Finding the Consonants of Khuzdul
To determine Khuzdul's consonant inventory, I started by taking the existing corpus of Khuzdul words and listing all of the consonants that could be identified in isolation. This was a fairly straightforward process, and yielded the following chart (listed in IPA for comparison to other languages).
Attested Khuzdul Consonants in IPA:
Next, I decided to look at the consonants found in the related languages: Adunaic, Biblical Hebrew, Modern Hebrew, and Classical/Quranic Arabic. The comparison to Adunaic and Hebrew would probably be more important. Adunaic is the closest language to Khuzdul found in Middle-earth, and Tolkien stated that Khuzdul resembled Hebrew in its phonology. The charts shown are from The Treason of Isengard (for Adunaic) and Wikipedia (for Hebrew and Arabic). For Hebrew and Arabic, I also referenced the books I have, and for the sake of a rough comparison, the charts from Wikipedia are close enough.
Adunaic Consonants in IPA:
Biblical Hebrew Consonants in IPA:
Modern Hebrew Consonants in IPA:
Classical/Quranic Arabic Consonants in IPA:
After doing this side by side comparison, some interesting things pop up:
1) The existing examples of Khuzdul end up with the same set of labials as Arabic. Often, LotR fans wonder if or assume that Khuzdul will have / p /, but this shows that it is equally likely that it does not.
2) The two aspirated consonants that Tolkien mentions in the books rergarding Khuzdul are / th kh /. There is no mention of / ph /. It's noteworthy that these parallel the emphatics of both Hebrew & Arabic, which are the velarized or pharyngealized consonants / tʕ kʕ / or / tʕ qʕ /.
Generally, though, you can tell from the comparison that Khuzdul is within the ballpark of Adunaic and Semitic languages.
Adunaic, Hebrew, and Arabic all had more consonants than what is evidenced in Khuzdul. The question was where to look and see if Khuzdul might have more. I decided that the logical place to look was at their writing system, the Angerthas (runes).
I went to the table of the Angerthas runes in Appendix E of LotR to find potential sounds and select which of those may or may not be found in Khuzdul. . I figured that since the Dwarves wrote with those letters, if they didn't have a rune for a given sound, it probably didn't exist in Khuzdul. Sorting through all of the information surrounding the Angerthas was a bit tricky though, since it was used differently by the different races and for several different languages. However, I combed through that information, which is mostly from LotR Appendix E, Treason of Isengard, and Sauron Defeated. By comparing the sounds listed in the Angerthas with the consonant inventories of Adunaic, Hebrew, and Arabic, I was able to arrive at a list for Khuzdul that seems to make a lot of sense.
For reference, here's the consonant inventory I decided on for Quasi-Khuzdul, which also found on the Phonology - Consonants page.
Consonant Phonemes in Tolkien's Orthography:
Consonants in IPA:
The sounds in bold blue are the sounds that I added to the inventory based on the evidence.
Rationale for Added Consonants
Here are the reasons why I added each consonant to the inventory (using orthography):
p, ph: I was a bit hesitant to add these. The attested consonants of Khuzdul ended up matching its labial stops & fricatives with Arabic, which I though was pretty nifty. However, Hebrew, several other Semitic languages, and Yiddish all have <p>, and Tolkien said that Khuzdul's phonetics resembled those of Hebrew. I was even more hesitant to add <ph> since even Biblical Hebrew doesn't have an "emphatic [p]", which would be velarized or pharyngealized. This would match the attested aspirates in Khuzdul. However, Adunaic does have both <p> and <ph>. Also, Appendix E in LotR says that aspirates were "common" in Khuzdul, so it would seem odd to only have 2 in <th> and <kh>. The Dwarves of the Lonely Mountain (Erebor) eventually added runes for the consonant combinations <ps ts ks>, which were common in Quenya (and maybe Sindarin?). Khuzdul already has <khs> (which is <kh> + <s>). I figured if both Khuzdul and surrounding Mannish languages of the Northmen had these, then it would have made even more sense for the Dwarves to add them since their use of Quenya would have been fairly limited as far as I can tell.
v: Since I added <p> on the basis that Hebrew and Adunaic had it, I went ahead and added this sound as well.
w: Again, Hebrew and Adunaic have this sound, and so does Arabic. It also adds another "weak consonant" (in addition to < ' y h >) that will make for some irregular word forms. Khuzdul was noted as complex, so this will add so that.
c j ch: This combination might be controversial for those who have read about Khuzdul a bit. In Appendix E, Tolkien states that "ch is only used to represent the sound heard in bach (in German and Welsh), not that in English church." However, when he says that, he is only referring to the orthography used in the narratives; that is, in the stories. The <ch> found on the table of Angerthas values actually represents [ tʃ ] in English church. The Angerthas use a different transliteration scheme, which is intended more to represent different sounds somewhat accurately, as opposed to the start of Appendix E which is how those sounds are presented to readers in the stories. I have started to call this difference "Tolkien's Narrative Orthography" vs. "Tolkien's Phonetic Orthography".
