Forums

Another Eb/No Question

Started by Necronomicon October 2, 2007
Does Eb/No, as used in digital communication
calculations for signal to noise ratios, take into
account the VCO/PLL/Reference phase noise of the carrier(s)?

I ask because the phase noise is never really flat, and the Noise
Power Spectral Density, or "No", is in Watts/Hz, which is assuming
that the noise is flat across all frequencies (white noise).

Certainly one could use the integrated phase noise, and just
add this to the integrated power of the noise floor, which would
be the (Watts/Hz) x (receiver noise band-width).

But it seems like Eb/No is only considering an assumed spectrally
flat, white-noise, noise floor.


Thanks for any clarity on this....

The problem is when you ask for 
>> as used in digital communication calculations
Of course, both need to be considered at some point, but not necessarily at the same time. IMO phase noise and Eb/No are quite "far from each other", and it may be not the most efficient approach to tackle them both at the same time. Of course, sometimes one needs to make simplifying assumptions. It is OK for some questions to treat phase-noise induced errors simply as AWGN, and add them to the noise budget. But keep in mind, phase noise enters the signal through multiplication. It is by nature neither additive nor white. Phase noise becomes really interesting, once there is a narrowband blocker on an adjacent frequency. -mn
Necronomicon wrote:
> Does Eb/No, as used in digital communication > calculations for signal to noise ratios, take into > account the VCO/PLL/Reference phase noise of the carrier(s)? > > I ask because the phase noise is never really flat, and the Noise > Power Spectral Density, or "No", is in Watts/Hz, which is assuming > that the noise is flat across all frequencies (white noise). > > Certainly one could use the integrated phase noise, and just > add this to the integrated power of the noise floor, which would > be the (Watts/Hz) x (receiver noise band-width). > > But it seems like Eb/No is only considering an assumed spectrally > flat, white-noise, noise floor. > > > Thanks for any clarity on this.... >
Unless you have a really crappy PLL the phase noise of your carrier won't directly affect the decoding of your signal. What phase noise _does_ do is to raise the noise floor in the presence of nearby carriers that wouldn't otherwise cause any problems. Ideally your carrier is at exactly one frequency, forever. When you mix your desired signal with your carrier the signal's frequency is shifted to your IF frequency. Other signals that you don't want to receive also have their frequencies shifted, but they aren't (barring other design problems) shifted to your IF frequency. If your PLL has noise, then the carrier energy is spread out. Normally this will take the form of a big spike at the intended frequency with 'tails' (either smooth, or with a series of spikes at your reference frequency). In the presence of an adjacent signal, these carrier tails are shifted over, and some of their energy may land in your IF passband. When this happens, the noise floor at your demodulator will rise and your system will respond to it just as it would to thermal or other system noise. So phase noise is more or less a 2nd-order phenomenon: it is only a problem if you are trying to receive a weak signal in the presence of adjacent strong signals, and the degree to which it is a problem depends on the spectrum of the phase noise, the strength of your intended signal, the strength of your adjacent signal, the frequency spacing of the adjacent signal, and the characteristics of the filters that you have in front of your mixer. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
On Oct 2, 3:38 pm, Tim Wescott <t...@seemywebsite.com> wrote:
> Necronomicon wrote: > > Does Eb/No, as used in digital communication > > calculations for signal to noise ratios, take into > > account the VCO/PLL/Reference phase noise of the carrier(s)? > > > I ask because the phase noise is never really flat, and the Noise > > Power Spectral Density, or "No", is in Watts/Hz, which is assuming > > that the noise is flat across all frequencies (white noise). > > > Certainly one could use the integrated phase noise, and just > > add this to the integrated power of the noise floor, which would > > be the (Watts/Hz) x (receiver noise band-width). > > > But it seems like Eb/No is only considering an assumed spectrally > > flat, white-noise, noise floor. > > > Thanks for any clarity on this.... > > Unless you have a really crappy PLL the phase noise of your carrier > won't directly affect the decoding of your signal. What phase noise > _does_ do is to raise the noise floor in the presence of nearby carriers > that wouldn't otherwise cause any problems. >
But what constitutes a crappy PLL is not the same for BPSK as 256QAM, in other words the modulation type matters. John
