On 14/02/16 23:56, Philip Pemberton wrote: Naturally, I figured out what I was doing wrong an hour or so after I hit send... Enjoy! // val is a FFP float, returns IEEE float format // basically a direct conversion of the motorola FFPTIEEE function. // comments are from there. // you are not expected to understand this horrendous mess. float ffpieee(const uint32_t val) { uint32_t x = val; union { float f; uint32_t i; } _fcast; x = x + x; // delete mantissa high bit if (x == 0) { // if zero, branch zero as finished return (float)x; } uint8_t k = x & 0xff; k ^= 0x80; // to two's complement exponent k >>= 1; // form 8-bit exponent k -= 0x82; // adjust 64 to 127 and excessize x = (x & ~0xff) | k; x = (x << 16) | (x >> 16); // swap for high byte placement x <<= 7; // set sign+exp in high byte _fcast.i = x; return _fcast.f; } -- Phil. classiccmp at philpem.me.uk http://www.philpem.me.uk/