How to generate a high-precision sine source for ad test?

[wateror]

sine source

I need to test a 16bits ADC, but I don't have the proper sine source?
Is there any idea?

 

[Mehdi1357]

precision sine wave generator

hi
There are many different ways to generate a precision sine wave , but by using a DDS (direct digital synthesizer) you can achieve your goals. That’s an effective and inexpensive way.
A basic DDS circuit consists of an electronic controller, a random access memory, a frequency reference (usually a crystal oscillator), a counter and a digital to analog converter (DAC) all these are gathered in a small chip.
I suggest you to use AD9834 and so on, The AD9834 is a 75 MHz DDS device capable of producing high performance sine and triangular outputs. The AD9834 is written to using a 3-wire serial interface. This serial interface is compatible with DSP and microcontroller standards.
See data sheet for more information

 

[dmilosd]

ad9834

I know that I am out of date, but isnt this AD9834 a way too bad to characterize 16bit ADC??? We are talking about ~96db SNR and SFDR (if the adc is good) of smaller than 100dB or I missed something?

Regards,
Milos

 

[Petre Petrov]

bandwidth generating test

Hello!
You should specify at least:
* needed accuracy
* test frequency
* parameters of the ADC to test
* the tests methods and goals
If you wish an accuracy up to approximately 10 bits you could you use the methods in:
**broken link removed**
But there is no easy solution if you need 16 bits accuracy +-0.5LSB
May be the only solution is to use industry standard source from the best manufactures.
It that case you are in need approx 18 bits accuracy of the generated test signals.
BR

 

[rfsystem]

high precision signal generator

If on the edge of technology it could mean a 100MS/s 13-14 ENOB 16bit ADC

Which source generate 15-16bit ENOB at 50MHz ???

I think all ADC's are testet wich high purity generators with less than -50dB spurs which are passive filtered down further. there is no instrument to my knowledge.

But others should prove that I am wrong.

 

[eecs4ever]

high precision adc test

Maybe try cascading multiple bandpass filters after your signal generator.

**broken link removed**

 

[Petre Petrov]

how to generate sin wave

It depends on test goals and available resources.

There are relative cheep solutions, eg.
**broken link removed**
https://www.testequipmentdepot.com/bk-precision/signal-generators/4017b.htm
**broken link removed**
and many others.

I am sure you could find several solutions after you precise your task., but in all cases a test bench for quality testing of 16-bits devices is not an easy and cheep task.

Also pay attention to home made amplifiers and filters.
Usually they are not suitable for testing 16-bits devices.

 

[wateror]

how to adtest

thanks. The bandwidth of my AD is 50kHz. I tried to use a forth order active bandpass filter to filter the output of an sine wave generator (which is only 70dB snr). But it ended up in getting worse distortion then. I am trying to test it with an Agilent ATE system. Its spec shows that it can reach 24bits accuracy. I hope it will work. Thanks all.

 

[FvM]

adtest how to

I tried to use a forth order active bandpass filter to filter the output of an sine wave generator (which is only 70dB snr). But it ended up in getting worse distortion then.

It's necessary to use high quality, high GBW OPs for the filter. Passive filters may achieve low distortion more easily, if you don't worry about manufacturing the inductors by hand.

 

[Petre Petrov]

sine source

Hi!
The term “band width” with ADC has different and sometimes unclear definition with different manufacturer.
The test methods of defining the bandwidth of a “sampler” (ADC) are more or less controversial and not fully explained.
I preferred my own method that states

"If everything else is neglected you could divide the sampling rate Fd at factor of four (4) in order to find the guaranteed bandwidth (-3dB) from your ideal ADC in the worst case sampling of a sine wave without direct current component (DC= 0)."

Example:
With 44.1 kHz sampling rate you have guaranteed bandwidth of approx. 11 kHz.

Also there is a relation between the Signal Sampling Factor N = Fd/Fsmax and the number of the required bits n and this is not fully explained in the publications.

But the subject is complicated and a lot of people will disagree with the written above.
It is really hard to guarantee 16 bits accuracy with 50 kHz sine wave testing especially at low signal levels.
Normally there should be good synchronization between the generator and the sampler and appropriate reference testing system.
Good luck!

 

[FvM]

howto adtest

What you discuss as ADC bandwidth (e.g. Fs/4) is a property of the sampling process rather than the device.

