[Sls-slp] [EXTERNAL] Re: Your LFSR OID frame RIDs affecting USLP, TM, AOS SDLPs

Tue Jun 1 13:16:11 UTC 2021

Hi Victor,

The resolution to adding the LFSR in TM, AOS, and USLP was to state the requirements in the book, but then add both representations (Galois and Fibonacci) into a new “non-normative” annex, since SLP WG blue books can’t provide normative info about coding in our SLP blue books. “King Solomon approach”

In the normative text in Chapter 4, the requirements are:

  1.  The TFDZ of an OID Transfer Frame shall be generated by use of a 32-cell Linear Feedback Shift Register (LFSR) with polynomial 1 + D + D2 + D22 +D32.
NOTE – See Annex H which contain diagrams describing the LFSR.

  1.  The LFSR shall be initialized at device start-up with an all-one seed and shall not be restarted. This requirement pertains exclusively to figure H-1 in Annex H.
NOTE – The first 10 bytes of the OID data pattern, in hexadecimal, are: FF FF FF FF 6D B6 D8 61 45 1F ....

So here are the diagrams that would go into a non-normative Annex in TM, AOS, and USLP:

[cid:image001.png at 01D756AD.9D8A5AA0]

From: SLS-SLP <sls-slp-bounces at mailman.ccsds.org> on behalf of "Sank, Victor J. (GSFC-567.0)[SCIENCE SYSTEMS AND APPLICATIONS INC] via SLS-SLP" <sls-slp at mailman.ccsds.org>
Reply-To: "Sank, Victor J. (GSFC-567.0)[SCIENCE SYSTEMS AND APPLICATIONS INC]" <victor.j.sank at nasa.gov>
Date: Monday, May 31, 2021 at 7:10 PM
To: "Andrews, Kenneth S (US 332B)" <kenneth.s.andrews at jpl.nasa.gov>
Cc: "Kazz, Greg J (US 312B)" <greg.j.kazz at jpl.nasa.gov>, "Kazz, Greg J (US 312B) via SLS-SLP" <sls-slp at mailman.ccsds.org>, "Fong, Wai H. (GSFC-5670)" <wai.h.fong at nasa.gov>, "Rodriguez, Shannon (GSFC-5670)" <shannon.rodriguez-1 at nasa.gov>, "Lee, Wing-tsz (GSFC-5670)" <wing-tsz.lee-1 at nasa.gov>
Subject: [EXTERNAL] Re: [Sls-slp] Your LFSR OID frame RIDs affecting USLP, TM, AOS SDLPs

Ken,
              Did you have any comment about the use of Fibonacci vs Galois for maximal pattern randomizers?  If so I think I missed it.  As I remember Greg asked you to explain the difference in the seeds for the two representations and said that both representations would be shown in his book.
My question is general CCSDS standards.  I would like to see a standard, MatLab, IEEE, or whatever.  For coding like BCH, that uses a polynomial that is not irreducible (only 4 terms and it is “reducible”, g(x) = x^7+x^6 + x^2 + 1 can be factored into (x + 1)(x^6 + x + 1)), I do not know if a Fibonacci equivalent can generate the parity.  Can it?   If it can, then for the following reason, I suggest that for any future CCSDS book, we use Fibonacci for all shift registers diagrams.
A maximal pattern register diagram can be either in the Fibonacci or Galois type representation, as you have submitted for the long randomizer.  I am recommending Fibonacci be used in CCSDS books at least where maximal and Gold codes are covered, since the output pattern is easily seen based on the register contents.   Since CCSDS does not recommend implementation, the user is free to use Galois for implementation if that is their preference.    It would be nice if we can help the understanding of engineers in the field that are not PhDs.  The fact that you were asked to explain why the seeds for the Fibonacci and Galois are different is telling.
              I do not agree with the comment made by Enrico? or was it GP?, that since the BCH or other code in the book used a Galois type format, that the maximal or Gold code representation used should therefore be in Galois format.
              I would love to hear your comments.  (and can the BCH be shown in a Fibonacci format?)
Thanks,
Victor

