Searching \ for '[EE]:A2d and PGA? Theory? Answer!' in subject line. ()
Help us get a faster server
FAQ page: www.piclist.com/techref/io/atod.htm?key=a2d
Search entire site for: 'A2d and PGA? Theory? Answer!'.

Exact match. Not showing close matches.
'[EE]:A2d and PGA? Theory? Answer!'
2000\08\10@102851 by

Thanks to everyone who responded.

I have discussed the theory with the guy who wants me to do this, I will try
to describe
what I will do.

Sorry I got confused yesterday, as Oliver said it is all about increasing
resolution at small signal values.

For simplicity lets say we have a 0 to 10 Volt input.  The PGA has gains of
1,2,4,8 (binary values).  It has two gain selection inputs. 00 for 1, 10 for
2, 01 for 4, 11 for 8.  The A2D has a resolution of 16 bits.  Here is the
theory:

In the software we use the 16 bits from the A2D as bits 0-15, then create
bits 16,17 from the input to the PGA.  E.g. If we are at full scale input
with a gain of one we would have:

1111 1111 1111 1111 00

Once the input signal is less that 5 Volts the MSB becomes zero for ever,
now we only have 15 bits resolution.  With a gain of one we would have:

1111 1111 1111 1110 00

If we set the gain of the PGA to two with the same input we would have:

1111 1111 1111 1111 10                  We have just 'created'
another bit!

Now when we get to 2.5 Volts, the same happens.  So we set the gain to four
and get:

1111 1111 1111 1111 01

At 1.25 Volts set the gain to eight and get:

1111 1111 1111 1111 11                  Effectively 18 bit
resolution at small signals!

Any comments on this technique?

If this does not make sense (I'm sure this is badly described!), feel free
to ask any questions I will do my best to answer them.

BTW

Olin the reason I didn't reply to you question about the speed requirement
is because I had gone home from work. The speed requirement of this is to
measure every 2 ms max.  1 ms would be better.  This is the reason I am
steering way from  Delta Sigma A2Ds, due to them being increasingly noisy at
fast sample rates.  That and the fact that we are measuring DC signals, are
they not better suited to AC (audio etc)?

If you know of on that would do (fast, 18+ bits, +-10V input range) please
let me know as I also have to investigate whether that would be a better
method.

Thanks

Graham North

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]: PIC only [EE]: engineering [OT]: off topic [AD]: advertisements
>        1111 1111 1111 1111 00
>
>Once the input signal is less that 5 Volts the MSB becomes zero for ever,
>now we only have 15 bits resolution.  With a gain of one we would have:
>
>        1111 1111 1111 1110 00

I would want to do a calibration run at each gain setting of the PGA, and
possibly have the software do a fiddle on the value actually sent out, to
minimize non-linearity's at the change over points.

It is also wise to remember that you are not actually getting 18 bit resolution.
You are getting 16 bit resolution on 4 different scales!!!! It is like changing
the selector on a voltmeter to change the voltage range. Unless you are very
careful with how you cross over between PGA gain selections you will get
inconsistencies in your results.

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]: PIC only [EE]: engineering [OT]: off topic [AD]: advertisements
Like one of the first replys says, taking more samples over time effectively
gives you more resolution.
We managed to get 21 bits (+/- 1v to 1uv resolution) using a 16 bit
processor- though the thoroughput was 1sample/sec.
I find it hard to beleive that the linearity will not get affected by level
shifting.( just an opinion)
There are a few audio specific adc out there made by Analog Devices, TI,
Burr-brown, Crystal - maybe Maxim even.
As far as I recall the output was faster than 1ksps.

Regards,

robertf

{Original Message removed}
> From: "Graham North" <graham.northLANDINST.COM>

> Olin the reason I didn't reply to you question about the speed requirement
> is because I had gone home from work. The speed requirement of this is to
> measure every 2 ms max.  1 ms would be better.  This is the reason I am
> steering way from  Delta Sigma A2Ds, due to them being increasingly noisy
at
> fast sample rates.  That and the fact that we are measuring DC signals,
are
> they not better suited to AC (audio etc)?
>
> If you know of on that would do (fast, 18+ bits, +-10V input range) please
> let me know as I also have to investigate whether that would be a better
> method.

I don't know if they're fast enough but Cirrus do some neat A/D up to 24 bit
with built in Instrumenation amplifier, auto-calibration and switchable
input sensitivity. Bear in mind that 24 bits of precision is not the same as
24 bit accuracy, as others have mentioned you can easily get the wrong
result with great precision.

They're the C55** range, if my memory servers me well...

.

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]: PIC only [EE]: engineering [OT]: off topic [AD]: advertisements
Ok,

What you are doing here is increasing the dynamic range of your measurement,
not its precision.

If the PGA is accurrate to better than 1/65536 then you will be getting 16
bits resolution in the A/D (this is a _big_ if).

Relative to the _original_ 10V range, then:

For values in the range 0.0V to 1.25V you effectively have 19 bits of
resolution.
For values in the range 0.0V to 2.5V you effectively have 18 bits of
resolution.
For values in the range 2.5V to 5.0V you effectively have 17 bits of
resolution.
For values in the range 5.0V to 10.V you have 16 bits of resolution.

This can work, if the big ifs are all satisfied.

Bob Ammerman
RAm Systems
(contract devlelopment of high function, high performance, low-level
software)

{Quote hidden}

four
{Quote hidden}

at
> fast sample rates.  That and the fact that we are measuring DC signals,
are
{Quote hidden}

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]: PIC only [EE]: engineering [OT]: off topic [AD]: advertisements
At 03:28 PM 8/10/00 +0100, you wrote:
>Thanks to everyone who responded.
>
>I have discussed the theory with the guy who wants me to do this, I will try
>to describe
>what I will do.

<snip>

The issue with this plan is that the ratio of the gains will not likely be
exactly the values it is supposed to be, and it may not stay put due to
slight drifts in the relative values of the resistors. Say the error in
the PGA is only 0.1%, then there will be +/-65 counts difference in the
output above and below the PGA gain transition.

I don't know your application, but in the control applications I work on
it is more important that the A/D be monotonic (increasing counts) than
that the resolution be high and far more important than the linearity
and accuracy. If the converter is non-monotonic, it can cause reversal
of the control at that point and thus cause continous oscillation at a
low level around that point.

To get the error down to 1/2 LSB would require a match between the
gains of +/- 4ppm. This is very difficult to achieve and it would probably
be better for you to just buy or make a better A/D to begin with.

Best regards,

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Spehro Pefhany --"it's the network..."            "The Journey is the reward"
speffinterlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com
Contributions invited->The AVR-gcc FAQ is at: http://www.bluecollarlinux.com
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]: PIC only [EE]: engineering [OT]: off topic [AD]: advertisements

More... (looser matching)
- Last day of these posts
- In 2000 , 2001 only
- Today
- New search...