Author Topic: Load cell (Wheatstone bridge) excitation voltage hurdles  (Read 3490 times)

0 Members and 1 Guest are viewing this topic.

Offline V_KingTopic starter

  • Regular Contributor
  • *
  • Posts: 117
  • Country: gb
Load cell (Wheatstone bridge) excitation voltage hurdles
« on: July 11, 2020, 06:57:20 am »
Hello,

I am playing with my load cells and ran into an issue and wonder if anybody could help with some advice.

I usually use the INA125 load cell amplifier with in-build voltage reference for load cell excitation. For this particular project I want to use 10V to supply to the load cell. The INA125 amplifier can only supply ~10mA, whichis not enough for 350ohm load cell. The data sheet recommends to use TIP29C NPN transistor to increase current. But because I missed the current limit on the amplifier, first PCB I did was without the transistor. Just to get me going, I just connected the load cell excitation wire to the linear voltage regulator from the op-amp rail temporarily until I get the updated PCB. The load cell reading were very good, no unwanted noise.
I used a MMBTA42LT1G transistor as an equivalent for the TIP29C for my new PCB board. Once the board arrived and I connected everything, the first thing I noticed was a low frequency noise which was not there before.   

Am I doing something wrong with the transistor? Adding a 0.1uF capacitor between load cell excitation leads seem to have reduced the noise, but did not remove it completely. That noise was not there when I was using load cells directly driven from the INA125 at 5V or 2.5V. What is more, the PCB with load cell driven of a linear voltage regulator and a few caps seem to be working without any noise issue. That got me thinking whether those precision voltage references for excitation are necessary? I would ideally would like to use a better bandwidth amplifier and this discovery with simple linear voltage regulator got me thinking again. I am not worried about voltage drift, because the load cell is used for short periods of time.

I tried reading the application notes from the usual suspects but I just could not understand the complexity and the need of a number of expensive ICs just to read a load cell. It sounded more like a marketing brochure than advice.

Any input and feedback is much appreciated.

Thanks
« Last Edit: July 11, 2020, 06:59:14 am by V_King »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17429
  • Country: us
  • DavidH
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #1 on: July 11, 2020, 08:50:28 pm »
The usual and better solution is to make a ratiometric conversion which has the advantage of removing excitation noise and drift.  To do this, the reference for the analog-to-digital converter is made the same as the excitation for the load cell, or the reverse.  With this done, there is no particular need for the load cell excitation and analog-to-digital converter reference to be either noise free or low drift; the output from a common integrated fixed voltage regulator like a 7812 is perfectly adequate.

If you use a separate load cell excitation and load cell reference, then not only does this contribute another gain error term and noise source, but it prevents rejection of flicker noise which now must also be accounted for so a low drift and low noise design becomes more important.
« Last Edit: July 12, 2020, 02:21:18 am by David Hess »
 
The following users thanked this post: V_King

Offline srb1954

  • Super Contributor
  • ***
  • Posts: 1125
  • Country: nz
  • Retired Electronics Design Engineer
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #2 on: July 12, 2020, 01:42:04 am »
There is no way an MMBTA42LT1G is an acceptable substitute for a TIP29C.

The TIP29C is in a TO-220 package and is rated for up to 30W dissipation (depending on heatsinking) versus 225mW for the MMBTA42 in its SOT-23 surface mount package.

The TIP29C at 1A continuous collector current has twice the rating of the MMBTA42 at 500mA although the useful collector current range of the MMBTA42 is much less as its current gain drops off severely at collector currents more than 20mA.
 

Offline wraper

  • Supporter
  • ****
  • Posts: 17952
  • Country: lv
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #3 on: July 12, 2020, 01:55:58 am »
There is no way an MMBTA42LT1G is an acceptable substitute for a TIP29C.

The TIP29C is in a TO-220 package and is rated for up to 30W dissipation (depending on heatsinking) versus 225mW for the MMBTA42 in its SOT-23 surface mount package.

The TIP29C at 1A continuous collector current has twice the rating of the MMBTA42 at 500mA although the useful collector current range of the MMBTA42 is much less as its current gain drops off severely at collector currents more than 20mA.
Pretty dumb train of thought about suitability of replacement. He does not replace it in some device he bought. This is a question of suitability for particular application (driving 350 ohm load), not about what is a drop in substitute for TIP29C in general.
IMHO it should work but it's far from the best choices. There is no reason to use 300V transistor. And probably something with a thermal tab/pad should be used depending on how much voltage drops across it.
« Last Edit: July 12, 2020, 02:03:43 am by wraper »
 

Offline wraper

  • Supporter
  • ****
  • Posts: 17952
  • Country: lv
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #4 on: July 12, 2020, 02:09:45 am »
The question is where you connect collector of a transistor? to V+ of INA125? What is V+ voltage?
 

