Searching \ for '[EE] Precision ADC design.' in subject line. ()
Help us get a faster server
Search entire site for: 'Precision ADC design.'.

Exact match. Not showing close matches.
2010\12\02@230327 by

I'm thinking there's a trick I'm missing somewhere.

I'm trying to design a circuit which uses an ADC to measure 0-60V at least to the nearest 0.1V.

The obvious solution is to take a precision reference, and a voltage divider with high stability resistors, and a 12bit ADC (although a 10bit ADC may be close enough).

But this brute-force method seems to get expensive very quickly, especially since I really need this to work from -30 to +50C.

However, in looking at this closer, it appears I might not need that high of stability (or that high of precision resistors).   In a voltage divider situation, if both resistors exibit the same 'temperature coefficient error' (or whatever the correct term really is) at each temperature in the range, then the divider won't change over temperature.

Is it reasonable to expect say metal film resistors from the same manufacturer (and type) to be pretty consistent with how many ppm the resistance drifts at any given temperature?  For instance if they were sitting right next to each other on the board, perhaps with a gob of heatsink adhesive encapsulating both of the units?

The other thing that's bothering me is the lack of a precison reference with extremely low ppm/C rating.   With the 80*C application range I'm talking about, and typical voltage references in the 50ppm/C range, I end up with +-0.4% overall temperature drift.   Or 0.24V at full scale error - which isn't going to work.

Short of building an oven, is there something else I can do to get a decent reference voltage for the ADC?  I haven't seen a voltage-reference equivalent of a TXCO, and even if I did I suspect that it would be horribly priced.   Plus, I'm trying to keep low power here (this product will be used at times in  a solar powered application), and burning power to heat up a voltage reference doesn't sound like low-power to me.

And what else am I likely to be missing?

-forrest

The dreaded Maxim but they do exist. Digikey shows stock.
http://www.maxim-ic.com/datasheet/index.mvp/id/4212

MAX6143
High-Precision Voltage Reference with Temperature Sensor

On 12/2/2010 11:03 PM, Forrest W Christian wrote:
{Quote hidden}

>
For a reference, there's always the lm399h
(http://www.national.com/opf/LM/LM399.html#Overview). I first used it more
than 30 years ago. I always thought they should put a voltage divider
inside the chip to give different precision voltages or at least low
tempco.

Harold

-- FCC Rules Updated Daily at http://www.hallikainen.com - Advertising
opportunities available
Wow.. it's amazing what the correct magic incantation will do at google.

I think I Was trying everything *but* that one.

-forrest

On 12/2/2010 9:17 PM, Charles Craft wrote:
{Quote hidden}

>>

> -----Original Message-----
> From: piclist-bouncesmit.edu [piclist-bouncesmit.edu] On
Behalf
> Of Forrest W Christian
> Sent: 03 December 2010 04:03
> To: Microcontroller discussion list - Public.
> Subject: [EE] Precision ADC design.
>
>
> The other thing that's bothering me is the lack of a precison
reference
> with extremely low ppm/C rating.   With the 80*C application range I'm
> talking about, and typical voltage references in the 50ppm/C range, I
> end up with +-0.4% overall temperature drift.   Or 0.24V at full scale
> error - which isn't going to work.
>
> Short of building an oven, is there something else I can do to get a
> decent reference voltage for the ADC?  I haven't seen a
> voltage-reference equivalent of a TXCO, and even if I did I suspect
that
> it would be horribly priced.   Plus, I'm trying to keep low power here
> (this product will be used at times in  a solar powered application),
> and burning power to heat up a voltage reference doesn't sound like
> low-power to me.
>
Would it be possible to apply some calibration/temperature compensation
to the ADC value?  This could allow you to overcome the initial resistor
accuracy and the tempco of the resistors and reference.

Regards

Mike

=======================================================================
This e-mail is intended for the person it is addressed to only. The
information contained in it may be confidential and/or protected by
law. If you are not the intended recipient of this message, you must
not make any use of this information, or copy or show it to any
received this e-mail, and return the original to us. Any use,
forwarding, printing or copying of this message is strictly prohibited.
No part of this message can be considered a request for goods or
services.
=======================================================================
At 11:03 PM 12/2/2010, you wrote:
>I'm thinking there's a trick I'm missing somewhere.
>
>I'm trying to design a circuit which uses an ADC to measure 0-60V at
>least to the nearest 0.1V.

Why would you want to do that? Are you sure you don't require measurements
over a 0..60V range that don't differ by more than 0.1V when they are, say,
2V from each other? That's a much easier and MUCH cheaper spec to meet.

>The obvious solution is to take a precision reference, and a voltage
>divider with high stability resistors, and a 12bit ADC (although a 10bit

Maybe not good enough if you're mostly looking at voltage differences.

{Quote hidden}

Only what is in the data sheets is guaranteed. If you buy 25ppm/K resistors
then one could be +25 and the other -25. It's 'likely' they will be less,
and 'likely' that one will be relatively similar to the other, but can you
gamble on it? Is your spec real or not?

