send q VRF-Frequency_Generator vrf

from: Greg Goebel / HP-MXD

gvg@lvld.hp.com / 800-452-4844

website: ftp://fcext3.external.hp.com/dist/mxd/index.html

to: VRF-Mike Mancuso

date: Thursday, 23 April 1998 1245 MDT

Regarding accuracy of the Frequency Generator object:

I don't know the precise algorithms and we've never characterized this --

we've never been asked about this before -- but generation of simple

waveforms is a trivial operation that I could concoct in a few short BASIC

programs.

We use 64-bit floating-point math and it's just a small set of normal

arithmetic and trig operations -- meaning that with 64-bit computations the

amount of error in the computation is without doubt far smaller than the

amount of error in your ARB.

If this remains a concern, I suggest generating a simple waveform with a

known period with the Frequency Generator and using a pocket calculator to

doublecheck the data. However, the pocket calculator's level of error is

probably only slightly better than VEE's ... an uncertainty principle

applies.

Of course with any sampled waveform you can have uncertainties due to the

sample period. For example, if you have a square wave, one sample will be

a 1 and the next will be a 0 ... the actual transition could be somewhere

in between those values. But this is purely theoretical as well, since that

would apply to an ARB as well as a VEE-generated waveform.

[<>] regards -- gvg

from: Greg Goebel / HP-MXD

gvg@lvld.hp.com / 800-452-4844

website: ftp://fcext3.external.hp.com/dist/mxd/index.html

to: VRF-Mike Mancuso

date: Thursday, 23 April 1998 1245 MDT

Regarding accuracy of the Frequency Generator object:

I don't know the precise algorithms and we've never characterized this --

we've never been asked about this before -- but generation of simple

waveforms is a trivial operation that I could concoct in a few short BASIC

programs.

We use 64-bit floating-point math and it's just a small set of normal

arithmetic and trig operations -- meaning that with 64-bit computations the

amount of error in the computation is without doubt far smaller than the

amount of error in your ARB.

If this remains a concern, I suggest generating a simple waveform with a

known period with the Frequency Generator and using a pocket calculator to

doublecheck the data. However, the pocket calculator's level of error is

probably only slightly better than VEE's ... an uncertainty principle

applies.

Of course with any sampled waveform you can have uncertainties due to the

sample period. For example, if you have a square wave, one sample will be

a 1 and the next will be a 0 ... the actual transition could be somewhere

in between those values. But this is purely theoretical as well, since that

would apply to an ARB as well as a VEE-generated waveform.

[<>] regards -- gvg