I have been generating a simple square + dc signal in arb. mode. As seen both from an oscilloscope and a lock-in triggered by the SYNC signal, the resulting signal seems to move erratically between two phases (as a result, the lock-in is periodically unlocked). This does not happen for specific frequencies (like 2.0, 1.6, 1.0 ... MHz). Apparently, no matter the number of points in the arb. signal. Going back to the built-in functions, I do not encounter this problem. It is reproducible on another 33120.
Another type of waveform error visible in the frequency domain is
phase truncation error. This error results from time quantization of the
output waveform. Whenever a waveshape is described by a finite number
of horizontal points (length), it has been sampled in time (or quantized)
causing a phase truncation error. Spurious signals caused by phase
truncation introduce jitter into the output waveform. This may be
regarded as time (and phase) displacement of output zero crossings.
Phase truncation causes phase modulation of the output signal which
results in spurious harmonics (see the equation below). For lower output
frequencies, the phase accumulator periodically does not advance RAM
addresses, causing the DAC to deliver the same voltage as recorded on
the previous clock cycle. Therefore, the phase “slips” back by 360 / points
before continuing to move forward again. When RAM address increments
are the same on each cycle of the output, phase truncation error (and
jitter) are essentially zero. All standard waveshapes in the 33120A are
generated with at least 16,000 waveform points which results in
spurious signals below the wide-band noise floor of the DAC.
Phase Truncation Harmonics –20 x log10 (P) dBc
where “P” is the number of waveform points in RAM.