Adunaic had a "C-series" of consonants, which was apparently post-alveolar, palatal, or both. Like the other series, it had a voiced, unvoiced, and unvoiced-aspirated trio of stops. In Sauron's Defeat, they are described as "front or palatal consonants orginally; that is roughly consonants of the K-series in the extreme forward or y-position." It's hard to tell exactly what that means, but it sounds like they may have been palatalized [ kj gj kjh] or just palatal [c ɟ ch ]. I suspect the former, because Tolkien adopted the notation < c j ch > "because their later development was to simple consonants". To my understanding, [c ɟ] are often realized more as affricates. Tolkien also uses the orthography <j>, which leads me to believe that his "simple consonants" may actually be more in the line of [ç ʝ çh ], [ tɕ dʑ tɕh ], or [ tʃ dʒ tʃh ]. He often describes alveolars as "dental", which makes it seem like he broke front-to-back locations into roughly front (dental/alveolar), middle (post-alv. to palatal), and back (velar to glottal). There is nothing I can find in the source material that describes a "hard palatal C" [c], as opposed to an affricate.
Another thing that leads me to include <c j ch> is that both Biblical Hebrew and Classical Arabic have [ s z sʕ ] (according to my sources). It may be debatable whether [ sʕ ] is actually pharyngealized. The point is that they have an "emphatic [s]". I didn't like that there was no good way for Khuzdul to have an emphatic <s>, since the Khuzdul "emphatics" are apparently aspirates. I did note that some scholars believe Biblical Hebrew's [ s z sʕ ] may have actually been [ts dz tsʕ]. That is just a step away from [ tʃ dʒ tʃʕ ]! So, because of that and that Adunaic has the series, I felt comfortable adding them.
The Dwarves modified the cirth (rune) for <j> not once, but twice. One of those instances was by the Dwarves of Moria when they were still there. This is just one more indicator that the <j> sound may have been in Khuzdul, as well as other languages with which they were in contact.
One final note: adding this sound allows the name Telchar to be a Khuzdul word, rather than a very stretched attempt to explain it via Sindarin. This works well, because the two other Dwarves of the 1st Age that are named also have apparently Khuzdul names: Azaghâl and Gamil Zirak. The assumption is that Dwarves didn't give out their inner, Khuzdul names. To me, it seems obvious that these names are probably outer names. In the 1st Age, there wouldn't have been as much contact with Men yet, so the Dwarves may have used Khuzdul words as outer names. Also, Khuzdul somehow influenced the languages of Men (Adunaic), and I doubt that would have only been through a few place-names. Instead, it seems far more likely that the Dwarves had outer Khuzdul names that provided more samples to loan to Men.
zh: Like <j> above, this cirth for this sound was modified twice by the Dwarves. It doesn't appear in Biblical Hebrew, Arabic, or Adunaic, but does show up in Modern Hebrew and Yiddish. However, Biblical Hebrew and Classical Arabic may have had the laterals [ ɬ ] and [ ɮ ] respectively. Since Middle-earth is a sort of prehistoric Earth, I could see Khuzdul's <zh> shifting into those slots. This sound actually doesn't appear in any of the languages of Middle-earth as they are described, so Khuzdul may well have had it. The cirth for this sound was added by the Elves of Eregion to represent sounds not found in Sindarin, so Khuzdul seems like the likely candidate since there are no others.
hy: This was added by the Dwarves of Moria, and it is also found in Adunaic. Arabic and Hebrew have a bunch of fricatives in the velar to glottal range. Having this sound strengthens Khuzdul's inventory in that area, or at least right next door. See the comments below about <kh> and <gh>.
