Type of ADC and sampling frequency for IF signals.

[Osawa_Odessa]

Hi, I need to choose ADC to sample these IF bands with their bandwidth respectively.
Could you suggest me some type of ADC that is good for this application?
I mean what type of ADC that I should use: Pipelined ADCs, Flash ADCs, Time-interleaved / parallel converter, Delta-Sigma A/D,...
What sampling frequencies I can use?
Thanks for help.

 

[barry]

It all depends on what your requirements are. You've said nothing about resolution, so it's impossible to even begin to suggest an ADC for you. You do know that you need a sampling frequency at least twice your bandwidth, right? Basically, once you've determined your resolution and sampling rate you can look at different types of ADCs. Depending on what your system requirements are, one type may be better or worse.

 

[Osawa_Odessa]

Hi, thanks for the reply!

It all depends on what your requirements are. You've said nothing about resolution, so it's impossible to even begin to suggest an ADC for you.

Okay, I want to use 4 bit ADC.

You do know that you need a sampling frequency at least twice your bandwidth, right?

Yes, but I am confused between Undersampling and Oversampling. Don't know what type should I choose.
Here is the paper in which ADCs that I want to use are mentioned. I am lost here.

 

[barry]

That paper is too scholarly for me, I'm a working engineer

Undersampling in this context means that you are sampling your signal with a lower sample rate than twice your carrier (IF) frequency. However, and this is where it gets tricky, you can sample with this low frequency if you're looking a narrow bandwidth signal (e.g. 94.86 MHz with 6MHz BW). When you under-sample, you get images (aliases) of your sampled signal appearing where they don't really exist, but they LOOK just like the original. In other words, you can get your bandlimited signal around 94.86 MHz +/- 3MHz to appear at 0-6MHz after sampling. I hope this is somewhat clear.

Oversampling is something that is used mostly in audio applications (I think). The reason for this is that if you sample at a much higher frequency than your Nyquist frequency, then the reconstruction filter that you need when you convert your digital back to analog can have less stringent requirements (less expensive!). There may be other reasons for oversampling, but that's all I know.

 

[Osawa_Odessa]

Thanks barry,
I got the paper from IEEE but I don't understand that much.:sad:
Now could you suggest some ideas about ADC such as sampling frequency, adc's architecture?

 

[yuvan]

Hi barry,
In under-sampling we can get back the aliases centered @ DC(or atleast close to DC) & still get proper separation between the aliased bands (by proper selection of sampling frequency) considering the input to be a proper band-limited signal. So, even in this case the re-construction filter can be less stringent. Correct me if I'm wrong.

 

[barry]

Yuvan,

You are partially right; for IF applications what you say is true. But I don't think undersampling is used in audio applications; there's no way to sample a 20KHz bandwidth with less than a 40KHz sampler and then be able to reconstruct 20KHz (especially since DC is close to the band of interest for audio). The point of oversampling is that if your sampling frequency is much higher than your frequency of interest, your reconstruction filter can have a higher cutoff frequency and won't have to be as sharp (i.e., fewer poles) to keep sampling artifacts out of your analog signal.