Reference Manual

Turbo PMAC/PMAC2 Software Reference
Turbo PMAC Global I-Variables 253
Excitation Address Setup Word: The second setup line in a resolver conversion entry contains $4 in the
first hex digit, and the Y-address of the excitation value register in the low 19 bits (bits 0 – 18), used to
correlate the excitation and the feedback values. Multiple resolver channels can use the same excitation
register. For example, if the excitation address is at Y:$078C10, the second setup line would be set to
$478C10
Sine/Cosine Bias Setup Word: The third setup line in a resolver conversion entry contains bias-correction
terms for the sine and cosine ADC values. The high twelve bits (the first three hex digits) contain the
bias-correction term for the sine input; the low twelve bits (the last three hex digits) contain the bias-
correction term for the cosine input. Each 12-bit section should be treated as a signed 12-bit value (so if
the most significant of the 12 bits is a 1, the bias value is negative).
Each 12-bit bias-correction term should contain the value opposite that which the high 12 bits of the
matching A/D converter report when they should ideally report zero. In action, the bias term will be
added to the high 12 bits of the corresponding ADC reading before subsequent calculations are done.
In most cases, the bias-correction word will be determined automatically by a high-resolution
“diagnostic” entry (format $F/$1) in the conversion table. The result of that diagnostic entry, containing
both bias corrections, can simply be copied into this setup word.
For example, if the bias-correction word were set to $004FFA, the sine bias correction would be +4 LSBs
of a 12-bit ADC, and the cosine bias correction would be -6 LSBs ($FFA = -6) of a 12-bit ADC. In use, 4
12-bit LSBs would be added to the sine reading, and 6 12-bit LSBs would be subtracted from the cosine
reading each cycle before further processing.
In most cases, the bias-correction word will be determined automatically by an analog “diagnostic” entry
in the conversion table (method $F/$1). The result of that diagnostic entry, containing both bias
corrections, can simply be copied into this setup word.
Result Word: The output value of the resolver conversion is placed in the 24-bit X-register of the third
line of the conversion table entry. The values in bits 5 – 16 of the result word contain the high 12 bits of
the calculated arctangent of the bias-corrected sine and cosine values from the resolver. Because PMAC
software considers the value in bit 5 to be a “count” for its scaling purposes, this conversion returns
resolver position values of a 12-bit conversion (4096 “counts” per cycle of the resolver).
However, because the conversion uses dual 14-bit converters and the arctangent calculations are valid to
15 bits, the result contains additional resolution in bits 0 – 4 that PMAC software considers to have
“fractional”, but still real, count resolution. If the electromagnetic noise levels are low and the signals use
near the full scale of the ADCs, a repeatable 14-bit resolution (16,384 states per cycle of the resolver) can
be achieved.
Bits 17 – 23 of the result contain cycle data from software extension of the result to multiple resolver
cycles. If the result is then used for feedback or master data, it will be further extended in the motor
algorithms.
This resolver conversion is a direct, and not a tracking, conversion. As such, it is more dynamically
responsive, but also more susceptible to measurement noise. If a more noise-immune result is desired, at
the cost of some dynamic responsiveness (but still no steady-state tracking errors), a digital tracking filter
can be implemented on this result with another conversion table entry (format $D8). The result of that
filter entry can then be used as the feedback or master data.