DSPRelated.com
Forums

Equalization without a training sequence

Started by HardySpicer May 16, 2010
I have seen many IEEE papers on what is now called Blind Equalization
where a PBRS sequence is not needed for training. This is for muliple-
channel (multivariable) problems. A colleague of mine says this is
just academic rubbish and not used in practice. true or false and if
true would there be any advantage?


Hardy
On Sun, 16 May 2010 13:20:32 -0700 (PDT), HardySpicer
<gyansorova@gmail.com> wrote:

>I have seen many IEEE papers on what is now called Blind Equalization >where a PBRS sequence is not needed for training. This is for muliple- >channel (multivariable) problems. A colleague of mine says this is >just academic rubbish and not used in practice. true or false and if >true would there be any advantage?
Blind equalization is an accepted and robust technique by now. Its use may be somewhat more difficult in packet based systems. Checkout 1000Base-T and 10GBase-T for multi-channel examples where channel is is continously modulated. -- Muzaffer Kal DSPIA INC. ASIC/FPGA Design Services http://www.dspia.com
On May 16, 4:20=A0pm, HardySpicer <gyansor...@gmail.com> wrote:
> I have seen many IEEE papers on what is now called Blind Equalization > where a PBRS sequence is not needed for training. This is for muliple- > channel (multivariable) problems. A colleague of mine says this is > just academic rubbish and not used in practice. true or false and if > true would there be any advantage? > > Hardy
Blind equalization is used in practice in some contexts. One example of an algorithm of practical use is the constant-modulus algorithm (CMA). It can be useful when your expected signal has constant envelope (like PSK), but you don't know any details about the channel. Jason
On 5/16/2010 1:20 PM, HardySpicer wrote:
> I have seen many IEEE papers on what is now called Blind Equalization > where a PBRS sequence is not needed for training. This is for muliple- > channel (multivariable) problems. A colleague of mine says this is > just academic rubbish and not used in practice. true or false and if > true would there be any advantage? > > > Hardy
In a broad sense many, many systems use "blind" equalization, if by that you mean that no preamble or formal training sequence is used. Single-carrier systems with continuous signals (e.g., trunking, satellite, backhaul, point-to-point, whatever) tend to do this. TV broadcast tends to be done this way, too, although in that case there may be some pilot and/or framing signals (although arguably not not formal training sequences) that can be used to assist equalization. -- Eric Jacobsen Minister of Algorithms Abineau Communications http://www.abineau.com
HardySpicer <gyansorova@gmail.com> writes:

> I have seen many IEEE papers on what is now called Blind Equalization > where a PBRS sequence is not needed for training. This is for muliple- > channel (multivariable) problems. A colleague of mine says this is > just academic rubbish and not used in practice. true or false and if > true would there be any advantage?
For old-style single-carrier systems like QPSK, QAM, etc., it is NOT rubbish at all. John Treichler was a "codiscoverer" of the constant-modulus algorithm and it was (is?) used extensively in the comm systems developed there at Applied Signal Technology, and I'm sure many other places as well. -- Randy Yates % "Rollin' and riding and slippin' and Digital Signal Labs % sliding, it's magic." mailto://yates@ieee.org % http://www.digitalsignallabs.com % 'Living' Thing', *A New World Record*, ELO
Randy Yates  <yates@ieee.org> wrote:

>HardySpicer <gyansorova@gmail.com> writes:
>> I have seen many IEEE papers on what is now called Blind Equalization >> where a PBRS sequence is not needed for training. This is for muliple- >> channel (multivariable) problems. A colleague of mine says this is >> just academic rubbish and not used in practice. true or false and if >> true would there be any advantage?
>For old-style single-carrier systems like QPSK, QAM, etc., it is NOT >rubbish at all. John Treichler was a "codiscoverer" of the >constant-modulus algorithm and it was (is?) used extensively in the comm >systems developed there at Applied Signal Technology, and I'm sure many >other places as well.
Of course, but I think the question had to do with multiple-channel systems. There are plenty of single-channel systems doing blind equalization -- V.x telephone modems, single-carrier modes of 802.11/11b ... almost any single-carrier single-channel system has a tracking equalizer although some rely on pilots and are therefore not "blind". The question, I think, if I project correctly how HardySpicer may be paraphrasing his friend, is if you have MIMO or similar do you give up on trying to track it. The 802.3 examples referenced upthread may be an example of cases where you do in fact do this, on a multichannel system. Steve
spope33@speedymail.org (Steve Pope) writes:

