High res strain gage inputs

B

Thread Starter

Bill Sturm

I am looking for a ISA or PC/104 card that has 16 to 20 bit resolution for a strain gage or just a +-10VDC signal. Speed needs to be at least 50 samples per second for three channels.

I have several questions:

1. Is it feasible to have > 16bit res for such a small signal source. Each bit would be way under 1 microvolt. It seems that noise would be a limiting factor.

2. Is it generally useful or legitimate to use averaging to increase my effective resolution? Can I get 20 bits from 16 by averaging?

3. Can anyone suggest a good quality analog input card for PC/104 or ISA bus?

Thanks,

Bill Sturm
 
C

Curt Wuollet

> I have several questions:
>
> 1. Is it feasible to have > 16bit res for such a small signal
> source. Each bit would be way under 1 microvolt. It seems that
> noise would be a limiting factor.

Yes, but with qualification. Noise is indeed a problem but can often be averaged out. Whether those bits provide meaningful information is another matter. Keeping even 16 bits of dynamic range implies mechanical conditions that are unusual also. No vibration, etc. And time to settle to that degree of accuracy may make 50 samples per second unrealistic. This is mechanical and electronic settling.

> 2. Is it generally useful or legitimate to use averaging to increase
> my effective resolution? Can I get 20 bits from 16 by averaging?

If it's noise and not valid signal variation that you are averaging out, yes! You can mathmatically derive 20 bits from 16, that doesn't guarantee that it means anything. For a stable DC voltage, it may mean something. for anything dymnamic it is most likely a lie. These are some of the most
misunderstood and misused aspects of data acquisition.

> 3. Can anyone suggest a good quality analog input card for
> PC/104 or ISA bus?

ComputerBoards, National Inst. or Kiethly are all reputable. I use CB because they are somewhat cheaper and seem equivalent on speed, noise, etc.

Regards

cww
 
R

Rodney Stevens

Bill

You need to look closely at your requirements.

Why do you need such high resolution?
What is this resolution in relation to the parameter you are trying to measure?
Is the noise of the parameter your trying to measure going to be greater than the value you are trying to resolve?

16 Bits gives a resolution of 300uV for a +-10 not less than 1uV or do you mean the signal from the strain gauge directly. Usually you would have some signal conditioning on the strain gauge.

What do you mean can you get 20 bits resolution by averaging 16 bits.

http:/www.DataTranslation.com have an isa board. Most a/d board suppliers should have some older style boards also.

Rodney Stevens
CSIRO Minerals
http://minerals.csiro.au

Ph. 61 2 97106701
Fax 61 2 97106789

Elan SS s/e 45/7616
http://sites.netscape.net/rodjohnstevens/homepage
 
N
ComputerBoards www.computerboards.com has a variety of both ISA and PC/104 cards with 16 bit resolution and bi-polar 10Vdc input, however like you stated I'd be concerned about the signal to
noise ratio if dealing with strain gauge inputs. Perhaps a separate device like a load cell indicator would provide the A/D conversion you are looking for with much better signal conditioning circuitry then you would typically find on your average A/D card.

Best Regards,

Nick Pirog
CPU Automation, Inc.
Voice: 978.692.5404
Fax: 978.692.8844
email: [email protected]
web: www.cpuauto.com
 
B
At 09:43 PM 7/19/2000 -0500, you wrote:

>Yes, but with qualification. Noise is indeed a problem but can often be
>averaged out. Whether those bits provide meaningful information is another
>matter. Keeping even 16 bits of dynamic range implies mechanical
>conditions that are unusual also. No vibration, etc. And time to settle to
>that degree of accuracy may make 50 samples per second unrealistic.
>This is mechanical and electronic settling.

The process does not ever settle but it changes slowly. I am planning to oversample at as high of a rate as I can, then average these samples.
Maybe sample at 10Khz and average every 100 samples. This should give me an effective sample rate of 100 Hz.


>ComputerBoards, National Inst. or Kiethly are all reputable. I use CB
>because they are somewhat cheaper and seem equivalent on speed
>noise , etc.

I'm glad that you like Computerboards, they were also mentioned by several others. I had proposed a Computerboards PC/104 DAS16jr to my customer. It looks like a very impressive board.

Bill Sturm
 
OK experts about the 16/20 bits resolution of the proposed card, and about noise.
What puzzles me here from the point of view of the meaning of the information gathered (whatever the nature of the project) is the maner to arrive at relying that information to the reality.
Simple suggestion:
1 . sample at will and filter numericaly the data.
2 . adapt the sampling rate and the order of filtration to match the need of information.

This is the best way to obscure the noise and validate the resolution of the card. It will kind of delay the information: add hysteresis to the measurement. One could lab test, then evaluate the overall hysteresis of the total system.
[email protected]
 
B
I am planning to use a signal conditioner to amplify the signal from the strain gage to +- 10VDC. Dataforth makes some very nice looking
DIN rail mounted signal conditioners.

Is this what you had in mind?

Bill Sturm
 
Top