PicoScope 7 Software
Available on Windows, Mac and Linux
Sample rate confusion
Sample rate confusion
Hi,
I have a 3423 automotive scope and after a bit of reading the help topics I found out why it appeared to have a 5ns sample rate while being specced at 20MS/s.
I'm now staring at screenshots of two similar waveforms at 500ns/div and 10x zoom, one with Sin(x)/x and one linear. The Sin(x) has an effective sample rate of 4.8ns while the linear has 50ns.
My question is, just how accurate/inaccurate is Sin(x)/x interpolation? It seems too good to be true..
I have a 3423 automotive scope and after a bit of reading the help topics I found out why it appeared to have a 5ns sample rate while being specced at 20MS/s.
I'm now staring at screenshots of two similar waveforms at 500ns/div and 10x zoom, one with Sin(x)/x and one linear. The Sin(x) has an effective sample rate of 4.8ns while the linear has 50ns.
My question is, just how accurate/inaccurate is Sin(x)/x interpolation? It seems too good to be true..
- Attachments
-
Re: Sample rate confusion
Hi Iscw,
Apologies for the delay in answering. Coronavirus has had its toll on our ability to respond to posts on the forum.
Your display using Sinx/x interpolation is actually very inaccurate. Unfortunately, Sin/x can't multiply the sample rate.
As our documentation points out, Sinx/x interpolation creates many intermediate values when the number of sample you are you using is not enough to create a smooth curve, but the display resolution that you have is enough to create a smooth curve. This works fine when the points to interpolate are a relatively small part of a waveform and the waveform is clearly defined, as shown by the waveforms I have captured with and without Sinx/x below (I have had to create them in another Scope as I don't have a PicpScope 3423 to hand):
However, what you have captured is either a waveform with too large a gap between sample points, because the sample rate that you have used is too slow for the frequency of the waveform, or because what you have captured doesn't have enough dynamic range. So the captured signal then doesn't resemble a waveform at all (i.e. looks more like low resolution noise) and, because the Sinx/x calculation doesn't have one unique frequency to lock onto, it just adds a ringing effect to what you have captured.
So, this is what has happened in your case, and what you should do is check that:
1/ You have an actual signal going into the Scope input (not just noise).
2/ The frequency of the waveform is no more than 1/5th of the sample rate that you are using (see Sample Rate below).
3/ The Channel is switched to a small enough Input Range to give you a well defined signal when using Linear Interpolation.
You should then find that SInx/x should work as intended.
Sample Rate
In our manual for the PicoScope 3000 series (see Page 2 here: https://www.picotech.com/download/datas ... aSheet.pdf) under 'High Bandwidth, High Sample Rate', we recommend using a sample rate of 4 to 5 times the Signal bandwidth. However, to avoid any signal reconstruction problems, it should be at least greater than 4 times the bandwidth. So, for simplicity, what we recommend now is that the sample rate should be at least 5 times the signal bandwidth.
Regards,
Gerry
Apologies for the delay in answering. Coronavirus has had its toll on our ability to respond to posts on the forum.
Your display using Sinx/x interpolation is actually very inaccurate. Unfortunately, Sin/x can't multiply the sample rate.
As our documentation points out, Sinx/x interpolation creates many intermediate values when the number of sample you are you using is not enough to create a smooth curve, but the display resolution that you have is enough to create a smooth curve. This works fine when the points to interpolate are a relatively small part of a waveform and the waveform is clearly defined, as shown by the waveforms I have captured with and without Sinx/x below (I have had to create them in another Scope as I don't have a PicpScope 3423 to hand):
However, what you have captured is either a waveform with too large a gap between sample points, because the sample rate that you have used is too slow for the frequency of the waveform, or because what you have captured doesn't have enough dynamic range. So the captured signal then doesn't resemble a waveform at all (i.e. looks more like low resolution noise) and, because the Sinx/x calculation doesn't have one unique frequency to lock onto, it just adds a ringing effect to what you have captured.
So, this is what has happened in your case, and what you should do is check that:
1/ You have an actual signal going into the Scope input (not just noise).
2/ The frequency of the waveform is no more than 1/5th of the sample rate that you are using (see Sample Rate below).
3/ The Channel is switched to a small enough Input Range to give you a well defined signal when using Linear Interpolation.
You should then find that SInx/x should work as intended.
Sample Rate
In our manual for the PicoScope 3000 series (see Page 2 here: https://www.picotech.com/download/datas ... aSheet.pdf) under 'High Bandwidth, High Sample Rate', we recommend using a sample rate of 4 to 5 times the Signal bandwidth. However, to avoid any signal reconstruction problems, it should be at least greater than 4 times the bandwidth. So, for simplicity, what we recommend now is that the sample rate should be at least 5 times the signal bandwidth.
Regards,
Gerry
Gerry
Technical Specialist
Technical Specialist