Venkat.Vijay.Kumar@gmail.com wrote:

   ...

>        But if you go from basic priciple of sampling that it is the DC
> that samples the input signal and not the clock frequency and it is the
> phase noise around the DC that corrupts the input signal and not the
> phase noise around the clock frequency.

   ...

This is an expression of such profound misunderstanding that I have no 
idea how to address it. I will think about it for a while. For one 
thing, DC has no phase, but that's the least of it.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Hi John,
          Thanks for your response. I agree with your comment that we
may not need other waveform to do sampling. But this basic point is
critical if you are trying to undestand the basic sampling and do
further study on non-idealities.

       For example I am currently looking at the impact of jitter on
sampling. If you look at the PLL spectrum in all text books they show
it as phase noise around the clock frequency. And the impact of this
jitter is very small if you use for sampling (Especially if you have
high over sampling).

       But if you go from basic priciple of sampling that it is the DC
that samples the input signal and not the clock frequency and it is the
phase noise around the DC that corrupts the input signal and not the
phase noise around the clock frequency.

      To understand these kind of things I felt it is essential to
understand the basic process and not to have any assumptions..

Just my openion.

    I have a follow up question on jitter that I will post it as a
seperate question.. Thanks for your help.

Venkat.Vijay.Kumar@gmail.com wrote:
> John, Jerry,
>      Thanks for the excellent discussion. I was frustrated in the
> beginning when I posted the question becuase I don't seem to explain it
> in a very good way and nobody seems to follow it very well what I was
> asking.
> 
>      Just to clarify the answer I was looking for was what John Moore
> gave and later seconded by you.
> 
>      I have done my own bit of reading and it is clear that if you
> sample a sine wave(Fin) with a clock(+1, 0) signal only then you can
> retain the baseband and not when if the clock is (+1, -1). Perhaps it
> is called modulation but the point that you guys clarified, which is
> 
> "IF THE SAMPLING CLOCK HAS NO DC COMPONENT YOUR OUTPUT DOES NOT HAVE
> THE BASEBAND(Fin)"
> 
> It is very frustrating that none of the books (I refered atleast 3
> books that are very popular in analog and digital singal processing
> world). explain this phenomnon and everyone assumes that it is obvious.
> 
> Also another thing none of them clearly distinguishes is the fact that
> "Sample of Hold" operation is different from Simple sampling process
> because Sample and Hold has inband droop (Sinx/x droop as someone
> mentioned) due to the hold operation.  I was confused by the fact that
> sampling clock  (50% duty cycle clock (+1, 0)) also has a Sinx/x droop
> for its harmonics however since the baseband signal (Fin) is sampled by
> DC and not by the fundemental clock frequency or its harmonics there is
> no inband droop.
> 
> Is there a better way of communicating in this forum may be attaching
> some pictures or something that I can use to quickly explain what my
> question is about instead of so confusing so many people??
> 

Vencat,
I suspect the reason the textbooks don't mention the relationship 
between the DC component and the presence of the baseband signal is that 
  this issue only arises when the feasibility of other sampling 
waveforms is investigated.

In analysing the effect of sampling, a description of the sampling 
process: "measure the instantaneous signal value at regular intervals" 
pretty well defines the sampling waveform that is needed.  This gives 
the results we want, so there is no actual need to investigate other 
waveforms. In any case, it is hard to see how any other waveform could 
accomplish this particular task.  Of course it may be an interesting 
exercise to do the investigation, as you have found.

Regarding the sin(x)/x droop, the mere fact that a sample-and -hold 
circuit is present does not introduce sin(x)/x droop.  Once a sample is 
captured it does not matter how long it is held while the DAC does its 
work, as long as the process finishes in time for the next sample. 
Droop does not occur until we translate this instantaneous measurement 
into a broad pulse at the output of the ADC.

Regards,
John

John, Jerry,
     Thanks for the excellent discussion. I was frustrated in the
beginning when I posted the question becuase I don't seem to explain it
in a very good way and nobody seems to follow it very well what I was
asking.

