Communications - out of band suppression?

I realize this is a bit of a fuzzy question that warrants a nebulous
answer, but I'm looking for orders of magnitude.

In most real  world communications systems, when filtering in on a
band of interest, what is the typical level of attenuation of
neighbouring bands? How much does this vary for different modulation
schemes?

The reason I ask is that I'm playing around with implementing a radio
on (eventually) an FPGA and I'm curious about the tradeoff between out
of band attenuation and filter complexity. I realize this will vary
greatly based on my desired outcome, modulation, etc. but what is the
ballpark that I should be looking at.

Right now I'm just looking at audio AM/FM, but I think I can
generalize this to various digital modulations.

20dB, 40dB, 60dB? For a neighbouring radio station on the AM band?
What about between channels in an 802.11 system?

Does anyone have any ballpark numbers they can share?

Chris

Chris Maryan wrote:

> I realize this is a bit of a fuzzy question that warrants a nebulous
> answer, but I'm looking for orders of magnitude.
> 
> In most real  world communications systems, when filtering in on a
> band of interest, what is the typical level of attenuation of
> neighbouring bands? How much does this vary for different modulation
> schemes?

For a receiver of the communication quality, the typical number would be 
at least 60dB of attenuation in the middle of the adjacent channel. For 
the consumer applications, somewhat 40dB suffice. You can also take a 
look at the performance charts of the monolite IF filters.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

Some numbers from WCDMA (European 3G cellular):
Noise floor in 3.84 MHz: ~ -108.1 dBm
Signal: around -106 dBm (absolute power level, including other users)
Interferer at 1 channel offset: -46 dBm
Interferer at 2 channel offset: -46 dBm
around 60 dB, allright.

All the current numbers are here:
http://www.3gpp.org/ftp/Specs/archive/25_series/25.101/25101-810.zip

mnentwig wrote:

> Some numbers from WCDMA (European 3G cellular):
> Noise floor in 3.84 MHz: ~ -108.1 dBm
> Signal: around -106 dBm (absolute power level, including other users)
> Interferer at 1 channel offset: -46 dBm
> Interferer at 2 channel offset: -46 dBm
> around 60 dB, allright.
> 
> All the current numbers are here:
> http://www.3gpp.org/ftp/Specs/archive/25_series/25.101/25101-810.zip

You seem to mix the spillage into adjacent channel in the transmitter 
and the adjacent channel attenuation in the receiver.

VLV

>
>You seem to mix the spillage into adjacent channel in the transmitter 
>and the adjacent channel attenuation in the receiver.

Well, I was referring to the two-tone intermodulation test from table 7.9
in the 25101 document.
BTW, the numbers are slightly different than what I wrote, please take
from the reference in case it matters. 
Also, it's not necessarily the most stringent test on the filtering.

Speaking about intermodulation: suppression of blockers is only part of
the story. They may (as in above test) still create spurious products that
fall into the wanted band... plus possible crossmodulation, phase noise
mixing etc. But nonetheless, a "ballpark number" for adjacent channel
rejection is a start.

>> BTW, the numbers are slightly different 
Actually the numbers are VERY different because the first tone in said
test appears at two channel offsets, not one. Possibly not the best
example for the filter, after all.

On Jan 10, 3:22 pm, "mnentwig" <mnent...@elisanet.fi> wrote:
> >> BTW, the numbers are slightly different
>
> Actually the numbers are VERY different because the first tone in said
> test appears at two channel offsets, not one. Possibly not the best
> example for the filter, after all.

Some applications require > 100 dB. I'm thinking of HF radio.