About the sounds that I have not included:
hw, lh, rh: These are voiceless <w l r>. Hw appears in Westron, and I think maybe the Elvish languages as well. The others were used by the Elves. Hebrew, Arabic, and Adunaic do not have them, so it seems likely that they come from the Elves, not Khuzdul.
mb, nj, ng, nd: These are all a result of "nasal infixion", which is a process that occurs in Quenya and Adunaic. In Sauron Defeated, Lowdham's Report says that Khazadian (Khuzdul) does not have this feature at all. Therefore, I have left them out as phonemes. That said, the Dwarves apparently did use them to write simple consonant clusters, even though they would probably cross syllable boundaries (given Semitic syllable structures).
th, dh, kh: These are the sounds "th" in "thin", dh as "th" in "there", and "kh" is the same as German's "ach-laut" found in "bach". Tolkien says in Appendix E that Khuzdul does not have these sounds. It's a bit surprising that Khuzdul doesn't have this <kh> sound (note that this <kh> is different than the <kh> seen in Khuzdul, which is an aspirated stop). Adunaic and Biblical Hebrew both have it, and Modern Hebrew and Arabic have the voiceless uvular fricative instead. The uvular fricatives don't show up in the Angerthas chart, so Khuzdul probably doesn't have that. My guess is that Tolkien was concerned with orthography (he seems to have been quite particular about how words look on the printed page) as he was already using <kh> for an aspirated stop. Using <ch> might have been an alternate route, just like in Sindarin, but as I mention above that is probably used for the aspirated post-alveolar affricate <ch> in English "church". This leaves Khuzdul empty in a range where all related languages have one or more sounds. I think this is where <hy> enters the picture (German's "ich-laut"). Like uvulars, palatals are right next to velars, and its presence in German may have reminded Tolkien of the Jews he was familiar with. They probably would have been Yiddish speakers, but even Yiddish doesn't have the <hy> sound; it has a voiceless uvular fricative like Modern Hebrew and Arabic. Still, they would have also commonly spoken German, so that may have made Tolkien feel comfortable in adding it.
gh: This is the voiced fricative paired with <kh> above. There is one Khuzdul name, Azaghâl, where this sound potentially shows up. However, in Appendix E, Tolkien notes that this sound is seen in Orcish and Black Speech, but there is no mention of Khuzdul, Westron, or Adunaic. I think that is telling. Also, I have found no vocalization in Hebrew or Arabic similar to Azaghâl, where a glottal stop is prefixed with a vowel and the first radical has a vowel as well. You can find 'aCCâC commonly in Arabic, but nothing like 'aCaCâC. The -âl could conceivably be the same as the adjectival suffix -ul seen elsewhere in Khuzdul, but Tolkien used a <u> or <û> everywhere else. Although Adunaic and Biblical Hebrew have this sound, Modern Hebrew and Yiddish do not. They only have a voiceless fricative in this range with no paired voiced fricative. From all of this, I can see Tolkien leaving this sound out of Khuzdul. It's not a clear call, but I lean in this direction.
kw, gw, khw, ghw, ngw, nw: These are labialized consonants, and for the most part are found as either phonemes in Quenya or consonant clusters in Sindarin. Hebrew, Arabic, and Adunaic do not have them at all. I can't picture Khuzdul having them at all, either.
ps, ts, x (ks): These are said to be common consonant clusters in Quenya and I think Sindarin. Even there they are not phonemes. The Dwarves added these runes, and I suspect they were found in Khuzdul, Adunaic, and other languages of Men in addition to Quenya and Sindarin. Since the Dwarves used the "nasal infixion" runes (above) for simple consonant clusters, they probably found it useful to add these since they were common. They were NOT added by the Dwarves of Moria, so I doubt they are considered phonemes. They are consonant combinations, not affricates.
ŋ: Hebrew, Arabic, and Adunaic all lack this sound, unless it's an allophone of <n> when before a velar consonant. Sindarin and Quenya have it, and it was said that where Adunaic differed from Quenya was where it most resembled Khuzdul. Hence, Khuzdul doesn't get this sound.
bh, dh, gh ("aspirated" voiced stops): One of the common assumptions by linguistic novices (and myself at one point!) is that because there are aspirated voiceless stops <th kh> there must be voiced ones as well. Adunaic did not have these at all. Biblical Hebrew and Arabic have the voiceless - voiced - "emphatic" voiceless trio of stop. Like Adunaic, Khuzdul's "emphatics" are apparently aspirated consonants, which gives it a 3-way distinction of voiceless - voiced - aspirated voiceless, which is common in the world's languages. Also, Tolkien routinely uses <gh> to represent [ɣ], a fricative, so that takes away the orthography for an "aspirated <g>". The only place that even hints at "aspirated voiced stops" (which would actually be "breathy voiced") is the word sharbhund. However, I decided to deal with this example by breaking it into the syllables sharb + hund.