Recently I purchased an DG1062Z with the 16MByte option. Owning already an DG4162Z
the only reason to do this was the possibility to generate 2 arbitrary waveforms with reasonable
length and reasonable costs to feed the I/Q-Modulation Inputs of a Vector Signal Generator.
During my experiments I observed a certain number of characteristics which could be
interesting to know before buying a DG1062Z as arbitrary generator. Here are my
findings with an instrument with firmware 01.07.00.02 installed:
In all documents where the specifications of the instrument are listed we can read
under the section 'Arbitrary Waveform' that the Sample Rate is specified to 200 MS/s.
To my surprise, the actual sampling rate setting on the DG1062 in the mode
[Arb] -> [ArbMode-> [SRate] is limited to 60 MS/s. Indeed on page 2-18 of
the User Guide I could find a mention of this.
Generating long arbitrary sequences is only possible by downloading
data with the :SOURCE<n>:TRACE:DATA:DAC16 VOLATILE,<flag>,<data>
command, because this is the only SCPI-command which has the <flag> = {CON|END}
argument which allows the download of multiple blocks to the DDRII internal memory.
But as opposed to the Programming Guide (p. 2-173) the packet size must not
exceed 8192 samples, otherwise an error occurs.
The output can be generated in 2 modes:
[ArbMode] -> [Freq] and [ArbMode] -> [SRate]
A. [ArbMode] -> [Freq]
* Even if a long sequence (> 8192 points) is downloaded with the SCPI-command
mentioned above, the sequence restarts from the beginning after 8192 samples.
* If a shorter sequence is loaded, the sequence length remains
8192 but to complete the sequence to 8192 points, previously
loaded, old memory contents are inserted.
* The sample rate is always 8192 * f, where f is the frequency setting.
* There are apparently 2 asynchronouly superposed (cascaded) sampling
systems; one with a frequency of 8192 * f and a second one fixed at
200 MS/s. Cascading these samplers produces a jitter of 5 ns.
This behavior is similar to the DG4162 whose sampling rate is 500 MS/s
and therefore produces a 2ns jitter.
* No interpolation takes place between the samples.
So, the sequence 0,0.1,0.2,0.3,0.4,0.5.... produces a staircase.
B. [ArbMode] -> [SRate]
* Generating long arbitrary sequences (> 8192 samples) is only possible
in this mode.
* The sampling rate is limited to 60 MS/s which is not in accordance with
the specification.
* A linear interpolation is done between the samples. I find that this is
good behaviour in that it gives smoother waveforms with less undesired
harmonics. The drawback is that the rise time of a signal cannot be
lower than the sample interval.
* With the 16M memory option one can effectively generate waveforms with
16 Msamples on each channel simultaneously.
* As opposed to the mode ArbMode->Freq there is no jitter due to interferences
between samplers.
* Downloading a sequence of 16 Msamples in one channel takes
125 s over the LAN (120 kS/s or 2Mbit/s). This is not very exciting on a
100 Mb/s Network but for my application it is acceptable.
* While downloading data, the output of the concerned channel is
stopped while the other continues to work.
* As expected, pressing the function [Align Phase] synchronizes both channels
if both channels are set to the same sampling rate and are loaded with
sequences of same length.