High quality sine wave source.

[David_]

Hello.

I trying to construct a evaluation platform for data converters, which should be capeable of supplying ADCs with different kind of signals to be read and the converters characteristics would be calculated and plotted with the resulting output.

I'll need a good high fidelity sine wave source and I want to ask, how pure or how fine a sine wave can be produced through digital means?

when you hear "pure sine wave" digital isn't the first that comes to mind perhaps, but a analog oscillator would not be as manageable as a digital source. I don't want to have a sine wave of one frequency but it should be tailored to specific converter but I don't know if I can go through digital synthesis at all?

Providing its clean enough for the purpose of characterizing converters I guess that accurate frequency choices and amplitude adjustments is what is needed, analog solutions does provide these but what do you people think about this?

Regards
David

 

[horace1]

DDS modules can give very high quality sine waves across a large frequency range
https://www.analog.com/en/rfif-components/direct-digital-synthesis-dds/products/index.html

 

[David_]

If we assume that the DAC will be 14bit, would there be any difference in the spectral purity of a 10kHz sine wave if a DDS with a 400MHz clock was compared to a DDS with a 1GHz clock?
Assume that all other parameters are the same.

- - - Updated - - -

I might use a DDS with a number of filters, say that I choose 5 frequency's to be available to the test set up then I could design filters for those freq's to produce really clean sine waves.
Or maybe a adjustable filter.
I have not done anything like this before so I don't know if that is a possible solution, jet.

Does DDS chips produce a sine wave with adjustable frequency period or is there any solution to get triangle wave shape?
other than some op-amp sine-triangle converter.

 

[horace1]

the DDS output frequency is adjustable (In our application we were generating frequencies between 100KHz and 10MHz)
some DDS devices have a SRAM when in which you can store complex waveforms to be generated

 

[D.A.(Tony)Stewart]

You need to define spectral purity. RF uses very high SNR LO or low phase noise & THD.

With high resolution sigma-delta DAC's and bandpass filters or harmonic notch filters and non RF can be synthesized. For RF , PLL with linear VCO's are used.

 

[David_]

Honestly I don't know what I mean with spectral purity as such, I know that it relates to the frequency content of the sine wave and I know(assumes) that I need a very clean sine wave to be used to characterize data converters.

ADCs will read a sine wave of appropriate amplitude and frequency for that specific converter(a couple of amplitudes and freq's will be used for every converter) and the result will be processed in matlab to determine specifications through different methods.

I really don't know what the quality of the sine wave will actually be demanded but the following circuit seams to be a good choice, it is built with a DDS that's clocked at 75MHz, but the circuit is suggested with a alternative DDS with a 16MHz clock:
https://www.analog.com/static/imported-files/circuit_notes/CN0304.pdf

Is there any more difference than that the 16MHZ obviously can't output a 18MHz sine wave as the 75MHZ can. But if both would output a 1kHz wave, would the it be a advantage to have a lower or higher clock?

I'm not sure that this is a valid question even.

 

[Dan Mills]

DDS spurs are odd, what you usually find is that there are some ratios that give rise to horrible spurs but a ratio just very different substantially clears them up.

In general running at a low output frequency relative to the clock is cleaner and using a part with a bigger DA word length gives rise to lower background noise levels.

Be careful of the AD9910 & 9912, they lack any way to bypass the internal bandgap and as a result have significant AM noise sidebands.

AD have plenty of good app notes on DDS and the various issues they have.

For an audio source I personally would forgo the DDS chip and use the best audio DAC I could find (Which in a good one might be good for 20, maybe 21 bits actual) and clock it with the best low phase noise source I could afford, designing the anlogue section to preserve the noise performance of such parts is **HARD**.
Multiple DACs can be paralled to get a few extra Db on the SNR numbers, but the major noise source is almost always resistor thermal noise.

Regards, Dan.

 

[D.A.(Tony)Stewart]

Honestly I don't know what I mean with spectral purity as such, I know that it relates to the frequency content of the sine wave and I know(assumes) that I need a very clean sine wave to be used to characterize data converters.
...

Purity is defined by %THD or -dB levels on harmonics and phase noise dB/√Hz.

For a crude sine wave an 8 DAC can be used with a lookup sine table.

 

[David_]

I've found AD9102, Low Power, 14 Bit, 180MSPS, Digital to Analog Converter and Waveform Generator.
It appears as a nice solution:
The AD9102 TxDAC® and Waveform Generator is a high performance DAC integrating on-chip pattern memory for complex waveform generation with a Direct Digital Synthesizer (DDS).

The DDS is a 14-bit output, up to 180Mhz master clock sinewave generator with a 24-bit tuning word allowing 10.8Hz/LSB frequency resolution.

SRAM data can include directly generated stored waveforms, amplitude modulation patterns applied to DDS outputs or DDS frequency tuning words.
SFDR to Nyquist 87 dBc @ 10 MHz output

Phase Noise @ 1 kHz offset, 180 MSPS, 8 mA: −150 dBc/Hz

Differential current outputs: 8 mA max @ 3.3 V

I'm considering trying this, but then to the clock source.
I have to read about this as I have never ever touched on clock generation, but I have trouble finding any literature on the subject which I'm sure is because I don't even know what search term to use.
Digital clock sources didn't work, any pointers?
I'm looking for some introduction.

Am I about to search for a clock generator IC that produces the frequency I want or can I get a higher or lower clock speed and double/half it or what?
Looking at AD products there seams to be some very good choices regarding jitter but the ones that would be for my application(freq range) is very bad in comparison.