Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency



Leo,

               I wanted to give you an update based on the discussion at the meeting.

               A number of people believe that since EPON and Ethernet use 100 ppm oscillators, we should be able to work with that in EPoC.  Avi thought the current design could be made to work with that oscillator accuracy, and there was interest in confirming that.

               I know that is less accurate than in other organizations but in 802.3 that is the value most often used, so if it can be made to work with 100 ppm oscillators in both the CLT and the CNU that would be best.

               My assumption is that the impact would only be on acquisition time, but acquisition would be possible.

Regards,
Steve

From: Leo Montreuil [mailto:leo.montreuil@xxxxxxxxxxxx]
Sent: Monday, March 17, 2014 6:47 AM
To: Shellhammer, Steve; STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: RE: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency

Steve,

To acquire the PLC (without a separate receiver) you have first to determine the Cyclic Prefix and align the FFT RX window with the incoming signal. In 802.11, there is a preamble (short in time) that occupies the whole channel. The 802.11 preamble is used to estimate the synchronization and mark the beginning of the burst.

In EPoC, we do not have such preamble to synchronize the receiver but we have continuous transmission. There are 8 continuous pilots surrounding the PLC. These pilots are BPSK, the pilot modulation (+1+0*j or -1 +0*j) is static but position dependent, i.e. the subcarrier number determine if it is a +1 or a -1.

Once you know the CP, you can search for this pilots pattern after the RX FFT. Unlike the PLC preamble that repeat every 128 symbols, the pilot pattern is always there. After you found the pilot pattern surrounding the PLC, you wait for the PLC preamble for frame synchronization

Thanks
Leo

From: Shellhammer, Steve [mailto:sshellha@xxxxxxxxxxxxxxxx]
Sent: Monday, March 17, 2014 4:38 AM
To: Leo Montreuil; STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: RE: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency


Leo,



    Thanks for your email.



    I appreciate that the DS is continuous and hence even with a frequency offset the CNU will be able to connect to the CLT through the PLC, so this is just an issue of how long it takes to make a connection.



    I am glad to hear the TG decided to select a single FFT size (4K) that makes the entire standard simpler and in this case, less sensitive to frequency offset.  See what happens when I miss a meeting, I miss some of these technical decisions.



    Your comment about using the continuous pilots to add in initial acquisition is an interesting idea.  I would be interested to hear more about how that can help speed up the process.



    See you tomorrow a the EPoC meeting.



Regards,

Steve



________________________________
From: Leo Montreuil [leo.montreuil@xxxxxxxxxxxx]
Sent: Friday, March 14, 2014 6:56 AM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: Re: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency
Steve,
The downstream is continuous (broadcast). In 802.11 the transmission in both direction are burst.

In cable plant, headend equipment usually requires better than 5-10 ppm accuracy. For subscriber equipment, it is better than 30-50 ppm. First acquisition can go along these steps (it is vendor specific):


1.       Find FFT size 4K or 8K and CP size by correlation. Note: For simplification, we removed some CP options and the 8K FFT in Indian Wells.

2.       The PLC is surrounded by a number of continuous pilots arranged in a specific pattern. This pattern can be used to find the PLC and get the internal time reference synchronized to the TX (within few Hz).

3.       Receive 1 or more PLC preamble and message to reset your internal clock to the network clock.

At this time there is no special preamble to accelerate this process. I do not think it is needed. It is relatively fast and this is done only the first time CNU join a network.

Thanks
Leo

From: Jim Farmer [mailto:jofarmer@xxxxxxxxxxxxxx]
Sent: Friday, March 14, 2014 8:41 AM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: Re: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency

Yeah, I was not advocating that we specify something that will work for 1588 - I pointed out that more than just tighter-tolerance oscillators would be needed.  I was just making the comment that 100 ppm oscillators will not do for that specialized application.  If we can make them work for routine applications, I have no problem with that.

Thanks,
jim

Please ignore all the changes in font in my message

  - my email client does that just to mess with my head.

Jim Farmer, K4BSE

Mobile 678-640-0860

jofarmer@xxxxxxxxxxxxxx<mailto:jofarmer@xxxxxxxxxxxxxx> (spam blocked)

Personal web site: http://home.mindspring.com/~jofarmer/index.html

Boss lady: http://www.kathysflute.com/

Youngest daughter: http://www.joyfarmerclary.com/Sites/Joy_Farmer-Clarys_Welcome.html



"A people that values its privileges above its principles soon loses both."

        Dwight D. Eisenhower, Inaugural Address, January 20, 1953
On 3/14/2014 2:03 AM, Duane Remein wrote:
So for EPON, we just adopted the Ethernet 100 ppm for both OLT and ONU. Of course it didn’t make a great deal of difference there either. While it might be nice to specify something better than that for the CLT 802.3 WG is pretty attached to the 100 ppm spec.
As Marek points out most equipment vendors do much better than than.
Best Regards,
Duane

FutureWei Technologies Inc.
duane.remein@xxxxxxxxxx<mailto:duane.remein@xxxxxxxxxx>
Director, Access R&D
919 418 4741
Raleigh, NC

From: Marek Hajduczenia [mailto:marek.hajduczenia@xxxxxxxxx]
Sent: Thursday, March 13, 2014 10:29 PM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: Re: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency

Jim,

Bear in mind that any implementer is free to choose components with *better* precision, tolerances etc. when manufacturing their products - it is called product differentiation. The standard specifies the minimum that has to be met. So if I were a product designer and wanted to develop a product (CNU) that supports IEEE1588v2, I would certainly use more precise oscillators. However, in cases of products that do not need to meet such tight requirements, 100ppm oscillators would do just fine.

