This is taken from Coxe’s 1801 ‘An Historical Tour in Monmouthshire’:
To the west of the new passage inn, near the ruins of Sudbrook or Trinity Chapel, are remains of an entrenchment, which are usually supposed to be Roman; they occupy a flat surface on the edge of a perpendicular cliff, and are nearly in the form of a stretched bow, whose cord is the sea coast. The entrenchment is formed by a triple rampart of earth, and two ditches; the two exterior ramparts are low, and in many places destroyed; the interior is in greater preservation, and not less than twenty feet in height [...]
It is generally imagined that this entrenchment, in its present state, is not perfect, and that half of it has been destroyed by the sea, which has likewise carried away part of the church-yard. It is likewise by many supposed to have been a maritime fortress, erected by the Romans to cover the landing of their troops, adn their first station in Siluria; an opinion grounded on the erroneous description [as a square] of Harris, and on the discovery of a single coin struck by the city of Elaia in honour of the Emperor Severus. For notwithstanding repeated enquiries among the farmers and labourers of the vicinity, I could not learn that any coins or Roman antiquities had been found within the memory of the present generation. It has been also attributed to the British, Saxons, and Danes; but was occupied, if not constructed by Harold during his invasion of Gwent.
The ruins of a 12th century chapel lie among the ramparts to the south east, and Coxe mentions “Within the memory of several persons now living, divine service was performed therein; and a labourer whom I met on the spot, assisted forty years ago as pall-bearer, and pointed out the half of a dilapidated grave stone, under which the corpse was interred.”
The Rev Sayer's diagram is quite far off the mark: look at the thing from the air (in Google or Bing or the photographs at the Royal Commission on the Ancient and Historical Monuments of Wales) and you'll see that what's left of it appears to be one corner of a rather baggy irregular square, not a semi-circle. Since probably more than half of it is missing it's impossible to say if it really was square, or triangular, or pear-shaped but what's left comes to a blunt point. Not Roman, though - it's supposed to date from 600BC.
I'm taking a particular interest in it because for complicated reasons I think it was probably the model for Caer Sidi - beginning with the fact that it was almost certainly at some point called something like Caer Sudd.
The stories of the rescue of Mabon from Gloucester in Culhwch ac Olwen and the rescue of Gweir from Caer Sidi in the Llyfr Taliesin are so similar that they are probably the same story, with drift. In Culhwch ac Olwen a rescue party set out from the lake called Llyn Llyw and swim to Gloucester where they rescue the prisoner, but the introduction of Gloucester makes no internal sense here - the prisoner is meant to be plainly audible through the walls of his prison and yet to have been kept in complete secrecy for centuries, which makes no sense if his prison was in a city, so in the original version Mabon was probably imprisoned in Caer Sidi and it was to there that the rescuers swam from Llyn Llyw.
Llyn Llyw is almost certainly the same as Llyn Liuan in the Historia Brittanum, and Llyn Liuan has been identified with a high degree of probability as a siphoning lake with three whirlpools which used to exist between Caerwent and Caldicot, until the "Great Spring" which maintained it was diverted during the driving of the Severn Tunnel. The outlet from this very probable Llyn Llyw was at Sudbrook, next to the fort.
Caer Sidi is usually translated as "Spiral Castle", and the stripes of ditches and banks along the sides of Sudbrook Camp must have given it a vaguely spiral appearance when it was complete. Caer Sidi was said to have four towers, and whether Sudbrook Camp was really square when complete or not, it's enough that it should look as if it might have been square, for the storytellers to imagine it as having had four corners and four towers, and as then having been battered down by Arthur's warships to account for why half of it had ended up on the beach below.
Most strikingly, one of the things said about Caer Sidi is that "The fruitful fountain is above it // Sweeter than white wine is the liquor thereof". Before the engineers working on the Severn Tunnel diverted and culverted the waterways in the area, just upstream from Sudbrook Camp (probably = Caer Sudd) there was an upwelling whirlpool which was actually a "swallow hole" fed by the Great Spring - which even nowadays is known for the exceptional purity of its water.
This story by Nennuis (9th century) is presumably of the same lake, can't remember if he was describing a real happening of the roman soldiers and horses being drowned.. but I suspect a lot of stories and myths must have built up over time.
"The Mouth of Llyn Lliwan 'Another wonder is the mouth of Llyn Lliwan. Its estuary is in the Severn and when the Severn is flooded in the bore and the sea also floods up the estuary of the aforesaid river, the river is received into the estuary waters like a whirlpool, and the sea does not go up; and there is a shore by the river, and whenever the Severn is flooded in the bore, that shore is not touched, and when the sea ebbs from the Severn, then Lake Lliwan spews up everything that is devoured from the sea and that shore is touched, and, like a hill, breaks and spews up in one wave.And if the army of the whole country where it is should be there, and should front the wave, the force of the wave would drag down the army, its clothing filled with water, and the horses would also be dragged down. But if the army should turn its back on the wave, the wave does not harm it, and when the sea ebbs, then the whole of the shore that the wave covered is laid bare again, and the sea ebbs from it."
The monk Nennius in Historia Brittonum