As a device parameter, the term bandwidth is usually designating the analog bandwitdh of the sampler and can be expected by a factor of 5 - 10 higher than Fs. It gets effective in pulse signal acquisition or undersampling digital receiver systems.

 

[wateror]

dds ad9834

Yeah, actually the SNR will drop as the input signal frequency increases. For the bandwidth I mentioned here, it refered to the Nyquist bandwidth.
I have found some precision sine wave generator on the internet (as stanford ds360), but it's expensive for me. I just wanna some economic way to build a circuit with sine signal output.

 

[FvM]

A Nyquist bandwith frequency of 50 kHz implies, that your test frequency has to be considerably lower to see distortion in the ADC output, unless you don't go for a two-tone intermodulation test. At frequencies of e.g. 1 - 10 kHz high end audio OPs provide really low THD values, the said passive filter option may be still more easy, however.

Generally, it's always easier to generate a low distortion fixed frequency by a medium quality generator and a specific filter than trying to get the signal from a variable frequency generator directly. It's true in RF and AF measurements as well.

 

[Petre Petrov]

Re: How to generate a high-precision sine source for ad test

FvM : “What you discuss as ADC bandwidth (e.g. Fs/4) is a property of the sampling process rather than the device.”
PP: Yes You are right. This is for the sampling process and with ideal ADC, almost infinite number of bits, etc. But if you add the reality to this ideal situation usually you are getting something worse.
Of course with synchronization, etc. you could get something better, but I am not sure that this is the situation.

FvM : “As a device parameter, the term bandwidth is usually designating the analog bandwitdh of the sampler and can be expected by a factor of 5 - 10 higher than Fs. “
PP: I am not sure that this is a clear and testable from the users point of view.

FvM : “It gets effective in pulse signal acquisition or under sampling digital receiver systems.”
PP:
“pulse” is highly unclear term for me.
“unders sampling” is also meaningless for me. May be you mean with “signal sampling factor” (SSF) N lower than 2?

Wateror :....”Nyquist bandwidth.....”
PP: I am not sure that there is relevant quotation(s) from the paper(s) of H Nyquist concerning any “bandwidth” applicable to sampling, ADC, DAC and signal processing. I cannot find any understanding from the side of Nyquist to these processes.
May be you mean SSF N=2 and the frequency of the 100% amplitude modulation when SS, CS and SBLS are sampled?
If you have F100% = 50 kHz it is good to try 100 times lower frequency.
Try to calculated try frequency using method with amplitude errors from
http://www.ieindia.org/pdf/88/88ET104.pdf
or from
**broken link removed**
But you should have really flat amplitude characteristic of the tested and testing equipment and a good software. 16 bits means approx +-0.001%?

Wateror : “I just wanna some economic way to build a circuit with sine signal output.”
PP: Again: If you wish an accuracy up to approximately 10-12 bits you could you use the methods in:
**broken link removed**
I have tested most of them many years ago and they words very well.
There are something to change of course
But you should specify the technical requirements of the desired solution.

Wateror :... SNR...
PP: This is really difficult and discussion.

FvM: ... of e.g. 1 - 10 kHz ...
PP: May be it is good first to tests into 250-1000Hz range and after that go to to higher frequency.
But how to guarantee the needed 16 bits accuracy of the test bench?

May be if you contact a manufacturer or independent testing firm that will be good.
There is no enough information about your problem.
I hope that will help

 

[wateror]

Thank you, I should try anyway.

 

[dmilosd]

Re: How to generate a high-precision sine source for ad test

thanks. The bandwidth of my AD is 50kHz. I tried to use a forth order active bandpass filter to filter the output of an sine wave generator (which is only 70dB snr). But it ended up in getting worse distortion then. I am trying to test it with an Agilent ATE system. Its spec shows that it can reach 24bits accuracy. I hope it will work. Thanks all.

Hi,

I also used the 4th order active filter, but got quite good results. I used NS LME49720 and LME49600 in order to maintain THD very low. Perhaps you could try with these components? I measured harmonics at -90dBc and before they were at -65dBc. I also had to play around with many parameters like power supply, termination resistors and so on, but it is achievable!

BTW, I would just like to point out, that SFDR is quite hard to measure as you need really a PURE input sine wave (it is easier to measure SINAD).

Best regards,
Milos