Does that make sense? I know it is a mantra we repeat over and over again, but we are supposed to develop a standard and not a product spec.

Regards

Marek Hajduczenia, PhD
Network Architect, Principal Engineer
Bright House Networks
Office +1-813-295-5644
Cell +1-813-465-0669

From: Jim Farmer [mailto:jofarmer@xxxxxxxxxxxxxx]
Sent: March 13, 2014 9:41 PM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: Re: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency

100 ppm seems to be about where most inexpensive oscillators are.  I have not measured any, but my experience is that they will likely not be any better than that.  It is reasonable to specify something tighter, especially at the CLT.  For cable TV headend equipment, we used to specify 25 ppm and we beat that pretty easily.  But it did take a somewhat more "complex" crystal and an individual adjustment, both of which add that-of-which-we-are-not-to-speak.  And we didn't have to cover the outdoor temperature range.  If a computer gains or looses 1 minute in a week (between synchronization to a time server), then its oscillator is in error by 99 ppm.  And that is at room temperature.  If you have two independent oscillators, one on each end, you have to conservatively assume that they are off indifferent directions, which could double Steve's estimate.  But you can specify a much better oscillator, just that you have to exchange, uh, "complexity" for accuracy.  Probably a good thing to do at the CLT.

This presupposes that we will not have to support IEEE 1588 through the system - that requires very tight frequency tolerances.  Or we would have to re-specify for 1588 - will likely take different hardware anyway from what I know (which is not all that much).

jim

Please ignore all the changes in font in my message

  - my email client does that just to mess with my head.

Jim Farmer, K4BSE

Mobile 678-640-0860

jofarmer@xxxxxxxxxxxxxx<mailto:jofarmer@xxxxxxxxxxxxxx> (spam blocked)

Personal web site: http://home.mindspring.com/~jofarmer/index.html<http://home.mindspring.com/%7Ejofarmer/index.html>

Boss lady: http://www.kathysflute.com/

Youngest daughter: http://www.joyfarmerclary.com/Sites/Joy_Farmer-Clarys_Welcome.html



"A people that values its privileges above its principles soon loses both."

        Dwight D. Eisenhower, Inaugural Address, January 20, 1953
On 3/13/2014 6:48 PM, Shellhammer, Steve wrote:
Marek,

               In that case the frequency error would be around +/- 100 kHz (for 1 GHz RF) and so there would be even larger frequency uncertainty.  So it would be even more important for the PLC preamble to have some short training fields that can be used to disambiguate the correct subcarrier.

               Is 100 PPM what you would expect for the CNU?  I do not think we have specified the oscillator accuracy yet for the CLT or the CNU.  Maybe something we should figure out.  My preference is for a low-cost oscillator in the CNU and maybe a more accurate one in the CLT, where the cost may be less of an issue.

Regards,
Steve

From: Marek Hajduczenia [mailto:marek.hajduczenia@xxxxxxxxx]
Sent: Wednesday, March 12, 2014 6:00 PM
To: Shellhammer, Steve; STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: RE: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency

Steve,

Once you consider that we might have to work with 100ppm oscillators in CNUs, the resulting value gets 5 times larger. Is that a big problem ?

Regards

Marek Hajduczenia, PhD
Network Architect, Principal Engineer
Bright House Networks
Office +1-813-295-5644
Cell +1-813-465-0669

From: Shellhammer, Steve [mailto:sshellha@xxxxxxxxxxxxxxxx]
Sent: March 12, 2014 8:02 PM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: [STDS-802-3-EPOC] PLC Preamble and Initial Frequency Offset Frequency

EPoC Group,

               I was thinking about the PLC preamble and the initial frequency error due to use of a low-cost and low-accuracy oscillator in the CNU.  In 802.11 they have a way of dealing with initial frequency offset due to low-accuracy oscillators.  I was wondering if this make sense in EPoC.   I did a few calculations below.  I also attached the Word document since I was not sure what the email reflector would do to the equations.

               I would be interested in knowing if this approach used in 802.11 would be useful in EPoC.

               Any comments would be appreciated.

Thanks,
Steve
-----------

PLC Preamble Frequency Calculations

Assume low-cost crystal with 20 ppm oscillator in CNU.  Assume accuracy of oscillator in CLT is much better so we will ignore that oscillator error.

Assume the maximum carrier frequency around 1 GHz.  Could be a little higher but for these calculations this is good enough.

Frequency Error

Δf=±fc×20×10-6=±109×20×10-6=±20×< m:e>103=±20 kHz

Initial frequency error at CNU can be up to 20 kHz.

PHY supports subcarrier spacing of 25 kHz and 50 kHz.  The worst case situation from a frequency error perspective is the 25 kHz PHY.

For a 25 kHz PHY the initial frequency error is up to one subcarrier on each side.  So there are three possible tones that represent the middle tone when first acquiring the PLC.  This would triple the acquisition time, since the CNU would need to search over three times as many cases.

In the 802.11 OFDM PHY the preamble includes several short training fields (STFs) where only one out of every four tones is used, which disambiguates the subcarrier selection due to frequency offset.

Does this make sense for EPoC?



________________________________

________________________________

<="" p="">


________________________________

________________________________

<="" p="">

________________________________

<="" p="">


________________________________

<="" p="">

________________________________

________________________________________________________________________

To unsubscribe from the STDS-802-3-EPOC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC&A=1