     Just to clarify the answer I was looking for was what John Moore
gave and later seconded by you.

     I have done my own bit of reading and it is clear that if you
sample a sine wave(Fin) with a clock(+1, 0) signal only then you can
retain the baseband and not when if the clock is (+1, -1). Perhaps it
is called modulation but the point that you guys clarified, which is

"IF THE SAMPLING CLOCK HAS NO DC COMPONENT YOUR OUTPUT DOES NOT HAVE
THE BASEBAND(Fin)"

It is very frustrating that none of the books (I refered atleast 3
books that are very popular in analog and digital singal processing
world). explain this phenomnon and everyone assumes that it is obvious.

Also another thing none of them clearly distinguishes is the fact that
"Sample of Hold" operation is different from Simple sampling process
because Sample and Hold has inband droop (Sinx/x droop as someone
mentioned) due to the hold operation.  I was confused by the fact that
sampling clock  (50% duty cycle clock (+1, 0)) also has a Sinx/x droop
for its harmonics however since the baseband signal (Fin) is sampled by
DC and not by the fundemental clock frequency or its harmonics there is
no inband droop.

Is there a better way of communicating in this forum may be attaching
some pictures or something that I can use to quickly explain what my
question is about instead of so confusing so many people??

Mark wrote:
> 
>>In common parlance, a sample has a single value. It may in fact be an
>>average about an instant, or an average leading up to an instant, but it
>>is a single value.
> 
> 
> 
> That is becasue "in common parlance"  sampling commonly means sampling
> ___and quantizing____.

I don't agree. In common parlance, it means sampling and measuring. In 
CCD image sensors, there are as many samplers as pixels. Sometimes the 
samples are quantized to produce a digital data stream, sometimes the 
analog samples are used directly. I suppose one could argue that the 
charge is quantized because it consists of a theoretically countable 
number of electrons, but that's a petty quibble in our context.

We have gone beyond discussing definitions and into the realm of 
consensus usage. That's my doing, and I regret it.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

>
> In common parlance, a sample has a single value. It may in fact be an
> average about an instant, or an average leading up to an instant, but it
> is a single value.


That is becasue "in common parlance"  sampling commonly means sampling
___and quantizing____.

But the pure definition of sampling requires only a switch or chopper
or if you prefer a modulator or a multiplier and a 1 and 0 sampling
signal, nothing more.

I agree this discussion has drifted far off topic of the OPs question.

thanks

Mark

glen herrmannsfeldt wrote:
> Jerry Avins wrote:
> 
> (snip, I wrote)
> 
>>> The distinction between sampling and modulation is complicated,
>>> though, especially if filtering is involved.
> 
> 
>> That's the point. However you want to classify Venkat's process, the 
>> usual expectations of the nature of sampled signals don't apply to its 
>> products, nor do the text-book ways of dealing with them.
> 
> 
>> Communities sometimes declare roads long before they plan to build 
>> them, in order to reserve the paths they take. Sometimes those roads 
>> appear on maps. To accept a balanced modulator as a sampler in the 
>> Shannon sense is to follow one of those phantom roads into a marsh.
> 
> 
> I don't disagree with that.
> 
> But do you still consider it sampling if it averages over a finite time 
> instead of a zero width delta function?   (As all real samplers do.)
> As the time approaches the sample rate is it still sampling?

In common parlance, a sample has a single value. It mat in fact be an 
average about an instant, or an average leading up to an instant, but it 
is a single value. A musician's demo CD is a sample of his work; that is 
not a sample in the signal-processing sense. If a way to express a 
full-period sample number is specified, then it is real sampling. As the 
actual sample width becomes greater than about a tenth of the sample 
period, accurate reconstruction requires frequency compensation.

> Also, even agreeing that a balanced modulator isn't a sampler, before
> it can be used for FM stereo generation the signals must be band 
> limited.  Once they are band limited, all the sampling rules apply.
> 
> If I have the math right,  the output of a stereo multiplexer
> (without any other subcarriers) can be properly sampled it at 76kHz.
> (That is just a little surprising since it goes up to 53kHz, though I
> am not so sure about the 19kHz pilot.)

