Here's how I do it;

Each "cell" in the paradigm has a mandatory template or "transfix", an optional 
prefix, and an optional suffix.

Roots are mostly 3 consonants; some are 2 consonants and some are 4 
consonants. (Order counts, of course: b-c-d isn't the same as c-b-d nor as b-
d-c.)

Templates include; which (if any) of the root consonants gets doubled; and 
which vowels appear where. (That is, baced is different from bacid or biced or 
becad; babecid is different from bacecid is different from bacedid.  Though all 
of those are the b-c-d root.)

Prefixes and suffixes mostly have one consonant; some have two consonants.

There are 7 places and 6 manners of articulation.  

All of the consonants of any root are in the 5 most-common PoAs and the 
4 most-common MoAs.

The last (or only) consonant of any prefix, and the first (or only) consonant of 
any suffix, must come from one of the 3 least-common PoAs and/or one of the 
3 least-common MoAs.

The language has the following "constraints" (in an "optimality-theory" sense".)

In any word, the last consonant of the prefix can’t be the same as the first 
consonant of the root.

In any word, the last consonant of the root can’t be the same as the first 
consonant of the prefix.

There are high-priority constraints against any root, any prefix, or any suffix, 
containing two consecutive appearances of any consonant.

There are very-high-priority constraints against any word containing three 
consecutive appearances of any consonant.

There is a high-priority constraint against any word containing two 
consecutive consonants at the same place-of-articulation.

There is a low-priority constraint against any word containing two consecutive 
consonants in the same manner-of-articulation.

There is a high-priority constraint against using both a two-consonant prefix 
and a two-consonant suffix in the same word.

There is a medium-priority constraint against using both a prefix and a two-
consonant suffix in the same word.

There is a medium-priority constraint against using both a suffix and a two-
consonant prefix in the same word.

There is a  low-priority constraint against using both a prefix and a suffix in 
the same word.

There is a low-priority constraint against using a two-consonant affix.

There are low-priority constraints against the first consonant of the root 
matching either consonant of a two-consonant prefix; against the last 
consonant of a prefix matching either of the first two consonants of the root; 
against the last consonant of the root matching either of the consonants of a 
suffix; and against the first consonant of a suffix matching either of the last 
two consonants of the root.

(Remember the last consonant of a prefix and the first consonant of a root 
can't be the same; and the first consonant of a suffix and the last consonant 
of a root can't be the same.)

---------------------------------------------------------------------

There is one primary paradigm that is used for all roots that satisfy all of the 
following;
•	it’s exactly three consonants long
•	the first and second consonants are different from each other
•	the  last and next-to-last consonants are different from each other
•	the first consonant of the root is not the same as the last consonant 
of any existing prefix
•	the last consonant of the root is not the same as the first consonant 
of any existing prefix
•	none of the consonants is a semivowel, or a glottal, or a laryngeal, 
or a pharyngeal.

(Such roots are “regular”.)

There are several secondary paradigms that are used for roots that satisfy all 
but one of those rules, depending on which rule it violates, which positions the 
consonants that violate it are in, and which consonant is involved.  These 
might be called “regularly irregular” or “quasi-regular” roots, I guess.  Anyway, 
these roots have non-defective and non-suppletive paradigms.  They’re not 
uncommon.

If a root breaks exactly two of the above rules, it usually can have a non-
defective paradigm only if there is some suppletion (a phonologically different 
root is used to fill in the missing cells); and it usually can have a non-
suppletive paradigm only if there are some cells left empty (deficient), 
resulting in a defective paradigm.  They’re less common than the regular 
and “quasi-regular” roots, but there are several of them.

Few roots break three of the above rules.  Those that do usually have 
paradigms that are suppletive, and yet are still defective even after suppletion.


The reasons for the rules are to make it clear when the prefix ends and the 
root-and-template begins, and when the root-and-template ends and the 
suffix begins.

Also, if two consecutive positions in a root are occupied by the same 
consonant, it will be difficult to tell whether that has occurred because some 
template doubled one of the consonants; or, if one of them is doubled, 
whether it was the earlier one or the later one which was doubled.

Also, if a semivowel occurs in a word, it may be difficult to tell whether it’s a 
consonant of the root or a vowel of the template.

Also, if a glottal or laryngeal or pharyngeal consonant occurs intervocalically 
(between two vowels), it’s likely to be “lost”; either unheard by the 
addressee, or unarticulated by the speaker.


--------------------------------------------

I'm not sure if the above morphology is exactly "self-segregating".  I think it's 
pretty clear that it could easily be "self-segregating"; but I think with 
creativity and a little difficulty one could make it be _not_ "self-segregating".

Word-boundaries are often easy; especially between a word ending in a one-
consonant suffix and a word beginning in a one-consonant prefix.  Most 
occasions where two consecutive consonants have one of the less-common 
places or manners of articulation will be such word-boundaries.

Telling the difference between the root and the template is usually pretty easy.

Since there's at most one prefix, it's usually easy to tell where the prefix ends 
and the root-and-template begins.

Since there's at most one suffix, it's usually easy to tell where the root-and-
template ends and the suffix begins.

By properly managing and introducing the "quasi-regular" paradigms, one can 
create a large collection of roots, templates, prefixes, and suffixes, and still 
have a "self-segregating morphology".

I don't know if anyone else's self-segregating morphology has had any 
irregularity in it.  A perfectly-regular morphology that's also self-segregating 
probably has to have either fewish roots or fewish cells-per-paradigm.  That's 
my bet; I suppose I could be wrong.

Anyway, I think that difficulty (if indeed it actually exists, as I expect) can 
be "cured" by the gentle introduction of some "regular" irregularity.

------------------------------------

I hope that helps; does it?