A short code sample for your SERIES structure:

LOCAL nLen, nHorizon, nPpc
nLen = 1200
nHorizon = 4
nPpc = 54
cBuffer = num2dword(m.nLen) + num2dword(m.nHorizon) +;
	num2dword(m.nPpc) + Repli(Chr(0), 32)  && 8 times by 4 bytes

FUNCTION num2dword(lnValue)
#DEFINE m0  256
#DEFINE m1  65536
#DEFINE m2  16777216
	IF lnValue < 0
		lnValue = 0x100000000 + lnValue
	ENDIF
	LOCAL b0, b1, b2, b3
	b3 = Int(lnValue/m2)
	b2 = Int((lnValue - b3*m2)/m1)
	b1 = Int((lnValue - b3*m2 - b2*m1)/m0)
	b0 = Mod(lnValue, m0)
RETURN Chr(b0)+Chr(b1)+Chr(b2)+Chr(b3)

I'm not completely sure about float * values. It might be that instead of Repli(Chr(0),32) you should allocate 8 memory blocks 4 bytes each. And write into each memory block a float value. For that you need a procedure converting Fox numeric to float format.

I do not have this conversion procedure availble right now. Peter Easson once posted a set of them on Wikis site. You can reach it as soon as this site will be available.