Prev: Simple Hack TO Get $1500 To Your PayPal Account.
Next: pow/log functions in fixed point
From: Eric Jacobsen on 20 Jul 2010 15:47
On Tue, 20 Jul 2010 12:13:31 -0500, Vladimir Vassilevsky
<nospam(a)nowhere.com> wrote:

>
>
>Eric Jacobsen wrote:
>
>> I'm still trying to wrap my head around "omnivorous" carrier recovery.
>> I'm not sure whether the methods I've used previously are carnivorous
>> or vegetarian or what.
>> As mentioned by others, there is no universal method. IF, and it's a
>> significant IF, you can assume that there are no signals in adjacent
>> spectrum you can just take the power difference in both halves of the
>> extended baseband channel. If one side has more power than the
>> other, and the spectrum is symmetric, then there's an offset.
>
>This is essentially a discriminator; the output is going to be something
>like a centroid of the signal + noise in the bandwidth.
>
>> That'll work with most modulation methods, but it's generally not
>> accurate enough for fine tuning. It won't work if there's an
>> adjacent channel within the ambiguity range. It won't work if the
>> spectrum isn't symmetric.
>
>1. Perform the FFT in the sliding window of one symbol.
>2. Interpolate the FFT to get fractional bins.
>3. Perform N-th power operation for each fractional bin.
>4. Perform a long FFT on the result.
>
>Pretty heavy, but it works for any single- or multi-carrier signal.
>
>Vladimir Vassilevsky
>DSP and Mixed Signal Design Consultant
>http://www.abvolt.com

It's much easier than that. The ratio of the output of two filters,
one on each side of the desired center frequency, provides the
discriminator.


Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com
From: Vladimir Vassilevsky on 20 Jul 2010 16:06


Eric Jacobsen wrote:

> On Tue, 20 Jul 2010 12:13:31 -0500, Vladimir Vassilevsky
> <nospam(a)nowhere.com> wrote:
>
>
>>
>>Eric Jacobsen wrote:
>>
>>
>>>I'm still trying to wrap my head around "omnivorous" carrier recovery.
>>>I'm not sure whether the methods I've used previously are carnivorous
>>>or vegetarian or what.
>>>As mentioned by others, there is no universal method. IF, and it's a
>>>significant IF, you can assume that there are no signals in adjacent
>>>spectrum you can just take the power difference in both halves of the
>>>extended baseband channel. If one side has more power than the
>>>other, and the spectrum is symmetric, then there's an offset.
>>
>>This is essentially a discriminator; the output is going to be something
>>like a centroid of the signal + noise in the bandwidth.
>>
>>
>>>That'll work with most modulation methods, but it's generally not
>>>accurate enough for fine tuning. It won't work if there's an
>>>adjacent channel within the ambiguity range. It won't work if the
>>>spectrum isn't symmetric.
>>
>>1. Perform the FFT in the sliding window of one symbol.
>>2. Interpolate the FFT to get fractional bins.
>>3. Perform N-th power operation for each fractional bin.
>>4. Perform a long FFT on the result.
>>
>>Pretty heavy, but it works for any single- or multi-carrier signal.
>
> It's much easier than that. The ratio of the output of two filters,
> one on each side of the desired center frequency, provides the
> discriminator.

Recently I had to come up with efficient blind algorithm to estimate
large carrier offset (several times of baud rate) in conditions of low
SNR. Discriminator is trashed once the SNR in the full band falls below
~10dB; hence, this approach won't get you very far.


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





From: Eric Jacobsen on 20 Jul 2010 18:02
On Tue, 20 Jul 2010 15:06:33 -0500, Vladimir Vassilevsky
<nospam(a)nowhere.com> wrote:

