N5182B选通信号为什么不是预期的500ksa

我正在使用N5182B矢量信号发生器，其RF输出直接连接到频谱分析仪。
我正在使用Agilent N7622B Signal Studio Toolkit 1.3.0.0版将matlab波形写入VSG。
我使用Toolkit附带的MATLAB示例代码生成1000个样本音 - 相关代码总结如下：phase =（2 * pi / 100）*  [0：999 ];
Idata = cos（阶段）;
Qdata =罪（阶段）;
这非常有效;
出现预期的语气。
VSG表示该段在BBG内存中占用1000sa，如预期的那样。
但是，当我将音调调整为500,000个样本时（在设备的32Msa限制内），VSG表示该段占用488ksa，而不是预期的500ksa。
噪声基底升高，出现音调的副本（镜像中心频率），并且音调每秒打开和关闭许多次。
对MATLAB代码的唯一更改是使用“phase =（2 * pi / 100）*  [0：499999 ];”
而不是“phase =（2 * pi / 100）*  [0：999 ];”
谁能告诉我可能是什么原因导致的？
我附上了截图，显示频谱分析仪在预期频率下设置为零跨度，而500ksa音调以50MHz时钟频率播放。
这种模式的周期恰好是10ms。
模式开始时的较高幅度步长为1.3ms。
模式的总时间为7.76ms。



     以下为原文

  I'm using a N5182B vector signal generator, with its RF output connected directly to a spectrum analyzer.  I'm using the Agilent N7622B Signal Studio Toolkit version 1.3.0.0 to write Matlab waveforms to the VSG.

I used the MATLAB sample code that comes with Toolkit to generate a 1000 sample tone - relevant code summarized below:
phase = (2*pi/100)*[0:999];
Idata = cos(phase);
Qdata = sin(phase);

This works perfectly; a tone appears where expected.  The VSG indicates that the segment occupies 1000sa in BBG memory, as expected.  However, when I make the tone 500,000 samples long (well within the device's 32Msa limit), the VSG indicates the segment occupies 488ksa, rather than the expected 500ksa.  The noise floor becomes elevated, a copy of the tone appears (mirrored about the center frequency), and the tone gates on and off many times per second.  The only change to the MATLAB code was using "phase = (2*pi/100)*[0:499999];" instead of "phase = (2*pi/100)*[0:999];"

Can anyone tell me what might be causing this?

I've attached screenshot showing the spectrum analyzer set to zero span at the expected frequency, while the 500ksa tone is being played back at a 50MHz clock rate.  The period of this pattern is exactly 10ms.  The higher-amplitude step at the start of the pattern is 1.3ms long.  The total on time of the pattern is 7.76ms.   
附件

• 500ksa tone at 50MHz.png71.0 KB