Gaussmeter - hall-effect sensors

[Melodolic]

My gaussmeter project is moving along. I dropped the multimeter idea and I'm
doing one with a PIC. It reads the voltage from a hall-effect sensor through
the ADC, crunches the numbers and puts a gauss value on a 4-digit display.
(The multimeter idea went for a burton last night, when my shiny new 16F876A
sprang into ADC life - I wasn't looking forward to doing deltaV * 320 in my
head, outdoors, anyway!) I've migrated most of the bits, other than the
chip, from the PIC development board to a breadboard.

I'd like to use it to measure very small quantities - milligauss. Can I do
something to the sensor's output to amplify the voltage changes within a
small range, while still presenting 0-5Vdc to the PIC's ADC, with a null
around 2.5V? Per the earlier idea with the multimeter, I'd like to use the
same supply that runs the PIC and LEDs.

Can I do something like: apply a -2.5V DC offset, amplify the resultant,
apply a +2.5V offset, and send that to the ADC? (If my single-rail 5V supply
precludes this, can it be done if that's disregarded?)

Do these sensors have the resolution to match a display that can show 9.999
gauss? If not, what's the best high sensitivity range I can hope for? 99.99?

 

[Bob Eld]

My gaussmeter project is moving along. I dropped the multimeter idea and I'm
doing one with a PIC. It reads the voltage from a hall-effect sensor through
the ADC, crunches the numbers and puts a gauss value on a 4-digit display.
(The multimeter idea went for a burton last night, when my shiny new 16F876A
sprang into ADC life - I wasn't looking forward to doing deltaV * 320 in my
head, outdoors, anyway!) I've migrated most of the bits, other than the
chip, from the PIC development board to a breadboard.

I'd like to use it to measure very small quantities - milligauss. Can I do
something to the sensor's output to amplify the voltage changes within a
small range, while still presenting 0-5Vdc to the PIC's ADC, with a null
around 2.5V? Per the earlier idea with the multimeter, I'd like to use the
same supply that runs the PIC and LEDs.

Can I do something like: apply a -2.5V DC offset, amplify the resultant,
apply a +2.5V offset, and send that to the ADC? (If my single-rail 5V supply
precludes this, can it be done if that's disregarded?)

Do these sensors have the resolution to match a display that can show 9.999
gauss? If not, what's the best high sensitivity range I can hope for? 99.99?

I don't believe you have enough bits in the PIC ADC's to do what you want.
These ADC's are 10 bits and if you use all 10 bits you have one part in 1024
as the maximum resolution. You are asking for one part in 10,000 which at
minimum would take 14 bits. In order to provide head room and less fooling
around with precise analog range, gain and offset, I'd go for a 16 bit ADC.
You can still use a PIC for linearization and display taking two bytes at a
time and processing as a 16 or 14 bit word.
Bob

 

[Melodolic]

I don't believe you have enough bits in the PIC ADC's to do what you
want. These ADC's are 10 bits and if you use all 10 bits you have one
part in 1024 as the maximum resolution. You are asking for one part
in 10,000 which at minimum would take 14 bits.

Good point. They also do some 12-bit ones, but they may be relatively
pricey.

In order to provide
head room and less fooling around with precise analog range, gain and
offset, I'd go for a 16 bit ADC. You can still use a PIC for
linearization and display taking two bytes at a time and processing
as a 16 or 14 bit word.

Okay, I'll look into that a bit more.

Cheers,

 

[Mike]

Good point. They also do some 12-bit ones, but they may be relatively
pricey.



Okay, I'll look into that a bit more.

Cheers,

Take a look at the Analog Devices AD7705/7706. It's pretty easy to use
and has some nice amplifiers built in.

Mike

 

[Melodolic]

Take a look at the Analog Devices AD7705/7706. It's pretty easy to use
and has some nice amplifiers built in.

Been looking at the 7705 datasheet, and I suspect that this is a bit more
than I can chew just now - all that inter-chip serial stuff. Methinks I'll
go for finishing the basic gaussmeter I've got, try it out, and come back to
the increased sensitivity later. Thanks for the pointer, though - A/D is
something I want to learn more about.


I wonder how they work that out.

 

[Mike]

Been looking at the 7705 datasheet, and I suspect that this is a bit more
than I can chew just now - all that inter-chip serial stuff. Methinks I'll
go for finishing the basic gaussmeter I've got, try it out, and come back to
the increased sensitivity later. Thanks for the pointer, though - A/D is
something I want to learn more about.

I used to make cheap little Gauss meters and had pretty good results
with an analog front end driving a 3 1/2 digit DPM. Only 2000 count
tho. They measured -1999 to +1999 Gauss with a 1 Gauss resolution.
After warmup they would drift about +-1 Gauss, the best I could tell.
I later switched to using a PIC and LCD using an LTC1298 dual 12bit
serial A/D because it was cheaper and easier to build. The serial
stuff isn't that hard after you've done it once or twice.
I'm not sure I would attempt a full 4 digit accuracy, Not sure, but I
don't believe standard Hall sensors are that stable.

I wonder how they work that out.

Here's a brief description of how it was calculated with a reference
to the original work.

This is an easy exercise in combinatorials. Suppose the chain is about
two hundred amino acids long; this is , if anything, rather less than
the average length of proteins of all types. Since we have just twenty
possibilities at each place, the number of possibilities is twenty
multiplied by itself some two hundred times. This is conveniently
written 20^200, that is a one followed by 260 zeros!

This number is quite beyond our everyday comprehension. For
comparison, consider the number of fundamental particles (atoms,
speaking loosely) in the entire visible universe, not just in our own
galaxy with its 10^11 stars, but in all the billions of galaxies, out
to the limits of observable space. This number, which is estimated to
be 10^80, is quite paltry by comparison to 10^260. Moreover, we have
only considered a polypeptide chain of a rather modest length. Had we
considered longer ones as well, the figure would have been even more
immense.(Life Itself, its origin and nature, Francis Crick, 1981, pp
51-52)

Mike