# Effect of oversampling on Signal Band width

Started by August 12, 2009
```Can anyone tell that what is the effect of doing oversampling before pulse
shaping on the transmission bandwidth?

My concern is that as books describe that doing upsampling and then
filtering results in reduction in the Bandwidth but at the same time if we
view it from different angle, we are throwing more samples per second so
so shouldn't Bandwidth also be increased ?

```
```Ali A Nasir wrote:
> Can anyone tell that what is the effect of doing oversampling before pulse
> shaping on the transmission bandwidth?
>
> My concern is that as books describe that doing upsampling and then
> filtering results in reduction in the Bandwidth but at the same time if we
> view it from different angle, we are throwing more samples per second so
> so shouldn't Bandwidth also be increased ?
>
>

That is the point of the upsampling - any filtering
or aliasing artifacts would tend to be more out of band.

How you get back down to the desired bandwidth depends on
things not in evidence in the system. You may just be able
to LPF the output.

--
Les Cargill
```
```"Ali A Nasir" <aliarshad46@hotmail.com> wrote in message
> Can anyone tell that what is the effect of doing oversampling before pulse
> shaping on the transmission bandwidth?
> My concern is that as books describe that doing upsampling and then
> filtering results in reduction in the Bandwidth but at the same time if we
> view it from different angle, we are throwing more samples per second so
> so shouldn't Bandwidth also be increased ?

Actually, from the theoretical viewpoint, increasing the sample rate does
_NOT_ increase the bandwidth, but merely increases the number of
harmonics present in any given band.

The reason is because the theoretical sampling waveform, a Shah function
of multiple delayed Diracians (cue : Airy R Bean?) contains every harmonic
up to infinity, all at the same level.

The bandwidth of the sampled waveform is infinite and is independent of
the sampling frequency. (Sequency?)

Decimating***** reduces the processing workload, but not the bandwidth.

***** Means deleting every 10th, according to Roman tradition, and not
deleting 9 out of every 10!

```
```Computer Man wrote:

> Decimating*****

...

> ***** Means deleting every 10th, according to Roman tradition, and not
> deleting 9 out of every 10!

When it comes to soldiers (legionairs?), the word is not usually "delete".

Jerry
--
Engineering is the art of making what you want from things you can get.
&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
```
```Jerry Avins wrote:
> Computer Man wrote:
>
>
>> Decimating*****
>
>  ...
>
>> ***** Means deleting every 10th, according to Roman tradition, and not
>> deleting 9 out of every 10!
>
> When it comes to soldiers (legionairs?), the word is not usually "delete".
>
> Jerry

I dunno - the "class Spartan" soldier had a "delete" method,
as required by "with your shield or on it."

This method has been deprecated since the abandonment
of Bushido.

--
Les Cargill
```
```Ali A Nasir wrote:
> Can anyone tell that what is the effect of doing oversampling before pulse
> shaping on the transmission bandwidth?
>
> My concern is that as books describe that doing upsampling and then
> filtering results in reduction in the Bandwidth but at the same time if we
> view it from different angle, we are throwing more samples per second so
> so shouldn't Bandwidth also be increased ?
>
>

It would help if you were clearer...

Oversampling implies original sampling.
Upsampling implies increasing the sample rate after sampling has already
been done.
So, which is it?

If you oversample in the original sampling process then that just means
there is bandwidth margin in the sample spectrum - around fs/2 there is
very little energy and this extends for some distance around fs/2
depending on the oversample ratio.

If you upsample, there are a few ways to do this.  It depends on whether
there is a fixed array of samples to start with or whether you're
streaming data.  Fundamentally it's an interpolation process in time.
A common way to do this is to inject zeros between the samples .. thus
increasing the sample rate without doing much else.
Then, lowpass filter so that the repeated spectra are removed and the
zeros are generally filled in with nonzero values.
None of this does anything to the bandwidth of the samples you started with.

Fred
```