From: Sank, Victor J. (GSFC-567.0)[SCIENCE SYSTEMS AND APPLICATIONS INC]
Sent: Monday, May 17, 2021 1:16 PM
To: Andrews, Kenneth S (JPL-332B)[JPL Employee] <kenneth.s.andrews at jpl.nasa.gov>
Cc: Kazz, Greg J (US 312B) via SLS-SLP <sls-slp at mailman.ccsds.org>; Kazz, Greg J (US 312B) <greg.j.kazz at jpl.nasa.gov>
Subject: RE: Your LFSR OID frame RIDs affecting USLP, TM, AOS SDLPs

Ken,
              Maximal pattern shift registers.  LFSR
              CCSDS is not supposed to make recommendations about implementation.  For implementing I understand that the Galois shift register is sometimes preferred over the Fibonacci form.  However, for understanding and for setting an initial seed, Fibonacci has tremendous advantage because WYSIWYG, if you remember that phrase from the early Apple computers, what you see is what you get.
              You might also remember that we had a recent go-around in the code and synch WG about how to show these shift registers and we (GSFC) were recommending Fibonacci with particular labeling of the cells so that the polynomial is directly related to the cells that are tapped.  Yes, I realize that this can also be done for the Galois form.  Another value of the Fibonacci with the particular cell numbering is that it is what MatLab uses.  So when people are doing simulations, no conversion is needed.
              As you might remember there is an ambiguity between the associated polynomial and the shift register labeling.  A particular pattern can be generated in one direction or the reverse.  We had a particular proposal and it turned out that MatLab used the same convention.     Our proposal was discarded since it would be too much editing to get all the books to follow the same convention, and maybe other reasons.
              But in any case, seems better from an understanding point of view to use Fibonacci.

Victor

From: Kazz, Greg J (US 312B) <greg.j.kazz at jpl.nasa.gov<mailto:greg.j.kazz at jpl.nasa.gov>>
Sent: Monday, May 17, 2021 12:31 PM
To: Andrews, Kenneth S (JPL-332B)[JPL Employee] <kenneth.s.andrews at jpl.nasa.gov<mailto:kenneth.s.andrews at jpl.nasa.gov>>
Cc: Sank, Victor J. (GSFC-567.0)[SCIENCE SYSTEMS AND APPLICATIONS INC] <victor.j.sank at nasa.gov<mailto:victor.j.sank at nasa.gov>>; Kazz, Greg J (US 312B) via SLS-SLP <sls-slp at mailman.ccsds.org<mailto:sls-slp at mailman.ccsds.org>>
Subject: Your LFSR OID frame RIDs affecting USLP, TM, AOS SDLPs

Hi Ken,

The SLP WG decided to put both of your diagrams into a new non-normative annex in USLP, TM, and AOS.
The text in the normative part of the link layer books will state:

The TFDZ of an OID Transfer Frame shall be generated by use of a 32-cell Linear Feedback Shift Register (LFSR) with polynomial 1 + D + D2 + D22 +D32, see Annex X.

Where Annex X is TBD and will be added by me to the document.

We have some questions and requests for you:

  1.  Why does the initialization data differ from one figure to the other one?
  2.  The ‘1’ in the figures are misleading to some WG members, please replace ‘1’ with D sub zero i.e., D0 in the figures.
  3.  Consensus of WG was to specify the initialization vector

Thanks!
Greg

Greg Kazz
Principal Engineer
Technical Group Supervisor,
PSSE/EEISE/PPSE (312B)
Jet Propulsion Laboratory
4800 Oak Grove Dr., M/S 301-490
Pasadena, CA 91109
1+(818)393 6529(voice)
1+(818)393 6871(fax)
email: greg.j.kazz at jpl.nasa.gov<mailto:greg.j.kazz at jpl.nasa.gov>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ccsds.org/pipermail/sls-slp/attachments/20210601/8794c881/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 57926 bytes
Desc: image001.png
URL: <http://mailman.ccsds.org/pipermail/sls-slp/attachments/20210601/8794c881/attachment-0001.png>