Online mikerj

  • Super Contributor
  • ***
  • Posts: 3383
  • Country: gb
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #5 on: July 12, 2020, 08:33:40 am »
Noise from the reference will be proportional to the reference voltage you use, e.g. at 10V you can expect 4x the noise compared to 2.5v.  At low frequencies you will be fighting against 1/f noise.

However have you ruled out more fundamental issues such as your circuit oscillating?
 

Offline V_KingTopic starter

  • Regular Contributor
  • *
  • Posts: 117
  • Country: gb
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #6 on: July 12, 2020, 03:17:27 pm »
The usual and better solution is to make a ratiometric conversion which has the advantage of removing excitation noise and drift.  To do this, the reference for the analog-to-digital converter is made the same as the excitation for the load cell, or the reverse.  With this done, there is no particular need for the load cell excitation and analog-to-digital converter reference to be either noise free or low drift; the output from a common integrated fixed voltage regulator like a 7812 is perfectly adequate.

If you use a separate load cell excitation and load cell reference, then not only does this contribute another gain error term and noise source, but it prevents rejection of flicker noise which now must also be accounted for so a low drift and low noise design becomes more important.

Thanks. I will keep that in mind.
The ADC uses 2.5Vref, but ADC Vin is +-10V, so I would loose a lot of resolution that way. But I will try such idea in the future.
 

Offline V_KingTopic starter

  • Regular Contributor
  • *
  • Posts: 117
  • Country: gb
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #7 on: July 12, 2020, 03:23:52 pm »

IMHO it should work but it's far from the best choices. There is no reason to use 300V transistor. And probably something with a thermal tab/pad should be used depending on how much voltage drops across it.

My knowledge in analogue circuits is very limited. The analog devices equivalent transistor selection guide advises to match the gain in such circuits and MMBTA42LT1G was in the basic JLCPCB parts list (eg available without surcharges) with the same hfe.

Are there any other practical rules of thumb to try to match transistors?

Thanks


The question is where you connect collector of a transistor? to V+ of INA125? What is V+ voltage?

Yes, the collector is connected to V+, which is +12V with 10uF, 1uF and 0.1uF caps next to MC78L12 V reg.
 

Offline V_KingTopic starter

  • Regular Contributor
  • *
  • Posts: 117
  • Country: gb
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #8 on: July 12, 2020, 03:27:37 pm »
Noise from the reference will be proportional to the reference voltage you use, e.g. at 10V you can expect 4x the noise compared to 2.5v.  At low frequencies you will be fighting against 1/f noise.

However have you ruled out more fundamental issues such as your circuit oscillating?

The v ref is internal in the INA125.

In regards to the circuit oscillation, that is where my knowledge and experience runs out. The noise looks as if the transistor is turned on and off at a very low frequency, instead of following the voltage reference. I did calcs for natural frequencies in uni, but never had a chance to convert that into practical knowledge in the field, because I only work with digital circuits.   
 

Offline wraper

  • Supporter
  • ****
  • Posts: 17952
  • Country: lv
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #9 on: July 12, 2020, 05:47:14 pm »
Yes, the collector is connected to V+, which is +12V with 10uF, 1uF and 0.1uF caps next to MC78L12 V reg.
Very likely this is your problem. You want 10V load cell excitation voltage from 12V V+ on top of that using emitter follower. So INA125 needs to output about 10.7V VrefOut voltage for circuit to work properly. Therefore it's marginal at best. Did you measure actual excitation voltage you get? Try powering the circuit from 15V.
« Last Edit: July 12, 2020, 05:51:04 pm by wraper »
 
The following users thanked this post: V_King

Offline wraper

  • Supporter
  • ****
  • Posts: 17952
  • Country: lv
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #10 on: July 12, 2020, 06:20:19 pm »

IMHO it should work but it's far from the best choices. There is no reason to use 300V transistor. And probably something with a thermal tab/pad should be used depending on how much voltage drops across it.

