHz so
>> after ADC we get around +-40 counts. but then since the DDC has a huge
>> reduce to the bandwidth, after DDC the noise due to the analog device
>> is too small to show on the spectrum.
>> 
>> Here is my question:
>> 1. when we talk about noise figure for a digital receiver, it is the
>> RF+ADC, or the DDC will also contribute to the noise figure even though
>> it is all digital? If DDC contributes, in what manner?
>> 
>> 2. the noise floor at the DDC output is too low to see, is there any
>> other experiment setup method that I can get around this?
>> 
>> 3. we can cascade two RF to increase the gain, but then the measurement
>> become the noise figure for RF+RF+ADC+DDC, from this measurement how
>> can I estimate the actual receiver noise figure?
>> 
>> Thanks a lot.
> 
> If after the DDC step the noise falls out, then your biggest contributer
> to "noise" is the DDC's quantization noise.  Why in heavens name did you
> give your DDC too-narrow data paths?
> 
> To measure the noise figure as-is you may consider injecting a sine wave
> into your system, at a frequency that falls into some easy-to-deal-with
> portion of your DDC's output spectrum and an amplitude that's high
> enough so that the DDC output amplitude cleanly follows changes in
> amplitude of the input.  Then do spectral analysis on the DDC output for
> intended signal power vs. noise power, or signal vs. signal+noise, or
> some such. The sinusoid will give you a signal that will "carry" the
> noise (most of it quantization noise in your DDC), which should give you
> a setup that you can measure with.
> 
> Note that you'll have a noise figure that's much worse than it should
> be: your noise figure in this case is going to be largely a consequence
> of quantization noise in the DDC, which should be the easiest part of
> the design to control.

Oh -- when you redesign your DDC for sufficient dynamic range, reduce 
your RF gain!  You can certainly receive perfectly adequately with a 
noise variance of 1LSB at the ADC input, and you can probably go two or 
three times below that, even if you're trying to receive signals that are 
significantly weaker than your broadband noise.

-- 
http://www.wescottdesign.com