Which floating-point representation did Konrad Zuse's Z1 use?

score:8

Accepted answer

There is version he rebuilt himself on display at the German Museum of Technology.

According to mrob.com, he used a base-2 system with 24 bits of data (I'm not sure why the discrepency with those who claim 22. Perhaps its an inadvertent paste from the Z3 line?) with 7 bits for the exponent and 16 for the mantissa (and presumably one bit for sign), which resulted in a range of 9.9999×10^8 .

There's a discussion of his system in detail here. That page says he actually used 14 bits for the mantissa (with of course one extra bit "implied", just like in IEEE). If this reference is right, then that may explain the discrepency. Thus the format would have been 22 bits with 14 mantissa bits (with one extra implied), a 7-bit exponent, and a sign bit.

I haven't found anything spelling out where in his data word he kept the mantissa, exponent, and sign.

More post

Search Posts

Related post