My knowledge in analogue circuits is very limited. The analog devices equivalent transistor selection guide advises to match the gain in such circuits and MMBTA42LT1G was in the basic JLCPCB parts list (eg available without surcharges) with the same hfe.
Something like MMBT4401, BC817, MMBT2222 is better suited for the circuit. Gain specs of TIP29C and MMBTA42LT1G are specified at very different conditions. Also gain does not matter that much since this is an emitter follower. And to get so low current gain spec for small current transistor you need to dig into outskirts, like high voltage part as MMBTA42LT1G.
Quote
Are there any other practical rules of thumb to try to match transistors?
At least keep rated voltage in the same ballpark. Going for higher voltage will increase price and likely will negatively affect other specs.
« Last Edit: July 12, 2020, 06:21:58 pm by wraper »
 
The following users thanked this post: V_King

Offline V_KingTopic starter

  • Regular Contributor
  • *
  • Posts: 117
  • Country: gb
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #11 on: July 13, 2020, 08:39:39 am »
Thank you, wraper, for all your help. Transistors do seem to be a black art. Maybe an idea for Dave's videos about practical ins-outs of transistors for young players?!


IMHO it should work but it's far from the best choices. There is no reason to use 300V transistor. And probably something with a thermal tab/pad should be used depending on how much voltage drops across it.

My knowledge in analogue circuits is very limited. The analog devices equivalent transistor selection guide advises to match the gain in such circuits and MMBTA42LT1G was in the basic JLCPCB parts list (eg available without surcharges) with the same hfe.
Something like MMBT4401, BC817, MMBT2222 is better suited for the circuit. Gain specs of TIP29C and MMBTA42LT1G are specified at very different conditions. Also gain does not matter that much since this is an emitter follower. And to get so low current gain spec for small current transistor you need to dig into outskirts, like high voltage part as MMBTA42LT1G.
Quote
Are there any other practical rules of thumb to try to match transistors?
At least keep rated voltage in the same ballpark. Going for higher voltage will increase price and likely will negatively affect other specs.

I will order the transistors you are recommending and see if they make any difference. Would it be a good idea to add a small cap of 0.1uF on the output as well? At the moment I added one within load cell wiring and it seem to have reduced the noise levels. But it is more like treating symptoms not causes I guess.

Yes, the collector is connected to V+, which is +12V with 10uF, 1uF and 0.1uF caps next to MC78L12 V reg.
Very likely this is your problem. You want 10V load cell excitation voltage from 12V V+ on top of that using emitter follower. So INA125 needs to output about 10.7V VrefOut voltage for circuit to work properly. Therefore it's marginal at best. Did you measure actual excitation voltage you get? Try powering the circuit from 15V.

The voltage measured with transistor was 10.01V sharp, the overheads seem to be sufficient. The INA125 datasheet recommends min of 1.25V above the reference voltage for correct operation. 

I would like to keep the 12V rail, so I will get some of the transistors you recommend and see if there is any improvement.

 

Online mikerj

  • Super Contributor
  • ***
  • Posts: 3383
  • Country: gb
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #12 on: July 13, 2020, 08:40:34 am »
Noise from the reference will be proportional to the reference voltage you use, e.g. at 10V you can expect 4x the noise compared to 2.5v.  At low frequencies you will be fighting against 1/f noise.

However have you ruled out more fundamental issues such as your circuit oscillating?

The v ref is internal in the INA125.

The noise level from the device will be increased according to the reference multiplier you select.  The datasheet does states this.
 

Offline voltsandjolts

  • Supporter
  • ****
  • Posts: 2549
  • Country: gb
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #13 on: July 13, 2020, 09:24:46 am »
Maybe this is slightly off-topic since you are looking for analog output but the reference input of SD converters can float (above ground).
So, you can use the voltage across the bridge as the reference which makes everything nicely ratiometric.
As a bonus, you can also measure resistance variation of the bridge itself which might be useful for temperature drift compensation, e.g. in silicon strain gauges.
Refering to the attached schematic: Measuring VR2 (versus GND, single-ended) gives ratio of bridge resistance (perhaps temperature dependant) to some temperature stable resistance R2.
Using the bridge itself to measure temperature provides better compensation than using some external temperature sensor (PRT or whatever), particularly when temperature is changing (no thermal lag).
R1 helps with keeping in the common mode range of the input diff amp.
If you add capacitors for filtering, keep the time constants on the measurand and the reference roughly the same.

So, no accurate voltage references required.



