Page 1 of 1

Maximum bandwidth

PostPosted: Tue Aug 19, 2014 2:09 am
by Rainer
Hi,

the specification shows that it is possible to sample with a maximum of 400MHz when using up to 4 channels.
But on the techspec page you state a maximum input bandwidth of 50MHz for each channel at the interface. So does this mean we can only sample digital signals up to around 50MHz?

I can understand that for 16 channels@100MHz the input bandwidth would be 50MHz. But the sentence sounds like a general limit and not that it goes up with less channels (8 channels@200MHz or 4 channels@400MHz)

I hope someone can clarify this for me. :)

Ciao,
Rainer

Re: Maximum bandwidth

PostPosted: Tue Aug 19, 2014 6:03 pm
by Andy
Hi Rainer,

The input bandwidth is different with the frequency of digital signals. even a 1KHz square signal may contain 100MHz or even 1GHz frequency component. It's decided by the slope of the digital signals.
The input bandwidth indicates 3db cutoff frequency. In simple words, There would be a 3db attenuation for a pure 50MHz sine signal.
According to the Shannon’s sampling theorem, a two times frequency will be needed to rebuild a signal. But this is just a theoretical value. In practice, 4 times would be safe, 8 times would be better.

You can use DSLogic to sample digital signals higher than 50MHz, but this would introduce some phase problems. The sample result can only be used as a reference.

Re: Maximum bandwidth

PostPosted: Wed Aug 20, 2014 3:16 am
by electronic_n00b
Andy wrote:
According to the Shannon’s sampling theorem, a two times frequency will be needed to rebuild a signal. But this is just a theoretical value. In practice, 4 times would be safe, 8 times would be better.

You can use DSLogic to sample digital signals higher than 50MHz, but this would introduce some phase problems. The sample result can only be used as a reference.



Correct me if I'm wrong, I am trying to use simpler language:
You have a max sampling rate of 400 MHz on each of the four channels or 400 million samples per second. For reconstructing a single bit, you need 4-8 samples. This depends on the frequencies which bulid the flank/slope of the bit. Therefore you have to divide the theoritical 400 MHz by 4 or 8 to get reliable measurements.

In other words: I can NOT reliably measure signals on older 100Mbps Ethernet networks, right?

electronic_n00b

Re: Maximum bandwidth

PostPosted: Thu Aug 21, 2014 2:16 am
by Rainer
I wouldn't say that it's not possible. As long as the sampling rate is a multiple of the digital signal to measure it should work up to a 200MHz signals without great timing errors. I created a few pictures to illustrate the sampling. The red lines are the sampling points for 400MHz sampling rate and the red circles are the taken samples and the timeline is in ns.

---------------------------------------------------------------------------------------------------------------
Here is an 100MHz signal shown where the sampling starts after the first rising edge:
Image

The reconstructed signal on the pc looks like this:
Image

---------------------------------------------------------------------------------------------------------------
The same for a 200MHz signal:
Image

The reconstructed signal on the pc looks like this:
Image

---------------------------------------------------------------------------------------------------------------
But as soon as the sampling rate is no multiple of the signal frequency like for a 166MHz signal with 400MHz sampling rate the timing gets large offsets.

Here is an 166MHz signal shown where the sampling starts after the first rising edge:
Image

The reconstructed signal on the pc looks like this:
Image

This clearly shows that a sampling rate which is an multiple of the signal frequency (especially for digital signals) lead to quite good samples. As soon as the sampling rate is no multiple of the signal frequency the sampling produces timing offsets when reconstructing the original signal. To compensate for this you need a much higher sampling rate than the signal frequency.

Ciao,
Rainer

Re: Maximum bandwidth

PostPosted: Thu Aug 21, 2014 3:46 am
by electronic_n00b
Rainer wrote:I wouldn't say that it's not possible. As long as the sampling rate is a multiple of the digital signal to measure it should work up to a 200MHz signals without great timing errors. I created a few pictures to illustrate the sampling.


Aha, I see. The images do not load here, but I got the URLs and now everything is clear.

Thanks for your effort!

electronic_n00b