>The other thing that's bothering me is the lack of a precison reference
>with extremely low ppm/C rating.   With the 80*C application range I'm
>talking about, and typical voltage references in the 50ppm/C range, I
>end up with +-0.4% overall temperature drift.   Or 0.24V at full scale
>error - which isn't going to work.

You can buy 10ppm/K resistors (or a divider network that has a guaranteed
ratio drift, but they tend to be expensive). Also a 10ppm/K reference,
for less than \$10 total. You may also need a precision low-drift op-amp
buffer. 30ppm/K total worst-case leaves you with a 90mV error due to those
causes at -30C (assuming it's calibrated at 20C), probably not quite
good enough to allow for drift and so on. 5ppm/K resistors are several
times more expensive.

Of course, you could add a temperature sensor and a lookup table and
calibrate each unit over temperature in an environmental chamber. That
would be expensive and time consuming, but could be automated. Then
you'd be left with typically smaller errors (initial calibration error,
drift over time, hysteresis etc.)

>Short of building an oven, is there something else I can do to get a
>decent reference voltage for the ADC?  I haven't seen a
>voltage-reference equivalent of a TXCO, and even if I did I suspect that
>it would be horribly priced.

Around \$80 for 0.6ppm/K and 0.01% initial tolerance. That's the equivalent
of a TXCO, not an OXCO.

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

Forrest W Christian wrote:
> I'm trying to design a circuit which uses an ADC to measure 0-60V at
> least to the nearest 0.1V.

So you need 1 part in 600 accuracy.  That means a minimum of a 10 bit A/D,
so in reality you want a 12 bit A/D.

> But this brute-force method seems to get expensive very quickly,
> especially since I really need this to work from -30 to +50C.

10 bits from -30C to +50C is not trivial.

One possibility is firmware temperature compensation.  We're doing that for
a customer now that wants to control something to .2% over a wide
temperature range.  The feedback sensor doesn't just have a single offset
and gain calibration, but a bunch of offsets and gains in a table indexed by
a temperature measurement.  Individual parts can drift with temperature, but
that's OK as long as they do it repeatably.  The downside is that each unit
needs to be calibrated at a few different temperatures in production.  The
production fixture will most likely have a Peltier device in it, and there
will be tens of these fixtures to get reasonable bandwidth.

> However, in looking at this closer, it appears I might not need that
> high of stability (or that high of precision resistors).   In a
> voltage divider situation, if both resistors exibit the same
> 'temperature coefficient error' (or whatever the correct term really
> is) at each temperature in the range, then the divider won't change
> over temperature.

Exactly.  That's why you can get "tracking" resistors at various ratios in
the same package.  I haven't done this in a while, but Vishay used to be big
in that area.  They probably still are.

> Is it reasonable to expect say metal film resistors from the same
> manufacturer (and type) to be pretty consistent with how many ppm the
> resistance drifts at any given temperature?

Sortof.  They will likely track, but there is no guarantee.

> The other thing that's bothering me is the lack of a precison
> reference with extremely low ppm/C rating.   With the 80*C
> application range I'm talking about, and typical voltage references
> in the 50ppm/C range, I end up with +-0.4% overall temperature drift.
> Or 0.24V at full scale error - which isn't going to work.

That's another reason we are doing total unit temperature compensation as I
describe earlier.  All that stuff cancels out.

********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000
On 03-Dec-10 05:03, Forrest W Christian wrote:
> I'm thinking there's a trick I'm missing somewhere.
>
> I'm trying to design a circuit which uses an ADC to measure 0-60V at
> least to the nearest 0.1V.

The easiest way would be to use MCP3421(MCP3425) delta-sigma converter which has internal 2.048V 5ppm reference and two low temperature coefficient resistors for the divider.

You could use standard resistors for the divider and auto-calibrate the division ratio before each measurement but that would add much hardware and software complexity for little money saved.

> The obvious solution is to take a precision reference, and a voltage
> divider with high stability resistors, and a 12bit ADC (although a 10bit
> ADC may be close enough).
>
> But this brute-force method seems to get expensive very quickly,
> especially since I really need this to work from -30 to +50C.

Define expensive?

{Quote hidden}

No, the temperature coefficients between different batches will NOT match, I know this from experience.

It will match in the same batch, some better manufacturers specify this in the datasheet, I remember one specified 5ppm match.

Djula

{Quote hidden}

At 09:20 AM 03/12/2010, you wrote:
>On 03-Dec-10 05:03, Forrest W Christian wrote:
> > I'm thinking there's a trick I'm missing somewhere.
> >
> > I'm trying to design a circuit which uses an ADC to measure 0-60V at
> > least to the nearest 0.1V.
>
>The easiest way would be to use MCP3421(MCP3425) delta-sigma converter
>which has internal 2.048V 5ppm reference and two low temperature
>coefficient resistors for the divider.

Well, except that it's not "5ppm" but rather +/-15ppm/°C "_typical_",
with NO maximum specified. If we are willing to gamble on a 3-sigma
distribution 99.7% of units should be within +/-45ppm/K.

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

More... (looser matching)
- Last day of these posts
- In 2010 , 2011 only
- Today
- New search...