Tim Wescott wrote:
> Necronomicon wrote: > > Does Eb/No, as used in digital communication > > calculations for signal to noise ratios, take into > > account the VCO/PLL/Reference phase noise of the carrier(s)? > > > > I ask because the phase noise is never really flat, and the Noise > > Power Spectral Density, or "No", is in Watts/Hz, which is assuming > > that the noise is flat across all frequencies (white noise). > > > > Certainly one could use the integrated phase noise, and just > > add this to the integrated power of the noise floor, which would > > be the (Watts/Hz) x (receiver noise band-width). > > > > But it seems like Eb/No is only considering an assumed spectrally > > flat, white-noise, noise floor. > > > > > > Thanks for any clarity on this.... > > > Unless you have a really crappy PLL the phase noise of your carrier > won't directly affect the decoding of your signal.
This is incorrect. The PLL phase noise will directly affect the bit error rate of your demodulation, especially if it's a 16-QAM or 64-QAM signal. You'll be able to see the phase or amplitude shift directly off a displayed constellation. It's less critical if it's a QPSK or BPSK, but the phase noise will still affect your BER. And my original question is if the Eb/No takes the phase noise into account.
John wrote:
> On Oct 2, 3:38 pm, Tim Wescott <t...@seemywebsite.com> wrote: > > Necronomicon wrote: > > > Does Eb/No, as used in digital communication > > > calculations for signal to noise ratios, take into > > > account the VCO/PLL/Reference phase noise of the carrier(s)? > > > > > I ask because the phase noise is never really flat, and the Noise > > > Power Spectral Density, or "No", is in Watts/Hz, which is assuming > > > that the noise is flat across all frequencies (white noise). > > > > > Certainly one could use the integrated phase noise, and just > > > add this to the integrated power of the noise floor, which would > > > be the (Watts/Hz) x (receiver noise band-width). > > > > > But it seems like Eb/No is only considering an assumed spectrally > > > flat, white-noise, noise floor. > > > > > Thanks for any clarity on this.... > > > > Unless you have a really crappy PLL the phase noise of your carrier > > won't directly affect the decoding of your signal. What phase noise > > _does_ do is to raise the noise floor in the presence of nearby carriers > > that wouldn't otherwise cause any problems. > > > > But what constitutes a crappy PLL is not the same for BPSK as 256QAM, > in other words the modulation type matters. >
Correct! Thank you! Last i heard, they still haven't decided on the standard for WiMax, or 802.16, but they were considering 256-QAM. We'll need some serious linearity for that!
Necronomicon wrote:
> Tim Wescott wrote: >> Necronomicon wrote: >>> Does Eb/No, as used in digital communication >>> calculations for signal to noise ratios, take into >>> account the VCO/PLL/Reference phase noise of the carrier(s)? >>> >>> I ask because the phase noise is never really flat, and the Noise >>> Power Spectral Density, or "No", is in Watts/Hz, which is assuming >>> that the noise is flat across all frequencies (white noise). >>> >>> Certainly one could use the integrated phase noise, and just >>> add this to the integrated power of the noise floor, which would >>> be the (Watts/Hz) x (receiver noise band-width). >>> >>> But it seems like Eb/No is only considering an assumed spectrally >>> flat, white-noise, noise floor. >>> >>> >>> Thanks for any clarity on this.... >>> >> Unless you have a really crappy PLL the phase noise of your carrier >> won't directly affect the decoding of your signal. > > > This is incorrect. The PLL phase noise will directly > affect the bit error rate of your demodulation, especially if it's > a 16-QAM or 64-QAM signal. You'll be able to see the phase > or amplitude shift directly off a displayed constellation. > > It's less critical if it's a QPSK or BPSK, but the phase > noise > will still affect your BER. > > And my original question is if the Eb/No takes the > phase noise into account. >
I reserve the right to define "crappy PLL" as one that changes the phase enough, within one bit time, to mess up the demodulation. So, by definition, I'm right :). At any rate, I would expect that a textbook Eb/No curve doesn't take this into account, at least if it's a basic textbook. You certainly _could_ calculate the effect of the phase shift of the signal due to the PLL, and probably _should_ if you expect the phase noise to be significant. Whether an Eb/No curve that you find in a data book or a paper takes this into account is something that you'd have to verify on a case-by-case basis. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
>> We'll need some serious linearity for that!
Maybe it's only my imagination, but I think I recognize your handwriting. -mn
On Tue, 02 Oct 2007 11:21:28 -0700, Necronomicon <radio913@aol.com>
wrote:

>Does Eb/No, as used in digital communication >calculations for signal to noise ratios, take into >account the VCO/PLL/Reference phase noise of the carrier(s)? > >I ask because the phase noise is never really flat, and the Noise >Power Spectral Density, or "No", is in Watts/Hz, which is assuming >that the noise is flat across all frequencies (white noise). > >Certainly one could use the integrated phase noise, and just >add this to the integrated power of the noise floor, which would >be the (Watts/Hz) x (receiver noise band-width). > >But it seems like Eb/No is only considering an assumed spectrally >flat, white-noise, noise floor. > > >Thanks for any clarity on this....
The short answer is no. Eb/No is a power efficiency metric, used often to analyze the efficiencies of FEC and modulation in a theoretical or simulation sense. For those cases phase noise is not considered since it is implementation specific. So, in general Eb/No does not consider phase noise, nor does it consider imperfections in the filters or amplifiers. That doesn't mean that when you make a lab measurement to calibrate a system or measure performance, that phase noise doesn't affect the result. Clearly a measurement made on a practical system will include the effects of all impairments, including phase noise. Usually this gets lumped into an "implementation loss" term that will include phase noise, distortions due to filters and amplifiers, etc., etc. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org
Tim Wescott wrote:
> Necronomicon wrote: > > Tim Wescott wrote: > >> Necronomicon wrote: > >>> Does Eb/No, as used in digital communication > >>> calculations for signal to noise ratios, take into > >>> account the VCO/PLL/Reference phase noise of the carrier(s)? > >>> > >>> I ask because the phase noise is never really flat, and the Noise > >>> Power Spectral Density, or "No", is in Watts/Hz, which is assuming > >>> that the noise is flat across all frequencies (white noise). > >>> > >>> Certainly one could use the integrated phase noise, and just > >>> add this to the integrated power of the noise floor, which would > >>> be the (Watts/Hz) x (receiver noise band-width). > >>> > >>> But it seems like Eb/No is only considering an assumed spectrally > >>> flat, white-noise, noise floor. > >>> > >>> > >>> Thanks for any clarity on this.... > >>> > >> Unless you have a really crappy PLL the phase noise of your carrier > >> won't directly affect the decoding of your signal. > > > > > > This is incorrect. The PLL phase noise will directly > > affect the bit error rate of your demodulation, especially if it's > > a 16-QAM or 64-QAM signal. You'll be able to see the phase > > or amplitude shift directly off a displayed constellation. > > > > It's less critical if it's a QPSK or BPSK, but the phase > > noise > > will still affect your BER. > > > > And my original question is if the Eb/No takes the > > phase noise into account. > > > I reserve the right to define "crappy PLL" as one that changes the phase > enough, within one bit time, to mess up the demodulation. So, by > definition, I'm right :). >
When you say: "Unless you have a really crappy PLL the phase noise of your carrier won't directly affect the decoding of your signal"...... ......you make it sound like phase noise is not a VCO design issue when this is clearly not the case, as people are constantly looking for ways to increase the Q of their designs, to minimize phase noise, especially for 802.11a/g, with the higher OFDM data rates. Also, you really meant to say "within one symbol period", because you can have more than one bit per symbol.