How to determine multi band sampling frequency?

[Osawa_Odessa]

Hi, could anyone help me. I need to determine sampling frequency for a system including many bands simultaneously. Here are bands with their center frequency and bandwidth. fc1= 117.18MHz and BW = 10MHz; fc2=168.33MHz and BW=20MHz; fc3=66.03MHz and BW=25MHz; fc4 = 106.72MHz and BW= 25MHz. I got stuck here, don't know what sampling frequency or frequency range that i should use to avoid alaising beween bands.

 

[crutschow]

I don't really see how you can simultaneously sample all bands with a single sampler. :???: How would you keep the signals separate after they are sampled?

 

[Osawa_Odessa]

I don't really see how you can simultaneously sample all bands with a single sampler. :???: How would you keep the signals separate after they are sampled?

I think we can seperate them by choosing an appropriate sampling frequency. However i don't know how to figure out the frequency.

 

[crutschow]

I think we can seperate them by choosing an appropriate sampling frequency. However i don't know how to figure out the frequency.

"I think we can separate them" is not a valid engineering approach, no matter how high the sampling frequency. Specify how you would do that first, otherwise it makes little sense to discuss the sampling rate.

 

[Osawa_Odessa]

"I think we can separate them" is not a valid engineering approach, no matter how high the sampling frequency. Specify how you would do that first, otherwise it makes little sense to discuss the sampling rate.

Okay, sorry i am posting this from my phone and cannot attach my reference.
i read "Design and practical implementation of multi RF frequency front ends using direct sampling". They mentioned about a direct RF sampling receiver front end designed to simultaneously receive multiple frequency bands. The front end consists of a LNA, a multiband bandpass filter and an ADC with sampler. They give the example: the frequency bands of interest are centered at the three widely seperated carrier frequencies fc1, fc2 , and fc3. The ADC sampling frequency fs is chosen so that it aliases each of the frequency bands to non-overlapping portions. I think that we can use a similar technique with my signal above( i knew that they are not widely seperated but it may work). With the multiband signal above, i should use each ADC for each band?

 

[Technar]

I think in worst case you will have sum of signals with frequencies below:

fc1 - 107.18MHz, 117.18MHz, 127.18MHz (central and +- BW);
fc2 - 148.33MHz, 168.33MHz, 188.33MHz;
fc3 - 41.03MHz, 66.03MHz, 91.03MHz;
fc4 - 81.72MHz, 106.72MHz, 131.72MHz.

Nearest frequencies are: 106.72MHz and 107.18MHz - you have consider it when will filter frequencies.

Highest fc is 188.33MHz. To analise you mix you have to use samling rate 188.33 MHz x 2 = 376.66 MHz or greater. See theory: https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem

If I understand you right. Give more information about your task. Will you make some DSP or not? Is it some digital radio?

 

[FvM]

Highest fc is 188.33MHz. To analise you mix you have to use samling rate 188.33 MHz x 2 = 376.66 MHz or greater.

Yes. That's the general answer. In addition, the bands are already overlapping. So they can be only separated if special modulation features allow it.

 

[Osawa_Odessa]

Thanks for help.
I need to choose sampling frequency for the ADC in the attached paper.
The two IF down-converted channels are filtered and then undersampled by two on chip ADC (Analog to Digital Converter).
If I understand the paper right, L2, E6, E5, E5a, E5b signals are down-converted with a double stage (two mixers) conversion then use only one ADC to sample them.

Could you suggest me some ideas? I want to use ADC 4 bits, but at presence I can't determine sampling frequency and type of ADC (flash adc, sigma-delta ADC,...).

 

[Osawa_Odessa]

And could you help me to choose ADC specification such as sampling frequency? For example, with fc=168.33MHz and BW=20MHz how can I know if one ADC is appropriate for this?
There are many sampling technique such as nyquist sampling rate, oversampling, undersampling,... what should I choose?
What kind of ADC architectures is appropriate for this?
And can I use the ADC in the paper: A 42 mW 2 GS/s 4-bit flash ADC in 0.18-μm CMOS to sample the signal above?
(I want to use 4 bit ADC instead 8 bit's)

 

[Technar]

And could you help me to choose ADC specification such as sampling frequency? For example, with fc=168.33MHz and BW=20MHz how can I know if one ADC is appropriate for this?
There are many sampling technique such as nyquist sampling rate, oversampling, undersampling,... what should I choose?
What kind of ADC architectures is appropriate for this?
And can I use the ADC in the paper: A 42 mW 2 GS/s 4-bit flash ADC in 0.18-μm CMOS to sample the signal above?
(I want to use 4 bit ADC instead 8 bit's)

You have to use oversampling
Fc is (168.33+20/2)*4
Little later I'll make an illustration why.

You ADC is right, but give concretic model for sure .

 

[Osawa_Odessa]

Thanks Technar, I am eager to know why.
You know I need to find ADC's architecture in paper that is appropriate for this project. However, not knowing sampling frequency is hindering me.

 

[Technar]

Osawa_Odessa I wrote from phone and didn't see your ADC description.

Look at the picture:

On the half (1) I draw signal (above) and what you will see (below) if you will use just Nyquist frequency (or less more). I make not equal frequencies to make aliasing evident.

On the half (2) you can see result of samling with frequency Nyquist * 2 (highest frequency of signal * 4). Aliasing (amplitude-phase depending) will exist but substantively less.

For DSP usually advise make samling frequency "substantively more" than highest signal frequency. In physic "substantively more" is x10.

Resume:
Required sampling rate will depend your required signal to noise ratio. Nyquist frequency allow only detect signal. There is not speech about measurement. Situation worsen by alising. If you have got resourses (memory size and speed of PLD or other) you have to make as high as pissible. If it is not possible that Nyquist * 2 is acceptable frequency. Some compromise.

 

[yuvan]

Technar,
I accept that oversampling does reduce the noise floor (discontinuity) but even proper selection of sampling frequency in undersampling case can yield lesser discontinuity between sample to sample as done in general bandpass sampling. Correct me if I'm wrong.

 

[Technar]

Sorry for desinformation! Undersamling is good variant because we have limited bandpass. Have to choose right frequency. When I will on PC i will learn question and keep on take a part on discussion.

 

[FvM]

I accept that oversampling does reduce the noise floor (discontinuity) but even proper selection of sampling frequency in undersampling case can yield lesser discontinuity between sample to sample as done in general bandpass sampling. Correct me if I'm wrong.

Yes. As a prerequisite, these undersampling schemes usually involve I/Q sampling, so no in-band component get's lost, independent of it's phase. But undersampling requires filters to prevent aliasing of off-band components.

For the present thread, much essential information is still missing.

 

[Technar]

Osawa_Odessa, the aim must be specified.

There are swithable IF LC on the circuit. Will be choosed only one frequency at the one time? If yes, case of narrow band will be (10 - 30 MHz) and can talk about bandpass sampling.

You once told about "many bands simultaneously" and you would like to choose samling frequency to avoid of aliasing spectras. What do you need?

And what is AGS on the circuit?