> Randy Yates <yates@ieee.org> wrote: > >>HardySpicer <gyansorova@gmail.com> writes: > >>> I have seen many IEEE papers on what is now called Blind Equalization >>> where a PBRS sequence is not needed for training. This is for muliple- >>> channel (multivariable) problems. A colleague of mine says this is >>> just academic rubbish and not used in practice. true or false and if >>> true would there be any advantage? > >>For old-style single-carrier systems like QPSK, QAM, etc., it is NOT >>rubbish at all. John Treichler was a "codiscoverer" of the >>constant-modulus algorithm and it was (is?) used extensively in the comm >>systems developed there at Applied Signal Technology, and I'm sure many >>other places as well. > > Of course, but I think the question had to do with multiple-channel > systems. There are plenty of single-channel systems doing blind > equalization -- V.x telephone modems, single-carrier modes of > 802.11/11b ... almost any single-carrier single-channel system > has a tracking equalizer although some rely on pilots and are > therefore not "blind". > > The question, I think, if I project correctly how HardySpicer may > be paraphrasing his friend, is if you have MIMO or similar do > you give up on trying to track it.
The problem being computational complexity? Otherwise couldn't you just do single-channel blind on each of the channels?
> The 802.3 examples referenced upthread may be an example of > cases where you do in fact do this, on a multichannel system.
What, give up? -- Randy Yates % "Watching all the days go by... Digital Signal Labs % Who are you and who am I?" mailto://yates@ieee.org % 'Mission (A World Record)', http://www.digitalsignallabs.com % *A New World Record*, ELO
On 5/16/2010 6:19 PM, Randy Yates wrote:
> spope33@speedymail.org (Steve Pope) writes: > >> Randy Yates<yates@ieee.org> wrote: >> >>> HardySpicer<gyansorova@gmail.com> writes: >> >>>> I have seen many IEEE papers on what is now called Blind Equalization >>>> where a PBRS sequence is not needed for training. This is for muliple- >>>> channel (multivariable) problems. A colleague of mine says this is >>>> just academic rubbish and not used in practice. true or false and if >>>> true would there be any advantage? >> >>> For old-style single-carrier systems like QPSK, QAM, etc., it is NOT >>> rubbish at all. John Treichler was a "codiscoverer" of the >>> constant-modulus algorithm and it was (is?) used extensively in the comm >>> systems developed there at Applied Signal Technology, and I'm sure many >>> other places as well. >> >> Of course, but I think the question had to do with multiple-channel >> systems. There are plenty of single-channel systems doing blind >> equalization -- V.x telephone modems, single-carrier modes of >> 802.11/11b ... almost any single-carrier single-channel system >> has a tracking equalizer although some rely on pilots and are >> therefore not "blind". >> >> The question, I think, if I project correctly how HardySpicer may >> be paraphrasing his friend, is if you have MIMO or similar do >> you give up on trying to track it. > > The problem being computational complexity? Otherwise couldn't > you just do single-channel blind on each of the channels? > >> The 802.3 examples referenced upthread may be an example of >> cases where you do in fact do this, on a multichannel system. > > What, give up?
I don't have citations handy, but there've been some papers on what is essentially blind equalization for multi-carrier, done as you suggest using slicer error on non-pilot subcarriers. One issue is complexity, since a multi-tap EQ on a subcarrier in a multi-carrier system generally defeats the purpose of a multi-carrier system. If a multi-carrier system is being used as expected such that each subcarrier is flat-faded then the trick is just to sort out the gain constant for each subcarrier. If there's no AM in the signal this can be done without too much trouble, e.g., if the subcarriers are modulated with something like BPSK or QPSK. It's not that hard to run an independent AGC on each subcarrier if the signal is constant modulus. If there's AM, though, like with QAM, it's tough to tell gain deviations due to the channel from AM modulation. So I don't think it would be impossible to build a blind multi-carrier system if the modulation were restricted to something like QPSK. I don't think it'd necessarily be easy, especially if the coherence time of the channel is short. -- Eric Jacobsen Minister of Algorithms Abineau Communications http://www.abineau.com
For example, a blind criterion can easily be used in the FMT multi-carrier
modulation for subcarrier equalization or multiple inputs channel
estimation.