叹息…我已经经历过这样的事情,一次或三次。还有一点比NCO插入一个增量值要多。这里的协议……源时钟为32兆赫,NCO当然可以产生一个平均频率为12.2879 MHz的信号,但是由于NCO的DDS性质,NCO源的一个周期会有抖动。E时钟。换句话说,平均周期将是大约80纳秒,但抖动将是大约30纳秒。而且,这不是一个更短或更长的周期。抖动频率是输出频率的可观分数。我不能想象这将是有用的采样时钟的DSP(或任何其他类型的“时钟”)。(作为对感兴趣的读者的练习:做数学)另一个现实生活细节中的那些讨厌的现实:即使是很轻的负载(“范围探测”),假设抖动是可以接受的,PPS输出引脚在12 MHz时不会给出干净的信号。这些东西不具备高速数字输出的驱动能力。我停止了“范围”,并用我的PIC16F15355在32 MHz的NCOCON=805306下运行了附加的范围拍摄,这样你们就可以看到我在说什么了。(注意,对于方波输出,切换频率必须是方波频率的2倍)。底线:我认为从PIC期待一个有用的12.288 MHz时钟输出是不合理的。[编辑]当我进行范围拍摄时,MBRUBULIN把我和我正在制作的点(+ 1)打得一模一样。我仍然希望这是值得的带宽来显示他正在谈论什么,以及实际的局限性,产生一个时钟信号,这是一个可观的分数输入时钟与NCO。只是不实用(没有一吨的外部过滤,在这种情况下,它比使用一个真正的时钟生成器更昂贵)。(编辑)问候,戴夫
以上来自于百度翻译
以下为原文
Sigh...
I've been through this kind of thing a time or three. And there's a little more to it than just plugging an increment value into the NCO.
Here's the deal...
With a source clock of 32 MHz, the NCO can certainly generate a signal with
average frequency 12.2879 MHz, but due to the DDS nature of the NCO, there will be a jitter of one period of the NCO source clock. In other words, the average period will be around 80 ns, but the jitter will be about 30 ns. And, this isn't a shorter or longer period every now and then. The jitter frequency is an appreciable fraction of the output frequency. I can't imagine this would be useful as a sample clock for DSP (or any other kind of "clock"). (Left as an exercise for the interested reader: Do the math.)
Another one of those little nasty-realities-of-real-life details: Even with a very light load ('scope probe), and assuming the jitter would be acceptable, the PPS output pins will not give a clean signal at 12 MHz. These things just doesn't have the drive capability for high-speed digital outputs.
I stopped the 'scope and took the attached 'scope shot with my PIC16F15355 running at 32 MHz with NCOINC = 805306 so you can see what I'm talking about. (Note that for square wave output, the toggle frequency must be 2x the square wave frequency.)
Bottom line: I don't think it's reasonable to expect a usable 12.288 Mhz clock output from the PIC.
[Edit]
While I was taking the scope shot, mbrowning beat me to the punch with the point I'm making (+1). I still hope it's worth the bandwidth to show what he is talking about, and the practical limitations of generating a clock signal that is an appreciable fraction of the input clock with the NCO. Just not practical (without a ton of external filtering, in which case, it's more expensive than just using a real clock generator).
[
/Edit]
Regards,
Dave
Attached Image(s)