>
>
>Eric Jacobsen wrote:
>
>> On Tue, 20 Jul 2010 12:13:31 -0500, Vladimir Vassilevsky
>> <nospam(a)nowhere.com> wrote:
>>
>>
>>>
>>>Eric Jacobsen wrote:
>>>
>>>
>>>>I'm still trying to wrap my head around "omnivorous" carrier recovery.
>>>>I'm not sure whether the methods I've used previously are carnivorous
>>>>or vegetarian or what.
>>>>As mentioned by others, there is no universal method. IF, and it's a
>>>>significant IF, you can assume that there are no signals in adjacent
>>>>spectrum you can just take the power difference in both halves of the
>>>>extended baseband channel. If one side has more power than the
>>>>other, and the spectrum is symmetric, then there's an offset.
>>>
>>>This is essentially a discriminator; the output is going to be something
>>>like a centroid of the signal + noise in the bandwidth.
>>>
>>>
>>>>That'll work with most modulation methods, but it's generally not
>>>>accurate enough for fine tuning. It won't work if there's an
>>>>adjacent channel within the ambiguity range. It won't work if the
>>>>spectrum isn't symmetric.
>>>
>>>1. Perform the FFT in the sliding window of one symbol.
>>>2. Interpolate the FFT to get fractional bins.
>>>3. Perform N-th power operation for each fractional bin.
>>>4. Perform a long FFT on the result.
>>>
>>>Pretty heavy, but it works for any single- or multi-carrier signal.
>>
>> It's much easier than that. The ratio of the output of two filters,
>> one on each side of the desired center frequency, provides the
>> discriminator.
>
>Recently I had to come up with efficient blind algorithm to estimate
>large carrier offset (several times of baud rate) in conditions of low
>SNR. Discriminator is trashed once the SNR in the full band falls below
>~10dB; hence, this approach won't get you very far.
>
>
>Vladimir Vassilevsky
>DSP and Mixed Signal Design Consultant
>http://www.abvolt.com

That sort of performance will depend on the nature of the spectrum and
the discriminating filters. It's been used successfully quite often
for coarse acquisition, but fails in situations where there are
adjacent channels in use in a crowded spectrum.

So it's not all that universal or accurate, but is independent of
modulation type, assuming symmetric spectrum.


Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com
From: mtr on 21 Jul 2010 08:03
>
>
>Eric Jacobsen wrote:
>
>> I'm still trying to wrap my head around "omnivorous" carrier recovery.
>> I'm not sure whether the methods I've used previously are carnivorous
>> or vegetarian or what.
>> As mentioned by others, there is no universal method. IF, and it's a
>> significant IF, you can assume that there are no signals in adjacent
>> spectrum you can just take the power difference in both halves of the
>> extended baseband channel. If one side has more power than the
>> other, and the spectrum is symmetric, then there's an offset.
>
>This is essentially a discriminator; the output is going to be something
>like a centroid of the signal + noise in the bandwidth.
>
>> That'll work with most modulation methods, but it's generally not
>> accurate enough for fine tuning. It won't work if there's an
>> adjacent channel within the ambiguity range. It won't work if the
>> spectrum isn't symmetric.
>
>1. Perform the FFT in the sliding window of one symbol.
>2. Interpolate the FFT to get fractional bins.
>3. Perform N-th power operation for each fractional bin.
>4. Perform a long FFT on the result.
>
>Pretty heavy, but it works for any single- or multi-carrier signal.
>
>Vladimir Vassilevsky
>DSP and Mixed Signal Design Consultant
>http://www.abvolt.com
>
>
>

It's interesting, but how does it work in presence of ISI and raised root
cosine filter? The phase of base band signal through the symbol length
window would change significantly.
From: Vladimir Vassilevsky on 21 Jul 2010 10:12


mtr wrote:
>>
>>Eric Jacobsen wrote:
>>
>>
>>>I'm still trying to wrap my head around "omnivorous" carrier recovery.
>>>I'm not sure whether the methods I've used previously are carnivorous
>>>or vegetarian or what.
>>>As mentioned by others, there is no universal method. IF, and it's a
>>>significant IF, you can assume that there are no signals in adjacent
>>>spectrum you can just take the power difference in both halves of the
>>>extended baseband channel. If one side has more power than the
>>>other, and the spectrum is symmetric, then there's an offset.
>>
>>This is essentially a discriminator; the output is going to be something
>>like a centroid of the signal + noise in the bandwidth.
>>
>>
>>>That'll work with most modulation methods, but it's generally not
>>>accurate enough for fine tuning. It won't work if there's an
>>>adjacent channel within the ambiguity range. It won't work if the
>>>spectrum isn't symmetric.
>>
>>1. Perform the FFT in the sliding window of one symbol.
>>2. Interpolate the FFT to get fractional bins.
>>3. Perform N-th power operation for each fractional bin.
>>4. Perform a long FFT on the result.
>>
>>Pretty heavy, but it works for any single- or multi-carrier signal.
>
> It's interesting, but how does it work in presence of ISI and raised root
> cosine filter? The phase of base band signal through the symbol length
> window would change significantly.

It works very well, if you don't ask too much. If you need something
better, we can talk about that. This is going to be some other place
then here, and not for free.


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



First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4 5
Prev: Simple Hack TO Get $1500 To Your PayPal Account.
Next: pow/log functions in fixed point