From ear, despite what the conventional literature tells me, I often hear that the accented mora has a higher amplitude, not just coming before the downstep. This is particularly obvious in This Youtube fragment, where the recording equipment's low quality causes the audio to crackle on the ga-mora. Searching for it myself, I could only find this, referencing one study that finds it.
[...] and that since, in his data, the amplitude peak fell on the accented syllable in the words in which the F0 fall was delayed [...]
Obviously this is an argumentative piece, and I cannot find the actual named research anywhere though many others that also reference it, and I cannot find anything else that corroborates it and searching for anything just seems to draw a comparison to English where accent is indicated by amplitude — so does anyone else have anything that corroborates or disputes the notion that in Japanese, accented moræ tend to have a higher amplitude?