I think the 76 Kc works because baseband isn't needed. Apparently, both 
L and R can be extracted from the subcarrier and its sidebands.

> If I had a black box that sampled it at the appropriate frequency and 
> then reconstructed it again, you would not be able to tell that the box 
> had sampled it.

But you'd be hard pressed to reconstruct it accurately if the samples 
were too wide.

> I am not so sure about your road analogy, but it seem that there is at 
> least some similarity between the two.

I'm not so sure about it either. Is rang true when I wrote it.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins wrote:

(snip, I wrote)

>> The distinction between sampling and modulation is complicated,
>> though, especially if filtering is involved.

> That's the point. However you want to classify Venkat's process, the 
> usual expectations of the nature of sampled signals don't apply to its 
> products, nor do the text-book ways of dealing with them.

> Communities sometimes declare roads long before they plan to build them, 
> in order to reserve the paths they take. Sometimes those roads appear on 
> maps. To accept a balanced modulator as a sampler in the Shannon sense 
> is to follow one of those phantom roads into a marsh.

I don't disagree with that.

But do you still consider it sampling if it averages over a finite time 
instead of a zero width delta function?   (As all real samplers do.)
As the time approaches the sample rate is it still sampling?

Also, even agreeing that a balanced modulator isn't a sampler, before
it can be used for FM stereo generation the signals must be band 
limited.  Once they are band limited, all the sampling rules apply.

If I have the math right,  the output of a stereo multiplexer
(without any other subcarriers) can be properly sampled it at 76kHz.
(That is just a little surprising since it goes up to 53kHz, though I
am not so sure about the 19kHz pilot.)

If I had a black box that sampled it at the appropriate frequency and 
then reconstructed it again, you would not be able to tell that the box 
had sampled it.

I am not so sure about your road analogy, but it seem that there is at 
least some similarity between the two.

-- glen

glen herrmannsfeldt wrote:
> Jerry Avins wrote:
> (snip)
> 
>> He isn't sampling at all. As I wrote above, he's modulating. Elsewhere 
>> in this thread someone equated chopping and time-division multiplexing 
>> to sampling. These processes create sidebands ad infinitum, but they 
>> are not the same process. (Balanced modulation can be seen as TDM of a 
>> signal and it's inversion.)
> 
> 
> Somehow this reminds me of the claims that the device used to connect
> to a DSL line is not a modem, as it doesn't do any modulation, and
> that the output is digital not analog.  Well, yes, it does modulation,
> especially since it can't be a baseband signal.
> 
> There there was the claim that ethernet is digital, and not analog
> as it is not modulated but manchester encoded.  So I claimed that no,
> it isn't manchester coding but synchronous phase modulation with two 
> possible values for the phase.
> 
> The distinction between sampling and modulation is complicated,
> though, especially if filtering is involved.

That's the point. However you want to classify Venkat's process, the 
usual expectations of the nature of sampled signals don't apply to its 
products, nor do the text-book ways of dealing with them.

Communities sometimes declare roads long before they plan to build them, 
in order to reserve the paths they take. Sometimes those roads appear on 
maps. To accept a balanced modulator as a sampler in the Shannon sense 
is to follow one of those phantom roads into a marsh.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins wrote:
(snip)

> He isn't sampling at all. As I wrote above, he's modulating. Elsewhere 
> in this thread someone equated chopping and time-division multiplexing 
> to sampling. These processes create sidebands ad infinitum, but they are 
> not the same process. (Balanced modulation can be seen as TDM of a 
> signal and it's inversion.)

Somehow this reminds me of the claims that the device used to connect
to a DSL line is not a modem, as it doesn't do any modulation, and
that the output is digital not analog.  Well, yes, it does modulation,
especially since it can't be a baseband signal.

There there was the claim that ethernet is digital, and not analog
as it is not modulated but manchester encoded.  So I claimed that no,
it isn't manchester coding but synchronous phase modulation with two 
possible values for the phase.

The distinction between sampling and modulation is complicated,
though, especially if filtering is involved.


-- glen