06-08-2018, 12:22 PM
(This post was last modified: 06-08-2018, 12:24 PM by ppppenguin.)
I suspect that the error was a typo and in any case it's trivial. 1 part in 4 billion is a lot more accurate than any crystal you're likely to be using.
If you're trying to use a DDS as a constant amplitude signal generator you're probably aware of the sin(x)/x error. This is inherent in any sampled system using a conventional DAC and quite separate from any poor frequency response. For those who aren't familiar, it's due to the DAC maintaining it's output voltage between samples. Mathemtically this is a zeo order hold function. It causes the frequency response to fall off increasingly rapidly as you approach half the sampling frequency. It's about -1dB at a quarter of the sampling frequency. It's this same effect that is partly responsible for the poor movement protrayal on LCD monitors. The LCD pixels hold their previous value until the scan next passes them.
Brief explanation of sin(x)/x problem:
https://en.wikipedia.org/wiki/Zero-order_hold
This article is a reasonable introduction to sampling theory though some of the maths is a bit heavy for that sort of explanation. sin(x)/x is in section 5.1.3.
http://www.ti.com/lit/an/snaa079c/snaa079c.pdf
This is also quite useful:
https://m.eet.com/media/1051226/Sin(x)x_Agilent.pdf
I haven't dug them out but there are more advanced artciles that show how to predict spurii and suchlike. I think that Peter Scott knows alot about this since he worked for HP for many years on frequency synthesis.
If you're trying to use a DDS as a constant amplitude signal generator you're probably aware of the sin(x)/x error. This is inherent in any sampled system using a conventional DAC and quite separate from any poor frequency response. For those who aren't familiar, it's due to the DAC maintaining it's output voltage between samples. Mathemtically this is a zeo order hold function. It causes the frequency response to fall off increasingly rapidly as you approach half the sampling frequency. It's about -1dB at a quarter of the sampling frequency. It's this same effect that is partly responsible for the poor movement protrayal on LCD monitors. The LCD pixels hold their previous value until the scan next passes them.
Brief explanation of sin(x)/x problem:
https://en.wikipedia.org/wiki/Zero-order_hold
This article is a reasonable introduction to sampling theory though some of the maths is a bit heavy for that sort of explanation. sin(x)/x is in section 5.1.3.
http://www.ti.com/lit/an/snaa079c/snaa079c.pdf
This is also quite useful:
https://m.eet.com/media/1051226/Sin(x)x_Agilent.pdf
I haven't dug them out but there are more advanced artciles that show how to predict spurii and suchlike. I think that Peter Scott knows alot about this since he worked for HP for many years on frequency synthesis.
www.borinsky.co.uk Jeffrey Borinsky www.becg.tv
