On 3/7/2015 2:55 AM, rickman wrote:
On 3/6/2015 5:13 PM, Jerry Stuckle wrote:
On 3/6/2015 3:48 PM, rickman wrote:
On 3/6/2015 3:11 PM, Jerry Stuckle wrote:
Plus, DSPs do not look at amplitude. They measure the instantaneous
slope of the signal and store it as a digital value depending on the
number of bits, i.e. 16 bit samples would have 2^15 negative slope
values and 2^15-1 positive slope values (plus zero slope). By
recreating the instantaneous slope that is stored digitally, the DAC
converts the digital signal back to an analog signal.
This is just plain wrong. I'm not sure why you make a distinction
between DSP's [sic] and any other digital device since a DSP is not
needed at all to digitize or compress a signal, but the sample produced
by an ADC *is* the instantaneous value of the signal and not the slope.
If you were to compare adjacent ADC samples and calculate the slope
that would be a form of ADPCM. The DAC in turn converts this
instantaneous value back into analog followed by filtering to remove the
higher frequency images if important.
Once again you are wrong, Rick. Integrating ADCs have been used at
least since the 70's and are much more accurate and noise immune than a
simple level ADC. ADPCM isn't even closely related.
I'm only going to point out your error and then I won't argue with you
further. No one is talking about integrating ADCs. You said, "They
measure the instantaneous slope of the signal and store it as a digital
value". That is not what an integrating ADC does, nor does any other ADC.
The integrating ADC uses the input to charge up a capacitance (the
integrator) for some period of time, then a reference is used to
discharge the "integrated" voltage and the time this takes is measured.
This is *not* measuring the "instantaneous slope" of the input signal.
In fact "integrating" and "instantaneous" are contradictory since
"integrating" takes time and "instantaneous" is... well, instantaneous.
Also I will mention that although integrating ADCs are good for noise
rejection, they are *very* slow and only used in such low sample rate
apps as volt meters and the like. More accurate systems like weight
scales typically use sigma-delta converters for low noise, low power and
high resolution or in the case of and high end audio sigma-delta
converters offer high linearity and low distortion.
I think one reason integrating converters are used in volt meters is
that they can be designed to always display 0 for a 0 input voltage
which is important to consumer confidence.
ADPCM is a form of compression comparing adjacent ADC samples to
calculate the differential of the signal which is the closest thing to
what you are describing by the "instantaneous slope".
Sorry - I used the wrong term. The integration is done by the DAC, to
invert the actions of the ADC.
But no, if you understood ANY calculus, you would understand that
"integrating" and "instantaneous" are not contradictory. But then
"instantaneous" is only a theoretical concept, not possible in the real
world. But the word is still in common usage. I wonder why that is?
ADPCM (Adaptive Differential Pulse Code Modulation) is something
completely different.
Slope ADCs are used because they can more accurately recreate the
waveform. To make it simple - let's see the ADC is sampling at twice
the frequency being sampled, i.e. 10kHz signal and 20kHz sampling rate.
If the sample happens to be at the zero crossing point, your ADC will
show zero volts - IOW, no signal. But a slope detecting ADC will show a
fairly high positive slope on one sample and an equally negative slope
on the next sample. By integrating these, the DAC can closely recreate
the signal because it can estimate the maximum amplitude by the slopes
of the samples. No, it won't be perfect - but it will be a lot closer
than your simple ADC.
Now I know you're going to find all kinds of problems with this example
- but I made the example simple so that even you might be able to
understand it. As you increase the sample rate relative to the
frequency of the signal being sampled, the difference becomes less. But
the slope detecting ADC will always provide a more accurate signal
(until you get to an infinitely small sample anyway). The math is
somewhat complex, and I'm sure beyond anything you could possibly
understand. But it can be proven.
As for them not existing. I guess the whole quarter we spent on ADCs in
my EE classes were wrong then. Of course, this was over 40 years ago.
But I doubt physics has changed in that time.
--
==================
Remove the "x" from my email address
Jerry, AI0K
==================