« Last Edit: August 09, 2020, 02:01:48 pm by voltsandjolts »
 
The following users thanked this post: V_King, mycroft

Offline V_KingTopic starter

  • Regular Contributor
  • *
  • Posts: 117
  • Country: gb
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #14 on: July 13, 2020, 09:47:11 am »
Maybe this is slightly off-topic since you are looking for analog output but the reference input of SD converters can float (above ground).
So, you can use the voltage across the bridge as the reference which makes everything nicely ratiometric.
As a bonus, you can also measure resistance variation of the bridge itself which might be useful for temperature drift compensation, e.g. in silicon strain gauges.
Measuring VR2 below gives ratio of bridge resistance to some temperature stable resistance R2.
R1 helps with keeping in the common mode range of the input diff amp.
So, no accurate voltage references required.

(Attachment Link)

Thanks for the input.

Following David Hess's comment it got me thinking whether I could use two V-ref outputs from the INA125. The 10V for the load cell excitation and 2.5 to my current ADC.




I just need to understand the implications
 
 

Offline wraper

  • Supporter
  • ****
  • Posts: 17952
  • Country: lv
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #15 on: July 13, 2020, 09:52:42 am »
Following David Hess's comment it got me thinking whether I could use two V-ref outputs from the INA125. The 10V for the load cell excitation and 2.5 to my current ADC.
Note that you need an opamp buffer unless ADC has one built in.
Quote
The voltage measured with transistor was 10.01V sharp, the overheads seem to be sufficient. The INA125 datasheet recommends min of 1.25V above the reference voltage for correct operation.

I would like to keep the 12V rail, so I will get some of the transistors you recommend and see if there is any improvement.
I didn't say you need to keep it this way. But checking if it will solve the problem can be very useful. Also, say, your MC78L12 (4% tolerance) has output voltage 2% on the low side (11.76V). And INA125 is already out of dropout voltage spec.
« Last Edit: July 13, 2020, 10:03:20 am by wraper »
 
The following users thanked this post: V_King

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17429
  • Country: us
  • DavidH
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #16 on: July 14, 2020, 02:44:52 pm »
The usual and better solution is to make a ratiometric conversion which has the advantage of removing excitation noise and drift.  To do this, the reference for the analog-to-digital converter is made the same as the excitation for the load cell, or the reverse.  With this done, there is no particular need for the load cell excitation and analog-to-digital converter reference to be either noise free or low drift; the output from a common integrated fixed voltage regulator like a 7812 is perfectly adequate.

If you use a separate load cell excitation and load cell reference, then not only does this contribute another gain error term and noise source, but it prevents rejection of flicker noise which now must also be accounted for so a low drift and low noise design becomes more important.

Thanks. I will keep that in mind.
The ADC uses 2.5Vref, but ADC Vin is +-10V, so I would loose a lot of resolution that way. But I will try such idea in the future.

The whetstone bridge excitation just needs to use the same reference as the ADC.  Nothing prevents multiplying the 2.5 volt ADC reference to produce a higher excitation voltage, or dividing the excitation voltage to produce a lower voltage reference for the ADC.  This introduces a gain error term but removes the difference in excitation and ADC reference terms and rejects low frequency noise.

Check out Linear Technology application note 43 for the various bridge measurement configurations.  I particularly like the last one shown in figure 4 on page 4 because of simplicity and doing away with the instrumentation amplifier while allowing ratiometric operation and several examples of this configuration are shown later in the application note.

« Last Edit: July 14, 2020, 02:57:14 pm by David Hess »
 
The following users thanked this post: V_King

Offline gogoman

  • Regular Contributor
  • *
  • Posts: 123
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #17 on: August 09, 2020, 04:51:47 am »
hello  David,   can you help me understand the statement:
"If you use a separate load cell excitation and load cell reference, then not only does this contribute another gain error term and noise source, but it prevents rejection of flicker noise which now must also be accounted for so a low drift and low noise design becomes more important.[/i]"

What is the reason that the flicker noise is not  being rejected and is the drift and noise degeneration the result of the load cell reference?   
Is the Ratio-metric  an  improvement over the absolute measurement technique. 
     
thanks gogo
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17429
  • Country: us
  • DavidH
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #18 on: August 09, 2020, 08:30:37 am »
Quote
Is the Ratio-metric  an  improvement over the absolute measurement technique.

A ratiometric measurement is absolutely an improvement over an absolute measurement technique, but it is not always feasible.

"If you use a separate load cell excitation and load cell reference, then not only does this contribute another gain error term and noise source, but it prevents rejection of flicker noise which now must also be accounted for so a low drift and low noise design becomes more important.[/i]"

What is the reason that the flicker noise is not  being rejected and is the drift and noise degeneration the result of the load cell reference?

