Spin Semiconductor

I have a number of questions relating to the accuracy and resolution and
the values of some of the internal operations.

1. What are the accuracy, resolution and dynamic range of the LOG2
operation? I did not see this stated anywhere. The LOG2 output
is divided by 16, indicating at least a 16 bit dynamic range, but the
accuracy and the precision do not appear to be stated anywhere.
It would be useful to know these in order to be able to establish an
upper error bound on the computations involving LOG2 and EXP2.
Also, is the divide by 16 done with rounding or with truncation?

2. What are the accuracy, resolution and dynamic range of the EXP2
operation? I did not see this stated anywhere. Also, Is taking the
LOG2 of an in-range value and then following that with an EXP2
guaranteed to exactly reproduce the original input value to the
LOG2? The answer surely has to be no, but I would like to know
how close to the original input value is to be expected.

3. What are the sizes of the immediate values in the Boolean operations?
The AND, OR and XOR operation examples show up to 23 bits in binary
constant values. Are the Boolean immediate values really 23 bits long
or are they 24 bits long, or some other size? This should be specified
within the Boolean operation explanations.

4. The NOT instruction actually appears to be a NEG instruction based on
the description. A NOT would perform ones' complement while a NEG
performs twos' complement unless the FV-1 uses ones' complement for
the internal representations. Other examples suggest that the FV-1 uses
twos' complement rather than ones' complement arithmetic, hence the
question: is NOT performing a ones

For 1 and 2 I will need to see if I have the notes on that.

For 3 the instruction set PDF states "the 24bit MASK specified within the instruction word" for AND/OR/XOR so I am unclear why you feel it is not stated in the documentation.

For 4, the instruction set PDF states "performing a 1

Frank:

Sorry, I now see that I failed to indicate where number 3 was relevant.
I was referring to this location:
Knowledge Base => SpinAsm instructions and syntax => AND
(or OR or XOR).

I also noticed that PACC is mentioned in the AND and XOR explanations
but it is absent from the OR explanation. I assume that PACC is updated
in all 3 instructions, although the documentation does not indicate this.

Likewise, for number 4, the missing path is:
Knowledge Base => SpinAsm instructions and syntax => NOT
which states "Change sign of ACC value". Changing only the
sign would be twos' complement, while the name suggests the
other alternative of ones' complement.

Lastly, could you indicate where the "instruction set PDF" is located,
since it is apparently not the same as what I was viewing on the SPIN
website.

Thank you.

PS: I sent Keith email earlier today with a number of errors, some of
which are not included within this thread.

Ah, you were reading Keith's "cheat sheet" for the part. The full PDF is on the product page, a direct link to it is http://www.spinsemi.com/Products/datash ... Manual.pdf