John

On Jan 10, 12:58&#4294967295;pm, Chris Maryan <kmar...@gmail.com> wrote:
> I realize this is a bit of a fuzzy question that warrants a nebulous
> answer, but I'm looking for orders of magnitude.
>
> In most real &#4294967295;world communications systems, when filtering in on a
> band of interest, what is the typical level of attenuation of
> neighbouring bands? How much does this vary for different modulation
> schemes?
>
> The reason I ask is that I'm playing around with implementing a radio
> on (eventually) an FPGA and I'm curious about the tradeoff between out
> of band attenuation and filter complexity. I realize this will vary
> greatly based on my desired outcome, modulation, etc. but what is the
> ballpark that I should be looking at.
>
> Right now I'm just looking at audio AM/FM, but I think I can
> generalize this to various digital modulations.
>
> 20dB, 40dB, 60dB? For a neighbouring radio station on the AM band?
> What about between channels in an 802.11 system?
>
> Does anyone have any ballpark numbers they can share?
>
> Chris


As Vlad gave, a common value is 60dBc for the OOBE in the adjacent
band, but look at DTV, now we are talking about 130dBc!!!!! by the
time you are 6MHz away from the DTV band edge.  I know many here will
not believe this value, but yes it is true. Imagine putting a cellular
uplink band adjacent to a DTV band (This may happen in the newly
allocated 700MHz band) where the DTV's transmitter's OOBE will swamp
the cell site receiver.

Clay

On Jan 10, 10:58 pm, Chris Maryan <kmar...@gmail.com> wrote:
> In most real  world communications systems, when filtering in on a
> band of interest, what is the typical level of attenuation of
> neighbouring bands? How much does this vary for different modulation
> schemes?
> What about between channels in an 802.11 system?
>

For 802.11, the numbers specified depends on the modulation used. In
(a) DSSS/CCK it is around -50dBr
(b) OFDM it is around -40dBr

Krishna
http://www.dsplog.com

On Thu, 10 Jan 2008 14:58:12 -0800 (PST), clay@claysturner.com wrote:

>On Jan 10, 12:58&#4294967295;pm, Chris Maryan <kmar...@gmail.com> wrote:
>> I realize this is a bit of a fuzzy question that warrants a nebulous
>> answer, but I'm looking for orders of magnitude.
>>
>> In most real &#4294967295;world communications systems, when filtering in on a
>> band of interest, what is the typical level of attenuation of
>> neighbouring bands? How much does this vary for different modulation
>> schemes?
>>
>> The reason I ask is that I'm playing around with implementing a radio
>> on (eventually) an FPGA and I'm curious about the tradeoff between out
>> of band attenuation and filter complexity. I realize this will vary
>> greatly based on my desired outcome, modulation, etc. but what is the
>> ballpark that I should be looking at.
>>
>> Right now I'm just looking at audio AM/FM, but I think I can
>> generalize this to various digital modulations.
>>
>> 20dB, 40dB, 60dB? For a neighbouring radio station on the AM band?
>> What about between channels in an 802.11 system?
>>
>> Does anyone have any ballpark numbers they can share?
>>
>> Chris
>
>
>As Vlad gave, a common value is 60dBc for the OOBE in the adjacent
>band, but look at DTV, now we are talking about 130dBc!!!!! by the
>time you are 6MHz away from the DTV band edge.  I know many here will
>not believe this value, but yes it is true. Imagine putting a cellular
>uplink band adjacent to a DTV band (This may happen in the newly
>allocated 700MHz band) where the DTV's transmitter's OOBE will swamp
>the cell site receiver.
>
>Clay

Exactly, i.e., it depends on the application.   For TV-like systems
where the transmit powers are extremely high the near-far problem can
be pretty bad (well, extremely bad) and filtering needs to be able to
handle it.

But in general I wouldn't disagree with Vlad's answer in that 60dBc to
40dBc is pretty typical, also depending on how you interpret the
problem.   It is also possible to work from a known spec or
performance requirement and sort out how good it needs to be from
there.   e.g., if you want a certain BER, sort out how many bits you
need to get that, work backwards to the front end to assure that level
of SINR all the way back to the input.   Often the hardest part can be
done in the analog front end with ceramic or SAW IF filters or such,
which then eases the pain in the digital processing.





Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org