Reference clock signal.

Hello.

I'm lost, I am about to design my first DDS circuit but...
The chip I will use have a max frequency for the clock of 180MHz,
I'm not sure I will use 180MHz but how do I get a clock with the right frequency?

I do understand that PLL's(Phase Locked Loop) have something to do with it but the more I read the less unclear it all becomes.

A simpe crystal oscillator will not do but I understand that a crystal will be the reference clock for the PLL or something.

Please can someone clarify, this regards to fixed frequency application ei a clock for a DDS chip.

Regards

A simple crystal oscillator, e.g. 100 or 150 MHz would be my first choice.

Hm...
Okey, I think I have miss-understood. Or I followed a very old introduction, I thought that I would need a crystal/crystal oscillator and then use that as a reference for another circuit to generate the frequency needed.

But searching for crystal oscillator I quickly found suitable products.

Thanks

Better study more before you attempt to design https://www.ieee.li/pdf/essay/dds.pdf

Does a high performens low frequency crystal oscillator cost less than a high performens high frequency crystal oscillator(in general)?

And if so, is it correct that I can achieve a better performing clock by buying a low freq oscillator and with a PLL generate a higher frequency?

- - - Updated - - -

Thanks for that document, I have not been able to find the info I am searching for but this appears great.

David_ said:

Does a high performens low frequency crystal oscillator cost less than a high performens high frequency crystal oscillator(in general)?

And if so, is it correct that I can achieve a better performing clock by buying a low freq oscillator and with a PLL generate a higher frequency?

- - - Updated - - -

Thanks for that document, I have not been able to find the info I am searching for but this appears great.

Click to expand...

Crystals above 20MHz or so tend to be weak in fundamental and stronger in overtones or harmonics and this becomes,a limiting factor due to the small size and tolerance of cutting the angle of the crystal in minutes or 1/60th of a degree.

You can see for yourself the cost (<2$)of a high performance 2.5ppm TCXO vs a 1ppm vs a raw uncompensated crystal at 50 ppm ($0.50) in 1k qty.

Ultralow phase noise SC cut TCXO's are 0.01ppm cost >$200.

TCXO or OCXO's often use a standard 10MHZ Crystal then PLL up to,required frequency but also multiply phase noise are adequate for most applications but not all.

You can expect that a simple (respectively cheap) PLL will considerably reduce the performance of a clock signal, particularly add phase noise which is "copied" to the DDS output. In so far I would clearly prefer an industry standard high frequency crystal oscillator for usual DDS designs. It typically comes with LVDS outputs.

SunnySkyguy said:

.....
Ultralow phase noise SC cut TCXO's are 0.01ppm cost >$200.

TCXO or OCXO's often use a standard 10MHZ Crystal then PLL up to,required frequency but also multiply phase noise are adequate for most applications but not all.

Click to expand...

There is a big difference between an AT cut $1 crystal and $200 SC cut for phase noise dB/Hz @1kHz offset at ~ 20MHz .

Even at this input multiplying phase noise means adding dB/Hz by either harmonic injection or PLL scaling by adding dB/Hz with rising 10logN at a minimum on thus $45 chip. This assumes optimal input and >=7th order LPF feedback filters are also implemented,

Read the fine print.

I'm thinking of buying:
1pcs TCXO 50MHz 50.000M Ultra precision Crystal Oscillators DIP 0.5ppm(NT)
**broken link removed**

There are some info in a little picture on the add but I don't understand what 0.5ppm(NT) refers to, in any case it ether this or some 50ppm oscillator.

I have been looking at different DDS chips and I have found several that I do like the looks of, I am particularity interested in those with a 14bit DAC such as AD9102, but it would be much simpler to go with AD9834 but that only has a 10bit DAC.
1024 vs 16384 values.

If I am looking for a as pure a sine wave as possible I will filter the output with a couple of different filter perhaps but do you think I will regret going for a easier 10bit chip rather than digging in with a 14 bit?

I'm very tempted by AD9102 and there are some people online that has gotten in trouble with those but solved it so I'm not completely without references.

Personally I would avoid buying electronic components off of eBay. Especially ones that are coming from China, there is a thriving business in grey market components. As such the parts may be rejects, non-ROHS, removed from scrap (and cleaned up), handled improperly and have been subjected to ESD, etc...

Besides the description on that page shows +/- 1ppm not the 0.5ppm claimed (if that is even the datasheet for the part).

Here is a selection of parts from a actual distributor:
https://www.digikey.com/product-search/en?pv183=2633&pv139=243&pv253=46&FV=fff4000d%2Cfff8016e&k=tcxo&mnonly=0&newproducts=0&ColumnSort=0&page=1&quantity=0&ptm=0&fid=0&pageSize=25

Yeah they may cost more, but the part will at least be new and will have been handled correctly (not guaranteed, but highly likely). Given you want such an accurate oscillator I wouldn't pick up something from eBay. Unless you're willing to test all it's parameters and how stable it real is, and then find out it's a 100ppm...

Regards

- - - Updated - - -

If you're willing to sacrifice some frequency accuracy, these 10ppm ones have some as low as $2 a piece.
https://www.digikey.com/product-search/en?pv253=19&FV=fff4000d%2Cfff8016e%2C22c00f3&mnonly=0&newproducts=0&ColumnSort=0&page=1&quantity=0&ptm=0&fid=0&pageSize=25

I didn't hear any substantial requirements for your project, except generating a "pure sinewave". In so far it's difficult to guess if you actually need an oscillator with low frequency tolerance or if a high frequency standard quality would be sufficient. Low jitter/phase noise is a different point, but not directly related to frequency tolerance and temperature drift.

Regarding reference clock and DAC resolution, they are determined by your output frequency and signal quality requirements which haven't been told yet.

I am reading and learning as I go but the subject of signal quality is somewhat unclear to me, I know what I will use it for but I don't actually know what that will require. This is by all means a learning exercise as much as anything else but I'll try to give a perspective.

This will be used to generate sine waves, but depending on the choice of DDS chip also triangle and square waves with the purpose to be the input signal to different ADCs to test for all relevant specifications, INL and DNL comes to mind first but offset and gain errors and all others as well.
DACs will also be tested but I have focused on ADCs to begin with.

So for this I thought I start with a sine wave as per all the examples and literature found online such as "ADC histogram measurements", I have always been confused with what people mean with "pure sine wave". Internet searches generally put up sites about sine wave inverters and one or two pages about oscillators, in the end I will build one of those analog quality sine wave oscillators to compare to the digital one.

But regarding signal quality, for one thing I would like frequency accuracy. So that 1kHz, 100kHz or 1MHz is as close to that as possible, but that is not the primary priority. When performing the data converter tests I can't see that frequency will matter in terms of accuracy at all.
Although I would like frequency accuracy, I don't want to spend money on something that will give me that if that is all I get and not also improvements in other aspects.
But I want to avoid harmonic content in the signal and as I have seen on various documents there are two pictures of a signals frequency content.
The first image is the raw output and it has a dominant frequency but around that a lot of weaker but still quite prominent spikes, then in the second picture which is of the same signal but after a 7th order elliptic filter or something as such there are one dominant frequency without all the surrounding spikes.

But I don't understand how you characterize the what I interpret as "noise floor", I find this subject with DDS sine waves and characteristics of the output signal to be particularly confusing and I fail to get a grip of it.

What are the relevant parameters involved in expressing the quality of a sine wave generated through DDS?