If the load cell excitation is separate from the analog-to-digital converter reference, then errors from the two references including drift and noise are uncorrelated so they add.  If they both use the same reference source, then drift and noise from the reference are correlated so gain errors cancel out.  In practice this means that using the same reference for excitation and the analog-to-digital converter places much less of a demand on the performance of the reference, so even the drifty and noisy output from say an integrated regulator like a 7815 is sufficient.  Of course there are still noise and drift errors from the circuits between the reference and excitation and reference and analog-to-digital converter, but those errors would exist anyway with separate references.  And the reference is often the largest source of drift and noise anyway.

Flicker noise is especially a problem because it limits performance by increasing at lower frequencies which also makes it difficult to filter, so rejecting it is a big advantage.  It is so named because it literally causes "flicker" in the least significant digit in a system which should otherwise be "noiseless".  People are often used to the last digit flickering but it is the symptom of a real limitation if it varies by more than one quantization level.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 15158
  • Country: de
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #19 on: August 09, 2020, 03:10:33 pm »
The INA125 is not such a good choice for a 350 Ohms bridge, as it is quite noisy.
Also most 350 Ohms bridged don't like 10 V excitation. This would be just too much power and thus significant heating and thermal drift.

For 350 Ohms bridge I would more think about 1 V, maybe 2.5 excitation and an amplifier that is really low noise. There are also ADCs that include bridge amplification that is often better than the INA125.
 

Offline gogoman

  • Regular Contributor
  • *
  • Posts: 123
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #20 on: August 09, 2020, 03:51:49 pm »
Hello David, I appreciate your help, few more questions;

 "ratiometric measurement is absolutely an improvement over an absolute measurement technique"
Can you suggest several usage cases where and absolute measurement technique has the advantage?

"In practice this means that using the same reference for excitation and the analog-to-digital converter places much less of a demand on the performance of the referenc"
Is there a usage case where using a different excitation and ADC reference is an advantage?


thanks 
« Last Edit: August 09, 2020, 04:04:31 pm by gogoman »
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 15158
  • Country: de
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #21 on: August 09, 2020, 04:15:47 pm »
There is hardly a real advantage of a separate reference. However than can be cases when the disadvantages of a separate reference may not be so bad and could be tolerated, e.g. to use an ADC that has an ADC internal reference only.  So would low demands it may be simpler, especially if one does not have access to the ADC ref.
 

Offline gogoman

  • Regular Contributor
  • *
  • Posts: 123
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #22 on: August 09, 2020, 04:34:04 pm »
There is hardly a real advantage of a separate reference. However than can be cases when the disadvantages of a separate reference may not be so bad and could be tolerated, e.g. to use an ADC that has an ADC internal reference only.  So would low demands it may be simpler, especially if one does not have access to the ADC ref.
It's advantages to have the two references from the identical reference source, but there are devices available which allow two separate references. Having limited
experience resulted in me asking  :-// why...   

If the excitation and ADC reference differ, then the common mode volt applied to the ADC is not centered in the ADC Span, resulting a asymmetrical
ADC limit around the common mode voltage, is this statement correct?
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 15158
  • Country: de
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #23 on: August 09, 2020, 05:44:58 pm »
The question of a 2 nd reference has nothing to do with the measurement range. if the used range is in the center mainly depends on the symmetry of the bridge.

With a 2 nd reference one has an additional source of gain drift, as the ratio of the 2 references can change (e.g. with time or temperature).  In the radiometric mode, with only 1 reference the drift of the reference does not have an effect and one can thus even use a very simple one, like a regulator.
The drift of the 2 refs relative to each other essentially only effects the gain, not the zero stability. Here zero is where the 2 nd ref. (ADC) has no direct effect, e.g. the ADC reading zero. This may be different from the mechanical zero.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17429
  • Country: us
  • DavidH
Re: Load cell (Wheatstone bridge) excitation voltage hurdles
« Reply #24 on: August 10, 2020, 12:07:59 am »
Can you suggest several usage cases where and absolute measurement technique has the advantage?

Not all circuit topologies and applications allow for the reference to be shared between the excitation and analog-to-digital converter.  The transducer may have its own internal excitation like an integrated hall device.  Galvanic isolation would mean crossing the isolation barrier twice with the reference and signal.  Not all instrumentation amplifier topologies support ratiometric measurements.

Quote
Is there a usage case where using a different excitation and ADC reference is an advantage?

None occur to me; it removes a major source of error.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf