upd: 17.ii'16, 9.iii'16
I figured out how to spell the Quenya names of the tengwar in the Quenya mode of the Tengwar script, and I made a tiny font to showcase my work.
font name: Nauca, stylized as (with the a-tehta transposed left as in caita)
x-height: 3 px
ascent: 2 px
descent: 2 px
letter-spacing: 1 px (strict)
word-spacing: minimum 2 px (3 px preferred)
Nauca was designed with Quenya in mind; design choices were made to accommodate the features of the Quenya language. Thus, Nauca may not be suitable for use in other Tengwar modes. Throughout this document, numbers to refer to the squares of the table. Individual letters are referred to by their names set in upright text between angle brackets (e.g. <tinco>). Bear in mind that Tolkien, in the quotations below, refers to the letters using numbers only. When referring to my rendering of a letter in Nauca, I use the term "glyph". Foreign words (here primarily words in Quenya) are written in italics. "<" appearing in squares 11 and 31 is to be read ", formerly", giving the past name of the letter. In squares 9 and 24 it is to be read "coming from", indicating etymology-dependent usage. The comma "," in 24 stands for "and".
From Appendix E in The Return of the King:
The twenty-four primary letters were each formed of a telco (stem) and a lúva (bow). The forms seen in 1–4 were regarded as normal. The telco could be raised, as in 9–16; or reduced, as in 17–24. The lúva could be open, as in Series I and III; or closed, as in II and IV; and in either case it could be doubled, as in 5–8 and 13–16.The system's uniformity, along with a love of small things, inspired me to make the font. Once I had decided on telco lengths and a shape for the lúvar, the primary letters came out quickly. Initially I considered a hard corner on the closed series (e.g. vs. (minque "eleven")), but this looked heavy, thick, and less suggestive.
Though the AppE table's doubled bows and modified stems had phonetic meanings lost to time, the modern allotment of sounds is not without order: m becomes mb on the lowering of a stem and mp on the raising: . Likewise, n becomes nd and nt.
The table is identical to the one in AppE except for square 34: <hwesta þindarinwa> has been replaced by <halla>. I did this because (i) <hwesta þindarinwa> is unsightly, (ii) <hwesta þindarinwa> is the only letter in the table not used in writing Quenya, and (iii) <halla> is used in writing Quenya, namely in hr and hl e.g. hríve "winter" and hlóce "snake" (incidentally, "snake" is also leuca). hr only ever occurs initially. The only instance of hl not initially is in ohlon "diphthong" and its derivatives. How is that spelled? My guess is .
I consider <halla> distinct from the long carrier <ára>. They differ in function, and also in typography where <halla> sits on the baseline while <ára> hangs from the meanline.
The table exemplifies the tengwar as most of them appear in the spelling of their own names. Those which do not are: <óre> (21), <anna> (23), those in the penultimate grade, <halla> (34), <yanta> (35), and <úre> (36). In fact, these latter three appear nowhere in the table, having very specific functions (<halla> for hl, hr; <yanta> in the diphthongs ai, ei in inimeite, oi, ui; <úre> in the diphthongs au, eu, possibly iu). The letter <óre>, standing for final or pre-consonantal r (except before d, where <arda> is used), is common, figuring in the names of <parma> (2), <umbar> (6), <formen> (10), <harma> (11, formerly), and <hyarmen> (33), and in the word "nuquerna" in 30 and 32. <anna> appears in the name of <yanta>.
For sa-rince ("following s" or "s-flourish") in ts, ps, and x = ks we have the glyphs . For the doubled consonants cc, ll, mm, nn, ññ (only in aññol "stench"), and tt, there is an underbar; ss is spelled with <esse>. Adding the underbar to a regular stem requires the stem to descend 1 px farther: compare t and tt. The underbar's length may vary.
For long vowels, I doubled only the tehtar for e, o, and u, because "two dots was more often used as a sign for following y" and a double circumflex/triple-dot would look ridiculous. These three long vowels are exhibited here: quén "someone", lóme "night", yúla "ember".
In the name of <númen> I chose to show u's length by doubling the u-tehta rather than putting it on a long carrier, because the symmetrical trigraph <númen><malta><númen> looks a lot prettier than . However, doing the same for <rómen> makes it look crowded: . As for <þúle>, I chose to deploy ára because all of the raised-stem, single-bow tengwar (i.e. those in the third grade) have a cramped attic. Moreover, I'm very fond of the s < th sound-change remaining reflected in the writing, so I wanted to exhibit <þúle> unadorned.
Turning <silme> in 29 to carry the i-tehta was my choice, following AppE: "30 and 32 [...] were mostly used as mere variants of 29 and 31, according to the convenience of writing, e.g. they were much used when accompanied by superimposed tehtar". Tolkien sometimes does combine upright <silme> with i and e, but there is no way to do this with my font. In squares 30 and 32 I used <silme> and <esse> on their own to represent them (as was common in Old English, though I have no evidence of this practise in Quenya), for if I'd spelled out the name of <silme nuquerna> in full it'd've been 46 px long, necessitating a 41% increase (48/34) in box size. Note that the word nuquerna, glossed with "reversed" in AppE, literally means "under-turned". The spelling nunquerna, which appears in this table and in a past version of this table, is incorrect.
According to AppE, if lb ever occurs in writing, it is written with <lambe><umbar> (as in tolbo "big toe"), "since lmb could not occur".
<nwalme> (20) is very rare, appearing in only a few Quenya words: the adjective nwalca "cruel", the nouns nwalma "pain" and nwalme "torment", and the verbs nwalya "to hurt" and nwar "to fret" (as in wear away). The apparent occurrence of nw only initially is explained here.
<hwesta> (12) isn't much more common, appearing only in the adjective hwarin "crooked", the nouns hwan "sponge, fungus", hwarma "crossbar", hwerme "gesture-code", hwesta "breeze", and hwinde "birch" and also "eddy", and the verbs hwesta "to puff" and hwinya "to swirl". Why does hw appear only initially?
<hyarmen> (33) is rather wide, but as it often sports a few tehtar (namely, vowel + following y) its width is better off odd: thus, 3 px or 5 px. I considered many alternatives
At first I had for <yanta> (35); now, it is thinner than <hyarmen>'s lower half, though they are the same size in the AppE table. Since <yanta> only ever appears with the tehtar for a, o, and u, it was imperative that it look good with them. Desiring also consistency with <óre> led me to choose the current width of 3 px.
When making a tiny font, my goal is to render the glyphs to be as small as possible while maintaining identifiability and resemblance. So, detailed little <esse> was a lot of trouble. It needed to (i) be short, and (ii) look good. It needed to be short to accommodate the vowel-carrying function of its rotated counterpart; to have <esse nuquerna> rise a few pixels above the meanline was unacceptable—ideally, it would hang from it. Originally I had but this was too tall. I compressed it to (the rather cramped) whose 1-pixel overshoot was more bearable. Upon inspecting [DTS 19] I noticed Tolkien's <esse nuquerna> looked like a cursive z. This was perfect: <esse> used to be called áre (and before that, áze) when it denoted r coming from z. So, I drew for <esse nuquerna> and was very pleased; and I broke symmetry by adding one dot of detail for <esse>: . The letters <esse> and <esse n.> are the only two which overshoot.
With a 10-pixel vertical dimension, one of the biggest glyph-combinations is <hyarmen> + following y + a-tehta, as in square 33. hye would also work; hyu does not occur in Quenya (not surprisingly, hyi doesn't either; then there's hyóla "trump(et)"). The tallest combination is <hyarmen> + u-tehta, as in huo "dog", standing at 9 px. The tengwar in the third tyelle do nicely accommodate the u-tehta e.g. as in Tarmasundar. The deepest glyph, also 10 px, is <lambe> + following y + any legal tehta (that is, a or e or o), as in 24, since <lambe>'s scoop is not large enough to contain the tixet. The trigraph ssy does not occur in Quenya; if it did, one might win with *ssya, 11 px.
The tehtar should be aligned with respect to the left-right center. Round an interpixel result to the glyph's "heavier" side. For the letters with doubled bows, this is the imaginary "×" between the bows, and for letters with single bows, this is the stem-side of the hole in the bow. For alignment purposes, the solitary pixel at the extreme edge of a bow may be disregarded, as can be <rómen>'s and <arda>'s leftmost pixels.
For the record, the word "tengwar" is
Björkman, M. "The Tengwar of Feanor" and "Tengwar - The Classical Mode".
Fauskanger, H. "Quettaparma quenyallo ar quenyanna", 2008.
Mellonath Daeron. "Index of Tengwar Specimina".
Strack, P. "Quenya Words".
Tolkien, J.R.R. "The Return of the King", 1955.