ADC Calculations and sampling

terface

Hi John, I'm not going to be much help to you, but I was wondeirng why you need an ADC if you have square waves as an input signal. What do you want to know besides the frequency?

George

Hi,

Yes, you are right! The Input signal could be discountinuous, square wave or any other type of signal. But I still need verification of my calculations.

Thanks John

terface

Actually the Nyquist is the first frequency you can't measure not the last one you can. You can have a sine wave a Nyquist crossing zero exactly at the sample times and thus read zero.

This also doesn't leave you any room for an anti-alias filter.

The LTC2203 or LTC2286 will go fast enough to do 5MHz. How many bits do you need?

That is the peak rate don't expect to maintain that rate for the whole time.

It is sort of correct. You will have some overhead.

Can you actually obtain the MAX 187?

Again expect some overhead.

Nyquist didn't say that.

If you are saying that you have a 5MHz square wave, then you'll have harmonics going well above 5MHz -- the sharper the edges, the higher the harmonics.

A big part of this is that you can't get all the bus bandwidth in one channel -- it's simply not in the standard. You want to do a bit of digging and find out the maximum data rate that you can sustain using isochronous transfers with your high-speed USB.

You never mentioned your precision requirement -- are you just assuming that 12 bits is both necessary and sufficient, or might you get away with less (or absolutely need more)?

nal

mpling.html.

d text -

If the OP wants to measure a square wave why does he need any more than one bit? I'm confused.

George

He explains it in a response somewhere in this thread.

So, if you measure a square wave with one bit, how do you tell the amplitude, the offset, the sharpness of the edges, the exact _timing_ of the edges, etc?

I have managed to get about half the 480Mbps over usb 2.0 using the cypress ez-usb fx2. The gnu software radio project has a lot of info on how to do it. I found that the custom drivers were not actually necessary and the user mode wrapper from cypress will do just fine. In my experience multiple end points don't get you any more bandwidth and isochronous transfers are also a waste of time. However, IO completion ports are an absolute must. The real gotcha is that some usb host controllers will do there best to reorder every single packet you send, while others won't.

Good luck, Jon Pry

Hello,

Sampling Calculations:

I have worked with an ADC from Maxim (MAX 187). =A0This chip can do 75Ksps. So, it might be ok with the input signal of frequency of 40KHz. The Nyquist theorem says that I can not read the frequency higher that (fs / 2). =A0So, the maximum output analog frequency that I can read if 37.5 KHz. Am I correct?

Tim wrote: Nyquist didn't say that.http://www.wescottdesign.com/articles/Sampling/sampling.html.

So, I am wrong about sampling the 40KHz signal with 75Ksps ADC. And the reason is ???

Can anybody suggest the part for the Frequency range # 1? My other question about sampling is that the square wave consists of many frequency harmonics, what harmonic should be chosen for sampling the fundamental, second or third?

Tim wrote:If you are saying that you have a 5MHz square wave, then you'll have harmonics going well above 5MHz -- the sharper the edges, the higher the harmonics.

So , what harmonic should I sample, The highest frequency range one, middle frequency range ??

Data Rate Calculations:

The Data rate of =A0the USB2.0 IS 480Mbps and the MAX 187 is 12 bit / sample ADC at 75Ksps. So, the digital output rate of the MAX187 will be 900,000 bits/sec. Is this correct? Now, if I divide 480 Mbps with 12 bits / sample, then USB2.0 can support 40 Mbps. Is this correct?

I will appreciate of someone suggest some parts and clear few designing principles regarding ADC for me.

Tim wrote: 480Mbps is the absolute maximum raw data transfer rate, not the rate that can be sustained by any one channel, and possibly not (I haven't read the fine print of the USB specification) something that you could ever actually achieve in practice. A big part of this is that you can't get all the bus bandwidth in one channel -- it's simply not in the standard. =A0You want to do a bit of digging and find out the maximum data rate that you can sustain using isochronous transfers with your high-speed USB.

Lets suppose the USB 2.0 has stable 480Mbps. Then are my calculations true??

Tim wrote: You never mentioned your precision requirement -- are you just assuming that 12 bits is both necessary and sufficient, or might you get away with less (or absolutely need more)?

Yes, I am looking for 12 bit precision.

Regards.