Note that while in modern Japanese the final ン may sound a better match for -ng than for -n, the Portuguese materials show that as late as 16th century ン was pronounced [n].

Indeed, we can see in the 1603 Nippo Jisho that words spelled with ん in the Japanese are spelled consistently with ⟨n⟩ in the Portuguese orthography of that time. See this page, for instance, where a word like 寒巌【かんがん】 is spelled as ⟨cangan⟩.

Over time, I've noticed that Japanese goon often correspond with modern Min Nan pronunciations. For 氷, modern Min Nan Hokkian is peng and Teochew is bian¹, corresponding more closely to Old Japanese pyō, where we would expect that more open vowel in the Chinese. As you note, it is puzzling that the oldest kana spellings show ひよう instead of the expected へう.

The pronunciation of ン as [n] is, I believe, shown by the (productive at least until Edo period) process of renjō (観 かん + 音 おん > 観音 かんのん), which would be somewhat strange assuming velar [N] at the end of the first element (something like かんごん being more expected as medial g was pronounced ŋ up to middle 20th century).

@aguijonazo The retention of [b] in the beginning merely means that the Middle Chinese consonant was [b]. As EMC b > Chang’an pɦ, words starting with b can have b- or h- in Japanese based on go-on vs. kan-on; words with original p (or ph) can only have h. Also, EMC m > Chang’an b in many cases, hence examples like 文 ブン, while EMC mjun.

@EiríkrÚtlendi As I said, I’m not sure old go-on spellings would exist in this case. Go-on was reformed heavily at least twice, with the Buddhist masters making it more predictable from kan-on and then the kokugaku scholars matching it to LMC rhyme tables exactly. Also, unlike kan-on, the actual go-on was never systematic or covering all the existing characters, so it just might be that until Chang’an 氷 happened not to be borrowed to Japanese.

Re: 連声, see also 天皇【てんおう】 → 天皇【てんのう】. As counterpoint, we still have 雰囲気【ふんいき】 and 反映【はんえい】, making me wonder if there's something specific about the phonetics of ん+お.

I note that they already had <ã> but didn’t use it to represent Japanese sounds. Not related to the question but I am also curious why some of the measure words that begin with /h/ or /f/ get voiced as /b/ after 三 /san/ (e.g. 本, 杯) while others are pronounced with /p/ (e.g. 分, 泊). Is this also due to the difference in the original pronunciation in Chinese?

Re: 雰囲気, thank you for dating that, I hadn't dug into the history of that yet.

Re: historical kana ひよう not へう, I suspect that the alternative Middle Chinese reconstructions with medial //-iə-// may be key.

See also en.wiktionary.org/wiki/%E5%86%B0#Pronunciation and expand the Middle Chinese section.

Of the seven reconstructions, four of them have a medial glide.

@aguijonazo, that came up a while back with regard to the name さぶろう. One sec, I'll get the link...

@aguijonazo, japanese.stackexchange.com/questions/57702/…

This may be related to the reconstructed pre-nasalization of voiced obstruents in Old Japanese, so ⟨b⟩ was theorized with the pronunciation of //mb//.

This kind of thing surfaces in some modern dialects, such as up in the Tōhoku where 度【たび】 is pronounced as たんび instead.

@AlexanderZ., looks like 反映 is pretty recent too, only appearing in the 1870s: kotobank.jp/word/%E5%8F%8D%E6%98%A0-606046

However, 範囲【はんい】 seems to show up in 1060: kotobank.jp/word/%E7%AF%84%E5%9B%B2-7518

Checking now in the Nippo Jisho to see if they have any listing.

No such luck. Fani would be on this page, but no such entry. Fai is on this page, but none of the entries are relevant, and I don't see any use of the tilde to mark nazalization for the Japanese -- although it is clearly already in use in Portuguese at that time.

Well, the example is just a Classical Chinese text, so we cannot be sure in what way would it ever be assumed to be pronounced.

As for tildes, we have things like ninde

nindẽ 人天, not sure about significance of that versus n.

I think tildes were at that moment nothing but a typographic device, a shortener of the following -n, which was sufficiently optional and not codified.

@aguijonazo Frellesvig implies that combinations n+voiceless did not exist up until very recently (late Muromachi at earliest), as previously the prenasalization of voiced would automatically turn any n+voiceless cluster into voiced. Thus, any np- etc. would just be recent formations, that heppened to stick to literary language due to prevalence in Tokyo Japanese. There is a lot of variation in those still. Such as what is the pronunciation of 三階.

Re: 範囲, ya, as you note, the quote in the 日国 entry is kanji-only. That's why I went to the Nippo Jisho, to try to confirm the phonetics from an older source.

Good find for the Nindẽ entry! Curious that the usage of the tilde is so limited for the Japanese terms. I see also that the Ninguen entry on the very next page doesn't use the tilde, and spells out the final ん.

Re: //-np-// combinations, I see a few in the Nippo Jisho, such as various entries beginning with ⟨Canp-⟩ on this page. I assume that the Nippo Jisho entries reflect Nagasaki Japanese as opposed to Tokyo, just given the history of Portuguese-Japanese relations, but I do not know for certain if this is the case.

Either way, 1603 is later than the supposed late-Muromachi (1560s? 1570s?) re-emergence of //-p-//, so the Nippo Jisho gives us no insight into what the "original" Japanese pronunciations may have been.